diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index 883849349661dda925879f6e836b17aa9f4a8bdd..3e15036e2c68c7f50ee341d3e4aa61055a972d7a 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": "20260506T000041Z", - "generated_at": "2026-05-06T00:10:47Z", + "snapshot_id": "20260506T060033Z", + "generated_at": "2026-05-06T06:08:42Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ @@ -2086,6 +2086,35 @@ ] } ] + }, + { + "cluster_id": "cluster-44024-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44076, + "issue_numbers": [], + "pr_numbers": [ + 44024, + 44076 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44024, + "right_pr_number": 44076, + "code_similarity": 0.269, + "size_similarity": 0.976, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.492, + "shared_filenames": [], + "shared_file_areas": [] + } + ] } ], "duplicate_issues": [], @@ -2162,6 +2191,15 @@ "target_issue_number": null, "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." }, + { + "cluster_id": "cluster-44024-2", + "canonical_pr_number": 44076, + "duplicate_pr_numbers": [ + 44024 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44024-2 are treated as duplicates because they share soft_similarity evidence." + }, { "cluster_id": "cluster-45081-3", "canonical_pr_number": 45317, diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index 08228e0c6459eb7a482c91db91fcb891f722b07b..b2a480868e3bcef2699f240bca7bfc9785ce4b23 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": "20260506T000041Z", - "generated_at": "2026-05-06T00:10:47Z", + "snapshot_id": "20260506T060033Z", + "generated_at": "2026-05-06T06:08:42Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -26,46 +26,46 @@ "issue:13244|issue:16618", "issue:15980|issue:16618" ], - "prepared_review_unit_hash": "6c3c5e65eff60cc5b20b5ab9c96923baf80e6e48dfb46342fc8a23c331967af0", + "prepared_review_unit_hash": "09c84118ccf2dde47d3770ebc7e6cccace1e296993b699cbf2300384c60f576f", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2476, - "estimated_input_tokens": 619, - "estimated_eval_tokens": 1494 + "serialized_chars": 2477, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2476, - "estimated_input_tokens": 619, - "estimated_eval_tokens": 1494 + "serialized_chars": 2477, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These appear to be three separate issues: a Tapas tokenization mismatch, an authentication/error-message problem for private model downloads, and a generic model-loading failure. I don\u2019t see strong evidence that any pair is the same underlying bug.", - "confidence": 0.87, - "canonical_issue_reason": null, + "summary": "This looks like a mixed cluster, not a duplicate set. Issue 13244 is about Tapas tokenization behavior and is unrelated to the other two. Issues 15980 and 16618 both involve model-loading/download failures, but they describe different concrete problems (private-model auth/error messaging vs. loading bert-base-uncased), so I would not merge them as duplicates.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 16618 is the broadest and most central model-loading report in the cluster, but it still does not match the other items closely enough to serve as a true duplicate anchor.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 16618 is the broadest model-loading report, but it still doesn\u2019t clearly subsume the other two and I would not use it as a duplicate target with confidence.", + "best_issue_reason": "16618 is the best representative only in the loose sense that it is the most general and discussion-heavy model-loading issue; the cluster as a whole is too heterogeneous for a strong canonical duplicate choice.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:13244", "right": "issue:16618", "accept": false, - "reason": "Tapas tokenization behavior and failure to load bert-base-uncased are different code paths and different problem classes." + "reason": "Different problem domains: Tapas tokenization vs. generic model loading. The titles do not indicate the same underlying bug." }, { "left": "issue:15980", "right": "issue:16618", "accept": false, - "reason": "One is specifically about the error message for private-model auth failures; the other is a generic model-loading failure for a public model, so they do not look like the same bug." + "reason": "Both mention model download/loading, but one is specifically a private-model login/error-message issue and the other is a failure to load bert-base-uncased; these are different failure modes." } ] }, @@ -87,7 +87,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:24643", "issue:30277", @@ -135,7 +135,8 @@ "issue:45362", "issue:45464", "issue:45507", - "issue:45685" + "issue:45685", + "issue:45800" ], "soft_pairs": [ "issue:44322|issue:45464", @@ -185,10 +186,10 @@ "issue:44560|issue:45685", "issue:43039|issue:43828", "issue:30277|issue:36331", - "issue:43039|issue:44805", "issue:34928|issue:36010", - "issue:41762|issue:43596", + "issue:43039|issue:44805", "issue:43901|issue:44512", + "issue:41762|issue:43596", "issue:39290|issue:41720", "issue:43039|issue:44560", "issue:43866|issue:44863", @@ -196,8 +197,9 @@ "issue:39290|issue:43531", "issue:41093|issue:43638", "issue:43425|issue:45070", - "issue:43792|issue:45070", + "issue:43596|issue:45800", "issue:45362|issue:45464", + "issue:43792|issue:45070", "issue:43404|issue:43866", "issue:45464|issue:45685", "issue:43541|issue:45685", @@ -206,502 +208,340 @@ "issue:44841|issue:45084", "issue:41093|issue:41762" ], - "prepared_review_unit_hash": "8fcb5656f880ad7789164ec2969facbcd7942c2e8d9e352213b2bb892c2c24ee", + "prepared_review_unit_hash": "ba8241aa3ecd1f24f14f2f526e66fdd33af86c1d9e0d662f64cb7c213489f6ac", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 67, - "serialized_chars": 47121, - "estimated_input_tokens": 11781, - "estimated_eval_tokens": 23818 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 68, + "serialized_chars": 47969, + "estimated_input_tokens": 11993, + "estimated_eval_tokens": 24242 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous. The only clear duplicate family is the DeepSpeed ZeRO-3 \"index 0 is out of bounds for dimension 0 with size 0\" loading bug around Bert/related model init; the rest are separate model-, docs-, or hardware-specific issues.", - "confidence": 0.76, - "canonical_issue_reason": "issue:43638 is the best canonical issue because it captures the shared ZeRO-3 empty-index failure in the broadest way and has the clearest duplicate neighborhood (notably issue:43596, and closely related to issue:41762).", + "summary": "This is not a single duplicate cluster: most items are unrelated, spanning model-specific runtime errors, backend/device issues, trainer API breakages, and docs regressions. Only one soft edge looks like a true duplicate pair (the ZeRO-3/Bert index-0 loading failure).", + "confidence": 0.91, + "canonical_issue_reason": "No single canonical issue; the cluster mixes many unrelated bugs across different models, backends, and documentation changes.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43638 is the strongest representative of the only real duplicate cluster in this set; the other issues are either different bug classes or too loosely related to merge confidently.", + "best_issue_reason": "No single best issue for deduplication; the only clear duplicate is a small sub-pair, while the rest are separate problems.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44322", "right": "issue:45464", "accept": false, - "reason": "Both are Qwen3.5-related, but one is a missing config attribute and the other is a streaming chat/completions failure; different code paths." + "reason": "Different Qwen3.5 chat/runtime failures: missing config field vs streaming API crash." }, { "left": "issue:45237", "right": "issue:45685", "accept": false, - "reason": "Different failures: AMD GPU runtime problem vs MPS histogram op not implemented for Int." + "reason": "Different hardware/backend bugs: AMD GPU failure vs MPS histogram op unsupported." }, { "left": "issue:43716", "right": "issue:45237", "accept": false, - "reason": "Different bugs and model paths; one is an autocast dtype mismatch, the other is GPT-OSS on AMD GPUs." + "reason": "Different model and code path: Mistral-3 image dtype mismatch vs GPT-OSS GPU issue." }, { "left": "issue:43828", "right": "issue:45237", "accept": false, - "reason": "Both mention dtype/GPU behavior, but the concrete failures and model paths differ." + "reason": "Different bug family; autocast dtype mismatch is not the same as GPT-OSS device failure." }, { "left": "issue:45237", "right": "issue:45507", "accept": false, - "reason": "Different model-specific failures; not the same underlying bug." + "reason": "Different root cause: AMD GPU runtime issue vs invalid method call in GraniteMoEHybrid." }, { "left": "issue:44292", "right": "issue:45237", "accept": false, - "reason": "Different models and symptoms; one is NVFP4 runtime error, the other is AMD GPU compatibility." + "reason": "Different model/precision failure vs AMD GPU device-mapping failure." }, { "left": "issue:30277", "right": "issue:35545", "accept": false, - "reason": "DeepSpeed collective mismatch vs ONNX export failure are unrelated." + "reason": "Different failures: DeepSpeed collective mismatch vs ONNX export error." }, { "left": "issue:35545", "right": "issue:42915", "accept": false, - "reason": "Different model/export/runtime issues; no shared concrete code path." + "reason": "Different model/subsystem: ModernBERT export vs Qwen3Moe FP8 load failure." }, { "left": "issue:35545", "right": "issue:41720", "accept": false, - "reason": "ONNX export bug vs device-mapping CUDA assert are unrelated." + "reason": "Different code paths: export-time failure vs auto device mapping assert." }, { "left": "issue:44509", "right": "issue:44512", "accept": false, - "reason": "Both are v5 docs cleanup topics, but they reference different removed docs content and are not the same issue." + "reason": "Both are docs bugs, but they cover different removed features and pages." }, { "left": "issue:35545", "right": "issue:36010", "accept": false, - "reason": "ONNX export issue vs GenerationMixin import error are unrelated." + "reason": "Different bug classes: export failure vs GenerationMixin import break." }, { "left": "issue:43638", "right": "issue:44805", "accept": false, - "reason": "Both are index/mask-shape errors, but the reported failures differ and do not clearly point to one bug." + "reason": "Similar index-error theme, but different model/context and no clear shared root cause." }, { "left": "issue:43596", "right": "issue:44805", "accept": false, - "reason": "Same as above: both are index-related errors, but not enough evidence they are the same underlying defect." + "reason": "Same broad symptom family, but ZeRO-3 Bert load bug vs generic mask-shape mismatch." }, { "left": "issue:44805", "right": "issue:45161", "accept": false, - "reason": "Different failures; one is a mask/index shape mismatch, the other is a MoE tensor-parallel issue." + "reason": "Generic mask/index error vs GPT-OSS tensor-parallel-only failure." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Different training bug: mask-shape mismatch vs Qwen3.5 embedding unpacking." }, { "left": "issue:44661", "right": "issue:44805", "accept": false, - "reason": "Tokenizer mapping/new-model utility bug vs a tensor mask/index mismatch; unrelated." + "reason": "Tokenizer-mapping helper bug is unrelated to the mask-shape index error." }, { "left": "issue:34634", "right": "issue:35545", "accept": false, - "reason": "BarkProcessor voice preset bug is unrelated to ModernBERT ONNX export." + "reason": "Different feature areas: BarkProcessor voice preset vs ModernBERT ONNX export." }, { "left": "issue:30277", "right": "issue:43638", "accept": false, - "reason": "Both involve distributed setups, but the concrete failures are different." + "reason": "Different DeepSpeed failures: collective mismatch vs Bert ZeRO-3 index error." }, { "left": "issue:43596", "right": "issue:43638", "accept": true, - "reason": "Same ZeRO-3 empty-index failure during Bert model initialization/loading; effectively the same bug." + "reason": "Same ZeRO-3/Bert empty-shard index-0 failure; only wording differs." }, { "left": "issue:43975", "right": "issue:44457", "accept": false, - "reason": "Different model/output correctness issues; not the same bug." + "reason": "Different issues: detokenization bug vs LoRA merge/save/load inconsistency." }, { "left": "issue:43039", "right": "issue:45161", "accept": false, - "reason": "Liger kernel cross-entropy routing vs GPT-OSS MoE tensor-parallel failure are unrelated." + "reason": "Liger Kernel cross-entropy routing bug is unrelated to GPT-OSS TP failure." }, { "left": "issue:44292", "right": "issue:44457", "accept": false, - "reason": "Different model/runtime problems; no shared code path evident." + "reason": "Different model/runtime failure vs LoRA weight merge mismatch." }, { "left": "issue:34928", "right": "issue:44805", "accept": false, - "reason": "Activation checkpointing/FSDP tensor-size mismatch is different from the mask/index error." + "reason": "FSDP activation-checkpointing mismatch is not the same as the mask-shape index error." }, { "left": "issue:24643", "right": "issue:30277", "accept": false, - "reason": "DeepSpeed weight-dimension error vs collective mismatch are different failures." + "reason": "Different DeepSpeed bugs: weight dimensionality error vs collective mismatch." }, { "left": "issue:43638", "right": "issue:45161", "accept": false, - "reason": "Bert/ZeRO-3 loading bug vs GPT-OSS MoE TP issue are unrelated." + "reason": "Bert ZeRO-3 loading bug is unrelated to GPT-OSS tensor-parallel failure." }, { "left": "issue:34928", "right": "issue:36331", "accept": false, - "reason": "FSDP checkpointing mismatch vs CustomTrainer compute_loss signature mismatch are unrelated." + "reason": "Activation-checkpointing/FSDP bug is unrelated to CustomTrainer API breakage." }, { "left": "issue:43596", "right": "issue:45161", "accept": false, - "reason": "Different model families and failure modes." + "reason": "Different model and failure mode: Bert ZeRO-3 load error vs GPT-OSS TP issue." }, { "left": "issue:43716", "right": "issue:45507", "accept": false, - "reason": "Mistral dtype mismatch vs GraniteMoEHybrid invalid method are unrelated." + "reason": "Different model/runtime problem vs GraniteMoEHybrid invalid-method bug." }, { "left": "issue:39290", "right": "issue:45081", "accept": false, - "reason": "Gemma3 vLLM config attribute issue vs Mistral tokenizer patch crash are different bugs." + "reason": "Different models and subsystems: Gemma3 config issue vs Mistral tokenizer regex crash." }, { "left": "issue:43828", "right": "issue:45161", "accept": false, - "reason": "Autocast dtype mismatch vs MoE tensor-parallel issue are not the same bug." - }, - { - "left": "issue:43425", - "right": "issue:44292", - "accept": false, - "reason": "Torch version incompatibility vs Qwen NVFP4 runtime error are unrelated." + "reason": "Different runtime bugs: autocast dtype mismatch vs GPT-OSS TP failure." }, { "left": "issue:44918", "right": "issue:45161", "accept": false, - "reason": "Different training/runtime issues; no common concrete defect." - }, - { - "left": "issue:43039", - "right": "issue:43828", - "accept": false, - "reason": "Different problem areas: Liger kernel path vs autocast dtype mismatch." + "reason": "Different issue classes: Qwen3.5 embedding unpacking vs GPT-OSS TP-only failure." }, { "left": "issue:41762", "right": "issue:43638", - "accept": true, - "reason": "Same ZeRO-3 \"index 0 out of bounds\" loading failure; the model differs, but the underlying code-path problem looks the same." + "accept": false, + "reason": "Both are ZeRO-3 load index errors, but Gemma3 and Bert loading paths are distinct." }, { - "left": "issue:43901", + "left": "issue:43827", "right": "issue:44512", "accept": false, - "reason": "Different docs regressions about different deprecated pipeline/run-command references." + "reason": "Both docs regressions, but they refer to different removed v5 topics." }, { - "left": "issue:39290", - "right": "issue:41720", + "left": "issue:43716", + "right": "issue:44560", "accept": false, - "reason": "Different model/runtime failures." + "reason": "Different model/runtime bugs: image dtype mismatch vs video StopIteration." }, { - "left": "issue:43039", - "right": "issue:44560", + "left": "issue:44661", + "right": "issue:44805", "accept": false, - "reason": "Liger kernel cross-entropy routing vs Qwen3-vl video StopIteration are unrelated." + "reason": "Different helper/API bug vs runtime mask-shape mismatch." }, { "left": "issue:43866", "right": "issue:44863", "accept": false, - "reason": "Corrupted checkpoint vs NemotronH load failure are not the same issue." + "reason": "Different model artifacts/issues: Ovis2 checkpoint corruption vs NemotronH checkpoint loading." }, { "left": "issue:43854", "right": "issue:43866", "accept": false, - "reason": "Unit-test model loading problem vs corrupted checkpoint are different causes." + "reason": "Different models and failure modes; unit-test load issue vs corrupted checkpoint." }, { "left": "issue:39290", "right": "issue:43531", "accept": false, - "reason": "Different Qwen/Gemma config issues and different error surfaces." + "reason": "Related sliding-window theme, but different model families and code paths." }, { "left": "issue:41093", "right": "issue:43638", "accept": false, - "reason": "Both are IndexError-style reports, but the mask shape mismatch is not the same as the ZeRO-3 empty-index loading bug." + "reason": "Different errors: mask/tensor shape mismatch vs Bert ZeRO-3 index-0 failure." }, { "left": "issue:43425", "right": "issue:45070", "accept": false, - "reason": "Torch compatibility issue vs pydantic PretrainedConfig field breakage are unrelated." + "reason": "Torch-version incompatibility is unrelated to the pydantic PretrainedConfig field break." }, { - "left": "issue:43792", - "right": "issue:45070", + "left": "issue:43596", + "right": "issue:45800", "accept": false, - "reason": "Whisper loading failure vs pydantic model regression are unrelated." + "reason": "Bert ZeRO-3 index error is unrelated to the torch/transformers version incompatibility." }, { "left": "issue:45362", "right": "issue:45464", "accept": false, - "reason": "Both involve Qwen3.5, but one is a chat crash and the other is streaming inference; not the same defect." - }, - { - "left": "issue:43404", - "right": "issue:43866", - "accept": false, - "reason": "Mistral3 weight-tying bug vs corrupted Ovis2 checkpoint are unrelated." - }, - { - "left": "issue:45464", - "right": "issue:45685", - "accept": false, - "reason": "Qwen3.5 streaming failure vs MPS histogram op limitation are unrelated." - }, - { - "left": "issue:43541", - "right": "issue:45685", - "accept": false, - "reason": "Mixtral torch-dynamo grouped_mm tracing error vs MPS Int histogram limitation are different bugs." - }, - { - "left": "issue:43541", - "right": "issue:43828", - "accept": false, - "reason": "Different model/runtime failures; not mergeable as one bug." - }, - { - "left": "issue:36010", - "right": "issue:39290", - "accept": false, - "reason": "ImportError for GenerationMixin vs Gemma3 vLLM config mismatch are unrelated." - }, - { - "left": "issue:44841", - "right": "issue:45084", - "accept": false, - "reason": "Voxtral processor failure vs template-node compilation error are unrelated." - }, - { - "left": "issue:41762", - "right": "issue:43596", - "accept": false, - "reason": "Very similar Zero-3 loading symptoms, but the models and reported contexts differ enough that I would not merge them without stronger evidence." - }, - { - "left": "issue:43901", - "right": "issue:44512", - "accept": false, - "reason": "Different stale-doc references, even though both are v5 doc cleanup items." - }, - { - "left": "issue:39290", - "right": "issue:41720", - "accept": false, - "reason": "Different code paths and failure modes." - }, - { - "left": "issue:43039", - "right": "issue:44805", - "accept": false, - "reason": "Different bugs: kernel routing vs mask/index mismatch." - }, - { - "left": "issue:34928", - "right": "issue:36010", - "accept": false, - "reason": "FSDP checkpointing mismatch vs import error are unrelated." + "reason": "Same model family, but different errors and surfaces; not enough for one duplicate." }, { - "left": "issue:41762", - "right": "issue:43596", + "left": "issue:43792", + "right": "issue:45070", "accept": false, - "reason": "Same symptom family, but I do not have enough evidence they are the exact same bug across different models." + "reason": "Whisper load failure is unrelated to the PretrainedConfig/pydantic regression." }, { - "left": "issue:43975", - "right": "issue:44457", + "left": "issue:43404", + "right": "issue:43866", "accept": false, - "reason": "Different language/model correctness issues; not the same underlying defect." + "reason": "Different problems: untied lm_head weights vs corrupted checkpoint." }, { - "left": "issue:45237", + "left": "issue:45464", "right": "issue:45685", "accept": false, - "reason": "Different hardware/runtime issues with no shared code path." + "reason": "Different model/API bug vs MPS backend op support failure." }, { "left": "issue:43541", "right": "issue:45685", "accept": false, - "reason": "Different runtime failures." + "reason": "Grouped_mm tracing failure is unrelated to the MPS histogram error." }, { "left": "issue:43541", "right": "issue:43828", "accept": false, - "reason": "Different model and error class." + "reason": "Different runtime bugs: torchdynamo tracing vs autocast dtype mismatch." }, { "left": "issue:36010", "right": "issue:39290", "accept": false, - "reason": "Import path regression vs config attribute regression are unrelated." + "reason": "ImportError for GenerationMixin is unrelated to Gemma3 sliding-window config." }, { "left": "issue:44841", "right": "issue:45084", "accept": false, - "reason": "Processor loading failure vs compiler/template-node failure are not the same bug." - }, - { - "left": "issue:43716", - "right": "issue:45685", - "accept": false, - "reason": "dtype mismatch on one model vs unsupported MPS int histogram on another." - }, - { - "left": "issue:43854", - "right": "issue:43866", - "accept": false, - "reason": "Test loading issue vs corrupted checkpoint are different." - }, - { - "left": "issue:43828", - "right": "issue:44292", - "accept": false, - "reason": "Different models and failure modes." - }, - { - "left": "issue:44560", - "right": "issue:45685", - "accept": false, - "reason": "Video StopIteration vs MPS histogram limitation are unrelated." - }, - { - "left": "issue:43039", - "right": "issue:43828", - "accept": false, - "reason": "Liger kernel cross-entropy path vs autocast dtype mismatch are different issues." - }, - { - "left": "issue:30277", - "right": "issue:36331", - "accept": false, - "reason": "Distributed collective mismatch vs custom trainer API mismatch are unrelated." - }, - { - "left": "issue:43039", - "right": "issue:44560", - "accept": false, - "reason": "Different subsystems and error classes." - }, - { - "left": "issue:43866", - "right": "issue:44863", - "accept": false, - "reason": "Different model-loading failures; one is corrupted checkpoint data, the other is an implementation/loading mismatch." - }, - { - "left": "issue:43854", - "right": "issue:43866", - "accept": false, - "reason": "Unit-test checkpoint loading vs corrupted checkpoint are not the same bug." - }, - { - "left": "issue:39290", - "right": "issue:43531", - "accept": false, - "reason": "Both are Qwen/Gemma config-related, but the actual missing fields and affected code paths differ." + "reason": "Processor failure and template compilation error are different bugs." }, { "left": "issue:41093", - "right": "issue:43638", - "accept": false, - "reason": "Mask-shape/index mismatch is not the same as zero-sized-index during ZeRO-3 init." - }, - { - "left": "issue:43425", - "right": "issue:45070", - "accept": false, - "reason": "Torch version compatibility vs pydantic config field regression are unrelated." - }, - { - "left": "issue:43792", - "right": "issue:45070", - "accept": false, - "reason": "Whisper load failure vs pydantic regression are unrelated." - }, - { - "left": "issue:45362", - "right": "issue:45464", - "accept": false, - "reason": "Both are Qwen3.5, but the reported failures are different enough that they should stay separate." - }, - { - "left": "issue:43404", - "right": "issue:43866", - "accept": false, - "reason": "Different models and failure modes." - }, - { - "left": "issue:45464", - "right": "issue:45685", + "right": "issue:41762", "accept": false, - "reason": "Different runtime subsystems and symptoms." + "reason": "Both are failure reports, but the shape-mismatch error and Gemma3 ZeRO-3 load bug are different roots." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet, and the soft-edge decisions are mostly conservative. The ZeRO-3 empty-index Bert/Gemma3 cluster around 43638 is the only clearly supported duplicate family, while the other pairings are reasonably rejected as different bug classes or too ambiguous. " + "feedback": "The summary is conservative and grounded in the packet. It correctly treats the cluster as mixed rather than one coherent duplicate set, and the only accepted soft edge (43596/43638) is plausibly the same ZeRO-3 Bert index-out-of-bounds bug from the titles. The other soft-edge rejections are suitably cautious given the available evidence." }, "evaluator_used": true, "retried": false, @@ -720,7 +560,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:39186", "issue:39290", @@ -741,7 +581,6 @@ "issue:43716", "issue:43742", "issue:43828", - "issue:43866", "issue:43873", "issue:43950", "issue:43975", @@ -750,7 +589,6 @@ "issue:44292", "issue:44360", "issue:44387", - "issue:44403", "issue:44492", "issue:44512", "issue:44530", @@ -761,14 +599,17 @@ "issue:44928", "issue:44960", "issue:45005", + "issue:45020", "issue:45070", "issue:45161", "issue:45399", "issue:45464", "issue:45507", + "issue:45542", "issue:45632", "issue:45685", - "issue:45698" + "issue:45698", + "issue:45800" ], "soft_pairs": [ "issue:43296|issue:43366", @@ -777,18 +618,20 @@ "issue:43541|issue:45161", "issue:43828|issue:45464", "issue:44291|issue:45070", + "issue:45020|issue:45800", "issue:43366|issue:45464", "issue:43531|issue:45070", "issue:39290|issue:43742", "issue:43541|issue:44589", "issue:41093|issue:43596", "issue:43054|issue:43994", + "issue:45542|issue:45800", "issue:43716|issue:44918", "issue:44918|issue:45685", "issue:45632|issue:45698", "issue:43404|issue:45005", - "issue:43645|issue:45399", "issue:40990|issue:44960", + "issue:43645|issue:45399", "issue:44530|issue:45005", "issue:42915|issue:45399", "issue:44360|issue:44512", @@ -802,241 +645,246 @@ "issue:43644|issue:43950", "issue:44292|issue:45507", "issue:39692|issue:43054", - "issue:43828|issue:44928", - "issue:43866|issue:44403" + "issue:43828|issue:44928" ], - "prepared_review_unit_hash": "d725889cd1093380e45579b603bb3366c3096ad6e5b732d79662ff401b82058a", + "prepared_review_unit_hash": "98256ba92f0f391306da3c6343c126a436c80ae212ff4fdc2d73d8527f944a1a", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 33, - "serialized_chars": 36618, - "estimated_input_tokens": 9155, - "estimated_eval_tokens": 18566 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37550, + "estimated_input_tokens": 9388, + "estimated_eval_tokens": 19032 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a loose similarity cluster of unrelated Transformers/vLLM/model-specific regressions. The only clear duplicate subcluster is the non-persistent-buffer regression around `from_pretrained()` / `register_buffer(persistent=False)` in v5.", - "confidence": 0.86, - "canonical_issue_reason": "issue:43950 is the most specific and complete description of the non-persistent-buffer regression; issue:43644 and issue:44534 are paraphrases of the same bug.", - "canonical_pr_reason": null, - "best_issue_reason": "issue:43950 is the best representative because it clearly states the regression, affected API, and root symptom in one report.", - "best_pr_reason": null, + "summary": "This cluster is mostly a set of unrelated transformer bugs across different models/features. The only clear duplicate-like group is the v5 non-persistent-buffer regression; the rest are too semantically distant to merge.", + "confidence": 0.91, + "canonical_issue_reason": "No single canonical issue is appropriate here: the items span many unrelated bugs, models, and regressions, so there is no global duplicate target.", + "canonical_pr_reason": "No pull requests are present in this cluster, so there is no canonical PR.", + "best_issue_reason": "No global best issue is suitable as a representative; at most, issue 43950 is the best local representative for the non-persistent-buffer regression subgroup.", + "best_pr_reason": "No pull requests are present in this cluster, so there is no best PR.", "soft_edge_verdicts": [ { "left": "issue:43296", "right": "issue:43366", "accept": false, - "reason": "Different bugs: PaddleOCR-VL loading failure in vLLM vs GGUF gpt-oss architecture support." + "reason": "Different problems: PaddleOCR-VL load failure in vLLM vs GGUF architecture support for gpt-oss." }, { "left": "issue:43572", "right": "issue:45070", "accept": false, - "reason": "Both are v5 config regressions, but they affect different fields and code paths." + "reason": "Both are v5 config regressions, but the concrete failures differ: missing pad_token_idx vs pydantic field handling." }, { "left": "issue:44387", "right": "issue:45005", "accept": false, - "reason": "Int4 OOM under quantization is unrelated to tied-weights handling in translation models." + "reason": "Different bug classes: int4 quantization OOM vs tied-weights behavior on translation models." }, { "left": "issue:43541", "right": "issue:45161", "accept": false, - "reason": "Different failures: torch dynamo grouped_mm tracing vs tensor-parallel support for GPT-OSS MoE." + "reason": "Different code paths and models: grouped_mm torch-dynamo tracing error vs GPT-OSS tensor-parallel support." }, { "left": "issue:43828", "right": "issue:45464", "accept": false, - "reason": "Different symptoms and paths: autocast dtype mismatch vs chat/completions streaming inference failure." + "reason": "Both are dtype-related, but one is Phi-tiny-MoE autocast mismatch and the other is Qwen3.5 streaming inference failure." }, { "left": "issue:44291", "right": "issue:45070", "accept": false, - "reason": "`init_empty_weights` TypeError is a distinct initialization bug from the pydantic `PretrainedConfig` regression." + "reason": "Different regressions: unexpected _is_hf_initialized argument vs PretrainedConfig pydantic field breakage." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "General version incompatibility is too broad; remote_code breakage and torch-2.4.1 incompatibility are not the same issue." }, { "left": "issue:43366", "right": "issue:45464", "accept": false, - "reason": "gpt-oss GGUF architecture support is unrelated to Qwen3.5 streaming inference." + "reason": "Different models and symptoms: GGUF gpt-oss support vs Qwen3.5 chat/completions streaming failure." }, { "left": "issue:43531", "right": "issue:45070", "accept": false, - "reason": "Qwen3-MoE sliding_window handling is a different model-specific bug than the pydantic config regression." + "reason": "Qwen3-MoE sliding_window issue is unrelated to the PretrainedConfig pydantic regression." }, { "left": "issue:39290", "right": "issue:43742", "accept": false, - "reason": "Gemma3 sliding_window_pattern absence and MobileLLM key error are unrelated model-loading issues." + "reason": "Different model-loading failures: Gemma3TextConfig missing sliding_window_pattern vs MobileLLM key error." }, { "left": "issue:43541", "right": "issue:44589", "accept": false, - "reason": "Grouped_mm tracing failure is not the same as missing Float8 storage type resolution." + "reason": "Grouped_mm tracing failure and missing Float8 storage are unrelated." }, { "left": "issue:41093", "right": "issue:43596", "accept": false, - "reason": "Mask shape mismatch during training is unrelated to the empty-dimension index error under deepspeed zero3 init." + "reason": "Different runtime errors and mechanisms: mask shape mismatch vs deepspeed zero3/BertModel indexing failure." }, { "left": "issue:43054", "right": "issue:43994", "accept": false, - "reason": "Both mention SigLIP2, but one is about embedding quality and the other about loading/inference correctness." + "reason": "Both involve SigLIP2 quality, but the reports describe different symptoms and no shared concrete failure is clear." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "Undefined tf backend error from tensorboard-only install is unrelated to torch/transformers version incompatibility." }, { "left": "issue:43716", "right": "issue:44918", "accept": false, - "reason": "Mistral-3 image preprocessor dtype mismatch is a different issue from Qwen3.5 embedding unpacking with TRL." + "reason": "Different models and failures: image preprocessor dtype mismatch vs Qwen3.5 embedding unpacking in TRL SFT trainer." }, { "left": "issue:44918", "right": "issue:45685", "accept": false, - "reason": "TRl SFT embedding unpacking bug and MPS histogram implementation error are unrelated." + "reason": "Unpacking input embeddings and MPS histogram on Int are unrelated bugs." }, { "left": "issue:45632", "right": "issue:45698", "accept": false, - "reason": "Both involve loading custom/remote modules, but one is a cache-path collision and the other is wrong module selection after save/load." + "reason": "Both touch remote-code/custom-module loading, but one is a leaf-directory cache collision and the other is wrong module selection after save_pretrained; not clearly the same bug." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Untied lm_head weights in Mistral3 are unrelated to translation-model tied-weight behavior in v5." + "reason": "One is a Mistral3 tied-weight bug affecting image-text generation; the other is a broader translation-model tied-weights issue." }, { - "left": "issue:43645", - "right": "issue:45399", + "left": "issue:40990", + "right": "issue:44960", "accept": false, - "reason": "Custom-model notebook initialization problems are unrelated to flash-attn2 fallback blocking." + "reason": "Different model/benchmark reports: GPT-OSS perplexity vs GLM5 issue." }, { - "left": "issue:40990", - "right": "issue:44960", + "left": "issue:43645", + "right": "issue:45399", "accept": false, - "reason": "High perplexity on gpt-oss-20b and the generic GLM5 issue do not describe the same failure." + "reason": "Notebook custom-model initialization and flash-attn fallback logic are separate issues." }, { "left": "issue:44530", "right": "issue:45005", "accept": false, - "reason": "PagedAttentionCache invalid group type on Qwen3.5 is unrelated to tied-weights regressions." + "reason": "PagedAttentionCache linear_attention crash is unrelated to tied weights on translation models." }, { "left": "issue:42915", "right": "issue:45399", "accept": false, - "reason": "Qwen3Moe fine-grained FP8 failure is not the same as flash-attn fallback gating." + "reason": "Qwen3-MoE FineGrainedFP8Config failure is unrelated to flash-attn fallback blocking." }, { "left": "issue:44360", "right": "issue:44512", "accept": false, - "reason": "DSA indexer discussion and a removed-docs command typo are unrelated." + "reason": "DSA indexer ReLU discussion and a removed docs command are unrelated." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "NemotronH checkpoint loading failure is unrelated to translation tied-weight handling." + "reason": "NemotronH checkpoint loading failure is unrelated to translation-model tied weights." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantization/offloading behavior is unrelated to tied weights on translation models." + "reason": "Quantization/offloading behavior is unrelated to tied-weight handling." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "FSDP 2-D weight error is unrelated to Qwen3.5 embedding unpacking." + "reason": "FSDP 2-D weight error is unrelated to Qwen3.5 embedding unpacking in TRL." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "Cache-strategy typo and removed-docs command issue are different documentation problems." + "reason": "A cache-strategy typo and a removed run-command doc issue are not the same underlying change." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "TRL SFT embedding issue and GraniteMoEHybrid invalid method call are different code paths." + "reason": "Qwen3.5 trainer failure and GraniteMoEHybrid invalid method are different code paths." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "Wrong detokenization and Qwen-3-8B-NVFP4 runtime failure are unrelated." + "reason": "DeepSeek detokenization regression and Qwen-3-8B-NVFP4 loading error are unrelated." }, { "left": "issue:43950", "right": "issue:44534", "accept": true, - "reason": "Same underlying regression: v5 mishandles non-persistent buffers during `from_pretrained()` / model init." + "reason": "Same regression class: v5 mishandling of non-persistent buffers/register_buffer(persistent=False) leading to junk values." }, { "left": "issue:43644", "right": "issue:43950", "accept": true, - "reason": "Both report the same non-persistent-buffer corruption bug in Transformers v5." + "reason": "Same underlying bug: Transformers 5.x corrupts non-persistent buffers during loading/initialization." }, { "left": "issue:44292", "right": "issue:45507", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime failure and GraniteMoEHybrid invalid method are unrelated." + "reason": "Different model-specific failures: Qwen-3-8B-NVFP4 runtime error vs GraniteMoEHybrid invalid method." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "SigLIP2 docs example errors and SigLIP2 embedding quality regression are distinct issues." + "reason": "SigLIP2 appears in both, but one is a documentation example with processor/model mismatch and quantization failure, not the same as embedding-quality degradation." }, { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Autocast dtype mismatch is not the same as the RLHF NaN / SDPA fallback instability." - }, - { - "left": "issue:43866", - "right": "issue:44403", - "accept": false, - "reason": "Ovis2 checkpoint corruption is unrelated to generic loading noise." + "reason": "Different bugs: autocast dtype mismatch on Phi-tiny-MoE vs RLHF NaNs from Qwen3.5 position_ids/SDPA fallback." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative overall. The duplicate cluster around non-persistent buffers (`43644`/`43950`/`44534`) is well supported, and the other soft-edge false verdicts are phrased narrowly enough to avoid overclaiming. No obvious evidence overstretching or mergeability issue appears here." + "feedback": "Grounded overall: the cluster is mostly heterogeneous, there are no PRs in the packet, and the only clearly duplicate subgroup is the non-persistent-buffer v5 regression (43644/43950/44534). The soft-edge verdicts are conservative and do not overclaim equivalence where the packet only shows superficial similarity." }, "evaluator_used": true, "retried": false, @@ -1055,11 +903,12 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:36331", "issue:38175", "issue:41553", + "issue:41628", "issue:41762", "issue:42491", "issue:42915", @@ -1074,7 +923,7 @@ "issue:43792", "issue:43824", "issue:43827", - "issue:43854", + "issue:43828", "issue:43856", "issue:43866", "issue:43872", @@ -1093,7 +942,6 @@ "issue:44509", "issue:44589", "issue:44661", - "issue:44863", "issue:44960", "issue:44991", "issue:45020", @@ -1103,19 +951,23 @@ "issue:45356", "issue:45399", "issue:45520", - "issue:45701" + "issue:45701", + "issue:45800" ], "soft_pairs": [ + "issue:43866|issue:44403", "issue:36331|issue:41762", + "issue:41628|issue:45800", "issue:41553|issue:43883", "issue:43872|issue:44589", "issue:38175|issue:43054", "issue:41553|issue:43881", + "issue:43828|issue:45800", "issue:43054|issue:45701", "issue:42947|issue:44387", "issue:44403|issue:44451", - "issue:43881|issue:43883", "issue:43643|issue:45020", + "issue:43881|issue:43883", "issue:43054|issue:43493", "issue:43975|issue:45520", "issue:43975|issue:45237", @@ -1129,8 +981,8 @@ "issue:44991|issue:45399", "issue:45020|issue:45356", "issue:43782|issue:44960", - "issue:45081|issue:45399", "issue:43827|issue:44509", + "issue:45081|issue:45399", "issue:43792|issue:44220", "issue:43404|issue:43950", "issue:43742|issue:45399", @@ -1138,258 +990,268 @@ "issue:43257|issue:43866", "issue:43742|issue:45020", "issue:41553|issue:44991", - "issue:43872|issue:44291", "issue:43940|issue:45313", - "issue:42915|issue:43872", - "issue:43854|issue:44863" + "issue:43872|issue:44291" ], - "prepared_review_unit_hash": "1c21cad5a86a4b276512344ff3d1c698bcc382fa3d387ea35278aa0b653d6e00", + "prepared_review_unit_hash": "44a60c011a8406f58640a6c7569be07f9741c3e805cec9c8c949eccae19643e3", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 36, - "serialized_chars": 37861, - "estimated_input_tokens": 9466, - "estimated_eval_tokens": 19188 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 37, + "serialized_chars": 38718, + "estimated_input_tokens": 9680, + "estimated_eval_tokens": 19616 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly distinct model-loading/regression reports. The only clear soft duplicates here are the `_is_hf_initialized` TypeError pair and the v5 docs cleanup pair; the SigLIP2 pair is related but not proven to share the same bug/root cause and should stay separate.", + "summary": "Mostly a mixed cluster of unrelated transformers issues. A few soft pairs look like true duplicates around remote_code config loading, docs for removed pipeline tasks, Whisper fbank extraction, bitsandbytes/init_empty_weights, and SigLIP2 regression reports; the rest are only broad subsystem matches and should stay separate.", "confidence": 0.74, - "canonical_issue_reason": "issue:45020 is the best umbrella because it is the broad, open report about recent Transformers versions breaking `remote_code`-based model loading, which matches the cluster\u2019s overall regression theme better than the model-specific bugs.", + "canonical_issue_reason": "45020 is the best umbrella among the related reports: it is open, broad, and captures the remote_code/loading regression subgroup better than the more specific issues.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the strongest single issue candidate: open, broad, and aligned with the recurring loading-regression pattern across several reports in the cluster.", + "best_issue_reason": "45020 is the strongest triage anchor for this cluster because it is open and broad enough to absorb several of the genuine duplicate-like reports.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "Different bugs: corrupted checkpoint vs generic loading noise." + }, { "left": "issue:36331", "right": "issue:41762", "accept": false, - "reason": "Different failures: custom trainer kwarg mismatch vs DeepSpeed/Gemma3 load-time index error." + "reason": "CustomTrainer loss-arg regression and Gemma3 ZeRO-3 loading failure are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:45800", + "accept": false, + "reason": "AutoImageProcessor import failure is unrelated to torch/transformers version incompatibility." }, { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "Voxtral tokenizer error-message issue vs Molmo tied-weights attribute error; different code paths." + "reason": "Bad Voxtral tokenizer error message vs Molmo tied-weights attribute error are different code paths." }, { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "Same general TypeError theme, but the root causes differ: `_is_hf_initialized` vs missing `Float8_e4m3fnStorage`." + "reason": "Both are loading/type errors, but bitsandbytes _is_hf_initialized and Float8 storage issues are distinct failures." }, { "left": "issue:38175", "right": "issue:43054", - "accept": false, - "reason": "Both mention SigLIP2, but one is zero probabilities and the other is embedding quality; not clearly the same bug." + "accept": true, + "reason": "Both report SigLIP2 output degradation/regression; likely the same underlying implementation issue." }, { "left": "issue:41553", "right": "issue:43881", "accept": false, - "reason": "Unrelated model-loading problems for Voxtral and GLM-4v-9b." + "reason": "Voxtral tokenizer messaging issue and glm-4v loading failure are not the same bug." + }, + { + "left": "issue:43828", + "right": "issue:45800", + "accept": false, + "reason": "autocast dtype mismatch is not the same as the torch 2.4.1 / transformers 5.8 incompatibility report." }, { "left": "issue:43054", "right": "issue:45701", "accept": false, - "reason": "SigLIP2 embedding quality vs broad tokenization-version changes; no evidence of the same concrete defect." + "reason": "Specific SigLIP2 embedding regression vs generic tokenization changes; too broad to merge." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "Different symptoms and subsystems: LoRA gradient checkpointing vs int4 CUDA reserved-memory OOM." + "reason": "Different memory problems: LoRA checkpointing ineffectiveness vs int4 CUDA reserved-memory growth." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "A broad loading-noise complaint vs a specific model load failure; no clear shared bug." - }, - { - "left": "issue:43881", - "right": "issue:43883", - "accept": false, - "reason": "GLM loading failure vs Molmo tied-weights attribute error; different problems." + "reason": "Generic loading noise and a specific ScandiBERT load failure are not the same underlying issue." }, { "left": "issue:43643", "right": "issue:45020", + "accept": true, + "reason": "Both concern remote_code / trust_remote_code breakage in recent transformers versions." + }, + { + "left": "issue:43881", + "right": "issue:43883", "accept": false, - "reason": "Related theme, but `trust_remote_code` missing fields is not the same concrete bug as the broader remote_code regression report." + "reason": "glm-4v loading failure and Molmo tied-weights attribute error are different models and failures." }, { "left": "issue:43054", "right": "issue:43493", - "accept": false, - "reason": "Related SigLIP2 territory, but the available evidence is too thin to prove the same root cause or duplicate bug." + "accept": true, + "reason": "Both point to the same SigLIP2 HF-vs-original implementation discrepancy." }, { "left": "issue:43975", "right": "issue:45520", "accept": false, - "reason": "Tokenizer detokenization regression vs Python 3.13 flash-attn import KeyError; unrelated." + "reason": "Detokenization bug and flash_attn import/keyerror are unrelated." }, { "left": "issue:43975", "right": "issue:45237", "accept": false, - "reason": "Tokenizer/detokenization bug vs AMD GPU runtime incompatibility; unrelated." + "reason": "Detokenization bug and AMD GPU runtime failure are unrelated." }, { "left": "issue:44403", "right": "issue:44488", "accept": false, - "reason": "Different model-loading reports with no evidence of the same underlying defect." + "reason": "Generic loading noise and a specific model load failure are not duplicates." }, { "left": "issue:42947", "right": "issue:43856", "accept": false, - "reason": "LoRA gradient checkpointing vs Qwen3 MoE memory usage; same broad area, different bug." + "reason": "Both mention memory, but LoRA checkpointing vs Qwen3 MoE training inefficiency are different bugs." }, { "left": "issue:42491", "right": "issue:44368", "accept": false, - "reason": "Qwen3 MoE compatibility break vs a weight-tying warning; not the same failure." + "reason": "Cross-version Qwen3_moe incompatibility is not the same as a tie_word_embeddings warning." }, { "left": "issue:41762", "right": "issue:42915", "accept": false, - "reason": "Different DeepSpeed load failures on different model/config combinations." + "reason": "Both are ZeRO/deepspeed-related, but Gemma3 missing-index and Qwen3 MoE weight-conversion failures are different issues." }, { "left": "issue:43975", "right": "issue:44451", "accept": false, - "reason": "Tokenizer detokenization bug vs a specific model load failure; unrelated." + "reason": "Detokenization bug and model loading failure are not the same underlying change." }, { "left": "issue:44661", "right": "issue:45020", "accept": false, - "reason": "Model-registration helper failure vs remote_code loading regressions; not the same issue." + "reason": "Tokenizer-mapping add-new-model-like bug is unrelated to remote_code loading regressions." }, { "left": "issue:43824", "right": "issue:43872", "accept": false, - "reason": "Missing import for a model class vs bitsandbytes `Int8Params` TypeError; different defects." + "reason": "Qwen2.5-VL import failure and bitsandbytes storage/type error are unrelated." }, { "left": "issue:44991", "right": "issue:45399", "accept": false, - "reason": "Tokenizer load failure vs flash-attn fallback gating; unrelated." + "reason": "Tokenizer load regression and flash-attn fallback blocking are different subsystems." }, { "left": "issue:45020", "right": "issue:45356", "accept": false, - "reason": "Broad remote_code regression report vs Kimi tokenizer codec regression; different concrete bugs." + "reason": "Remote_code regressions and Kimi tokenizer codec/fix_mistral_regex regression are different bugs." }, { "left": "issue:43782", "right": "issue:44960", "accept": false, - "reason": "Qwen3VL `weight_only` loading error vs GLM5 issue; unrelated." - }, - { - "left": "issue:45081", - "right": "issue:45399", - "accept": false, - "reason": "Mistral regex patch crash vs flash-attn fallback blocking; different code paths." + "reason": "Qwen3VL weight_only loading error is unrelated to GLM5." }, { "left": "issue:43827", "right": "issue:44509", "accept": true, - "reason": "Both are the same docs drift: lingering references to removed v5 pipeline tasks." + "reason": "Both report stale docs still referencing pipeline tasks removed in v5." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Mistral regex crash and flash-attn fallback gating are unrelated." }, { "left": "issue:43792", "right": "issue:44220", - "accept": false, - "reason": "Both concern Whisper/audio loading, but one is a generic runtime failure and the other a specific fbank helper issue; not enough to call duplicates." + "accept": true, + "reason": "Both are about the same Whisper fbank feature extraction path failing." }, { "left": "issue:43404", "right": "issue:43950", "accept": false, - "reason": "Weight-tying bug in Mistral3 vs non-persistent buffer corruption on `from_pretrained`; different regressions." + "reason": "Tied lm_head weights in Mistral3 and non-persistent buffer corruption are different bugs." }, { "left": "issue:43742", "right": "issue:45399", "accept": false, - "reason": "MobileLLM load KeyError vs flash-attn fallback checks; unrelated." + "reason": "MobileLLM load failure and flash-attn fallback logic are unrelated." }, { "left": "issue:42915", "right": "issue:45020", "accept": false, - "reason": "Qwen3 MoE + FineGrainedFP8 failure is model/config-specific, not the same as the broader remote_code regression." + "reason": "Qwen3 MoE ZeRO loading/weight conversion and remote_code breakage are different issues." }, { "left": "issue:43257", "right": "issue:43866", "accept": false, - "reason": "Different Qwen3/Ovis model-loading failures with no clear shared root cause." + "reason": "Qwen3 MoE weight-conversion failure and checkpoint corruption are unrelated." }, { "left": "issue:43742", "right": "issue:45020", "accept": false, - "reason": "Specific MobileLLM loading bug vs broad remote_code regression; not the same issue." + "reason": "MobileLLM key error and remote_code regression are unrelated." }, { "left": "issue:41553", "right": "issue:44991", "accept": false, - "reason": "Different tokenizer/loading failures for unrelated models." - }, - { - "left": "issue:43872", - "right": "issue:44291", - "accept": true, - "reason": "Same root regression around the new `_is_hf_initialized` argument causing loading-time TypeErrors." + "reason": "Voxtral tokenizer error messaging and EMBEDDIA tokenizer loading regression are different bugs." }, { "left": "issue:43940", "right": "issue:45313", "accept": false, - "reason": "Both are DeepSpeed ZeRO-3 load failures, but they affect different Qwen variants and do not show the same concrete bug." + "reason": "Both are ZeRO-3 load failures, but different Qwen model families and no concrete shared bug evidence." }, { - "left": "issue:43854", - "right": "issue:44863", - "accept": false, - "reason": "Different model-loading failures for GLM-4.7-Flash and NemotronH." + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "Both describe the same init_empty_weights / _is_hf_initialized incompatibility during model loading." } ] }, "evaluator_result": { - "accept": false, - "feedback": "Overall the summary is grounded and conservative, but I would not accept the cluster as-is because one soft-edge verdict is overstated: issue:43054 vs issue:43493 is related SigLIP2 territory, but the packet only gives titles/body lengths and no explicit evidence that they are the same bug or same root cause. That should be treated as a cautious maybe, not a duplicate. The other accepted pairs (_is_hf_initialized and v5 docs cleanup) look well supported." + "accept": true, + "feedback": "The summary is grounded in the packet and stays mostly conservative. The accepted soft edges are supported by clear title-level overlap or the same documented regression subgroup (remote_code, docs removal, Whisper fbank, bitsandbytes/init_empty_weights, SigLIP2). The rejected edges are generally reasonable separations rather than overclaims." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -1405,7 +1267,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:24643", "issue:30064", @@ -1428,6 +1290,7 @@ "issue:43646", "issue:43824", "issue:43828", + "issue:43854", "issue:43872", "issue:43950", "issue:43957", @@ -1456,21 +1319,23 @@ "issue:45588" ], "soft_pairs": [ + "issue:43854|issue:44863", + "issue:42915|issue:43872", "issue:44928|issue:45588", "issue:41553|issue:44488", "issue:43257|issue:45305", "issue:36331|issue:41093", "issue:43872|issue:45538", - "issue:41093|issue:44805", "issue:44530|issue:44863", + "issue:41093|issue:44805", "issue:44479|issue:44560", "issue:24643|issue:30064", "issue:43366|issue:43828", "issue:43366|issue:43531", "issue:43550|issue:44910", "issue:43645|issue:43646", - "issue:43366|issue:44589", "issue:43950|issue:45325", + "issue:43366|issue:44589", "issue:44560|issue:44918", "issue:42617|issue:43366", "issue:42915|issue:43957", @@ -1479,228 +1344,240 @@ "issue:38175|issue:43994", "issue:43646|issue:43950", "issue:44155|issue:45584", - "issue:44560|issue:44805", "issue:44877|issue:45030", + "issue:44560|issue:44805", "issue:43638|issue:44661", "issue:43596|issue:44661", "issue:43317|issue:44756", "issue:38175|issue:39692", "issue:45313|issue:45357" ], - "prepared_review_unit_hash": "20f1470df5dfd3215b832f53041040cecaac738e3d07acbb141d8261b2500627", + "prepared_review_unit_hash": "f0a94ceaff6bd7bbdb5d44190a659882ce543e3093b03832fd5193cbaa368cf5", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 30, - "serialized_chars": 36301, - "estimated_input_tokens": 9076, - "estimated_eval_tokens": 18408 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37463, + "estimated_input_tokens": 9366, + "estimated_eval_tokens": 18988 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The items are mostly unrelated, with only a few clear duplicate-like pairs. The strongest matches are the generic IndexError shape-mismatch reports, the custom model initialization regression, and the SigLIP2 inference regression.", - "confidence": 0.63, - "canonical_issue_reason": "Issue 43994 is the best representative issue among the duplicate-like candidates because it states the SigLIP2 failure in a broader, user-facing way and has more discussion/references than the symptom-only report.", + "summary": "The cluster is mostly noisy and contains several separate bug themes; only a few pairs look like true duplicates or near-duplicates. The clearest duplicate is the Transformers 5.0 custom model initialization regression, with a couple of other plausible same-bug pairs around SigLIP2, Qwen video input, and a mask-shape IndexError family.", + "confidence": 0.58, + "canonical_issue_reason": "Issue 43646 is the broadest, most canonical wording for the custom-model initialization regression; 43645 is the notebook-specific variant of the same bug.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43994 is the strongest single issue to anchor this set: it describes the underlying SigLIP2 regression clearly, is broadly framed, and is the better duplicate target for 38175.", + "best_issue_reason": "43646 is the best single representative issue in this set because it describes the underlying regression generically and is the cleanest anchor for the duplicate pair.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "Different failures: GLM-4.7 test loading vs NemotronH checkpoint loading." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "Different concrete bugs: Qwen3Moe FP8 failure vs bitsandbytes constructor incompatibility." + }, { "left": "issue:44928", "right": "issue:45588", "accept": false, - "reason": "Different bugs: one is Qwen3.5 gradient instability in RLHF/SDPA, the other is a flash-attention crash on sink-less models." + "reason": "Both mention Qwen/attention problems, but the symptoms and code paths differ (RLHF NaNs vs flash_attention s_aux handling)." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "Both are loading problems, but they involve different models and different failure modes; not the same underlying bug." + "reason": "Different loading issues: bad Voxtral tokenizer error vs a model load failure for cjvt/sleng-bert." }, { "left": "issue:43257", "right": "issue:45305", "accept": false, - "reason": "Different DeepSpeed issues: one is weight conversion/loading for Qwen3 MoE, the other is gradient averaging with GAS." + "reason": "Both involve DeepSpeed, but one is Qwen3 MoE weight conversion and the other is gradient averaging with GAS." }, { "left": "issue:36331", "right": "issue:41093", "accept": false, - "reason": "Unexpected compute_loss kwargs vs a tensor mask shape IndexError are unrelated code paths." + "reason": "Unexpected trainer kwarg vs a tensor-mask shape mismatch; not the same bug." }, { "left": "issue:43872", "right": "issue:45538", "accept": false, - "reason": "bitsandbytes parameter construction and CLIPTokenizer max_length are unrelated." - }, - { - "left": "issue:41093", - "right": "issue:44805", - "accept": true, - "reason": "Same error signature and same shape-mismatch failure pattern; these look like duplicate reports of the same indexing bug." + "reason": "Unrelated: bitsandbytes init kwarg error vs CLIPTokenizer max-length behavior." }, { "left": "issue:44530", "right": "issue:44863", "accept": false, - "reason": "PagedAttention cache/group-type failure and NemotronH checkpoint loading are different problems." + "reason": "Different subsystems and symptoms: PagedAttentionCache group-type crash vs NemotronH checkpoint loading." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": true, + "reason": "Same IndexError family with the same mask-vs-tensor shape mismatch; 44805 looks like a later report of the same off-by-one indexing bug." }, { "left": "issue:44479", "right": "issue:44560", - "accept": false, - "reason": "Both mention Qwen video paths, but the concrete failures differ and do not look like one fixable bug." + "accept": true, + "reason": "Same Qwen video-input regression theme; 44560 is a specific StopIteration manifestation of the broader 5.3.0 video bug." }, { "left": "issue:24643", "right": "issue:30064", "accept": false, - "reason": "Training with DeepSpeed weight-dimension error is unrelated to void segmentation map processing." + "reason": "Completely different bugs: DeepSpeed weight dimensionality vs void segmentation map processing." }, { "left": "issue:43366", "right": "issue:43828", "accept": false, - "reason": "GGUF gpt-oss support and a torch.autocast dtype mismatch are unrelated." + "reason": "GGUF gpt-oss support and Phi-tiny-MoE autocast dtype mismatch are unrelated." }, { "left": "issue:43366", "right": "issue:43531", "accept": false, - "reason": "Different Qwen/GGUF feature requests or bugs; no shared concrete failure." + "reason": "Different issues: GGUF architecture support vs Qwen3-MoE sliding_window behavior." }, { "left": "issue:43550", "right": "issue:44910", "accept": false, - "reason": "torch.compile/SDPA crash on Bamba is unrelated to the Qwen3.5 flash-attention illegal-memory-access issue." + "reason": "Different failures in different models/code paths: Bamba torch.compile SDPA vs Qwen3.5 flash-attention memory access." }, { "left": "issue:43645", "right": "issue:43646", "accept": true, - "reason": "These are essentially the same Transformers 5.0 custom model initialization regression, with one report being a more specific notebook variant." + "reason": "Same underlying custom-model initialization regression; 43645 is just the notebook-specific description of the 43646 bug." }, { - "left": "issue:43366", - "right": "issue:44589", + "left": "issue:43950", + "right": "issue:45325", "accept": false, - "reason": "GGUF architecture support and missing Float8 storage are different failures." + "reason": "Different regressions: non-persistent buffer corruption vs Qwen2.5-VL rope-index scaling." }, { - "left": "issue:43950", - "right": "issue:45325", + "left": "issue:43366", + "right": "issue:44589", "accept": false, - "reason": "Non-persistent buffer corruption in from_pretrained is unrelated to Qwen2.5-VL rope-index scaling." + "reason": "Unrelated: GGUF support request vs missing Float8 storage type." }, { "left": "issue:44560", "right": "issue:44918", "accept": false, - "reason": "Both involve Qwen3.x and video/text handling, but the concrete failures and code paths differ." + "reason": "Different Qwen3.5/video vs TRL embedding unpacking bugs." }, { "left": "issue:42617", "right": "issue:43366", "accept": false, - "reason": "Running 3d_parallel.py and GGUF gpt-oss support are unrelated." + "reason": "3d_parallel.py runtime failure is unrelated to GGUF model support." }, { "left": "issue:42915", "right": "issue:43957", "accept": false, - "reason": "Qwen3 MoE FP8 loading and meta-device loading regressions are different bugs." + "reason": "Different failures: Qwen3Moe FP8 vs meta-device loading regression." }, { "left": "issue:43824", "right": "issue:43957", "accept": false, - "reason": "Importing Qwen2.5-VL and meta-device model loading are unrelated failures." + "reason": "Importing Qwen2_5_VL is a different issue from meta-device model loading." }, { "left": "issue:36010", "right": "issue:42915", "accept": false, - "reason": "GenerationMixin import failure and Qwen3 MoE FP8 failure are unrelated." + "reason": "GenerationMixin import error is unrelated to Qwen3Moe FP8 loading." }, { "left": "issue:38175", "right": "issue:43994", "accept": true, - "reason": "Same model family and same inference correctness regression; the second report is the broader duplicate of the first symptom report." + "reason": "Same SigLIP2 bad-output bug reported through different APIs; both point to the same model/processor mismatch symptom." }, { "left": "issue:43646", "right": "issue:43950", "accept": false, - "reason": "Custom model initialization regression and buffer corruption in from_pretrained are different issues." + "reason": "Custom model init regression and buffer corruption are different bugs." }, { "left": "issue:44155", "right": "issue:45584", "accept": false, - "reason": "AudioFlamingo3 batch leakage and Whisper empty-transcription failure are unrelated." + "reason": "Different audio/Whisper failures with different code paths." }, { - "left": "issue:44560", - "right": "issue:44805", + "left": "issue:44877", + "right": "issue:45030", "accept": false, - "reason": "Qwen video StopIteration and mask shape mismatch are unrelated." + "reason": "Both are config-validation related, but they are not the same model or failure mode." }, { - "left": "issue:44877", - "right": "issue:45030", + "left": "issue:44560", + "right": "issue:44805", "accept": false, - "reason": "Both are config-loading failures, but they concern different models and do not clearly share the same root cause." + "reason": "Different error classes and domains: video StopIteration vs mask shape mismatch." }, { "left": "issue:43638", "right": "issue:44661", "accept": false, - "reason": "Zero3 Bert loading crash and add-new-model-like mapping failure are different code paths." + "reason": "Both are loading/registration issues, but the causes are different (DeepSpeed/Bert index error vs tokenizer-mapping restriction)." }, { "left": "issue:43596", "right": "issue:44661", "accept": false, - "reason": "ZeRO-3 Bert init IndexError and tokenizer-mapping failure are unrelated." + "reason": "Different underlying bugs: ZeRO-3 zero-sized tensor indexing vs tokenizer-mapping lookup failure." }, { "left": "issue:43317", "right": "issue:44756", "accept": false, - "reason": "Dequantized model offload failure and mmap/OOM on Strix Halo are different problems." + "reason": "Different loading problems: dequantized model offload failure vs mmap/OOM behavior on Strix Halo." }, { "left": "issue:38175", "right": "issue:39692", "accept": false, - "reason": "Both involve SigLIP2, but one is an inference regression while the other is a documentation/example issue with additional mismatches; not the same bug." + "reason": "Related SigLIP2 context, but 39692 bundles multiple documentation errors and is not clearly the same single bug." }, { "left": "issue:45313", "right": "issue:45357", "accept": false, - "reason": "DeepSpeed ZeRO-3 weight loading and save_pretrained visual-encoder key corruption are separate load/save regressions." + "reason": "Different Qwen3.5 DeepSpeed load/save issues: ZeRO-3 language_model weights vs incorrect visual encoder keys on save_pretrained." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet titles, and the accepted soft pairs are conservative duplicates or near-duplicates: the two mask-shape IndexError reports, the two custom model initialization reports, and the two SigLIP2 inference reports. The rejected pairs generally stay on the safe side by treating different models or failure modes as distinct. No obvious overstatement of evidence." + "feedback": "The summary is grounded in the packet and stays appropriately cautious. The accepted pairs are framed as plausible same-bug matches rather than overstated certainties, and the rejected pairs are mostly conservative distinctions between different failure modes. The canonical issue choice around 43645/43646 is also well supported by the titles." }, "evaluator_used": true, "retried": false, @@ -1719,7 +1596,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:36010", "issue:39404", @@ -1774,30 +1651,30 @@ "issue:44960|issue:44977", "issue:44451|issue:44488", "issue:42491|issue:43720", - "issue:43854|issue:43931", "issue:43708|issue:44075", + "issue:43854|issue:43931", "issue:41720|issue:42915", "issue:42491|issue:45276", "issue:43856|issue:43873", "issue:43257|issue:43854", - "issue:42915|issue:44756", "issue:39692|issue:43994", + "issue:42915|issue:44756", "issue:42491|issue:43931", - "issue:44661|issue:44683", "issue:43856|issue:44387", + "issue:44661|issue:44683", "issue:43644|issue:43645", "issue:44910|issue:45588", "issue:45362|issue:45406", - "issue:44488|issue:44991", "issue:42491|issue:45292", - "issue:43404|issue:43454", + "issue:44488|issue:44991", "issue:36010|issue:39404", + "issue:43404|issue:43454", "issue:45276|issue:45335", "issue:44164|issue:44291", "issue:44402|issue:44488", - "issue:43425|issue:43957", "issue:39692|issue:44960", "issue:43335|issue:45663", + "issue:43425|issue:43957", "issue:42947|issue:45704", "issue:45663|issue:45759", "issue:43873|issue:44387", @@ -1805,14 +1682,14 @@ "issue:44387|issue:44756", "issue:44291|issue:44756" ], - "prepared_review_unit_hash": "7d2b4fdaf172dfc10cef17781b8237c71f2414d5f0dfeeb4756e1b72fcd18ecc", + "prepared_review_unit_hash": "d61e1ead56283f20fff914da5e7df9c6c60ff67a47fe81c9c67a9df580b5bf38", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 47, @@ -1827,225 +1704,225 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of semantically similar but unrelated Transformers issues, so most soft links are false positives. I only see a few likely true duplicates: the non-persistent-buffer regression, the two BERT-like tokenizer/model-loading failures, and possibly the two resize_token_embeddings regressions in composite models.", - "confidence": 0.86, - "canonical_issue_reason": null, + "summary": "The cluster is mostly heterogeneous, with two separate duplicate-like families: one around non-persistent buffers being filled with junk, and one around `resize_token_embeddings` not propagating to output/decoder embeddings. The rest of the pairs are only loosely related and should stay separate.", + "confidence": 0.77, + "canonical_issue_reason": "Issue #45292 is the best canonical issue for the `resize_token_embeddings` family: it states the generic propagation bug in a model-agnostic way and cleanly subsumes the model-specific variants (#45276, #45335).", "canonical_pr_reason": null, - "best_issue_reason": "No single issue is a good global canonical because the cluster splits into several unrelated bug themes; at most, 43644 is the cleanest anchor for its small duplicate subpair.", + "best_issue_reason": "Issue #45292 is the strongest representative of the `resize_token_embeddings` subgroup because it is the broadest, most model-agnostic description. The cluster also contains a separate duplicate-like subgroup around non-persistent buffers (#43644/#44534), so #45292 is not representative of the entire cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43644", "right": "issue:44534", "accept": true, - "reason": "Same underlying Transformers 5 bug: non-persistent buffers being filled with junk." + "reason": "Same bug with near-identical wording: non-persistent buffers are filled with junk in Transformers v5." }, { "left": "issue:44960", "right": "issue:44977", "accept": false, - "reason": "Different models and different failures; both mention flash attention, but not the same concrete bug." + "reason": "Both involve Qwen/flash-attention, but the concrete failures and model behavior differ; not the same bug." }, { "left": "issue:44451", "right": "issue:44488", - "accept": true, - "reason": "Both report the same latest-version model/tokenizer loading failure for similar BERT-style models." + "accept": false, + "reason": "Both are loading failures, but they concern different models and no shared root cause is shown." }, { "left": "issue:42491", "right": "issue:43720", "accept": false, - "reason": "Different loading paths and symptoms; one is Qwen3 MoE LoRA compatibility, the other BitNet packed-weight handling." + "reason": "Different code paths and failure modes: Qwen3 MoE LoRA compatibility vs BitNet packed-weight loading." }, { - "left": "issue:43854", - "right": "issue:43931", + "left": "issue:43708", + "right": "issue:44075", "accept": false, - "reason": "Different models and different load errors; not the same bug." + "reason": "Unrelated bugs: checkpoint resume step calculation vs optimizer arguments not being used." }, { - "left": "issue:43708", - "right": "issue:44075", + "left": "issue:43854", + "right": "issue:43931", "accept": false, - "reason": "Completely different subsystems: checkpoint resume math vs SGD optimizer args." + "reason": "Different model-loading failures for different architectures; not the same underlying issue." }, { "left": "issue:41720", "right": "issue:42915", "accept": false, - "reason": "Both are Qwen3-related, but the failures are different code paths and different causes." + "reason": "Both are Qwen3-related, but one is auto device mapping/cuda assert and the other is FP8 config failure." }, { "left": "issue:42491", "right": "issue:45276", "accept": false, - "reason": "Different model families and different resizing/loading problems." + "reason": "Different bugs: LoRA model incompatibility vs token-embedding resize propagation." }, { "left": "issue:43856", "right": "issue:43873", "accept": false, - "reason": "Both involve memory/quantization, but the reported bugs are different." + "reason": "Both concern memory, but one is training memory usage and the other is quantization/offloading behavior." }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Different models and unrelated loading/weight-conversion issues." + "reason": "Different subsystems and failure modes: accelerate/deepspeed weight conversion vs model loading in tests." }, { - "left": "issue:42915", - "right": "issue:44756", + "left": "issue:39692", + "right": "issue:43994", "accept": false, - "reason": "Different failures: Qwen3Moe FP8 config vs mmap/OOM on Strix Halo." + "reason": "Same model family, but one is a broken docs example and the other is a pipeline/AutoModel result-quality issue." }, { - "left": "issue:39692", - "right": "issue:43994", + "left": "issue:42915", + "right": "issue:44756", "accept": false, - "reason": "Same model family, but one is a docs example problem and the other is an inference/output regression." + "reason": "No concrete overlap beyond being performance/loading issues; different models and mechanisms." }, { "left": "issue:42491", "right": "issue:43931", "accept": false, - "reason": "Different Qwen3 loading problems with different symptoms and root causes." + "reason": "Distinct bugs affecting different Qwen variants and different loading/training paths." }, { - "left": "issue:44661", - "right": "issue:44683", + "left": "issue:43856", + "right": "issue:44387", "accept": false, - "reason": "Unrelated areas: tokenizer mapping vs torch flex_attention compilation." + "reason": "Both concern memory, but one is Qwen3 MoE training memory usage and the other is int4 CUDA reserved-memory OOM." }, { - "left": "issue:43856", - "right": "issue:44387", + "left": "issue:44661", + "right": "issue:44683", "accept": false, - "reason": "Both are memory-related, but one is training memory usage and the other is int4 quantization reserved memory growth." + "reason": "Unrelated: tokenizer-mapping logic vs compiled flex_attention on newer torch." }, { "left": "issue:43644", "right": "issue:43645", "accept": false, - "reason": "One is non-persistent buffer junk, the other is notebook custom-model initialization; not the same bug." + "reason": "Different regressions: junk non-persistent buffers vs notebook custom-model initialization." }, { "left": "issue:44910", "right": "issue:45588", "accept": false, - "reason": "Both touch flash attention, but the failure modes are different (packed sequence misread vs s_aux=None crash)." + "reason": "Both mention flash attention, but the crash conditions and root causes are different." }, { "left": "issue:45362", "right": "issue:45406", "accept": false, - "reason": "Different models and different crashes; not a duplicate." - }, - { - "left": "issue:44488", - "right": "issue:44991", - "accept": false, - "reason": "Both are tokenizer-loading complaints, but about different models and likely different root causes." + "reason": "Different bugs in different model stacks: Qwen chat crash vs Gemma4Processor missing `_tokenizer`." }, { "left": "issue:42491", "right": "issue:45292", "accept": false, - "reason": "Different issues: LoRA model compatibility vs resize_token_embeddings not updating outputs." + "reason": "Different issues: versioned Qwen3 MoE LoRA incompatibility vs generic embedding-resize propagation." }, { - "left": "issue:43404", - "right": "issue:43454", + "left": "issue:44488", + "right": "issue:44991", "accept": false, - "reason": "Both mention lm_head tying, but they are distinct model-specific bugs and not clearly the same code path." + "reason": "Both tokenizer-loading failures, but for different model repos with no evidence of the same root cause." }, { "left": "issue:36010", "right": "issue:39404", "accept": false, - "reason": "ImportError for GenerationMixin is unrelated to Whisper return_language pipeline behavior." + "reason": "Import error for `GenerationMixin` is unrelated to Whisper pipeline `return_language` behavior." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both involve untied `lm_head`, but they are model-specific implementations and not clearly the same concrete bug." }, { "left": "issue:45276", "right": "issue:45335", "accept": true, - "reason": "Same resize_token_embeddings regression: updated embeddings are not propagated to model-specific child embeddings." + "reason": "Same underlying bug: `resize_token_embeddings` does not update model-specific downstream embeddings/output weights." }, { "left": "issue:44164", "right": "issue:44291", "accept": false, - "reason": "Different save/load failures: extra_state handling vs init_empty_weights unexpected argument." + "reason": "Different save/load regressions: extra_state handling vs unexpected `_is_hf_initialized` during `init_empty_weights`." }, { "left": "issue:44402", "right": "issue:44488", "accept": false, - "reason": "Tokenizer vocab-size mismatch vs outright model loading failure; not the same issue." - }, - { - "left": "issue:43425", - "right": "issue:43957", - "accept": false, - "reason": "Different incompatibilities: torch 2.10 vs meta-device model loading regressions." + "reason": "A tokenizer vocab mismatch is not the same as the separate model-loading failure in `sleng-bert`." }, { "left": "issue:39692", "right": "issue:44960", "accept": false, - "reason": "Different model families and different failure modes." + "reason": "Different model families and different reported failures; no shared code-path bug is evident." }, { "left": "issue:43335", "right": "issue:45663", "accept": false, - "reason": "Different architectures and different bugs: sparse-layer config vs Gemma-4 FSDP2 shared_kv_states." + "reason": "Different model architectures and failure modes: sparse-layer creation vs FSDP2/shared_kv_states rebuilding." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Torch-version compatibility is too broad here; one is general incompatibility, the other is meta-device loading breakage." }, { "left": "issue:42947", "right": "issue:45704", "accept": false, - "reason": "Gradient checkpointing/LoRA behavior is unrelated to the apex RMSNorm memory leak." + "reason": "Different training/runtime issues: gradient checkpointing with LoRA vs apex RMSNorm memory leak." }, { "left": "issue:45663", "right": "issue:45759", "accept": false, - "reason": "Different model internals and different failures; not a duplicate." + "reason": "Different composite-model problems: FSDP2 attention state handling vs `from_config` not unwrapping `text_config`." }, { "left": "issue:43873", "right": "issue:44387", "accept": false, - "reason": "Both involve quantization/offloading, but they describe different concrete regressions." + "reason": "Both are quantization/memory related, but the failure mechanisms are different." }, { "left": "issue:43854", "right": "issue:45216", "accept": false, - "reason": "Different Qwen3.5 bugs: unit-test loading mismatch vs incorrect save_pretrained checkpointing." + "reason": "Different Qwen versions and different lifecycle stages: loading in tests vs checkpoint saving correctness." }, { "left": "issue:44387", "right": "issue:44756", "accept": false, - "reason": "Quantization OOM vs mmap-related OOM on Strix Halo are separate problems." + "reason": "Different problems: int4 quantization memory regression vs mmap-related OOM on Strix Halo." }, { "left": "issue:44291", "right": "issue:44756", "accept": false, - "reason": "Meta-init TypeError is unrelated to the Strix Halo mmap/OOM issue." + "reason": "No meaningful overlap beyond both affecting model loading." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet: the cluster does split across several distinct issue themes, and the proposed true-duplicate pairs are the most defensible ones from the titles. The soft-edge rejections are conservative, and the accepted pairs are reasonably supported without overclaiming beyond the available evidence." + "accept": false, + "feedback": "The summary is mostly grounded, but `best_issue_reason` overstates the evidence by calling #45292 the representative of the \u201conly clearly duplicate issue family\u201d even though the packet also contains a clearly duplicate buffer-junk family (#43644/#44534). The resize_token_embeddings canonicalization looks reasonable, and the accepted soft-edge pairs are conservative, but that specific framing is not supported by the cluster." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -2061,9 +1938,10 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:39404", + "issue:40990", "issue:41720", "issue:42947", "issue:43257", @@ -2114,8 +1992,8 @@ "soft_pairs": [ "issue:44292|issue:44912", "issue:42947|issue:43381", - "issue:45276|issue:45292", "issue:45684|issue:45698", + "issue:45276|issue:45292", "issue:43381|issue:44928", "issue:44164|issue:44756", "issue:45588|issue:45663", @@ -2128,270 +2006,277 @@ "issue:44402|issue:44451", "issue:43645|issue:44164", "issue:43828|issue:43957", - "issue:43425|issue:43828", "issue:43257|issue:44863", + "issue:43425|issue:43828", "issue:44589|issue:44756", "issue:44479|issue:45663", "issue:39404|issue:45701", "issue:43906|issue:44530", "issue:43611|issue:43645", - "issue:43425|issue:43716", "issue:43931|issue:44863", + "issue:43425|issue:43716", "issue:44530|issue:45084", "issue:43901|issue:44509", "issue:39404|issue:44912", "issue:44863|issue:45663", "issue:44265|issue:44898", - "issue:44479|issue:44805", "issue:43299|issue:43931", + "issue:44479|issue:44805", + "issue:43856|issue:44756", "issue:43425|issue:44918", "issue:44589|issue:45084", - "issue:43856|issue:44756" + "issue:40990|issue:42947" ], - "prepared_review_unit_hash": "973271ab5dc1964c96bc2f08d5e4b533d122ea2380efdb74e602e16058d31b31", + "prepared_review_unit_hash": "33e7ca2ffc966510e98149188c689be5840cbd6880fa142255934837fd9f0553", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 35, - "serialized_chars": 37525, - "estimated_input_tokens": 9382, - "estimated_eval_tokens": 19020 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38389, + "estimated_input_tokens": 9598, + "estimated_eval_tokens": 19452 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated Transformers bug reports with a few loose topic overlaps. The only clear duplicate family I see is the `resize_token_embeddings` regression affecting output/decoder embeddings across model variants.", - "confidence": 0.66, - "canonical_issue_reason": "issue:45292 is the most generic and central description of the `resize_token_embeddings` problem, and it naturally covers the more specific Gemma-4 and t5gemma reports.", + "summary": "This cluster is mostly a mix of unrelated Transformers/model-specific regression issues that only look similar textually. The only clear duplicate family is the resize_token_embeddings/output_embeddings bug across Gemma4, generic output_embeddings, and t5gemma variants.", + "confidence": 0.93, + "canonical_issue_reason": "issue:45292 is the broadest, most general statement of the one real duplicate family: resize_token_embeddings not updating output embeddings. The other members are model-specific variants of the same underlying bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45292 is the best representative issue because it states the shared failure mode without narrowing to one model family.", + "best_issue_reason": "issue:45292 best represents the cluster because it captures the shared concrete bug without narrowing to a single architecture, making it the best canonical issue for the accepted duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44292", "right": "issue:44912", "accept": false, - "reason": "Both are quantization/model-loading complaints around Qwen/Git-Oss, but they describe different models and different symptoms; not the same concrete bug." + "reason": "Both are quantization/loading complaints, but they involve different models and different failures (runtime error vs bf16 fallback), so not the same bug." }, { "left": "issue:42947", "right": "issue:43381", "accept": false, - "reason": "Both mention gradient checkpointing, but one is about PEFT LoRA effectiveness and the other about eval-mode restrictions; different bugs." - }, - { - "left": "issue:45276", - "right": "issue:45292", - "accept": true, - "reason": "Same underlying bug: `resize_token_embeddings` fails to update tied/output embeddings, with 45276 just naming a more specific Gemma-4 case." + "reason": "PEFT LoRA gradient checkpointing ineffective is a different problem from checkpointing being disallowed in eval mode." }, { "left": "issue:45684", "right": "issue:45698", "accept": false, - "reason": "One is a read-only permission propagation issue during `save_pretrained`, the other is loading the wrong custom module after save/load; different failure modes." + "reason": "Save/pretrained permission propagation and loading the wrong custom module after save are different save/load bugs." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same resize_token_embeddings regression: embeddings/output heads are not updated; 45276 is a Gemma4-specific instance of the broader 45292 bug." }, { "left": "issue:43381", "right": "issue:44928", "accept": false, - "reason": "Both involve training behavior, but one is about checkpointing in eval mode and the other about Qwen3.5 RLHF NaNs from 3D position ids; unrelated causes." + "reason": "Eval-mode checkpointing restriction is unrelated to the Qwen3.5 RLHF NaN/SDPA fallback issue." }, { "left": "issue:44164", "right": "issue:44756", "accept": false, - "reason": "`extra_state` handling and disabling mmap for Strix Halo OOM are unrelated issues." + "reason": "extra_state save/load handling and Strix Halo mmap OOM are unrelated subsystems." }, { "left": "issue:45588", "right": "issue:45663", "accept": false, - "reason": "Flash-attention crash on `s_aux=None` and Gemma-4 FSDP2 `shared_kv_states` rebuild are different code paths." + "reason": "Flash-attention sink-less crash and Gemma4 FSDP2 shared_kv_states KeyError are different concrete code-path failures." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "A800 auto device-map assert on Qwen3 is unrelated to AudioFlamingo3 batched inference token leakage." + "reason": "cudaErrorAssert on auto device mapping and AudioFlamingo3 token leakage are unrelated bugs." }, { "left": "issue:44479", "right": "issue:44877", "accept": false, - "reason": "Video-input regression for Qwen VL models and strict config loading for granite_speech are different bugs." + "reason": "Video-input regression for Qwen VL models and strict config blocking granite_speech loading are different issues." }, { "left": "issue:43646", "right": "issue:43708", "accept": false, - "reason": "Custom model initialization in notebooks and Trainer `resume_from_checkpoint` step calculation are not the same defect." + "reason": "Custom model initialization in notebooks is not the same as resume_from_checkpoint max_steps being miscomputed." }, { "left": "issue:43854", "right": "issue:44960", "accept": false, - "reason": "Unit-test model loading failure for GLM-4.7-Flash is unrelated to the generic GLM5 issue." + "reason": "Different model families and failures: GLM-4.7-Flash test loading vs GLM5 loading." }, { "left": "issue:43957", "right": "issue:44756", "accept": false, - "reason": "Meta-device loading breakage and Strix Halo mmap OOM are separate problems." + "reason": "meta-device loading breakage and mmap/OOM are distinct loading-path problems." }, { "left": "issue:45292", "right": "issue:45335", "accept": true, - "reason": "Same core bug: `resize_token_embeddings` does not propagate to other embeddings, here `output_embeddings` versus `decoder.embed_tokens`." + "reason": "Same underlying resize_token_embeddings bug: output/decoder embeddings are not updated after resizing." }, { "left": "issue:44402", "right": "issue:44451", "accept": false, - "reason": "Tokenizer vocab-size mismatch for one model and inability to load another model are not clearly the same issue." + "reason": "Tokenizer vocab-size mismatch and inability to load ScandiBERT are not the same concrete defect." }, { "left": "issue:43645", "right": "issue:44164", "accept": false, - "reason": "Custom-model notebook initialization breakage and `extra_state` save/load handling do not describe the same code-path failure." + "reason": "Notebook custom-model initialization regression is unrelated to extra_state save/from_pretrained handling." }, { "left": "issue:43828", "right": "issue:43957", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE and meta-device loading failures are unrelated." + "reason": "Autocast dtype mismatch and meta-device loading failures are different code paths." }, { - "left": "issue:43425", - "right": "issue:43828", + "left": "issue:43257", + "right": "issue:44863", "accept": false, - "reason": "Torch 2.10 incompatibility and autocast dtype mismatch are different compatibility bugs." + "reason": "Qwen3 MoE conversion/load issue and NemotronH checkpoint loading failure are separate model-loading bugs." }, { - "left": "issue:43257", - "right": "issue:44863", + "left": "issue:43425", + "right": "issue:43828", "accept": false, - "reason": "Both are model-loading reports, but one is Qwen3 MoE conversion with accelerate+deepspeed and the other is NemotronH checkpoint loading; not the same bug." + "reason": "Torch 2.10 incompatibility is not the same as the Phi-tiny-MoE autocast dtype mismatch." }, { "left": "issue:44589", "right": "issue:44756", "accept": false, - "reason": "Missing Float8 storage and mmap OOM on Strix Halo are unrelated runtime issues." + "reason": "Float8 storage lookup failure and Strix Halo mmap OOM are unrelated runtime/storage problems." }, { "left": "issue:44479", "right": "issue:45663", "accept": false, - "reason": "Qwen VL video regression and Gemma-4 FSDP2 attention state issues are different defects." + "reason": "Qwen VL video regression and Gemma4 FSDP2 KeyError are different failures." }, { "left": "issue:39404", "right": "issue:45701", "accept": false, - "reason": "Whisper `return_language` pipeline behavior and tokenizer differences across versions are not the same underlying bug." + "reason": "Whisper return_language pipeline breakage and tokenizer changes across versions are not the same bug." }, { "left": "issue:43906", "right": "issue:44530", "accept": false, - "reason": "A reproduction of another issue and a PagedAttentionCache linear_attention crash are unrelated." + "reason": "A reproduction of a prior issue is not the same as the Qwen3.5 PagedAttentionCache linear_attention crash." }, { "left": "issue:43611", "right": "issue:43645", "accept": false, - "reason": "Base-model-prefix loading breakage and notebook custom-model initialization are both v5 regressions but not the same concrete failure." + "reason": "Both are Transformers 5 custom-model regressions, but one is base_model_prefix loading and the other is notebook model initialization; too different to merge." }, { - "left": "issue:43425", - "right": "issue:43716", + "left": "issue:43931", + "right": "issue:44863", "accept": false, - "reason": "Torch version incompatibility and Mistral-3 dtype mismatch are different issues." + "reason": "Qwen3-VL weight-shape mismatch and NemotronH checkpoint loading are different model-specific loading failures." }, { - "left": "issue:43931", - "right": "issue:44863", + "left": "issue:43425", + "right": "issue:43716", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch and NemotronH checkpoint loading failure are different model-loading bugs." + "reason": "Torch 2.10 incompatibility is unrelated to the Mistral-3 preprocessor/model dtype mismatch." }, { "left": "issue:44530", "right": "issue:45084", "accept": false, - "reason": "PagedAttentionCache `linear_attention` crash and `Can't compile non template nodes` are unrelated errors." + "reason": "PagedAttentionCache linear_attention crash and 'Can't compile non template nodes' are different errors in different subsystems." }, { "left": "issue:43901", "right": "issue:44509", "accept": false, - "reason": "Both are documentation-related, but they concern different pipeline/task deprecations and should not be merged." + "reason": "Docs mentioning return_all_scores and docs mentioning removed pipeline tasks are separate documentation issues." }, { "left": "issue:39404", "right": "issue:44912", "accept": false, - "reason": "Whisper pipeline language return and MXFP4 quantization fallback for git-oss-20b are unrelated." + "reason": "Whisper pipeline return_language and MXFP4 quantization fallback are unrelated." }, { "left": "issue:44863", "right": "issue:45663", "accept": false, - "reason": "NemotronH checkpoint loading and Gemma-4 FSDP2 training state handling are different bugs." + "reason": "NemotronH checkpoint loading and Gemma4 shared_kv_states rebuild are different bugs." }, { "left": "issue:44265", "right": "issue:44898", "accept": false, - "reason": "`torch.export.export` failing on `torch_compilable_check` and Perceiver interpolation/resolution failure are unrelated." + "reason": "torch.export failures with torch_compilable_check and Perceiver non-default resolution failures are unrelated." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Both involve Qwen3-VL-style models, but one is a MoE loading regression and the other is a weight-shape mismatch; not the same defect." }, { "left": "issue:44479", "right": "issue:44805", "accept": false, - "reason": "Qwen VL video regression and a tensor-mask shape mismatch are different code paths." + "reason": "Video-input regression and mask/tensor shape mismatch are different symptoms with no clear shared code path." }, { - "left": "issue:43299", - "right": "issue:43931", + "left": "issue:43856", + "right": "issue:44756", "accept": false, - "reason": "Both are Qwen3-VL loading failures, but one is a broad MoE loading regression and the other is a specific weight-shape mismatch; not enough to treat as the same bug." + "reason": "Qwen3 MoE training memory usage and Strix Halo mmap OOM are different memory problems in different contexts." }, { "left": "issue:43425", "right": "issue:44918", "accept": false, - "reason": "Torch incompatibility and Qwen3.5 embedding unpacking with TRL are unrelated." + "reason": "Torch 2.10 incompatibility is unrelated to Qwen3.5 embedding unpacking in TRL SFT trainer." }, { "left": "issue:44589", "right": "issue:45084", "accept": false, - "reason": "Float8 storage lookup failure and template-node compile error are unrelated." + "reason": "Float8 storage object lookup and template-node compilation errors are different failures." }, { - "left": "issue:43856", - "right": "issue:44756", + "left": "issue:40990", + "right": "issue:42947", "accept": false, - "reason": "Qwen3 MoE memory inefficiency and Strix Halo mmap OOM are different performance issues." + "reason": "High perplexity on gpt-oss-20b and gradient checkpointing ineffective with PEFT LoRA are unrelated issues." } ] }, "evaluator_result": { "accept": true, - "feedback": "The cluster summary is grounded in the packet and stays conservative. The only clearly supported duplicate family is the `resize_token_embeddings` trio (`issue:45292`, `issue:45276`, `issue:45335`), and the other soft-edge pairs are appropriately rejected as different bugs." + "feedback": "The summary is grounded in the packet: issue:45292 is the clearest canonical issue for the resize_token_embeddings/output_embeddings family, and the two accepted soft pairs are well-supported by the issue titles/descriptions. The other soft-edge rejections are conservative and do not overclaim shared bugs." }, "evaluator_used": true, "retried": false, @@ -2410,7 +2295,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:32090", "issue:34567", @@ -2419,7 +2304,6 @@ "issue:40444", "issue:40990", "issue:42915", - "issue:42947", "issue:43317", "issue:43404", "issue:43421", @@ -2462,14 +2346,13 @@ "issue:45759" ], "soft_pairs": [ - "issue:43596|issue:44683", - "issue:40990|issue:42947", "issue:43425|issue:45685", - "issue:44164|issue:45759", + "issue:39404|issue:44671", "issue:44671|issue:44991", "issue:43994|issue:44079", - "issue:39404|issue:44671", "issue:36032|issue:43452", + "issue:44164|issue:45759", + "issue:43596|issue:44683", "issue:44991|issue:45081", "issue:32090|issue:39404", "issue:44164|issue:44479", @@ -2486,8 +2369,8 @@ "issue:44811|issue:45584", "issue:43644|issue:43646", "issue:43883|issue:44991", - "issue:44451|issue:44987", "issue:36032|issue:42915", + "issue:44451|issue:44987", "issue:43541|issue:45084", "issue:43452|issue:43482", "issue:42915|issue:44530", @@ -2497,244 +2380,235 @@ "issue:36032|issue:43645", "issue:45084|issue:45507" ], - "prepared_review_unit_hash": "ababc18109d012b5bcba4a1587900d027e40f5608f69b437aef779a9285f472d", + "prepared_review_unit_hash": "98202ebf9447becc4f9e01933fe2749f2f0d39660ec3bb6fc78f8637804b8aca", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 34, - "serialized_chars": 37900, - "estimated_input_tokens": 9475, - "estimated_eval_tokens": 19206 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37019, + "estimated_input_tokens": 9255, + "estimated_eval_tokens": 18766 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous set of closed Transformers issues. There is some light thematic overlap around v5 loading/init regressions, tokenizer/model loading, and Whisper/Qwen/Mistral edge cases, but the soft pairs do not look like true duplicates of the same concrete bug.", - "confidence": 0.89, - "canonical_issue_reason": "issue:43611 is the broadest representative of the v5 model-loading regression family in this set; several nearby items are narrower variants, but most others are unrelated enough that it should serve only as a cluster anchor, not a true umbrella duplicate.", + "summary": "The cluster is mostly superficial similarity across unrelated regression reports. The only remotely close pair is 44206/44220, but the evidence is still too thin to treat it as a confirmed duplicate without shared targets, filenames, or explicit linkage.", + "confidence": 0.66, + "canonical_issue_reason": "Issue 44206 is the most concrete audio-preprocessing report in the set, with the clearest crash path and symptoms, but it should only be treated as the best representative of a weakly related pair, not as a proven duplicate anchor.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43611 is the best single issue to represent the cluster because it is broad, clearly regression-oriented, and closest to the shared v5 loading/init theme seen in a few nearby issues.", + "best_issue_reason": "Issue 44206 is the strongest issue candidate overall because it is specific and actionable, and it is the closest thing here to a shared underlying bug among the soft-similarity pairs.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43596", - "right": "issue:44683", - "accept": false, - "reason": "Different failures: DeepSpeed ZeRO3/BertModel init vs compiled flex_attention on torch>=2.9. No shared code path." - }, - { - "left": "issue:40990", - "right": "issue:42947", - "accept": false, - "reason": "One is high perplexity on a specific model/dataset; the other is LoRA gradient checkpointing behavior. Same training area, but not the same bug." - }, { "left": "issue:43425", "right": "issue:45685", "accept": false, - "reason": "Unrelated compatibility problem vs an MPS histogram kernel error in MoE code." + "reason": "Torch version compatibility and an MPS runtime TypeError are different failures." }, { - "left": "issue:44164", - "right": "issue:45759", + "left": "issue:39404", + "right": "issue:44671", "accept": false, - "reason": "Serialization of extra_state vs config unwrapping for composite multimodal CausalLM configs; different mechanisms and symptoms." + "reason": "Whisper pipeline language return and CamemBERT masked-LM predictions are unrelated code paths." }, { "left": "issue:44671", "right": "issue:44991", "accept": false, - "reason": "Both involve loading/tokenization regressions, but one is CamemBERT masked-LM predictions and the other is tokenizer loading for a specific RoBERTa model." + "reason": "Both are model/tokenizer regressions, but they affect different models and different failure modes." }, { "left": "issue:43994", "right": "issue:44079", "accept": false, - "reason": "One is wrong outputs for SigLIP2; the other is ModelOutput key assignment when a previous key is None. Not the same defect." - }, - { - "left": "issue:39404", - "right": "issue:44671", - "accept": false, - "reason": "Whisper pipeline return_language regression vs CamemBERT masked-LM prediction bug; unrelated models and code paths." + "reason": "Bad SigLIP2 outputs and ModelOutput key assignment are different underlying bugs." }, { "left": "issue:36032", "right": "issue:43452", "accept": false, - "reason": "Tokenizer method/name conflict during load vs GGUF/from_pretrained breakage. Both are loading issues, but not the same underlying bug." + "reason": "T5 tokenizer method conflicts do not match GGUF loading failures." }, { - "left": "issue:44991", + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "Extra_state save/load handling is distinct from composite Qwen config unwrapping." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "ZeRO3/BertModel init and compiled flex_attention on newer torch are unrelated regressions." + }, + { + "left": "issue:44991", "right": "issue:45081", "accept": false, - "reason": "Different tokenizer-loading failures for different models; no evidence they share the same tokenizer backend bug." + "reason": "Tokenizer load failure and Mistral regex/backend_tokenizer errors are different issues." }, { "left": "issue:32090", "right": "issue:39404", "accept": false, - "reason": "Trainer GPU broadcast TypeError vs Whisper pipeline language return regression. No meaningful overlap beyond both being runtime errors." + "reason": "Trainer GPU broadcast NoneType and Whisper return_language are unrelated." }, { "left": "issue:44164", "right": "issue:44479", "accept": false, - "reason": "State-dict/extra_state handling vs video-input regression for Qwen VL models. Different inputs and failure modes." + "reason": "Extra_state serialization and Qwen video-input regression are different failures." }, { "left": "issue:44488", "right": "issue:45081", "accept": false, - "reason": "Loading a specific BERT model vs a Mistral regex patch crash; unrelated model families and code paths." + "reason": "These are different model-loading bugs with different subsystems involved." }, { "left": "issue:40990", "right": "issue:43421", "accept": false, - "reason": "Perplexity regression on a model vs runtime special-token post-processor update behavior. Not the same bug." + "reason": "Perplexity behavior and runtime special-token post-processing are not the same bug." }, { "left": "issue:43582", "right": "issue:43632", "accept": false, - "reason": "Apple Silicon caching allocator warmup TypeError vs v5 _is_hf_initialized flag breakage. Different subsystems." + "reason": "Apple Silicon allocator warmup and `_is_hf_initialized` are unrelated." }, { "left": "issue:43883", "right": "issue:44488", "accept": false, - "reason": "Missing all_tied_weights_keys on Molmo vs model loading failure for cjvt/sleng-bert. No shared concrete failure." + "reason": "Missing tied-weights metadata and a specific model load failure are different issues." }, { "left": "issue:43611", "right": "issue:43644", "accept": false, - "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is non-persistent buffers serialization. Broad theme only, not one bug." + "reason": "Both are v5 regressions, but they affect different loading/initialization mechanics." }, { "left": "issue:44206", "right": "issue:44220", "accept": false, - "reason": "Unsupported center arg in LASRFeatureExtractor vs _torch_extract_fbank_features issue. Related audio area, but different concrete breakages." + "reason": "Both look audio-related, but this packet only shows title-level similarity and no shared target, file, or deterministic evidence of the same crash." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Mistral image preprocessor dtype mismatch vs templating/compiler error. Different models and failure layers." + "reason": "Image preprocessor dtype mismatch and template compilation errors are unrelated." }, { "left": "issue:44488", "right": "issue:44987", "accept": false, - "reason": "Both are model-load failures, but for different repos and likely different code paths; no same underlying defect is evident." - }, - { - "left": "issue:34567", - "right": "issue:40444", - "accept": false, - "reason": "TrainerState token counting vs Qwen2.5-VL iterable dataset with multiple images. Completely different functionality." + "reason": "These are different model-specific loading regressions." }, { "left": "issue:43643", "right": "issue:44530", "accept": false, - "reason": "trust_remote_code config field loss vs PagedAttentionCache linear_attention error. Different behaviors and subsystems." + "reason": "Config field loss with trust_remote_code is not the same as PagedAttentionCache linear_attention errors." }, { "left": "issue:44811", "right": "issue:45584", "accept": false, - "reason": "Whisper batch_decode skip_special_tokens handling vs empty-transcription failure after align_special_tokens. Same model family, but not the same bug." + "reason": "Both are Whisper-related, but decode flag handling and empty-transcription generation are different bugs." }, { "left": "issue:43644", "right": "issue:43646", "accept": false, - "reason": "Both are v5 custom-model regressions, but one is junk in non-persistent buffers and the other is custom model initialization. Too different to merge." + "reason": "Non-persistent buffer junk and custom model initialization are not the same concrete failure." }, { "left": "issue:43883", "right": "issue:44991", "accept": false, - "reason": "Molmo tied-weights attribute error vs tokenizer loading for EMBEDDIA/est-roberta. No common code path." + "reason": "Tied-weights attribute errors and tokenizer loading regressions do not share the same root cause." }, { - "left": "issue:44451", - "right": "issue:44987", + "left": "issue:36032", + "right": "issue:42915", "accept": false, - "reason": "Different model-loading failures for different models; one cannot be treated as a duplicate of the other without stronger evidence." + "reason": "T5 tokenizer loading and Qwen3Moe FP8 config failures are unrelated." }, { - "left": "issue:36032", - "right": "issue:42915", + "left": "issue:44451", + "right": "issue:44987", "accept": false, - "reason": "Tokenizer attribute conflict vs FineGrainedFP8Config failure on Qwen3Moe. Distinct bugs." + "reason": "Different checkpoint-loading regressions with no shared code-path evidence." }, { "left": "issue:43541", "right": "issue:45084", "accept": false, - "reason": "Grouped_mm tracing error in Mixtral float32 vs template compilation error. Unrelated." + "reason": "Grouped-mm tracing errors and non-template-node compilation errors are unrelated." }, { "left": "issue:43452", "right": "issue:43482", "accept": false, - "reason": "Both mention GGUF loading, but one is a broad from_pretrained regression and the other is a specific Qwen2.5-GGUF failure; same area, not clearly the same concrete bug." + "reason": "Both involve loading, but one is a broad API break and the other is a model-specific v5 failure." }, { "left": "issue:42915", "right": "issue:44530", "accept": false, - "reason": "Fine-grained FP8 config failure vs PagedAttentionCache linear_attention crash. Different mechanisms." + "reason": "FP8 configuration failure and linear_attention cache issues are different Qwen/MoE bugs." }, { "left": "issue:43317", "right": "issue:43482", "accept": false, - "reason": "device_map=auto dequantized model offload failure vs v5 GGUF loading failure. Different loading stacks." + "reason": "GPU+CPU offload dequantized loading and GGUF loading failures are different problems." }, { "left": "issue:44849", "right": "issue:45325", "accept": false, - "reason": "Qwen3.5 output_hidden_states regression vs Qwen2.5-VL rope index scaling bug. Different model families and symptoms." + "reason": "Qwen3.5 hidden-state behavior and Qwen2.5-VL rope index scaling are distinct regressions." }, { "left": "issue:43404", "right": "issue:44466", "accept": false, - "reason": "Tie-weight bug in Mistral3 conditional generation vs device-dependent lm_head serialization inconsistency. Related conceptually, but not the same issue." + "reason": "Untied lm_head weights and device-dependent serialization of tied weights are not the same bug." }, { "left": "issue:36032", "right": "issue:43645", "accept": false, - "reason": "Tokenizer method conflict vs Jupyter custom-model initialization breakage. Different failures." + "reason": "Tokenizer method conflicts and Jupyter custom-model initialization breakage are different issues." }, { "left": "issue:45084", "right": "issue:45507", "accept": false, - "reason": "Template compilation failure vs GraniteMoEHybrid calling an invalid method. Different code paths and symptoms." + "reason": "Template compilation errors and GraniteMoEHybrid invalid-method calls are unrelated." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, + "evaluator_result": { + "accept": false, + "feedback": "The review is mostly grounded, but it overstates the evidence for the one accepted soft pair. The packet only shows title-level soft similarity between issue:44206 and issue:44220, with no deterministic accept or shared targets/files, so calling it a true duplicate / the only real duplicate relationship is stronger than the evidence supports. A more conservative verdict would be to mark that pair as possible or uncertain duplicate rather than accepted." + }, + "evaluator_used": true, + "retried": true, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -2750,12 +2624,11 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:36032", "issue:40990", "issue:41553", - "issue:42175", "issue:42915", "issue:43262", "issue:43296", @@ -2763,7 +2636,7 @@ "issue:43322", "issue:43377", "issue:43404", - "issue:43425", + "issue:43421", "issue:43531", "issue:43540", "issue:43582", @@ -2777,6 +2650,7 @@ "issue:43856", "issue:43994", "issue:44062", + "issue:44079", "issue:44220", "issue:44265", "issue:44291", @@ -2784,6 +2658,7 @@ "issue:44451", "issue:44479", "issue:44534", + "issue:44568", "issue:44589", "issue:44610", "issue:44683", @@ -2798,254 +2673,254 @@ "issue:45216", "issue:45325", "issue:45357", - "issue:45542" + "issue:45663" ], "soft_pairs": [ "issue:43262|issue:43377", "issue:43646|issue:44534", - "issue:44987|issue:44991", "issue:43994|issue:44610", - "issue:44403|issue:45081", + "issue:44987|issue:44991", "issue:43322|issue:43540", + "issue:44403|issue:45081", "issue:43377|issue:43819", - "issue:44220|issue:44683", + "issue:45216|issue:45357", "issue:44451|issue:45081", - "issue:42915|issue:45084", "issue:43645|issue:44534", - "issue:45216|issue:45357", "issue:44265|issue:44610", + "issue:42915|issue:45084", + "issue:44220|issue:44683", "issue:44987|issue:45081", "issue:43404|issue:45127", "issue:41553|issue:43582", "issue:43632|issue:44756", "issue:43296|issue:43531", - "issue:42915|issue:44291", "issue:44062|issue:44987", "issue:36032|issue:43742", + "issue:42915|issue:44291", "issue:43531|issue:44877", - "issue:42915|issue:44589", "issue:43611|issue:44534", "issue:40990|issue:44811", + "issue:42915|issue:44589", "issue:44479|issue:45325", "issue:43317|issue:43856", - "issue:43425|issue:44403", - "issue:42175|issue:45542", - "issue:43818|issue:45357" + "issue:43421|issue:44568", + "issue:43818|issue:45357", + "issue:44079|issue:45663" ], - "prepared_review_unit_hash": "b59561a1e77af6d20a565df9f6d7bf9984dec5e0ba581438c7190ac6a71c55ff", + "prepared_review_unit_hash": "0a58aa9477d8e649a17c32d8692c24ea9f6ab27b98107ebead28dee6ad2225b5", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36040, - "estimated_input_tokens": 9010, - "estimated_eval_tokens": 18276 + "serialized_chars": 36779, + "estimated_input_tokens": 9195, + "estimated_eval_tokens": 18646 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a noisy mix of unrelated Transformers regressions. I found one clear duplicate pair (Qwen3.5 save_pretrained / visual encoder keys) and several other pairs that are only loosely similar, not true duplicates.", - "confidence": 0.78, - "canonical_issue_reason": "Issue 45216 is the best canonical anchor: it is the earlier, broader report for the Qwen3.5 save_pretrained regression, and 45357 reads like the same bug resurfacing with incorrect visual encoder keys.", + "summary": "This cluster is mostly a grab-bag of distinct model- or feature-specific regressions. The only clear duplicate thread is the Qwen3.5 `save_pretrained` checkpoint bug; most other soft pairs are too broad or too different in code path to merge.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 45216 is the best representative of the only clear duplicate thread here: it states the Qwen3.5 `save_pretrained` regression at a higher level, while 45357 narrows it to the visual-encoder key mapping symptom.", "canonical_pr_reason": null, - "best_issue_reason": "45216 is the cleanest representative because it is the original report in the only clearly duplicated subcluster.", + "best_issue_reason": "45216 is the strongest anchor for deduping this batch because it captures a concrete, reproducible serialization regression and has a later follow-up that looks like the same bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43262", "right": "issue:43377", "accept": false, - "reason": "Different bugs: audio processor chat-template sampling-rate default vs MIMI padding-mask batching mismatch." + "reason": "Different bugs: audio chat-template sampling-rate default vs MIMI batching/padding-mask mismatch." }, { "left": "issue:43646", "right": "issue:44534", "accept": false, - "reason": "Different failures: custom model init vs non-persistent buffer corruption in v5." - }, - { - "left": "issue:44987", - "right": "issue:44991", - "accept": false, - "reason": "Both are loading failures, but for different models and likely different root causes (physical-intelligence/fast vs EMBEDDIA/est-roberta tokenizer)." + "reason": "One is custom model initialization; the other is non-persistent buffer contents in v5." }, { "left": "issue:43994", "right": "issue:44610", "accept": false, - "reason": "Unrelated model behavior bugs: nonsensical outputs vs wrong processor input resolution." + "reason": "Different model families and symptoms: SigLIP2 nonsense outputs vs OmDet-Turbo input-size mismatch." }, { - "left": "issue:44403", - "right": "issue:45081", + "left": "issue:44987", + "right": "issue:44991", "accept": false, - "reason": "Different issues: generic loading noise vs Mistral tokenizer regex patch crash on a missing backend_tokenizer attribute." + "reason": "Both are loading failures, but on different models and likely different code paths; not the same concrete bug." }, { "left": "issue:43322", "right": "issue:43540", "accept": false, - "reason": "Different multimodal failures: Llava Next segfault vs Qwen3OmniMoe video-processing ValueError." + "reason": "Llava Next load-time segfault is unrelated to Qwen3OmniMoe video processing ValueError." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Generic load-time noise is not the same as a tokenizer patch crashing on Mistral regex fixing." }, { "left": "issue:43377", "right": "issue:43819", "accept": false, - "reason": "Different code paths: batched MIMI encoder padding-mask bug vs DAC.from_latents / STE mismatch." + "reason": "Batch-vs-single padding behavior in MIMI is unrelated to DAC latent/forward mismatch." }, { - "left": "issue:44220", - "right": "issue:44683", - "accept": false, - "reason": "Different subsystems: fbank feature extraction vs compiled flex_attention on newer torch." + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Same Qwen3.5 `save_pretrained` serialization regression; 45357 looks like a later, more specific manifestation of the same incorrect checkpoint-key handling." }, { "left": "issue:44451", "right": "issue:45081", "accept": false, - "reason": "Both are tokenizer/loading problems, but the affected models and failures differ." + "reason": "Different models and failures: ScandiBERT loading vs Mistral tokenizer regex patch crash." }, { - "left": "issue:42915", - "right": "issue:45084", + "left": "issue:43645", + "right": "issue:44534", "accept": false, - "reason": "Completely different problems: Qwen3Moe FineGrainedFP8Config vs template compilation error." + "reason": "Notebook custom-model initialization regression is not the same bug as junk-filled non-persistent buffers." }, { - "left": "issue:43645", - "right": "issue:44534", + "left": "issue:44265", + "right": "issue:44610", "accept": false, - "reason": "Different regressions: notebook custom-model initialization vs junk-filled non-persistent buffers." + "reason": "Export/checking logic failure is unrelated to processor image-size mismatch." }, { - "left": "issue:45216", - "right": "issue:45357", - "accept": true, - "reason": "Same underlying Qwen3.5 save_pretrained regression; 45357 is a later symptom about incorrect visual encoder keys." + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "Qwen3Moe FP8 loading failure and template-node compilation error are different issues." }, { - "left": "issue:44265", - "right": "issue:44610", + "left": "issue:44220", + "right": "issue:44683", "accept": false, - "reason": "Different failures: torch.export with torch_compilable_check vs OmDet-Turbo processor size mismatch." + "reason": "Audio feature extraction bug and compiled flex_attention regression do not share the same code path." }, { "left": "issue:44987", "right": "issue:45081", "accept": false, - "reason": "Different loading bugs affecting different models and tokenizer internals." + "reason": "Different tokenizer/model-loading regressions on different models." }, { "left": "issue:43404", "right": "issue:45127", "accept": false, - "reason": "Related theme (weights/tied embeddings) but not the same concrete bug: Mistral3 lm_head tying vs LoRA merge collapse with extended vocab." + "reason": "Weight-tying bug in Mistral3 is unrelated to LoRA merge collapse after vocabulary extension." }, { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "Different platforms and bugs: bad AutoTokenizer error message vs Apple Silicon warmup TypeError." + "reason": "Bad tokenizer error message vs Apple Silicon allocator TypeError are unrelated." }, { "left": "issue:43632", "right": "issue:44756", "accept": false, - "reason": "Different regressions: _is_hf_initialized flag handling vs mmap OOM on Strix Halo." + "reason": "_is_hf_initialized / init regression and Strix Halo mmap OOM are unrelated." }, { "left": "issue:43296", "right": "issue:43531", "accept": false, - "reason": "Different issues: vLLM PaddleOCR-VL loading vs Qwen3-MoE sliding_window behavior." - }, - { - "left": "issue:42915", - "right": "issue:44291", - "accept": false, - "reason": "Both involve model loading, but the concrete failure modes differ (FP8 config vs init_empty_weights/_is_hf_initialized argument)." + "reason": "PaddleOCR-VL loading failure and Qwen3-MoE sliding_window issue are distinct model bugs." }, { "left": "issue:44062", "right": "issue:44987", "accept": false, - "reason": "Different tokenizer/model loading errors: AddedToken special-arg conflict vs model-specific load failure." + "reason": "AddedToken special-arg duplication is unrelated to model/tokenizer loading regressions on named checkpoints." }, { "left": "issue:36032", "right": "issue:43742", "accept": false, - "reason": "Different tokenizer/model load failures; no shared concrete bug." + "reason": "T5Tokenizer special-token conflict and MobileLLM key error are different loading problems." }, { - "left": "issue:43531", - "right": "issue:44877", + "left": "issue:42915", + "right": "issue:44291", "accept": false, - "reason": "Different model/config problems: Qwen3-MoE sliding_window vs granite_speech strict config loading." + "reason": "Different failure modes: Qwen3Moe FP8 config issue vs `_is_hf_initialized` argument error." }, { - "left": "issue:42915", - "right": "issue:44589", + "left": "issue:43531", + "right": "issue:44877", "accept": false, - "reason": "Different bugs: FineGrainedFP8Config failure vs missing Float8 storage object." + "reason": "MoE sliding_window behavior and strict config preventing granite_speech loading are unrelated." }, { "left": "issue:43611", "right": "issue:44534", "accept": false, - "reason": "Different regressions: base_model_prefix loading breakage vs junk non-persistent buffers." + "reason": "base_model_prefix loading breakage is not the same as non-persistent buffer corruption." }, { "left": "issue:40990", "right": "issue:44811", "accept": false, - "reason": "Different evaluation/processor issues: perplexity anomaly vs batch_decode skip_special_tokens being ignored." + "reason": "Perplexity regression and Whisper batch_decode special-token handling are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "Qwen3Moe FP8 failure and missing Float8 storage type are separate low-level issues." }, { "left": "issue:44479", "right": "issue:45325", "accept": false, - "reason": "Related video/temporal-position topic, but 44479 is a broad multi-model video regression while 45325 is a specific Qwen2.5-VL rope-index scaling bug; not clearly the same duplicate." + "reason": "Both involve Qwen VL/video behavior, but the titles point to different symptoms and input modes; not clearly the same concrete bug." }, { "left": "issue:43317", "right": "issue:43856", "accept": false, - "reason": "Different memory/offload issues: device_map=auto dequantized offload vs Qwen3 MoE training memory usage." + "reason": "device_map offload failure and Qwen3 MoE training memory inefficiency are different problems." }, { - "left": "issue:43425", - "right": "issue:44403", + "left": "issue:43421", + "right": "issue:44568", "accept": false, - "reason": "Different topics: torch version compatibility vs noisy model loading." + "reason": "Runtime special-token post-processor sync is not the same bug as BOS/EOS not being added for mdeberta-v3-base." }, { - "left": "issue:42175", - "right": "issue:45542", + "left": "issue:43818", + "right": "issue:45357", "accept": false, - "reason": "Both touch TensorFlow packaging, but one is missing TF in the torch extra and the other is a tensorboard-only / undefined tf backend issue." + "reason": "Video-LLaVA temporal-attention/weight-sharing bug is unrelated to Qwen3.5 save_pretrained key serialization." }, { - "left": "issue:43818", - "right": "issue:45357", + "left": "issue:44079", + "right": "issue:45663", "accept": false, - "reason": "Different models and bugs: Video-LLaVA tower/weight-sharing issue vs Qwen3.5 checkpoint saving regression." + "reason": "ModelOutput key assignment bug and Gemma4 FSDP2 shared_kv_states KeyError are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The analysis is grounded in the packet and the verdicts are mostly conservative. The one accepted pair (45216/45357) is plausibly the same Qwen3.5 save_pretrained regression, and the rest are framed as non-duplicates without overstating evidence. No clear overclaim stands out." + "feedback": "The summary is grounded in the packet and stays conservative overall. The only accepted soft-edge pair, 45216/45357, is reasonably supported by the titles as the same Qwen3.5 save_pretrained regression with a more specific follow-up symptom. The remaining false verdicts are appropriately cautious and do not overstate duplicate evidence." }, "evaluator_used": true, "retried": false, @@ -3064,7 +2939,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:33666", "issue:35707", @@ -3075,7 +2950,7 @@ "issue:42915", "issue:43296", "issue:43377", - "issue:43421", + "issue:43425", "issue:43454", "issue:43482", "issue:43525", @@ -3097,7 +2972,6 @@ "issue:44466", "issue:44488", "issue:44534", - "issue:44568", "issue:44589", "issue:44617", "issue:44683", @@ -3113,18 +2987,19 @@ "issue:45479", "issue:45491", "issue:45538", - "issue:45663" + "issue:45542", + "issue:45800" ], "soft_pairs": [ - "issue:43421|issue:44568", - "issue:44079|issue:45663", + "issue:43425|issue:44403", + "issue:42175|issue:45542", "issue:36032|issue:41720", - "issue:44683|issue:45020", "issue:43632|issue:44534", + "issue:33666|issue:40444", + "issue:44683|issue:45020", "issue:43792|issue:44589", "issue:44403|issue:44991", "issue:43482|issue:45538", - "issue:33666|issue:40444", "issue:44829|issue:45479", "issue:40444|issue:45325", "issue:43482|issue:43525", @@ -3133,263 +3008,273 @@ "issue:41720|issue:45081", "issue:44265|issue:45072", "issue:42175|issue:43296", - "issue:43632|issue:45072", - "issue:43531|issue:44291", "issue:43611|issue:43646", + "issue:43632|issue:45072", "issue:44079|issue:44534", + "issue:43531|issue:44291", "issue:43994|issue:44898", "issue:43454|issue:44466", "issue:43881|issue:44488", - "issue:43828|issue:44683", "issue:42915|issue:43482", - "issue:43792|issue:43957", + "issue:43828|issue:44683", "issue:44079|issue:45072", "issue:43644|issue:44079", + "issue:43792|issue:43957", "issue:40444|issue:45381", "issue:45325|issue:45381", + "issue:43296|issue:45800", "issue:44387|issue:44617", "issue:35707|issue:40444", "issue:44291|issue:44877" ], - "prepared_review_unit_hash": "5c1606e1cb89007b4afe1b059f942e80721b50e691f95312a442732307e948ec", + "prepared_review_unit_hash": "7b6cec7d3b4f4222ed8ca3541091fe59e07ce299d9cb8c0a30d984f800b25d19", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 34, - "serialized_chars": 38078, - "estimated_input_tokens": 9520, - "estimated_eval_tokens": 19296 + "soft_pair_count": 35, + "serialized_chars": 38174, + "estimated_input_tokens": 9544, + "estimated_eval_tokens": 19344 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a loose cluster of mostly unrelated Transformers bug reports, many about v5-era loading/init regressions, but the soft pairs do not appear to be true duplicates or mergeable into one fix. No soft edges should be accepted.", - "confidence": 0.79, - "canonical_issue_reason": "issue:44291 is the best representative of the cluster\u2019s main theme: a concrete Transformers v5 model-loading/init regression. It is broad enough to stand in for the dominant pattern without overclaiming exact duplication.", + "summary": "This cluster is mostly a loose group of Transformers v5 regression reports, with one clear duplicate-like pair around TensorFlow/backend installation. Most other pairs are only thematically similar, not the same bug.", + "confidence": 0.78, + "canonical_issue_reason": "43632 is the best representative because it captures a concrete Transformers v5 model-initialization regression that sits near several other loading/init failures in the set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44291 is the strongest single issue to anchor the cluster because it is a clear, broadly relevant v5 initialization/loading failure and closer to the common pattern than the more model-specific reports.", + "best_issue_reason": "43632 is also the strongest global issue choice: it is concrete, clearly bug-focused, and closest to the recurring v5 loading/initialization breakages across the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43421", - "right": "issue:44568", + "left": "issue:43425", + "right": "issue:44403", "accept": false, - "reason": "Both are tokenizer/special-token reports, but one is runtime post-processor syncing and the other is BOS/EOS not being added; different code paths and symptoms." + "reason": "Unrelated failures: one is torch version incompatibility, the other is a loading-noise/reporting issue." }, { - "left": "issue:44079", - "right": "issue:45663", - "accept": false, - "reason": "ModelOutput key assignment vs Gemma-4 FSDP shared_kv_states rebuild is unrelated; no shared bug mechanism." + "left": "issue:42175", + "right": "issue:45542", + "accept": true, + "reason": "Both describe the same backend packaging bug where installing the torch extra leaves TensorFlow backend support broken/undefined." }, { "left": "issue:36032", "right": "issue:41720", "accept": false, - "reason": "T5 tokenizer method-name conflict and Qwen3 auto-device-map CUDA assert are entirely different problems." + "reason": "Different models and failures: T5 tokenizer load conflict vs Qwen3 auto device-map CUDA assert." }, { - "left": "issue:44683", - "right": "issue:45020", + "left": "issue:43632", + "right": "issue:44534", "accept": false, - "reason": "Torch 2.9 flex_attention compilation failure is unrelated to remote_code model loading regressions." + "reason": "Both mention v5 regressions, but they are different concrete bugs: `_is_hf_initialized` handling vs non-persistent buffer serialization." }, { - "left": "issue:43632", - "right": "issue:44534", + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Both are Qwen2-VL/Qwen2.5-VL training issues, but the reported failures are different and not clearly the same code-path bug." + }, + { + "left": "issue:44683", + "right": "issue:45020", "accept": false, - "reason": "Both touch v5 initialization behavior, but one is the _is_hf_initialized flag and the other is non-persistent buffer junk; not the same concrete bug." + "reason": "Different scope: torch>=2.9 flex_attention compilation failure vs remote_code loading regressions." }, { "left": "issue:43792", "right": "issue:44589", "accept": false, - "reason": "Whisper failing to run and Float8 storage lookup errors do not point to the same underlying issue." + "reason": "Different root causes and symptoms: Whisper model runtime failure vs missing Float8 storage type." }, { "left": "issue:44403", "right": "issue:44991", "accept": false, - "reason": "Loading noise is a logging/UI issue; tokenizer load failure for EMBEDDIA/est-roberta is a functional tokenizer bug." + "reason": "Unrelated: logging noise on load vs tokenizer loading failure for a specific model." }, { "left": "issue:43482", "right": "issue:45538", "accept": false, - "reason": "Qwen2.5-GGUF loading failure and CLIPTokenizer model_max_length behavior are unrelated." - }, - { - "left": "issue:33666", - "right": "issue:40444", - "accept": false, - "reason": "Qwen2-VL multi-GPU training and Qwen2.5-VL iterable-dataset multi-image finetuning are different training-path failures." + "reason": "Different components: GGUF/model loading under v5 vs CLIPTokenizer max length behavior." }, { "left": "issue:44829", "right": "issue:45479", "accept": false, - "reason": "FlashAttention3 sequence-classification degeneration and single-label num_labels=1 zero-loss are different causes of bad training." + "reason": "Different training pathologies: flash_attention_3 degenerate training vs binary classification zero-loss issue." }, { "left": "issue:40444", "right": "issue:45325", "accept": false, - "reason": "Multiple-images-per-prompt finetuning failure and still-image rope-index scaling are both Qwen2.5-VL, but not the same bug." + "reason": "Both concern Qwen2.5-VL, but one is multi-image iterable-dataset finetuning and the other is rope/index scaling for still images." }, { "left": "issue:43482", "right": "issue:43525", "accept": false, - "reason": "GGUF loading failure for Qwen2.5 and missing pad_token_id on Llama4Config are separate model-loading regressions." + "reason": "Different bugs: GGUF loading failure vs missing `pad_token_id` on Llama4Config." }, { "left": "issue:44291", "right": "issue:44589", "accept": false, - "reason": "Unexpected _is_hf_initialized argument and missing Float8 storage object are distinct v5 incompatibilities." + "reason": "Both are loading/init-related, but one is an unexpected `_is_hf_initialized` argument and the other is a float8 storage lookup failure." }, { "left": "issue:43377", "right": "issue:45491", "accept": false, - "reason": "MIMI padding-mask batching mismatch and Gemma3 mixed-length NaN embeddings have different failure modes and code paths." + "reason": "Different model families and different symptoms: MIMI padding-mask discrepancy vs Gemma3 NaN embeddings with sliding-window attention." }, { "left": "issue:41720", "right": "issue:45081", "accept": false, - "reason": "CUDA assert during Qwen3 auto device mapping is unrelated to Mistral tokenizer regex patching." + "reason": "Different code paths: Qwen3 auto device mapping CUDA assert vs Mistral tokenizer regex patch crash." }, { "left": "issue:44265", "right": "issue:45072", "accept": false, - "reason": "torch.export with torch_compilable_check and bfloat16 dtype mismatches in CI are unrelated." + "reason": "Different failures: torch.export with torch_compilable_check vs bfloat16 dtype mismatches in CI." }, { "left": "issue:42175", "right": "issue:43296", "accept": false, - "reason": "Packaging/backend selection for TensorFlow and PaddleOCR-VL loading in vLLM are unrelated." + "reason": "Different packages and failure modes: missing TensorFlow backend in install vs PaddleOCR-VL loading failure in vLLM." }, { - "left": "issue:43632", - "right": "issue:45072", + "left": "issue:43611", + "right": "issue:43646", "accept": false, - "reason": "_is_hf_initialized breakage and bfloat16 inference dtype mismatch are different v5 regressions." + "reason": "Both are v5 model-loading regressions, but the concrete bugs differ: base_model_prefix loading vs custom initialization behavior." }, { - "left": "issue:43531", - "right": "issue:44291", + "left": "issue:43632", + "right": "issue:45072", "accept": false, - "reason": "Sliding-window issue in Qwen3-MoE is unrelated to init_empty_weights/_is_hf_initialized handling." + "reason": "Different issues: initialization-flag breakage vs dtype mismatch in inference." }, { - "left": "issue:43611", - "right": "issue:43646", + "left": "issue:44079", + "right": "issue:44534", "accept": false, - "reason": "base_model_prefix loading and custom model initialization both involve v5 loading, but they are not the same specific bug." + "reason": "Different problems: ModelOutput key assignment vs junk-filled non-persistent buffers." }, { - "left": "issue:44079", - "right": "issue:44534", + "left": "issue:43531", + "right": "issue:44291", "accept": false, - "reason": "ModelOutput key assignment and buffer initialization junk are separate implementation bugs." + "reason": "Different failures: Qwen3-MoE sliding-window behavior vs init_empty_weights argument error." }, { "left": "issue:43994", "right": "issue:44898", "accept": false, - "reason": "SigLIP2 nonsensical outputs via AutoModel/pipeline and Perceiver non-default resolution failure are unrelated model-specific issues." + "reason": "Different models and symptoms: SigLIP2 nonsensical outputs vs Perceiver resize/interpolation failure." }, { "left": "issue:43454", "right": "issue:44466", "accept": false, - "reason": "AyaVision lm_head weight tying and inconsistent lm_head serialization depending on device are related only broadly by weights; not the same bug." + "reason": "Related to lm_head/tied weights, but one is broken tying causing garbage generation and the other is serialization inconsistency by device." }, { "left": "issue:43881", "right": "issue:44488", "accept": false, - "reason": "glm-4v-9b load failure and cjvt/sleng-bert tokenizer loading are unrelated." - }, - { - "left": "issue:43828", - "right": "issue:44683", - "accept": false, - "reason": "torch.autocast dtype mismatch in Phi-tiny-MoE is unrelated to compiled flex_attention failing on torch >=2.9." + "reason": "Different models and failures: glm-4v-9b load failure vs cjvt/sleng-bert load failure." }, { "left": "issue:42915", "right": "issue:43482", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure and Qwen2.5-GGUF loading failure are different model/config paths." + "reason": "Different model stacks and causes: FineGrainedFP8Config failure vs GGUF loading failure." }, { - "left": "issue:43792", - "right": "issue:43957", + "left": "issue:43828", + "right": "issue:44683", "accept": false, - "reason": "Whisper load/run failure and meta-device loading regression are both loading-related but not the same code-path problem." + "reason": "Different numerical-path bugs: autocast dtype mismatch vs compiled flex_attention failure on newer torch." }, { "left": "issue:44079", "right": "issue:45072", "accept": false, - "reason": "ModelOutput key bookkeeping and dtype mismatch inference issues are unrelated." + "reason": "Different subsystems: ModelOutput key handling vs inference dtype mismatch." }, { "left": "issue:43644", "right": "issue:44079", "accept": false, - "reason": "Non-persistent buffer junk and ModelOutput key assignment both concern object state, but not the same failure." + "reason": "Different regressions: non-persistent buffer values vs ModelOutput key assignment." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Different loading problems affecting different models and mechanisms." }, { "left": "issue:40444", "right": "issue:45381", "accept": false, - "reason": "IterableDataset multi-image finetuning failure and Qwen2.5-VL video position_ids bug are distinct Qwen2.5-VL regressions." + "reason": "Same model family, but one is image finetuning with multiple images and the other is video position-id handling; not the same bug." }, { "left": "issue:45325", "right": "issue:45381", "accept": false, - "reason": "Still-image temporal scaling and video vision_position_ids are separate Qwen2.5-VL rope/positioning bugs." + "reason": "Both are Qwen2.5-VL positional-id issues, but one is still-image temporal scaling and the other is video vision_position_ids; too different to merge as one bug." + }, + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "Different problems and versions: PaddleOCR-VL loading failure vs torch 2.4.1/transformers 5.8.0 incompatibility." }, { "left": "issue:44387", "right": "issue:44617", "accept": false, - "reason": "int4 quantization memory growth and Sam3Video OOM are not the same underlying issue." + "reason": "Both end in OOM, but one is reserved-memory growth under int4 quantization and the other is a model-specific CUDA OOM." }, { "left": "issue:35707", "right": "issue:40444", "accept": false, - "reason": "inputs_embeds/past_key_values generation and Qwen2.5-VL multi-image finetuning failure are unrelated." + "reason": "Different generation/training failures: progressive generation with inputs_embeds/past_key_values vs multi-image iterable finetuning." }, { "left": "issue:44291", "right": "issue:44877", "accept": false, - "reason": "Unexpected _is_hf_initialized argument and strict config preventing granite_speech loading are different model-loading regressions." + "reason": "Different init/config issues: unexpected `_is_hf_initialized` argument vs strict config rejecting granite_speech." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative about the soft-edge pairs. The single accepted pair (42175/45542) is plausibly the same TensorFlow-backend packaging bug, and the rest of the verdicts avoid overstating similarity. The representative issue choice is also reasonable for this cluster\u2019s v5 loading/init regressions." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -3404,7 +3289,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:34634", "issue:35707", @@ -3440,7 +3325,6 @@ "issue:44610", "issue:44683", "issue:44756", - "issue:44877", "issue:44898", "issue:44977", "issue:44991", @@ -3459,8 +3343,8 @@ "issue:43526|issue:43883", "issue:43606|issue:44265", "issue:43317|issue:43873", - "issue:43701|issue:45325", "issue:44186|issue:44265", + "issue:43701|issue:45325", "issue:43881|issue:44991", "issue:43632|issue:45563", "issue:43606|issue:43819", @@ -3468,254 +3352,240 @@ "issue:44610|issue:45325", "issue:43454|issue:43550", "issue:43377|issue:44610", - "issue:43454|issue:45072", "issue:40990|issue:45584", - "issue:43572|issue:45759", "issue:39692|issue:40444", + "issue:43454|issue:45072", "issue:43377|issue:44898", - "issue:43296|issue:44991", - "issue:43296|issue:44877", - "issue:43792|issue:44683", + "issue:43572|issue:45759", "issue:44466|issue:45357", + "issue:43296|issue:44991", "issue:43454|issue:43720", "issue:43873|issue:45701", + "issue:43792|issue:44683", "issue:41720|issue:44977", "issue:45632|issue:45684", "issue:43404|issue:45357", "issue:44186|issue:45072", "issue:44423|issue:45406", - "issue:43577|issue:43708", "issue:34634|issue:35707", + "issue:43577|issue:43708", "issue:35707|issue:41720", - "issue:44877|issue:44991", "issue:43873|issue:44756" ], - "prepared_review_unit_hash": "c4ab53eb821190a35861874750b49dba0479f651d1fb2ba0180b79b69d5ca372", + "prepared_review_unit_hash": "2081391a90b079d364efa34e72eb3b1b970c6f560c46639256b479c55f15d5ba", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 33, - "serialized_chars": 37837, - "estimated_input_tokens": 9460, - "estimated_eval_tokens": 19176 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36685, + "estimated_input_tokens": 9172, + "estimated_eval_tokens": 18600 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is heterogeneous: most items are distinct model-specific bugs, loader issues, or docs/examples regressions rather than duplicates. No soft pair looks like the same concrete bug/change under a strict duplicate standard.", - "confidence": 0.78, - "canonical_issue_reason": "issue:43873 is the strongest cluster representative: it is open, has the most discussion/inbound references, and sits in the recurring offloading/quantization area that several nearby issues loosely resemble.", + "summary": "This set is a heterogeneous mix of unrelated Transformers issues. The soft edges mostly reflect broad topical similarity (model loading, offloading, multimodal serving, tokenization, training) rather than the same underlying bug or change, so none should be merged.", + "confidence": 0.96, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 has the best triage signal and broadest overlap with the cluster\u2019s recurring model-loading/offload theme, making it the most useful representative issue.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43526", "right": "issue:43883", "accept": false, - "reason": "BeitImageProcessorFast label reduction vs Molmo tied-weights attribute error are unrelated bugs in different code paths." + "reason": "Different failures in different models: BEiT Fast label reduction vs Molmo tied-weights attribute error." }, { "left": "issue:43606", "right": "issue:44265", "accept": false, - "reason": "CPU-offload device mismatch for bark-small is unrelated to torch.export failures around torch_compilable_check." + "reason": "One is Bark CPU-offload device mismatch; the other is a torch.export/torch_compilable_check export failure." }, { "left": "issue:43317", "right": "issue:43873", "accept": false, - "reason": "Both touch offloading/quantization, but one is a dequantized model load failure and the other is a broader offloading behavior report; not the same concrete bug." + "reason": "Both mention offloading/quantization, but one is dequantized model loading on gpu+cpu offload and the other is a broader quantization offload complaint." }, { - "left": "issue:43701", - "right": "issue:45325", + "left": "issue:44186", + "right": "issue:44265", "accept": false, - "reason": "Checkpoint resume key mismatch is unrelated to Qwen2.5-VL rope index scaling." + "reason": "Unrelated code paths: LayoutLMv2 tokenization/padding vs torch.export failing on torch_compilable_check." }, { - "left": "issue:44186", - "right": "issue:44265", + "left": "issue:43701", + "right": "issue:45325", "accept": false, - "reason": "Tokenizer crashes on NER/padding are unrelated to export-time torch_compilable_check issues." + "reason": "Checkpoint resume key mismatch is unrelated to Qwen2.5-VL temporal rope indexing." }, { "left": "issue:43881", "right": "issue:44991", "accept": false, - "reason": "A Whisper model loading/runtime failure is not the same as a tokenizer loading regression for est-roberta." + "reason": "Different models and layers: GLM-4V loading failure vs tokenizer loading regression for est-roberta." }, { "left": "issue:43632", "right": "issue:45563", "accept": false, - "reason": "_is_hf_initialized flag regression and a stale paging warning are different bugs with different symptoms and fix paths." + "reason": "Transformers v5 init-flag regression is unrelated to a paged generate warning about num_return_sequences." }, { "left": "issue:43606", "right": "issue:43819", "accept": false, - "reason": "Both mention device-related failures, but one is CPU offload mismatch and the other is DAC latent/forward mismatch from missing STE." + "reason": "Bark CPU offload device mismatch is unrelated to DAC.from_latents versus forward mismatch." }, { "left": "issue:43262", "right": "issue:43746", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting and GraniteSpeech PEFT checkpoint loading are unrelated." + "reason": "Audio chat-template sample-rate default bug is unrelated to GraniteSpeech PEFT checkpoint loading." }, { "left": "issue:44610", "right": "issue:45325", "accept": false, - "reason": "Processor output resolution mismatch is not the same underlying issue as temporal position_id scaling in Qwen2.5-VL." + "reason": "Processor image-size mismatch is unrelated to Qwen2.5-VL rope index scaling." }, { "left": "issue:43454", "right": "issue:43550", "accept": false, - "reason": "lm_head weight tying on AyaVision is unrelated to Bamba SDPA/torch.compile failure." + "reason": "lm_head tying bug in AyaVision is unrelated to Bamba torch.compile/SDPA failure." }, { "left": "issue:43377", "right": "issue:44610", "accept": false, - "reason": "MIMI padding-mask batching bug and OmDet processor/model image-size mismatch are different multimodal preprocessing problems." - }, - { - "left": "issue:43454", - "right": "issue:45072", - "accept": false, - "reason": "Weight tying in AyaVision is unrelated to bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel." + "reason": "MIMI padding-mask batching bug is unrelated to OmDet-Turbo processor size mismatch." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "High perplexity on GPT-OSS and Whisper empty-transcription generation failure are unrelated model-specific issues." - }, - { - "left": "issue:43572", - "right": "issue:45759", - "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig and Qwen multimodal config unwrapping are different configuration bugs." + "reason": "High perplexity on GPT-OSS is unrelated to Whisper empty-transcription generation after align_special_tokens." }, { "left": "issue:39692", "right": "issue:40444", "accept": false, - "reason": "A SigLIP2 doc example mismatch/quantization failure is unrelated to Qwen2.5-VL finetuning with multi-image iterable datasets." + "reason": "SigLIP2 docs example errors are unrelated to Qwen2.5-VL finetuning on iterable multi-image batches." }, { - "left": "issue:43377", - "right": "issue:44898", + "left": "issue:43454", + "right": "issue:45072", "accept": false, - "reason": "Missing padding-mask support in MIMI is unrelated to Perceiver non-default-resolution failures." + "reason": "AyaVision weight-tying bug and SwitchTransformers/TimmWrapper bfloat16 dtype mismatch are different bugs." }, { - "left": "issue:43296", - "right": "issue:44991", + "left": "issue:43377", + "right": "issue:44898", "accept": false, - "reason": "PaddleOCR-VL loading in vLLM and est-roberta tokenizer loading are separate model-specific regressions." + "reason": "Different models and fixes: MIMI batching/padding mask vs Perceiver non-default-resolution interpolation." }, { - "left": "issue:43296", - "right": "issue:44877", + "left": "issue:43572", + "right": "issue:45759", "accept": false, - "reason": "PaddleOCR-VL loading failure and strict granite_speech config loading are different failures in different subsystems." + "reason": "StableLm pad_token_idx regression is unrelated to Qwen composite config unwrapping in from_config." }, { - "left": "issue:43792", - "right": "issue:44683", + "left": "issue:44466", + "right": "issue:45357", "accept": false, - "reason": "Whisper model runtime failure is unrelated to compiled flex_attention breakage on newer torch." + "reason": "Serialization of lm_head.weight by device is unrelated to Qwen3.5 visual encoder key saving." }, { - "left": "issue:44466", - "right": "issue:45357", + "left": "issue:43296", + "right": "issue:44991", "accept": false, - "reason": "Inconsistent lm_head serialization by device is not the same as incorrect visual encoder keys in Qwen3.5 save_pretrained." + "reason": "PaddleOCR-VL loading failure in vLLM is unrelated to tokenizer loading for est-roberta." }, { "left": "issue:43454", "right": "issue:43720", "accept": false, - "reason": "AyaVision lm_head tying and BitNet packed-weight unpacking during accelerate loading are different issues." + "reason": "Different tied-weight/loading issues: AyaVision lm_head tying vs BitNet packed-weight unpacking during accelerate loading." }, { "left": "issue:43873", "right": "issue:45701", "accept": false, - "reason": "Both mention loading/tokenization-adjacent behavior, but one is quantization/offloading and the other is a general tokenization-version regression." + "reason": "Quantization/offload behavior is unrelated to a tokenization-version regression." + }, + { + "left": "issue:43792", + "right": "issue:44683", + "accept": false, + "reason": "Whisper runtime failure and compiled flex_attention failure are unrelated code paths." }, { "left": "issue:41720", "right": "issue:44977", "accept": false, - "reason": "A800 cudaErrorAssert during auto device mapping is distinct from flash-attention generation issues in Qwen3.5." + "reason": "Both are Qwen-family reports, but one is auto device mapping cudaErrorAssert and the other is flash-attention generation corruption." }, { "left": "issue:45632", "right": "issue:45684", "accept": false, - "reason": "Cache-path collisions for trust_remote_code are unrelated to permission propagation in save_pretrained/register_for_auto_class." + "reason": "Cache-path collisions for trust_remote_code are unrelated to read-only permissions propagated by save_pretrained/register_for_auto_class." }, { "left": "issue:43404", "right": "issue:45357", "accept": false, - "reason": "lm_head weight tying in Mistral3 is unrelated to Qwen3.5 visual encoder key serialization." + "reason": "Mistral3 lm_head tying bug is unrelated to Qwen3.5 visual encoder key serialization." }, { "left": "issue:44186", "right": "issue:45072", "accept": false, - "reason": "Tokenizer batching/NER crashes and dtype mismatches in bfloat16 inference are different bug classes." + "reason": "Tokenizer crash on NER/padding is unrelated to dtype mismatches in bfloat16 inference." }, { "left": "issue:44423", "right": "issue:45406", "accept": false, - "reason": "Continuous-batching multimodal crash and Gemma4Processor missing _tokenizer are separate serve-time processor issues." - }, - { - "left": "issue:43577", - "right": "issue:43708", - "accept": false, - "reason": "Blip2 dtype propagation and Trainer max_steps recalculation on resume are unrelated." + "reason": "Both affect serve, but the concrete bugs differ: multimodal continuous batching str-to conversion vs Gemma4Processor missing _tokenizer." }, { "left": "issue:34634", "right": "issue:35707", "accept": false, - "reason": "BarkProcessor voice_preset failure is unrelated to progressive generation with inputs_embeds/past_key_values." + "reason": "BarkProcessor voice_preset failure is unrelated to progressive generation with inputs_embeds and past_key_values." }, { - "left": "issue:35707", - "right": "issue:41720", + "left": "issue:43577", + "right": "issue:43708", "accept": false, - "reason": "Progressive generation with cached KV is unrelated to Qwen3 auto device-mapping cudaErrorAssert." + "reason": "Blip2 dtype loading issue is unrelated to Trainer max_steps miscalculation on checkpoint resume." }, { - "left": "issue:44877", - "right": "issue:44991", + "left": "issue:35707", + "right": "issue:41720", "accept": false, - "reason": "Granite_speech strict-config loading and est-roberta tokenizer loading are different incompatibilities." + "reason": "Progressive generation with embeddings/past_key_values is unrelated to Qwen3 auto device mapping cudaErrorAssert." }, { "left": "issue:43873", "right": "issue:44756", "accept": false, - "reason": "Quantization/offloading behavior is not the same as disabling mmap on Strix Halo to avoid OOM." + "reason": "Quantization offload behavior is unrelated to disabling mmap on Strix Halo to avoid OOM." } ] }, @@ -3737,7 +3607,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:30064", "issue:34634", @@ -3760,7 +3630,6 @@ "issue:43550", "issue:43572", "issue:43606", - "issue:43653", "issue:43720", "issue:43749", "issue:43782", @@ -3777,9 +3646,11 @@ "issue:44610", "issue:44671", "issue:44743", + "issue:44877", "issue:44898", "issue:44913", "issue:44918", + "issue:44991", "issue:45059", "issue:45161", "issue:45406", @@ -3790,6 +3661,7 @@ "soft_pairs": [ "issue:43749|issue:44534", "issue:43526|issue:43881", + "issue:44877|issue:44991", "issue:42617|issue:42915", "issue:43450|issue:45701", "issue:43550|issue:43720", @@ -3800,14 +3672,14 @@ "issue:44387|issue:45704", "issue:44479|issue:44918", "issue:43316|issue:44913", - "issue:42915|issue:43317", - "issue:43550|issue:43606", "issue:40990|issue:45701", + "issue:43550|issue:43606", "issue:43295|issue:45701", + "issue:42915|issue:43317", "issue:41720|issue:43322", "issue:41720|issue:45161", - "issue:43531|issue:43572", "issue:44671|issue:44743", + "issue:43531|issue:43572", "issue:44292|issue:44479", "issue:41720|issue:42617", "issue:36010|issue:42617", @@ -3820,223 +3692,234 @@ "issue:44492|issue:44509", "issue:43388|issue:45059", "issue:43606|issue:43720", - "issue:42915|issue:43322", - "issue:43550|issue:43653" + "issue:42915|issue:43322" ], - "prepared_review_unit_hash": "ea25de4ae67ecda0cf67c03140fc57ce40281be0f5e323ee2e1654acd38f7ffb", + "prepared_review_unit_hash": "b74cd2a20fdcfe702b50c8b5de7b9a8c0c5099ffaff25dd2ec87c9a859754ad2", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 36962, - "estimated_input_tokens": 9241, - "estimated_eval_tokens": 18738 + "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": "All proposed soft edges look like false positives: the cluster spans unrelated bugs across processors, tokenizers, loading/offload paths, quantization, and docs/API regressions rather than one duplicated issue or PR.", - "confidence": 0.94, - "canonical_issue_reason": "No canonical issue: the items are a mixed set of distinct bug reports, so there is no single duplicate hub to canonicalize.", + "summary": "This cluster is mostly an over-grouping of unrelated Transformers issues that only share broad themes like loading, offloading, quantization, tokenizer/processor regressions, or model-specific failures. I would not merge any soft pairs as duplicates.", + "confidence": 0.45, + "canonical_issue_reason": "issue:43873 is the broadest representative of the shared offloading/quantization loading-regression theme and is more general than the model- or processor-specific reports around it.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43749 is the broadest loading/offload-related report and is the closest representative of the small offloading/quantization subgroup, but it still does not unify the cluster as a duplicate.", + "best_issue_reason": "issue:43873 is the best single anchor for this cluster because it captures the common loading/offloading problem without being tied to one narrow model or component.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43749", "right": "issue:44534", "accept": false, - "reason": "FSDP CPU RAM-efficient loading vs a docs-only pipeline task note; different problem and code path." + "reason": "Different bugs: FSDP CPU RAM efficient loading vs non-persistent buffers being filled with junk." }, { "left": "issue:43526", "right": "issue:43881", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug vs glm-4v-9b load failure; unrelated subsystems." + "reason": "Unrelated components: BeitImageProcessorFast label reduction vs glm-4v-9b loading failure." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Config strictness for granite_speech is not the same as tokenizer loading for EMBEDDIA/est-roberta." }, { "left": "issue:42617", "right": "issue:42915", "accept": false, - "reason": "3d_parallel.py execution failure vs Qwen3Moe FineGrainedFP8Config failure; different runtime issues." + "reason": "Different failures: 3d_parallel.py runtime issue vs Qwen3Moe FineGrainedFP8Config loading." }, { "left": "issue:43450", "right": "issue:45701", "accept": false, - "reason": "Batched video processor shape bug vs tokenization/version change report; not the same underlying bug." + "reason": "Video processor batching shape bug is separate from broad tokenization changes." }, { "left": "issue:43550", "right": "issue:43720", "accept": false, - "reason": "Bamba torch.compile/SDPA failure vs BitNet packed-weight loading bug; different model and failure mode." + "reason": "Different code paths: torch.compile/SDPA in Bamba vs packed-weight unpacking during accelerate loading." }, { "left": "issue:36331", "right": "issue:42915", "accept": false, - "reason": "Trainer compute_loss signature mismatch vs Qwen3Moe FP8 loading/config issue; unrelated." + "reason": "Trainer API signature mismatch is unrelated to Qwen3Moe FP8 configuration/loading." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "OmDet-Turbo processor image-size mismatch vs Perceiver non-default resolution issue; separate processor/model bugs." + "reason": "Both vision-related, but OmDet-Turbo 640px output mismatch is a different bug from Perceiver non-default-resolution handling." }, { "left": "issue:44186", "right": "issue:44610", "accept": false, - "reason": "LayoutLMv2Tokenizer NER/padding crash vs OmDet-Turbo processor output size mismatch; different code paths." + "reason": "Different models and symptoms: LayoutLMv2 tokenizer padding/truncation crash vs OmDet-Turbo processor size mismatch." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "CustomTrainer loss signature error vs 3d_parallel.py failure; no shared underlying defect." + "reason": "CustomTrainer.compute_loss() API break is unrelated to the 3d_parallel.py execution failure." }, { "left": "issue:44387", "right": "issue:45704", "accept": false, - "reason": "Int4 quantization OOM from reserved memory growth vs T5 apex RMSNorm memory leak; not the same bug." + "reason": "Int4 quantization OOM and apex FusedRMSNorm memory leak are both memory-related but not the same bug." }, { "left": "issue:44479", "right": "issue:44918", "accept": false, - "reason": "Video-input regression in Qwen VL family vs Qwen3.5 embedding unpacking with TRL SFT trainer; distinct failures." + "reason": "Separate problems: video-input regression in VL models vs Qwen3.5 embedding unpacking with TRL SFT trainer." }, { "left": "issue:43316", "right": "issue:44913", "accept": false, - "reason": "Gemma3TextConfig API discrepancy vs GPTNeoX rotary_pct reload regression; unrelated config issues." + "reason": "Both involve config behavior, but Gemma3TextConfig API mismatch and GPTNeoX rotary_pct reload are different fields and failures." }, { - "left": "issue:42915", - "right": "issue:43317", + "left": "issue:40990", + "right": "issue:45701", "accept": false, - "reason": "Qwen3Moe FP8 config failure vs dequantized model offload/load failure; different loading paths." + "reason": "High perplexity on a model is not the same as tokenizer version-dependent tokenization changes." }, { "left": "issue:43550", "right": "issue:43606", "accept": false, - "reason": "Bamba SDPA/compile issue vs Bark-small CPU offload device mismatch; different models and failure signatures." + "reason": "Bamba compile/SDPA failure and bark-small CPU offload device mismatch are different model/runtime bugs." }, { - "left": "issue:40990", + "left": "issue:43295", "right": "issue:45701", "accept": false, - "reason": "High perplexity on gpt-oss-20b vs tokenization version changes; not the same bug." + "reason": "Processor.tokenizer regression in v4.57.5 is distinct from general tokenization changes across versions." }, { - "left": "issue:43295", - "right": "issue:45701", + "left": "issue:42915", + "right": "issue:43317", "accept": false, - "reason": "Processor.tokenizer regression in v4.57.5 vs tokenization changes across versions; different symptoms and causes." + "reason": "Qwen3Moe FP8 config failure is separate from dequantized model loading with gpu+cpu offload." }, { "left": "issue:41720", "right": "issue:43322", "accept": false, - "reason": "Qwen3 auto device-mapping cuda assert vs Llava Next segfault; separate model-loading crashes." + "reason": "Different models and symptoms: Qwen3 auto device-map cuda assert vs Llava Next segfault on load." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "Qwen3 auto device-mapping assert vs GPT-OSS MoE tensor-parallel issue; different model families and bugs." + "reason": "Qwen3 auto device mapping and GPT-OSS MoE tensor-parallel failure are different distributed execution issues." }, { - "left": "issue:43531", - "right": "issue:43572", + "left": "issue:44671", + "right": "issue:44743", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior vs missing pad_token_idx in StableLmConfig; unrelated config/runtime problems." + "reason": "CamemBERT masked LM prediction regression is unrelated to Qwen3.5 recurrent-state caching." }, { - "left": "issue:44671", - "right": "issue:44743", + "left": "issue:43531", + "right": "issue:43572", "accept": false, - "reason": "CamemBERT masked-LM prediction regression vs Qwen3.5 recurrent state reset with cache; different model logic." + "reason": "Qwen3-MoE sliding_window behavior and StableLmConfig pad_token_idx are separate configuration issues." }, { "left": "issue:44292", "right": "issue:44479", "accept": false, - "reason": "Qwen-3-8B-NVFP4 load/runtime error vs video-input regression in Qwen VL models; separate issues." + "reason": "Qwen-3-8B-NVFP4 runtime error is not the same as the VL video-input regression." }, { "left": "issue:41720", "right": "issue:42617", "accept": false, - "reason": "Qwen3 device-mapping assert vs 3d_parallel.py failure; no shared code-path problem." + "reason": "Qwen3 auto device mapping assert is different from the 3d_parallel.py failure." }, { "left": "issue:36010", "right": "issue:42617", "accept": false, - "reason": "GenerationMixin import error vs 3d_parallel.py run failure; unrelated exceptions." + "reason": "GenerationMixin import error and 3d_parallel.py failure are unrelated issues." }, { "left": "issue:44186", "right": "issue:44361", "accept": false, - "reason": "LayoutLMv2Tokenizer crash vs MLukeTokenizer AttributeError; both tokenizer bugs but not the same underlying one." + "reason": "Different tokenizers and different failure modes: LayoutLMv2 crash vs MLuke tokenizer AttributeError." }, { "left": "issue:43540", "right": "issue:43782", "accept": false, - "reason": "Qwen3OmniMoe video ValueError vs Qwen3VL weight_only load error; different models and failure points." + "reason": "Qwen3OmniMoe video-input ValueError is unrelated to Qwen3VL weight_only=True loading error." }, { "left": "issue:45406", "right": "issue:45464", "accept": false, - "reason": "Gemma4Processor missing _tokenizer vs Qwen3.5 streaming inference failure; different stack traces and components." + "reason": "Gemma4Processor missing _tokenizer in serve is different from Qwen3.5 streaming inference failure." }, { - "left": "issue:44492", - "right": "issue:44509", + "left": "issue:41720", + "right": "issue:43873", "accept": false, - "reason": "Typo in cache strategies docs vs stale pipeline-task docs; both docs-related but not the same issue." + "reason": "Qwen3 auto device-map cuda assert is a model-specific crash, not the same as generic offloading-with-quantization behavior." }, { - "left": "issue:43388", - "right": "issue:45059", + "left": "issue:43421", + "right": "issue:45701", "accept": false, - "reason": "gather_for_metrics label truncation bug vs SAM3 PCS text/bbox behavior; unrelated training vs model-input bug." + "reason": "Runtime post-processor refresh for special tokens is much narrower than general tokenization changes." }, { - "left": "issue:43606", - "right": "issue:43720", + "left": "issue:30064", + "right": "issue:34634", "accept": false, - "reason": "Bark CPU offload device mismatch vs BitNet packed-weights unpacking bug; different loading/offload mechanisms." + "reason": "Image processor void segmentation maps and BarkProcessor voice_preset are unrelated." }, { - "left": "issue:42915", - "right": "issue:43322", + "left": "issue:44492", + "right": "issue:44509", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure vs Llava Next segmentation fault; distinct crashes." + "reason": "A typo in cache strategies is not the same as stale docs mentioning removed pipeline tasks." }, { - "left": "issue:43550", - "right": "issue:43653", + "left": "issue:43388", + "right": "issue:45059", "accept": false, - "reason": "Bamba compile/SDPA failure vs BigBirdTokenizer special-token decode issue; unrelated model/runtime layers." + "reason": "Metric gathering label truncation and SAM3 PCS text/bbox behavior are different bugs." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "CPU offload device mismatch and packed-weight unpacking during loading are different root causes." } ] }, @@ -4058,7 +3941,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:33666", "issue:34567", @@ -4072,7 +3955,7 @@ "issue:43421", "issue:43525", "issue:43531", - "issue:43596", + "issue:43550", "issue:43606", "issue:43653", "issue:43746", @@ -4083,7 +3966,6 @@ "issue:43994", "issue:44062", "issue:44186", - "issue:44220", "issue:44368", "issue:44451", "issue:44514", @@ -4110,258 +3992,258 @@ "issue:45759" ], "soft_pairs": [ + "issue:43550|issue:43653", "issue:44514|issue:44545", "issue:43531|issue:45759", "issue:43606|issue:43653", "issue:43421|issue:43927", "issue:34567|issue:35707", "issue:44368|issue:45357", - "issue:44451|issue:44991", - "issue:44898|issue:45072", "issue:43873|issue:45704", "issue:41720|issue:43746", + "issue:44898|issue:45072", + "issue:44451|issue:44991", "issue:43421|issue:43653", - "issue:44062|issue:44451", - "issue:44987|issue:45042", "issue:45588|issue:45759", + "issue:44062|issue:44451", "issue:33666|issue:35707", - "issue:43531|issue:44589", + "issue:44987|issue:45042", "issue:42886|issue:44843", + "issue:43531|issue:44589", "issue:43421|issue:44186", "issue:39692|issue:44610", "issue:43381|issue:43844", "issue:44918|issue:44936", - "issue:43525|issue:43531", "issue:45081|issue:45356", "issue:44625|issue:45479", + "issue:43525|issue:43531", "issue:43782|issue:45381", "issue:43994|issue:44671", - "issue:34634|issue:41720", "issue:43381|issue:45704", - "issue:43596|issue:44220", + "issue:34634|issue:41720", "issue:39692|issue:43873", "issue:42915|issue:45759" ], - "prepared_review_unit_hash": "276aab7a38f7f871880c0076b4be82e80443fb7d0b75cb4d3c2511dc32217618", + "prepared_review_unit_hash": "ca3a147c018812ccd5ffabd286071e49f03ba671e0593fe84fb93fad94fb2a9c", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 37258, - "estimated_input_tokens": 9315, - "estimated_eval_tokens": 18886 + "serialized_chars": 36675, + "estimated_input_tokens": 9169, + "estimated_eval_tokens": 18594 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "There is one strong duplicate pair: the two Qwen2_5_VLProcessor.apply_chat_template batched-input padding=False reports. The rest are only loosely related by broad subsystem (tokenizers, quantization, Qwen, etc.) but appear to be distinct bugs or changes.", - "confidence": 0.94, - "canonical_issue_reason": "issue:44545 is the cleaner canonical: it states the crash condition more precisely, matches the duplicate title almost verbatim, and has slightly more discussion/references than 44514.", + "summary": "Most items are unrelated transformer/model bugs. I found two real duplicate pairs: the Qwen2_5_VL batched padding=False crash, and the HF_HUB_OFFLINE tokenizer load failure tied to _patch_mistral_regex. The rest are distinct code paths and should stay separate.", + "confidence": 0.79, + "canonical_issue_reason": "issue:44843 is the clearest root-cause report for the offline tokenizer regression: it names the offending _patch_mistral_regex/model_info path and the HF_HUB_OFFLINE failure mode, making it the best canonical issue among the accepted duplicates.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44545 is the best representative because it is the most exact description of the underlying batched-input padding=False crash and is already the clearer tracking thread.", + "best_issue_reason": "issue:44843 is the most actionable representative bug in the set because it pinpoints the concrete failing code path rather than only the symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "Different bugs: torch.compile/SDPA model failure vs tokenizer special-token registration/empty decode." + }, { "left": "issue:44514", "right": "issue:44545", "accept": true, - "reason": "Same Qwen2_5_VLProcessor.apply_chat_template batched-input crash with padding=False; clearly the same bug." + "reason": "Same processor and same batched apply_chat_template crash when padding=False; wording differs but underlying bug matches." }, { "left": "issue:43531", "right": "issue:45759", "accept": false, - "reason": "Both involve Qwen models, but one is a sliding_window issue and the other is config unwrapping for multimodal from_config; different bugs." + "reason": "Different Qwen config issues: sliding_window behavior vs from_config not unwrapping text_config." }, { "left": "issue:43606", "right": "issue:43653", "accept": false, - "reason": "Unrelated models and symptoms: CPU offload device mismatch vs tokenizer mask-token decode behavior." + "reason": "CPU offload device mismatch is unrelated to tokenizer mask-token registration." }, { "left": "issue:43421", "right": "issue:43927", "accept": false, - "reason": "Different layers and failures: runtime special-token post-processing vs save/load token ID loss causing generation IndexError." + "reason": "One is a feature/runtime-update request for tokenizer post-processing; the other is a save/load custom token ID regression." }, { "left": "issue:34567", "right": "issue:35707", "accept": false, - "reason": "TrainerState token counting and progressive generation with inputs_embeds/past_key_values are separate issues." + "reason": "TrainerState token counter bug is unrelated to progressive generation with inputs_embeds/past_key_values." }, { "left": "issue:44368", "right": "issue:45357", "accept": false, - "reason": "A warning about tie_word_embeddings in LoRA training is unrelated to incorrect visual encoder keys in save_pretrained." + "reason": "A warning about tie_word_embeddings is not the same as incorrect visual-encoder keys being saved." }, { - "left": "issue:44451", - "right": "issue:44991", + "left": "issue:43873", + "right": "issue:45704", "accept": false, - "reason": "Both are tokenizer loading failures, but for different models and likely different root causes." + "reason": "Quantization/offload behavior is unrelated to the T5 apex RMSNorm memory leak." }, { - "left": "issue:44898", - "right": "issue:45072", + "left": "issue:41720", + "right": "issue:43746", "accept": false, - "reason": "Perceiver image-size interpolation failure and dtype mismatch in inference are distinct problems." + "reason": "Qwen3 auto device-map cuda assert and GraniteSpeech PEFT checkpoint loading are different failures." }, { - "left": "issue:43873", - "right": "issue:45704", + "left": "issue:44898", + "right": "issue:45072", "accept": false, - "reason": "Quantization/offloading behavior and an apex RMSNorm memory leak are unrelated bugs." + "reason": "Perceiver resolution/interpolation bug is unrelated to bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel." }, { - "left": "issue:41720", - "right": "issue:43746", + "left": "issue:44451", + "right": "issue:44991", "accept": false, - "reason": "Qwen3 device-mapping cuda assert and GraniteSpeech PEFT local checkpoint loading are different code paths." + "reason": "Different artifacts and failure modes: model loading vs tokenizer loading for different repositories." }, { "left": "issue:43421", "right": "issue:43653", "accept": false, - "reason": "Both mention special tokens, but one is runtime post-processor updates and the other is BigBirdTokenizer mask-token registration." + "reason": "Runtime special-token/post-processor update feature request is not the same as BigBirdTokenizer mask-token decode failure." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention sink-less crash and Qwen config unwrapping are separate code paths." }, { "left": "issue:44062", "right": "issue:44451", "accept": false, - "reason": "AddedToken keyword duplication and tokenizer loading failure for ScandiBERT are unrelated." + "reason": "AddedToken special-argument conflict is unrelated to loading a specific model repo." + }, + { + "left": "issue:33666", + "right": "issue:35707", + "accept": false, + "reason": "Multi-GPU training for Qwen2-VL is unrelated to progressive generation with inputs_embeds/past_key_values." }, { "left": "issue:44987", "right": "issue:45042", "accept": false, - "reason": "Loading a specific model failing and PIL image processors requiring torchvision are separate issues." + "reason": "Model load regression and PIL image-processor torchvision dependency are different problems." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": true, + "reason": "Same offline tokenizer failure: 42886 reports the symptom, 44843 identifies the exact _patch_mistral_regex/model_info cause." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Sliding_window behavior for Qwen3-MoE is unrelated to missing Float8 storage support." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Tokenizer post-processor update behavior is not the same as LayoutLMv2Tokenizer crashes on NER/padding." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "SigLIP2 doc example errors and OmDet-Turbo processor size mismatch are distinct issues." }, { "left": "issue:43381", "right": "issue:43844", "accept": false, - "reason": "Gradient checkpointing eval-mode restriction is not the same as ZeRO-3 gradient inflation during training." + "reason": "Gradient checkpointing in eval mode is unrelated to gradient blow-up under ZeRO-3." }, { "left": "issue:44918", "right": "issue:44936", "accept": false, - "reason": "TRl SFT embedding unpacking and trainer.evaluate() after trainer.train() are different failures." - }, - { - "left": "issue:43525", - "right": "issue:43531", - "accept": false, - "reason": "Missing pad_token_id on Llama4Config is unrelated to Qwen3-MoE sliding_window behavior." + "reason": "TRL SFT input-embedding unpacking and trainer.evaluate() after train are different trainer bugs." }, { "left": "issue:45081", "right": "issue:45356", "accept": false, - "reason": "Both touch fix_mistral_regex, but one is a hard crash loading Mistral tokenizer and the other is a Kimi-K2.5 regression with codec/warning issues." + "reason": "Both touch fix_mistral_regex, but one is a loading crash and the other is a Kimi-K2.5 codec/warning regression; not the same bug." }, { "left": "issue:44625", "right": "issue:45479", "accept": false, - "reason": "num_labels propagation in Qwen3.5 config and zero-loss classification across models are different bugs." + "reason": "Qwen3.5 num_labels propagation and single_label_classification zero-loss behavior are separate classification issues." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Missing pad_token_id on Llama4Config is unrelated to Qwen3-MoE sliding_window behavior." }, { "left": "issue:43782", "right": "issue:45381", "accept": false, - "reason": "Qwen3VL weight_only loading error and Qwen2.5-VL video vision_position_ids mismatch are separate multimodal issues." + "reason": "Qwen3VL weight_only loading and Qwen2.5-VL video vision_position_ids are different multimodal bugs." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "SigLIP2 bad outputs and CamemBERT masked-LM regressions are unrelated model-specific problems." - }, - { - "left": "issue:34634", - "right": "issue:41720", - "accept": false, - "reason": "BarkProcessor voice_preset failure and Qwen3 auto device mapping cuda assert are not the same bug." + "reason": "SigLIP2 incorrect outputs and CamemBERT masked-LM regression are different model-specific failures." }, { "left": "issue:43381", "right": "issue:45704", "accept": false, - "reason": "Gradient checkpointing eval-mode error and T5 apex memory leak do not share the same underlying defect." + "reason": "Gradient checkpointing eval-mode restriction is unrelated to the T5 apex RMSNorm memory leak." }, { - "left": "issue:43596", - "right": "issue:44220", + "left": "issue:34634", + "right": "issue:41720", "accept": false, - "reason": "DeepSpeed ZeRO-3/BertModel index error and _torch_extract_fbank_features() are unrelated." + "reason": "BarkProcessor voice_preset and Qwen3 auto device-mapping cuda assert are unrelated." }, { "left": "issue:39692", "right": "issue:43873", "accept": false, - "reason": "SigLIP2 docs/quantization problems and offloading with quantization are different issues." + "reason": "SigLIP2 documentation bugs are unrelated to quantization/offloading behavior." }, { "left": "issue:42915", "right": "issue:45759", "accept": false, - "reason": "Qwen3Moe FP8Config failure and Qwen multimodal config unwrapping are distinct configuration bugs." - }, - { - "left": "issue:43421", - "right": "issue:44186", - "accept": false, - "reason": "TokenizersBackend special-token updates and LayoutLMv2Tokenizer NER/padding crashes are different tokenizer problems." - }, - { - "left": "issue:43531", - "right": "issue:44589", - "accept": false, - "reason": "Qwen3-MoE sliding_window and missing Float8 storage are unrelated." - }, - { - "left": "issue:42886", - "right": "issue:44843", - "accept": false, - "reason": "Offline cache loading and unconditional model_info() calls in _patch_mistral_regex are different offline-mode issues." - }, - { - "left": "issue:44987", - "right": "issue:45042", - "accept": false, - "reason": "Model load failure and torchvision dependency in PIL image processors are not the same bug." - }, - { - "left": "issue:45081", - "right": "issue:45356", - "accept": false, - "reason": "Different symptoms and targets despite shared helper name; not safe to merge as one fix." - }, - { - "left": "issue:43653", - "right": "issue:45759", - "accept": false, - "reason": "BigBirdTokenizer mask-token decode bug and Qwen multimodal config unwrapping are unrelated." + "reason": "Qwen3Moe FineGrainedFP8Config failure is not the same as Qwen 3.5/3.6 text_config unwrapping." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded overall. The 44514/44545 pair is a genuinely near-identical duplicate, and the canonical choice of 44545 is supported by the packet (clearer title and slightly more references/discussion). The remaining soft-edge verdicts are conservatively negative and do not overclaim duplication from broad subsystem similarity alone." + "feedback": "The summary is grounded in the packet, and the soft-edge calls stay conservative. The two accepted duplicate pairs are well-supported by the issue titles, and the remaining rejections avoid overclaiming sameness across clearly different bugs." }, "evaluator_used": true, "retried": false, @@ -4380,7 +4262,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:30064", "issue:36010", @@ -4393,6 +4275,7 @@ "issue:43550", "issue:43575", "issue:43577", + "issue:43596", "issue:43618", "issue:43643", "issue:43653", @@ -4431,17 +4314,18 @@ "issue:45759" ], "soft_pairs": [ - "issue:44220|issue:44661", + "issue:43596|issue:44220", "issue:45183|issue:45479", - "issue:43540|issue:44008", "issue:43618|issue:43697", - "issue:44077|issue:44625", + "issue:43540|issue:44008", + "issue:44220|issue:44661", "issue:43720|issue:43819", + "issue:44077|issue:44625", "issue:43575|issue:45704", "issue:30064|issue:36010", - "issue:43334|issue:45759", "issue:43335|issue:43756", "issue:43653|issue:43720", + "issue:43334|issue:45759", "issue:43716|issue:44008", "issue:42915|issue:43746", "issue:43746|issue:45381", @@ -4451,218 +4335,224 @@ "issue:43550|issue:43819", "issue:43421|issue:44811", "issue:44843|issue:45081", - "issue:43643|issue:44877", - "issue:44062|issue:44488", "issue:43577|issue:43818", + "issue:44062|issue:44488", "issue:43531|issue:43643", + "issue:43643|issue:44877", "issue:43653|issue:43819", - "issue:43540|issue:45381", "issue:44448|issue:44671", + "issue:43540|issue:45381", "issue:43782|issue:44977", "issue:44561|issue:45020" ], - "prepared_review_unit_hash": "dcf7639cc1dca7e022d07730bdaf5460e95b167afdce55a1fe1b1e0883af9683", + "prepared_review_unit_hash": "0fa4e38e93c91273f6d19c3978e8ce17cab74ab9e3b5c4583916b69e401e5277", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 29, - "serialized_chars": 36139, - "estimated_input_tokens": 9035, - "estimated_eval_tokens": 18326 - }, - "overflow_policy": "truncate_then_skip", + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 37048, + "estimated_input_tokens": 9262, + "estimated_eval_tokens": 18780 + }, + "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is largely a set of superficially similar but actually distinct Transformers issues across different models and code paths. I found no soft-edge pair that is a true duplicate; most are only thematically related.", - "confidence": 0.92, - "canonical_issue_reason": "issue:45020 is the closest umbrella candidate because it is the broadest remote_code regression report, but the cluster is too heterogeneous to treat it as a true canonical duplicate.", + "summary": "The candidate pairs are mostly superficially similar but describe different models, components, or failure modes; none look like safe duplicates or mergeable PR-equivalents. The cluster is heterogeneous, with a few repeated themes around loading/configuration regressions, but the concrete bugs differ.", + "confidence": 0.95, + "canonical_issue_reason": "issue:45020 is the broadest, most central tracker in this set: it describes a general class of recent transformers regressions affecting remote-code models, with active discussion and open status, making it the best canonical issue candidate.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the best global issue candidate only as a broad umbrella; it has the widest scope among the listed issues, but it should not absorb the others as duplicates.", + "best_issue_reason": "issue:45020 is the best global issue candidate because it is high-level, user-facing, and broad enough to subsume related remote-code breakages better than the more model-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44220", - "right": "issue:44661", + "left": "issue:43596", + "right": "issue:44220", "accept": false, - "reason": "Audio feature extraction failure vs tokenizer-mapping generation bug; different subsystems and code paths." + "reason": "Both are runtime errors, but one is a Deepspeed Zero3/BertModel indexing bug and the other is an audio feature extraction issue; different code paths and fixes." }, { "left": "issue:45183", "right": "issue:45479", "accept": false, - "reason": "Input validation/XOR error handling vs sequence-classification loss degeneracy; not the same bug." + "reason": "These are both validation/shape-related, but one is about input_ids vs inputs_embeds XOR checks and the other is about single-label classification degenerating to zero loss; not the same bug." }, { - "left": "issue:43540", - "right": "issue:44008", + "left": "issue:43618", + "right": "issue:43697", "accept": false, - "reason": "Different models and failures: Qwen3OmniMoe video input processing vs Gemma3n variable collision." + "reason": "CLIP attentions assignment and RTDetrV2 output drift in v5 are unrelated model behaviors with different failure surfaces and likely different fixes." }, { - "left": "issue:43618", - "right": "issue:43697", + "left": "issue:43540", + "right": "issue:44008", "accept": false, - "reason": "Missing CLIPOutput attentions field vs RTDetrV2 output regression; unrelated symptoms." + "reason": "One is a Qwen3OmniMoe video-input ValueError, the other is a Gemma 3n variable-name collision in forward(); different models and distinct defects." }, { - "left": "issue:44077", - "right": "issue:44625", + "left": "issue:44220", + "right": "issue:44661", "accept": false, - "reason": "PatchTSMixer post-init policy vs Qwen3.5 num_labels propagation; different configuration bugs." + "reason": "Audio fbank extraction and add-new-model-like tokenizer mapping behavior are unrelated subsystems; no shared underlying bug." }, { "left": "issue:43720", "right": "issue:43819", "accept": false, - "reason": "Packed-weight loading issue vs DAC forward/from_latents mismatch; not the same code path." + "reason": "Both mention BitNet-like loading issues, but one is about packed weights during accelerate loading and the other is DAC.from_latents vs forward with missing STE; not the same change." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "PatchTSMixer post_init policy and Qwen3.5 num_labels propagation are separate configuration issues affecting different models." }, { "left": "issue:43575", "right": "issue:45704", "accept": false, - "reason": "Tensor-parallel OOM on Qwen2 vs T5 Apex RMSNorm memory leak; unrelated problems." + "reason": "Qwen2 TP loading OOM and T5 apex.FusedRMSNorm memory leak are both memory-related, but they concern different models and different failure mechanisms." }, { "left": "issue:30064", "right": "issue:36010", "accept": false, - "reason": "Image processor segmentation-map handling vs GenerationMixin import error; completely different areas." - }, - { - "left": "issue:43334", - "right": "issue:45759", - "accept": false, - "reason": "Both involve Qwen multimodal config loading, but one is a missing pad_token_id field and the other is text_config unwrapping; not the same bug." + "reason": "Image processor void segmentation map handling and GenerationMixin import failure are unrelated bugs in different parts of the stack." }, { "left": "issue:43335", "right": "issue:43756", "accept": false, - "reason": "SwitchTransformers sparse-layer construction vs Smollm3 RoPE-layer selection; distinct model/config issues." + "reason": "SwitchTransformers sparse-layer creation and Smollm3 RoPE layer dropping are both model-architecture issues, but they are not the same underlying defect." }, { "left": "issue:43653", "right": "issue:43720", "accept": false, - "reason": "Tokenizer special-token registration bug vs BitNet packed-weight unpacking bug; unrelated." + "reason": "BigBirdTokenizer special-token decode behavior and BitNet packed-weight loading are separate tokenizer vs model-loading regressions." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "Qwen3-VL missing pad_token_id and Qwen 3.5/3.6 from_config text_config unwrapping are related to composite configs, but they describe different attribute propagation problems." }, { "left": "issue:43716", "right": "issue:44008", "accept": false, - "reason": "Mistral image-preprocessor dtype mismatch vs Gemma3n audio-mask attribute collision; different models and failures." + "reason": "Mistral-3 image preprocessor dtype mismatch and Gemma 3n audio_mel_mask collision are unrelated multimodal issues with different root causes." }, { "left": "issue:42915", "right": "issue:43746", "accept": false, - "reason": "Qwen3Moe FP8 config failure vs GraniteSpeech checkpoint/adapters loading; not the same underlying issue." + "reason": "Qwen3Moe FineGrainedFP8Config failure and GraniteSpeech PEFT adapter loading are distinct model-loading problems." }, { "left": "issue:43746", "right": "issue:45381", "accept": false, - "reason": "Local checkpoint loading with PEFT adapters vs Qwen2.5-VL video position IDs; different loaders and bugs." + "reason": "GraniteSpeech PEFT checkpoint loading and Qwen2.5-VL video vision_position_ids are unrelated; one is checkpoint restoration, the other is video position indexing." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "Checkpoint/adapters loading vs offloading with quantization; related theme, but not the same concrete failure." + "reason": "GraniteSpeech local checkpoint loading and offloading with quantization are different loading paths and different bugs." }, { "left": "issue:43550", "right": "issue:45072", "accept": false, - "reason": "torch.compile+SDPA regression on Bamba vs bfloat16 dtype mismatches on other models; unrelated." + "reason": "Both involve Bamba/Transformer inference and dtype/compile issues, but one is torch.compile with SDPA and the other is bfloat16 mismatches in different models; not mergeable into one fix." }, { "left": "issue:44186", "right": "issue:44898", "accept": false, - "reason": "Tokenizer padding/truncation crash vs Perceiver image-classification resolution failure; different model families and paths." + "reason": "LayoutLMv2Tokenizer padding/truncation crashes and Perceiver interpolation failures are unrelated preprocessing/model-usage bugs." }, { "left": "issue:43550", "right": "issue:43819", "accept": false, - "reason": "Bamba compile/SDPA issue vs DAC latent conversion mismatch; no shared code path." + "reason": "Bamba torch.compile+SDPA failure and DAC.from_latents mismatch are different code paths and different model families." }, { "left": "issue:43421", "right": "issue:44811", "accept": false, - "reason": "Runtime special-token/post-processor sync bug vs Whisper batch_decode skip_special_tokens behavior; both tokenizer-related but not the same issue." + "reason": "Runtime special-token post-processor updates and Whisper batch_decode skip_special_tokens handling are tokenizer/processor behaviors, but they are different mechanisms and fixes." }, { "left": "issue:44843", "right": "issue:45081", "accept": false, - "reason": "Both touch _patch_mistral_regex, but one is an offline model_info call and the other is a backend_tokenizer AttributeError; different failure modes." + "reason": "Mistral regex offline hub access and Mistral tokenizer backend_tokenizer crash both touch Mistral loading, but one is hub/network access and the other is tokenizer API compatibility; not the same bug." }, { - "left": "issue:43643", - "right": "issue:44877", + "left": "issue:43577", + "right": "issue:43818", "accept": false, - "reason": "trust_remote_code config field loss vs strict-config rejection of granite_speech; separate config-loading bugs." + "reason": "BLIP2 dtype propagation and Video-LLaVA missing temporal attention/shared weights are separate multimodal model problems." }, { "left": "issue:44062", "right": "issue:44488", "accept": false, - "reason": "AddedToken constructor argument collision vs model loading failure for cjvt/sleng-bert; no evidence of the same root cause." + "reason": "AddedToken special-argument duplication and failing to load a specific model checkpoint are unrelated tokenizer construction vs model loading bugs." }, { - "left": "issue:43577", - "right": "issue:43818", + "left": "issue:43531", + "right": "issue:43643", "accept": false, - "reason": "BLIP2 dtype propagation bug vs Video-LLaVA missing temporal attention/weight sharing; unrelated model issues." + "reason": "Sliding_window behavior in Qwen3-MoE and trust_remote_code missing fields in AutoConfig are not the same underlying issue." }, { - "left": "issue:43531", - "right": "issue:43643", + "left": "issue:43643", + "right": "issue:44877", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior vs trust_remote_code config-field loss; different subsystems." + "reason": "Both are config-loading regressions, but one is trust_remote_code field loss and the other is strict config blocking granite_speech; different failure modes." }, { "left": "issue:43653", "right": "issue:43819", "accept": false, - "reason": "BigBird tokenizer special-token registration vs DAC latent conversion mismatch; unrelated." + "reason": "BigBirdTokenizer special-token registration and DAC latent reconstruction mismatch are unrelated tokenizer vs audio codec issues." }, { - "left": "issue:43540", - "right": "issue:45381", + "left": "issue:44448", + "right": "issue:44671", "accept": false, - "reason": "Both are video-related Qwen issues, but one is video input processing in Qwen3OmniMoe and the other is incorrect vision_position_ids in Qwen2.5-VL; not the same bug." + "reason": "Pegasus v4/v5 output drift and CamemBERT masked LM prediction changes are both version regressions, but they affect different model architectures and behaviors." }, { - "left": "issue:44448", - "right": "issue:44671", + "left": "issue:43540", + "right": "issue:45381", "accept": false, - "reason": "Different regression reports for different models and tasks (Pegasus generation vs CamemBERT masked LM)." + "reason": "Qwen3OmniMoe video ValueError and Qwen2.5-VL video position id bug are both video-related, but they are different models and different internal failures." }, { "left": "issue:43782", "right": "issue:44977", "accept": false, - "reason": "Qwen3VL loading with weight_only=True vs Qwen3.5 flash-attention generation; same family, different concrete problems." + "reason": "Qwen3VL weight_only loading error and Qwen3.5 flash-attention generation issue are separate Qwen-family bugs with different triggers." }, { "left": "issue:44561", "right": "issue:45020", "accept": false, - "reason": "Both concern remote_code breakages, but one is specifically is_torch_fx_available removal and the other is a broad umbrella report; not a single concrete bug." + "reason": "Both involve remote-code compatibility regressions, but one is specifically about is_torch_fx_available removal and the other is a broader recent remote_code breakage report; related theme, not the same bug." } ] }, @@ -4684,7 +4574,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:30064", "issue:34634", @@ -4736,14 +4626,14 @@ "issue:45701" ], "soft_pairs": [ - "issue:43322|issue:45685", "issue:43825|issue:43994", + "issue:43322|issue:45685", "issue:44821|issue:45084", "issue:30064|issue:36331", "issue:43825|issue:43827", "issue:44485|issue:44509", - "issue:43577|issue:43582", "issue:44186|issue:44442", + "issue:43577|issue:43582", "issue:43606|issue:45072", "issue:34634|issue:36010", "issue:43452|issue:44991", @@ -4759,201 +4649,201 @@ "issue:43582|issue:44857", "issue:43262|issue:43819", "issue:43526|issue:43697", - "issue:45084|issue:45538", "issue:43644|issue:43749", + "issue:45084|issue:45538", "issue:43329|issue:45600", "issue:43653|issue:44912" ], - "prepared_review_unit_hash": "e336cd086a54854a4afb23ca1774bf7e96cebbb27dc3974559d7091a500b05cf", + "prepared_review_unit_hash": "8fb7d7f9854c833947ca3feca987cb54ac82c115d19e066ca27c590703c34df4", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 27, - "serialized_chars": 35995, - "estimated_input_tokens": 8999, - "estimated_eval_tokens": 18254 + "serialized_chars": 35997, + "estimated_input_tokens": 9000, + "estimated_eval_tokens": 18256 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: the soft pairs share broad vocabulary or subsystem area, but not the same concrete bug/change. I would not merge any of the candidate soft edges. If forced to anchor the cluster, issue 43825 is the closest representative because it is the most central of the pipeline-v5-related items.", - "confidence": 0.84, - "canonical_issue_reason": "issue:43825 is the closest thing to a hub in this set: it is concrete, user-facing, and is the only item with nearby v5 pipeline-removal follow-ups in the soft links. It is still not a true duplicate of the others, but it is the best anchor among the issues.", + "summary": "This cluster is heterogeneous: most items are unrelated bug reports that only share broad transformer/model-loading vocabulary. None of the soft pairs look like the same concrete bug/change, so I would not deduplicate any of them.", + "confidence": 0.94, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43825 is the best overall issue to represent the cluster because it is specific, actionable, and closest to the only small thematic pocket here (pipeline/v5 migration).", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43322", - "right": "issue:45685", + "left": "issue:43825", + "right": "issue:43994", "accept": false, - "reason": "Different failures in different stacks: Llava Next loading segfault vs Apple Silicon MPS histogram error. Only broad ML/runtime similarity." + "reason": "v5 pipeline error-message cleanup vs a SigLIP2 model quality/reporting issue; different code paths and symptoms." }, { - "left": "issue:43825", - "right": "issue:43994", + "left": "issue:43322", + "right": "issue:45685", "accept": false, - "reason": "Both mention pipeline/model behavior, but one is an error-message regression and the other is bad model outputs. Not the same bug." + "reason": "Llava Next loading segfault vs an MPS histogram/int dtype limitation; unrelated failures." }, { "left": "issue:44821", "right": "issue:45084", "accept": false, - "reason": "Unrelated code paths: URL image processor loading vs Jinja/template compilation error." + "reason": "AutoImageProcessor URL loading vs tokenizer template compilation; different subsystems and fixes." }, { "left": "issue:30064", "right": "issue:36331", "accept": false, - "reason": "Different domains entirely: image segmentation-map processing vs a trainer compute_loss signature mismatch." + "reason": "Image processor void segmentation maps vs trainer.compute_loss signature mismatch; unrelated." }, { "left": "issue:43825", "right": "issue:43827", "accept": false, - "reason": "Related v5 pipeline-removal theme, but one is runtime error messaging and the other is docs lag. Too different to be one duplicate/change." + "reason": "Both concern v5 pipeline removals, but one is an error message and the other is docs cleanup, not the same bug." }, { "left": "issue:44485", "right": "issue:44509", "accept": false, - "reason": "Different topics: GLM-5 RoPE implementation vs documentation still mentioning removed pipeline tasks." + "reason": "GLM-5 RoPE implementation discussion vs docs mentioning removed pipeline tasks; unrelated." }, { - "left": "issue:43577", - "right": "issue:43582", + "left": "issue:44186", + "right": "issue:44442", "accept": false, - "reason": "Different bugs: dtype propagation in BLIP-2 loading vs Apple Silicon allocator warmup TypeError." + "reason": "Two different tokenizer crashes in different classes and load paths; too far apart to be the same bug." }, { - "left": "issue:44186", - "right": "issue:44442", + "left": "issue:43577", + "right": "issue:43582", "accept": false, - "reason": "Both are tokenizer issues, but with different tokenizers, symptoms, and code paths." + "reason": "BLIP2 dtype persistence bug vs Apple Silicon allocator TypeError; unrelated." }, { "left": "issue:43606", "right": "issue:45072", "accept": false, - "reason": "Different model/runtime mismatches: CPU offload device mismatch vs bfloat16 dtype mismatch." + "reason": "CPU offload device mismatch vs bfloat16 dtype mismatch; different concrete failures." }, { "left": "issue:34634", "right": "issue:36010", "accept": false, - "reason": "BarkProcessor voice_preset behavior is unrelated to the GenerationMixin import error." + "reason": "BarkProcessor voice preset bug vs GenerationMixin import error; unrelated." }, { "left": "issue:43452", "right": "issue:44991", "accept": false, - "reason": "Both involve loading/tokenizers, but the concrete regressions are different and do not look mergeable into one fix." + "reason": "gguf_file loading regression vs a specific tokenizer load failure in v5; different underlying causes." }, { "left": "issue:44805", "right": "issue:44936", "accept": false, - "reason": "Mask-shape IndexError during indexing is unrelated to evaluate() failing after train()." + "reason": "Mask shape IndexError vs trainer.evaluate() after train lifecycle bug; unrelated." }, { "left": "issue:44610", "right": "issue:45072", "accept": false, - "reason": "Processor image-size mismatch vs dtype mismatch; different code paths and fixes." + "reason": "Processor output size mismatch vs dtype mismatch; not the same issue." }, { "left": "issue:42175", "right": "issue:45081", "accept": false, - "reason": "Backend dependency packaging issue vs tokenizer regex patch crash; no shared underlying bug." + "reason": "Packaging/backend install problem vs Mistral tokenizer regex patch crash; unrelated." }, { "left": "issue:43720", "right": "issue:45072", "accept": false, - "reason": "Packed-weight unpacking during accelerate loading is a different failure mode from bfloat16 dtype mismatch." + "reason": "Packed-weight unpacking during accelerate loading vs bfloat16 dtype mismatch; different bugs." }, { "left": "issue:45081", "right": "issue:45406", "accept": false, - "reason": "Both are attribute errors in tokenizer/processor handling, but they affect different models and different code paths." + "reason": "Different attribute errors in different processor/tokenizer classes; not one shared fix." }, { "left": "issue:44671", "right": "issue:45701", "accept": false, - "reason": "Specific CamemBERT MLM regression vs generic tokenization changes across versions; too broad and not the same bug." + "reason": "Specific CamemBERT MLM regression vs a broad tokenization-version complaint; too vague to merge." }, { "left": "issue:43650", "right": "issue:43906", "accept": false, - "reason": "No meaningful duplicate relationship; one is a placeholder-like issue title and the other is an isolated reproduction report." + "reason": "Placeholder issue vs isolated reproduction of another issue; no common underlying bug." }, { "left": "issue:44060", "right": "issue:44368", "accept": false, - "reason": "They both mention tied-weights warnings, but one reports a concrete incorrect tying bug and the other is a user-visible warning in a different scenario; not clearly the same code-path defect." + "reason": "Both mention Qwen tied-weight warnings, but one is a concrete wrong tying bug and the other is a config warning in fine-tuning." }, { "left": "issue:43404", "right": "issue:43782", "accept": false, - "reason": "Different model-loading problems in different model families; not the same bug." + "reason": "Mistral3 tied-weights bug vs Qwen3VL weight_only loading error; unrelated." }, { "left": "issue:43582", "right": "issue:44857", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError and CUDA AMP loss crash are unrelated runtime issues." + "reason": "Apple Silicon allocator warmup TypeError vs CUDA AMP loss crash; unrelated." }, { "left": "issue:43262", "right": "issue:43819", "accept": false, - "reason": "Audio chat-template sampling-rate default vs DAC latent reconstruction mismatch are different bugs." + "reason": "Audio chat-template sampling-rate default vs DAC latent-forward mismatch; different models and code paths." }, { "left": "issue:43526", "right": "issue:43697", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug is unrelated to RTDetrV2 differing outputs in v5." + "reason": "BeitImageProcessorFast label reduction bug vs RTDetrV2 output drift in v5; not the same concrete defect." }, { - "left": "issue:45084", - "right": "issue:45538", + "left": "issue:43644", + "right": "issue:43749", "accept": false, - "reason": "Template compilation error vs CLIPTokenizer model_max_length constant issue; different subsystems." + "reason": "Non-persistent buffer initialization issue vs FSDP RAM-efficient loading regression; related to loading broadly, but not the same bug." }, { - "left": "issue:43644", - "right": "issue:43749", + "left": "issue:45084", + "right": "issue:45538", "accept": false, - "reason": "Both are loading-related, but one is non-persistent buffer initialization and the other is FSDP CPU RAM-efficient loading." + "reason": "Template compilation error vs CLIPTokenizer model_max_length constant; unrelated." }, { "left": "issue:43329", "right": "issue:45600", "accept": false, - "reason": "Undefined video-token helper bug vs removed-config reference in auto_mappings; not the same underlying change." + "reason": "Multimodal token counting video-branch bug vs stale auto_mappings config references; different parts of the codebase." }, { "left": "issue:43653", "right": "issue:44912", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug is unrelated to MXFP4 quantization fallback." + "reason": "BigBirdTokenizer special-token registration bug vs MXFP4 quantization fallback; unrelated." } ] }, @@ -4975,7 +4865,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:34567", "issue:34634", @@ -5029,25 +4919,25 @@ "soft_pairs": [ "issue:43638|issue:44936", "issue:43746|issue:44811", - "issue:44589|issue:45042", - "issue:43454|issue:43577", "issue:38175|issue:44610", + "issue:44589|issue:45042", "issue:43854|issue:44977", + "issue:43454|issue:43577", "issue:36010|issue:43645", "issue:43526|issue:44554", - "issue:34634|issue:36331", "issue:44568|issue:45488", + "issue:34634|issue:36331", "issue:43819|issue:43927", - "issue:39692|issue:44912", "issue:43257|issue:43329", + "issue:39692|issue:44912", "issue:43720|issue:44610", "issue:43749|issue:44291", "issue:43550|issue:44912", - "issue:44568|issue:45356", "issue:43262|issue:44811", + "issue:44568|issue:45356", "issue:43531|issue:45042", - "issue:36010|issue:36331", "issue:43643|issue:45759", + "issue:36010|issue:36331", "issue:34567|issue:40990", "issue:43329|issue:43450", "issue:43329|issue:44479", @@ -5055,224 +4945,220 @@ "issue:43257|issue:43931", "issue:43697|issue:45357", "issue:43452|issue:43582", - "issue:42915|issue:43742", - "issue:44291|issue:45042" + "issue:42915|issue:43742" ], - "prepared_review_unit_hash": "52e6a02eab8b2c81fbce6a58a252d951bd59029a4d4b26acde00b79c1e46cc2f", + "prepared_review_unit_hash": "ad9a8a7a57f9faab700aed9bad61a1b5faf90dd128a667a4e281c5a830361de2", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 37020, - "estimated_input_tokens": 9255, - "estimated_eval_tokens": 18766 + "soft_pair_count": 29, + "serialized_chars": 36722, + "estimated_input_tokens": 9181, + "estimated_eval_tokens": 18618 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose semantic grouping of unrelated Transformers issues: trainer state, loading/config regressions, tokenizer changes, multimodal/video bugs, and processor/model mismatches. I don\u2019t see any true duplicate pairs; the similarities are generally only at the subsystem level.", - "confidence": 0.93, - "canonical_issue_reason": "issue:45701 is the broadest and most representative active issue around tokenizer behavior changes, with inbound references and a title that captures the shared theme better than the narrower variants.", + "summary": "The cluster is mostly heterogeneous. The only accepted duplicate edge is the Qwen video regression pair (43329 + 44479); the Qwen3 MoE loading issue (43257) is not video-related and should not be grouped with it.", + "confidence": 0.89, + "canonical_issue_reason": "issue:44479 is the best canonical issue for the only clear duplicate family: it describes the broad user-facing video-input regression across multiple Qwen VLM variants, while issue:43329 is the narrower helper-level diagnosis.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45701 is the best global representative because it is open, broadly worded, and sits closest to the recurring tokenizer-regression slice of this otherwise heterogeneous cluster.", + "best_issue_reason": "issue:44479 is the strongest global issue anchor because it captures the broadest symptom set in the accepted family and is the most user-visible report in the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43638", "right": "issue:44936", "accept": false, - "reason": "Both are trainer/runtime failures, but one is a DeepSpeed Zero3 indexing error and the other is an evaluate-after-train bug; different code paths and symptoms." + "reason": "Different training failures: deepspeed/zero3 index error vs trainer.evaluate() breaking after train()." }, { "left": "issue:43746", "right": "issue:44811", "accept": false, - "reason": "Local PEFT checkpoint loading for GraniteSpeech is unrelated to Whisper batch_decode skip_special_tokens behavior." + "reason": "Different audio bugs: GraniteSpeech PEFT checkpoint loading vs Whisper batch_decode skip_special_tokens handling." }, { - "left": "issue:44589", - "right": "issue:45042", + "left": "issue:38175", + "right": "issue:44610", "accept": false, - "reason": "Float8 storage lookup failure and PIL-image-processor torchvision dependency are unrelated problems." + "reason": "Different models and symptoms: SigLIP2 zero probabilities vs OmDet-Turbo processor size mismatch." }, { - "left": "issue:43454", - "right": "issue:43577", + "left": "issue:44589", + "right": "issue:45042", "accept": false, - "reason": "Weight tying in AyaVision and Blip2 dtype propagation are different model-loading bugs." + "reason": "Unrelated failures: float8 storage lookup error vs PIL backend image processors requiring torchvision." }, { - "left": "issue:38175", - "right": "issue:44610", + "left": "issue:43854", + "right": "issue:44977", "accept": false, - "reason": "Both mention model/processor mismatches, but one is SigLIP2 zero probabilities and the other is OmDet-Turbo image-size mismatch; not the same bug." + "reason": "Different model families and failure modes: GLM-4.7 load issue vs Qwen3.5 flash-attention generation regression." }, { - "left": "issue:43854", - "right": "issue:44977", + "left": "issue:43454", + "right": "issue:43577", "accept": false, - "reason": "GLM-4.7-Flash loading failures are not the same as Qwen3.5 flash-attention generation regressions." + "reason": "Different model-specific bugs: AyaVision weight tying vs Blip2 dtype preservation." }, { "left": "issue:36010", "right": "issue:43645", "accept": false, - "reason": "GenerationMixin import errors and Jupyter custom-model initialization issues are distinct import/runtime regressions." + "reason": "Both are runtime/loading-related, but one is a GenerationMixin import break and the other is a Jupyter custom-model init regression." }, { "left": "issue:43526", "right": "issue:44554", "accept": false, - "reason": "BeitImageProcessorFast label reduction and MPS attention correctness are unrelated." + "reason": "Completely different subsystems: Beit image label reduction vs MPS attention correctness." }, { - "left": "issue:34634", - "right": "issue:36331", + "left": "issue:44568", + "right": "issue:45488", "accept": false, - "reason": "BarkProcessor voice_preset and CustomTrainer.compute_loss kwarg handling are unrelated APIs and failures." + "reason": "Both are tokenizer regressions, but they affect different tokenizer classes and different concrete behaviors." }, { - "left": "issue:44568", - "right": "issue:45488", + "left": "issue:34634", + "right": "issue:36331", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different tokenizer behaviors and model families; not one concrete duplicate bug." + "reason": "No shared code path: BarkProcessor voice_preset vs CustomTrainer.compute_loss signature mismatch." }, { "left": "issue:43819", "right": "issue:43927", "accept": false, - "reason": "DAC from_latents mismatch and DiaConfig save/load token-id loss are different model/config bugs." + "reason": "Different bugs: DAC latent reconstruction mismatch vs DiaConfig losing custom token IDs on save/load." }, { - "left": "issue:39692", - "right": "issue:44912", + "left": "issue:43257", + "right": "issue:43329", "accept": false, - "reason": "SigLIP2 docs example errors and MXFP4 quantization fallback in git-oss-20b are separate issues." + "reason": "Different Qwen3-related problems: MoE weight conversion/loading with accelerate+deepspeed vs multimodal video token counting helper bug." }, { - "left": "issue:43257", - "right": "issue:43329", + "left": "issue:39692", + "right": "issue:44912", "accept": false, - "reason": "Qwen3 MoE weight conversion with accelerate/deepspeed and multimodal video token counting undefined names are different code paths." + "reason": "Both mention quantization/docs, but they are different models and different failure modes." }, { "left": "issue:43720", "right": "issue:44610", "accept": false, - "reason": "Packed-weight unpacking during accelerate loading and processor image-shape mismatch are unrelated." + "reason": "Unrelated: BitNet accelerate-loading packed weights vs OmDet-Turbo processor input-size mismatch." }, { "left": "issue:43749", "right": "issue:44291", "accept": false, - "reason": "FSDP CPU RAM efficient loading and init_empty_weights/_is_hf_initialized are both loading-related but not the same bug." + "reason": "Both are loading-related, but one is FSDP CPU RAM efficient loading and the other is init_empty_weights/_is_hf_initialized handling." }, { "left": "issue:43550", "right": "issue:44912", "accept": false, - "reason": "torch.compile+SDPA failure in Bamba and MXFP4 quantization fallback in git-oss are separate failures." + "reason": "Different problems: torch.compile with SDPA vs MXFP4 quantization fallback." }, { - "left": "issue:44568", - "right": "issue:45356", + "left": "issue:43262", + "right": "issue:44811", "accept": false, - "reason": "Both are tokenizer regressions, but the specific models and faulty behaviors differ." + "reason": "Both involve audio processors, but one is chat-template sampling-rate defaults and the other is batch_decode special-token stripping." }, { - "left": "issue:43262", - "right": "issue:44811", + "left": "issue:44568", + "right": "issue:45356", "accept": false, - "reason": "Audio chat-template sampling rate defaults and Whisper batch_decode token skipping are unrelated processor behaviors." + "reason": "Tokenizer regressions in different tokenizers with different concrete breakages." }, { "left": "issue:43531", "right": "issue:45042", "accept": false, - "reason": "Qwen3-MoE sliding_window and PIL backend torchvision dependency are unrelated." + "reason": "Completely different areas: Qwen3-MoE sliding_window vs PIL backend torchvision dependency." }, { - "left": "issue:36010", - "right": "issue:36331", + "left": "issue:43643", + "right": "issue:45759", "accept": false, - "reason": "GenerationMixin import resolution and CustomTrainer.compute_loss signature handling are different issues." + "reason": "Both are config-loading issues, but one is trust_remote_code missing fields and the other is composite Qwen config text_config unwrapping." }, { - "left": "issue:43643", - "right": "issue:45759", + "left": "issue:36010", + "right": "issue:36331", "accept": false, - "reason": "Both involve config/model construction, but one loses fields with trust_remote_code while the other fails to unwrap text_config in composite configs." + "reason": "Unrelated: GenerationMixin import failure vs CustomTrainer.compute_loss kwarg mismatch." }, { "left": "issue:34567", "right": "issue:40990", "accept": false, - "reason": "TrainerState token counting and high perplexity on gpt-oss-20b are unrelated." + "reason": "Trainer token-count bookkeeping is unrelated to the GPT-OSS perplexity regression." }, { "left": "issue:43329", "right": "issue:43450", "accept": false, - "reason": "Undefined video token-count helpers and batched video processor output shape are distinct video-path bugs." + "reason": "Both are video-related, but they describe different bugs: missing helper variables in multimodal token counting vs batched video processor shape mismatch." }, { "left": "issue:43329", "right": "issue:44479", - "accept": false, - "reason": "One is an undefined-helper bug in token counting; the other is a broader video-input regression across Qwen video models." + "accept": true, + "reason": "Same underlying Qwen video regression: 43329 identifies the broken video-path helper, and 44479 reports the broader user-visible failure across Qwen VLM variants." }, { "left": "issue:44849", "right": "issue:45701", "accept": false, - "reason": "Qwen3.5 hidden-state output behavior and generic tokenization changes are unrelated." + "reason": "Similar tokenization theme, but the reports concern different version/regression contexts and different failure descriptions." }, { "left": "issue:43257", "right": "issue:43931", "accept": false, - "reason": "Qwen3 MoE loading with accelerate/deepspeed and Qwen3-VL weight-shape mismatches are different model-loading failures." + "reason": "Different Qwen3 issues: MoE loading/conversion failure vs Qwen3-VL weight-shape mismatch." }, { "left": "issue:43697", "right": "issue:45357", "accept": false, - "reason": "RTDetrV2 output drift and Qwen3.5 save_pretrained visual-key regression are unrelated." + "reason": "Different regression types: RTDetrV2 output divergence vs Qwen3.5 visual encoder key saving bug." }, { "left": "issue:43452", "right": "issue:43582", "accept": false, - "reason": "gguf_file tokenizer/model loading and Apple Silicon caching_allocator_warmup TypeError are unrelated." + "reason": "Different loading/runtime bugs: gguf_file path handling vs Apple Silicon caching allocator warmup TypeError." }, { "left": "issue:42915", "right": "issue:43742", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure and MobileLLM-125M key error do not share a concrete bug." - }, - { - "left": "issue:44291", - "right": "issue:45042", - "accept": false, - "reason": "init_empty_weights argument handling and PIL backend torchvision dependency are unrelated." + "reason": "Different Qwen3-MoE-related failures: FineGrainedFP8Config vs MobileLLM key error during load." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, - "accepted_nontrivial_soft_edge": false, + "evaluator_result": { + "accept": false, + "feedback": "The cluster summary is mostly grounded, but at least one soft-edge rationale overstates shared evidence. In particular, the `issue:43257` vs `issue:43329` note says \"Both are video-related,\" but `43257` is about Qwen3 MoE weight conversion/loading with accelerate+deepspeed and is not video-related per the packet. That makes the verdict rationale non-conservative. The accepted Qwen video pair (`43329` + `44479`) looks reasonable, but the review should be rejected because not every soft-edge explanation stays strictly within the packet evidence." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -5287,7 +5173,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:33357", "issue:34567", @@ -5321,13 +5207,12 @@ "issue:44560", "issue:44737", "issue:44756", - "issue:44792", "issue:44821", "issue:44898", "issue:44912", - "issue:44964", "issue:44987", "issue:44991", + "issue:45042", "issue:45072", "issue:45081", "issue:45137", @@ -5340,17 +5225,18 @@ ], "soft_pairs": [ "issue:41720|issue:43540", + "issue:44291|issue:45042", "issue:33357|issue:34567", - "issue:44987|issue:45538", - "issue:43317|issue:45538", - "issue:42915|issue:44821", "issue:44479|issue:45381", + "issue:43317|issue:45538", + "issue:44987|issue:45538", "issue:43540|issue:44560", + "issue:42915|issue:44821", "issue:43927|issue:44912", "issue:43540|issue:44479", - "issue:43577|issue:45072", - "issue:42915|issue:45538", "issue:43257|issue:43756", + "issue:42915|issue:45538", + "issue:43577|issue:45072", "issue:43257|issue:43696", "issue:45137|issue:45405", "issue:38175|issue:44898", @@ -5360,237 +5246,236 @@ "issue:43906|issue:45381", "issue:43646|issue:43749", "issue:43749|issue:44756", - "issue:44062|issue:45081", "issue:34567|issue:36331", - "issue:36010|issue:42175", + "issue:44062|issue:45081", "issue:43643|issue:45375", "issue:43493|issue:43994", - "issue:42175|issue:43824", + "issue:36010|issue:42175", "issue:43452|issue:44488", - "issue:44291|issue:44821", + "issue:42175|issue:43824", "issue:43526|issue:45479", - "issue:44792|issue:44964" + "issue:44291|issue:44821" ], - "prepared_review_unit_hash": "f21f3bf86e9863844ca018d0c36ed334efd3db97d799a70c1e3fbc08cc699557", + "prepared_review_unit_hash": "06d530cb1f590d6d7a7f6bdfd8a7a58d6f0c7f0e17ff0b640088a22f75effb08", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 37139, - "estimated_input_tokens": 9285, - "estimated_eval_tokens": 18826 + "serialized_chars": 36566, + "estimated_input_tokens": 9142, + "estimated_eval_tokens": 18540 }, "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 based on the same concrete bug path. The strongest matches are the Qwen video-regression reports, the SigLIP2 output discrepancy reports, and the config-field-loss reports.", - "confidence": 0.74, - "canonical_issue_reason": "No single issue cleanly represents the whole cluster because the items span many unrelated bugs; if one must be chosen, issue 44479 is the best anchor because it is a clear regression report with a closely related duplicate in 45381.", + "summary": "This set is mostly a grab bag of unrelated closed issues across loading, tokenizer, dtype, quantization, and multimodal/video regressions. The only near-duplicate-looking subgroup is the Qwen video-input regression pair; everything else is too different to merge as duplicates.", + "confidence": 0.81, + "canonical_issue_reason": "issue:44479 is the broadest and most central bug report here: it describes a v5.3.0 video-input regression across several Qwen VL variants, and likely subsumes the more specific Qwen2.5-VL video-position-id report.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44479 is the best overall issue to anchor this cluster because it is a concrete, user-facing regression with the clearest duplicate signal and broad enough scope to represent the accepted Qwen video-input pair.", + "best_issue_reason": "issue:44479 is the best representative issue for the cluster because it covers the clearest shared regression theme and is the only one that plausibly generalizes to another item in the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41720", "right": "issue:43540", "accept": false, - "reason": "Different Qwen failures: auto device-map CUDA assert vs video-input ValueError; not the same bug." + "reason": "Both mention Qwen3, but one is auto device mapping cuda asserts and the other is a video-input ValueError; different failures and code paths." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "One is an init_empty_weights / unexpected argument bug, the other is a PIL image processor torchvision dependency regression." }, { "left": "issue:33357", "right": "issue:34567", "accept": false, - "reason": "Completely unrelated bugs: MacOS bus error in CLIP vs TrainerState token counter not updating." + "reason": "MacOS bus error with CLIP is unrelated to TrainerState.num_input_tokens_seen not updating." }, { - "left": "issue:44987", - "right": "issue:45538", - "accept": false, - "reason": "Tokenizer/model loading regression vs CLIPTokenizer max-length issue; different code paths." + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "Both are Qwen VL video-input regressions in v5.3.0, and the newer report looks like a specific symptom of the broader video-position-id regression." }, { "left": "issue:43317", "right": "issue:45538", "accept": false, - "reason": "device_map/offload loading failure is unrelated to CLIPTokenizer max_length handling." + "reason": "Dequantized model loading with device_map=auto is unrelated to CLIPTokenizer's model_max_length bug." }, { - "left": "issue:42915", - "right": "issue:44821", + "left": "issue:44987", + "right": "issue:45538", "accept": false, - "reason": "Qwen3MoE FP8 loading bug vs AutoImageProcessor-from-URL issue; different components." - }, - { - "left": "issue:44479", - "right": "issue:45381", - "accept": true, - "reason": "Both describe the same Qwen video-input regression in transformers 5.3.0, including wrong vision position handling." + "reason": "Loading physical-intelligence/fast on transformers>=5.1.0 is a model-loading regression, not the CLIPTokenizer max-length issue." }, { "left": "issue:43540", "right": "issue:44560", "accept": false, - "reason": "Both involve video inputs, but the models and reported failures differ; not clearly one concrete bug." + "reason": "Both are Qwen video-related, but one is a video-input processing ValueError and the other is a StopIteration during Qwen3-vl-embedding loading; not clearly the same concrete bug." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "FineGrainedFP8Config failure for Qwen3Moe is unrelated to AutoImageProcessor-from-URL loading." }, { "left": "issue:43927", "right": "issue:44912", "accept": false, - "reason": "DiaConfig token-ID loss and MXFP4 quantization fallback are unrelated." + "reason": "DiaConfig save/load token ID loss is unrelated to MXFP4 quantization falling back to bf16." }, { "left": "issue:43540", "right": "issue:44479", "accept": false, - "reason": "Shared video theme, but different Qwen model families and symptom details; too broad to merge." + "reason": "Related theme, but one is a specific video-processing ValueError in Qwen3OmniMoe while the other is a broader Qwen VL video regression; not enough evidence they are the same bug." }, { - "left": "issue:43577", - "right": "issue:45072", + "left": "issue:43257", + "right": "issue:43756", "accept": false, - "reason": "Separate dtype bugs in different models and inference paths; not the same issue." + "reason": "MOE weight conversion with accelerate+deepspeed is unrelated to Smollm3 RoPE layer dropping." }, { "left": "issue:42915", "right": "issue:45538", "accept": false, - "reason": "Qwen3MoE FP8 loading bug does not match CLIPTokenizer model_max_length behavior." + "reason": "Qwen3Moe FineGrainedFP8Config failure is unrelated to CLIPTokenizer max-length handling." }, { - "left": "issue:43257", - "right": "issue:43756", + "left": "issue:43577", + "right": "issue:45072", "accept": false, - "reason": "Qwen3 MoE weight-conversion bug is unrelated to SmolLM3 RoPE-layer mismatch." + "reason": "Both involve dtype, but one is BLIP2 loading with half/bfloat16 and the other is SwitchTransformers/TimmWrapperModel inference mismatch; different concrete failures." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Different failure modes: MoE weight conversion vs CUDA OOM." + "reason": "MOE weight conversion under accelerate+deepspeed is unrelated to GPT-oss-20b CUDA OOM." }, { "left": "issue:45137", "right": "issue:45405", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error is unrelated to peft version bump." + "reason": "DeepSpeed ZeRO3 deque underflow is unrelated to the MIN_PEFT_VERSION bump." }, { "left": "issue:38175", "right": "issue:44898", "accept": false, - "reason": "Different vision-model correctness bugs with different triggers and symptoms." + "reason": "Siglip2 zero probabilities and Perceiver non-default-resolution failure are different model-specific correctness bugs." }, { "left": "issue:44737", "right": "issue:45588", "accept": false, - "reason": "XLNet CPU tensor allocation bug and flash-attention AttributeError are unrelated." + "reason": "XLNet CPU arange-device bug is unrelated to flash_attention.py crashing on s_aux=None." }, { "left": "issue:44062", "right": "issue:44991", "accept": false, - "reason": "Tokenizer loading problems, but the concrete causes differ: AddedToken kwarg collision vs est-roberta load failure." + "reason": "AddedToken special-argument TypeError is unrelated to tokenizer loading failure for EMBEDDIA/est-roberta." }, { "left": "issue:43742", "right": "issue:44291", "accept": false, - "reason": "MobileLLM key error and init_empty_weights _is_hf_initialized TypeError are different load-path bugs." + "reason": "MobileLLM key error and init_empty_weights unexpected _is_hf_initialized argument are different loading failures." }, { "left": "issue:43906", "right": "issue:45381", "accept": false, - "reason": "One is an isolated reproduction of another issue; the title does not establish the same Qwen2.5-VL video bug." + "reason": "An isolated reproduction of another issue is not the same as the Qwen2.5-VL video-position-id regression." }, { "left": "issue:43646", "right": "issue:43749", "accept": false, - "reason": "Custom model init breakage and FSDP CPU RAM-efficient loading are different code paths." + "reason": "Custom model initialization breakage in Transformers 5.0.0 is unrelated to FSDP_CPU_RAM_EFFICIENT_LOADING." }, { "left": "issue:43749", "right": "issue:44756", "accept": false, - "reason": "Both mention memory/loading, but one is FSDP-specific and the other is mmap on Strix Halo; not duplicates." - }, - { - "left": "issue:44062", - "right": "issue:45081", - "accept": false, - "reason": "Different tokenizer errors: AddedToken duplication vs missing backend_tokenizer in Mistral regex patch." + "reason": "Both are memory/loading related, but FSDP CPU RAM efficient loading and Strix Halo mmap OOM are different problems." }, { "left": "issue:34567", "right": "issue:36331", "accept": false, - "reason": "TrainerState token-count update and compute_loss signature mismatch are separate Trainer API issues." + "reason": "TrainerState token counting not updating is unrelated to CustomTrainer.compute_loss receiving an unexpected keyword argument." }, { - "left": "issue:36010", - "right": "issue:42175", + "left": "issue:44062", + "right": "issue:45081", "accept": false, - "reason": "GenerationMixin import failure is unrelated to TensorFlow missing from the torch extra." + "reason": "AddedToken special kwarg duplication is unrelated to the Mistral regex patch crashing on backend_tokenizer." }, { "left": "issue:43643", "right": "issue:45375", - "accept": true, - "reason": "Both point to config fields being dropped on load; 45375 is a concrete model-specific instance of the same missing-fields problem." + "accept": false, + "reason": "AutoConfig missing fields with trust_remote_code=True is unrelated to a missing deepstack_visual_indexes field in Qwen3_5MoeVisionConfig." }, { "left": "issue:43493", "right": "issue:43994", - "accept": true, - "reason": "Both concern the same SigLIP2 model discrepancy causing incorrect outputs in HF compared with the expected behavior." + "accept": false, + "reason": "SigLIP2 JAX-vs-HF discrepancy and nonsensical AutoModel/pipeline results are different correctness reports, not the same bug." }, { - "left": "issue:42175", - "right": "issue:43824", + "left": "issue:36010", + "right": "issue:42175", "accept": false, - "reason": "Packaging/backend dependency issue and Qwen2_5_VL import error are unrelated." + "reason": "GenerationMixin import failure is unrelated to TensorFlow being absent from the torch extra backend." }, { "left": "issue:43452", "right": "issue:44488", "accept": false, - "reason": "GGUF tokenizer/model loading breakage and failure to load a specific BERT model are different problems." + "reason": "gguf_file breaking AutoTokenizer/AutoModel loading is unrelated to the cjvt/sleng-bert load failure." }, { - "left": "issue:44291", - "right": "issue:44821", + "left": "issue:42175", + "right": "issue:43824", "accept": false, - "reason": "init_empty_weights constructor error and AutoImageProcessor URL loading error are unrelated." + "reason": "Missing TensorFlow backend in a torch install is unrelated to Qwen2_5_VLForConditionalGeneration import failure." }, { "left": "issue:43526", "right": "issue:45479", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug is unrelated to sequence-classification zero-loss behavior." + "reason": "BeitImageProcessorFast reduce_labels returning one label is unrelated to the zero-loss classification bug." }, { - "left": "issue:44792", - "right": "issue:44964", + "left": "issue:44291", + "right": "issue:44821", "accept": false, - "reason": "Different multimodal model failures with different symptoms and code paths; not enough evidence of one underlying bug." + "reason": "Unexpected _is_hf_initialized argument during init_empty_weights is unrelated to loading AutoImageProcessor from URL." } ] }, "evaluator_result": { "accept": true, - "feedback": "The analysis is grounded in the packet and stays mostly conservative. The three accepted pairs are well supported by the issue titles: the Qwen video regression pair, the SigLIP2 discrepancy pair, and the config-field-loss pair. The cluster-level summary appropriately frames the rest as heterogeneous without overclaiming duplicates. No obvious overstatement stands out." + "feedback": "Grounded overall and conservatively phrased. The only clearly plausible duplicate/subsumption in the packet is the Qwen video-input pair (44479/45381); the rest are distinct by title and failure mode. Minor note: phrases like \"unrelated\" are a bit absolute, but not overstated here given the available evidence." }, "evaluator_used": true, "retried": false, @@ -5609,7 +5494,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:24643", "issue:28282", @@ -5630,8 +5515,6 @@ "issue:43742", "issue:43756", "issue:43819", - "issue:44361", - "issue:44442", "issue:44466", "issue:44484", "issue:44488", @@ -5641,12 +5524,14 @@ "issue:44589", "issue:44625", "issue:44704", + "issue:44792", "issue:44811", "issue:44821", "issue:44843", "issue:44877", "issue:44898", "issue:44936", + "issue:44964", "issue:44977", "issue:45005", "issue:45042", @@ -5660,230 +5545,223 @@ "issue:45600" ], "soft_pairs": [ + "issue:44792|issue:44964", "issue:28282|issue:34689", "issue:45042|issue:45081", - "issue:44484|issue:44704", "issue:45405|issue:45600", - "issue:43742|issue:44589", "issue:43756|issue:45325", + "issue:43742|issue:44589", "issue:28282|issue:29127", + "issue:44484|issue:44704", "issue:43531|issue:44821", "issue:43697|issue:45216", - "issue:43643|issue:44561", "issue:44514|issue:45381", + "issue:43643|issue:44561", "issue:44625|issue:44877", "issue:43582|issue:44977", "issue:43452|issue:43577", "issue:43720|issue:44898", - "issue:44514|issue:45290", "issue:43452|issue:44843", - "issue:42175|issue:44488", + "issue:44514|issue:45290", "issue:44466|issue:45005", "issue:43819|issue:44811", + "issue:42175|issue:44488", "issue:43295|issue:45356", "issue:41762|issue:44936", "issue:43653|issue:44568", "issue:24643|issue:29127", - "issue:38175|issue:43295", - "issue:44361|issue:44442", - "issue:43653|issue:44361" + "issue:38175|issue:43295" ], - "prepared_review_unit_hash": "1fd38f3b157a8dd2741cb5f50b8a82e973fd86b5b090549b3b37bbcabeca868c", + "prepared_review_unit_hash": "7f2d5c1e3fc5bc229cc7db2b39f5871f5203dcbc7392739182cd17c70d3bf2f1", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 27, - "serialized_chars": 35536, - "estimated_input_tokens": 8884, - "estimated_eval_tokens": 18024 + "soft_pair_count": 26, + "serialized_chars": 35262, + "estimated_input_tokens": 8816, + "estimated_eval_tokens": 17888 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a grab bag of unrelated bug reports. Only one soft pair looks plausibly like the same underlying v5 tied-weights serialization problem; the rest are distinct issues affecting different code paths or models.", - "confidence": 0.78, - "canonical_issue_reason": "issue:45005 is the best representative of the only plausible duplicate pair because it states the broader tied-weights-on-v5 problem directly, while issue:44466 is the more specific device-dependent manifestation.", + "summary": "Nearly all soft pairs are only superficially related by subsystem or model family and should stay separate. The one plausible duplicate is the tied-weights serialization/save_pretrained regression pair (44466/45005).", + "confidence": 0.84, + "canonical_issue_reason": "issue:44466 is the clearest concrete report of the v5 tied-weights serialization bug, with device-dependent behavior and a specific `lm_head.weight` symptom.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45005 is the strongest global representative: it is the broadest, most general statement of the tied-weights regression and fits the duplicate cluster better than the more specific report.", + "best_issue_reason": "issue:44466 is the best global issue candidate because it is specific, reproducible, and appears to capture the underlying tied-weights regression more precisely than the broader follow-up report.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Both are multimodal/model-loading problems, but they concern different models and failures: a Janus test case versus Phi-4 multimodal loadability." + }, { "left": "issue:28282", "right": "issue:34689", "accept": false, - "reason": "Different failures: missing PyTorch import vs a model-loading regression for Llama 3.2 Vision." + "reason": "One is a missing-PyTorch ImportError, the other is a Llama 3.2 vision loading regression; the symptoms and code paths are unrelated." }, { "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "Unrelated areas: torchvision dependency for PIL image processors vs a Mistral regex/tokenizer crash." + "reason": "These are different bugs: torchvision dependency handling for PIL image processors versus a Mistral regex/tokenizer crash." }, { - "left": "issue:44484", - "right": "issue:44704", + "left": "issue:45405", + "right": "issue:45600", "accept": false, - "reason": "One is about save_pretrained shard sizing, the other about AutoProcessor not forwarding kwargs to cached_file." + "reason": "A dependency-version bump and a removed-config reference in auto_mappings are separate CI/issues, not the same bug." }, { - "left": "issue:45405", - "right": "issue:45600", + "left": "issue:43756", + "right": "issue:45325", "accept": false, - "reason": "Unrelated: PEFT minimum version bump vs removed Sam3LiteText config references in auto_mappings." + "reason": "Both touch RoPE/position handling in different models, but one is about missing RoPE layers in Smollm3 and the other about Qwen2.5-VL temporal position scaling." }, { "left": "issue:43742", "right": "issue:44589", "accept": false, - "reason": "Different code paths and symptoms: MobileLLM loading key error vs float8 storage lookup failure." + "reason": "MobileLLM key lookup failure and a Float8 storage TypeError are distinct loading/runtime failures." }, { - "left": "issue:43756", - "right": "issue:45325", + "left": "issue:28282", + "right": "issue:29127", "accept": false, - "reason": "Both involve positional/rope logic, but they affect different models and different implementations." + "reason": "These involve completely different topics: PyTorch import detection versus clearer error messaging for LayoutLMv3 box inputs." }, { - "left": "issue:28282", - "right": "issue:29127", + "left": "issue:44484", + "right": "issue:44704", "accept": false, - "reason": "ImportError for missing PyTorch is unrelated to layoutlmv3 box-validation error messaging." + "reason": "Max shard size defaults and processor kwargs propagation are unrelated save/load API issues." }, { "left": "issue:43531", "right": "issue:44821", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior is unrelated to loading AutoImageProcessor from a URL." + "reason": "Qwen3-MoE sliding-window behavior and loading AutoImageProcessor from URL are not the same underlying problem." }, { "left": "issue:43697", "right": "issue:45216", "accept": false, - "reason": "Different regressions: RTDetrV2 output changes vs Qwen3.5 save_pretrained checkpoint corruption." + "reason": "RTDetrV2 output drift in v5 and Qwen3.5 checkpoint serialization are separate regressions." }, { - "left": "issue:43643", - "right": "issue:44561", + "left": "issue:44514", + "right": "issue:45381", "accept": false, - "reason": "Missing fields from AutoConfig trust_remote_code loading is separate from removal of is_torch_fx_available." + "reason": "Both are Qwen2.5-VL-related, but one is a chat-template batching crash and the other is incorrect video vision position IDs; different code paths." }, { - "left": "issue:44514", - "right": "issue:45381", + "left": "issue:43643", + "right": "issue:44561", "accept": false, - "reason": "Both are Qwen2.5-VL issues, but one is chat-template batching and the other is video vision_position_ids." + "reason": "Missing fields from `AutoConfig` with `trust_remote_code` and removal of `is_torch_fx_available` breaking remote-code models are different failures." }, { "left": "issue:44625", "right": "issue:44877", "accept": false, - "reason": "num_labels propagation for Qwen3.5 is a different config problem than strict granite_speech config loading." + "reason": "A Qwen3.5 `num_labels` propagation bug is not the same as strict config blocking granite_speech loading." }, { "left": "issue:43582", "right": "issue:44977", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError is unrelated to Qwen3.5 flash-attention generation bugs." + "reason": "Apple Silicon allocator warmup TypeError and Qwen3.5 flash-attention generation issues are unrelated." }, { "left": "issue:43452", "right": "issue:43577", "accept": false, - "reason": "Tokenizer/model loading via gguf_file is unrelated to BLIP2 dtype propagation." + "reason": "GGUF loading breakage and BLIP-2 dtype propagation are different model-loading bugs." }, { "left": "issue:43720", "right": "issue:44898", "accept": false, - "reason": "Packed-weight unpacking in BitNet loading is a different bug from Perceiver interpolation on non-default resolution." - }, - { - "left": "issue:44514", - "right": "issue:45290", - "accept": false, - "reason": "Same API surface, but different triggers: batched padding=False vs assistant tool-call messages with empty content." + "reason": "BitNet packed-weight loading and Perceiver non-default-resolution failures do not share the same code-path problem." }, { "left": "issue:43452", "right": "issue:44843", "accept": false, - "reason": "gguf_file loading failure is unrelated to mistral-regex offline-mode model_info calls." + "reason": "GGUF loading failures and offline Mistral-regex hub calls are different tokenizer/model-loading regressions." }, { - "left": "issue:42175", - "right": "issue:44488", + "left": "issue:44514", + "right": "issue:45290", "accept": false, - "reason": "Packaging/backend dependency issue is unrelated to a specific tokenizer/model loading failure." + "reason": "Both involve `apply_chat_template`, but the reported failures are different conditions: batching/padding versus assistant tool-call messages with no content." }, { "left": "issue:44466", "right": "issue:45005", "accept": true, - "reason": "These both describe the same tied-weights regression in v5/save_pretrained, with 45005 being the broader umbrella report and 44466 the more specific device-dependent symptom." + "reason": "Both point to the same tied-weights/save_pretrained regression in v5; 44466 is the more specific device-dependent serialization report." }, { "left": "issue:43819", "right": "issue:44811", "accept": false, - "reason": "DAC latent reconstruction mismatch is unrelated to Whisper batch_decode skip_special_tokens handling." + "reason": "DAC latent conversion mismatch and Whisper batch_decode special-token handling are unrelated bugs." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "A missing TensorFlow backend from a pip extra and a model-loading failure for `cjvt/sleng-bert` are not the same issue." }, { "left": "issue:43295", "right": "issue:45356", "accept": false, - "reason": "Different regressions: processor.tokenizer access/images-to-tokenizer breakage vs Kimi-K2.5 codec and warning issues." + "reason": "Processor/tokenizer regression and Kimi-K2.5 codec/fix_mistral_regex regression are different problems." }, { "left": "issue:41762", "right": "issue:44936", "accept": false, - "reason": "DeepSpeed ZeRO-3 loading error for Gemma3 is unrelated to trainer.evaluate() after train()." + "reason": "DeepSpeed ZeRO-3 loading IndexError and trainer.evaluate() after train are separate trainer/model loading issues." }, { "left": "issue:43653", "right": "issue:44568", "accept": false, - "reason": "Both concern special tokens, but BigBird mask-token registration and mdeberta BOS/EOS insertion are different bugs." + "reason": "BigBird special-token registration and MdeBERTa add_special_tokens behavior are different tokenizer bugs." }, { "left": "issue:24643", "right": "issue:29127", "accept": false, - "reason": "DeepSpeed training weight-dimension runtime error is unrelated to LayoutLMv3 error-message clarity." + "reason": "A DeepSpeed training `weight` dimensionality error and a LayoutLMv3 box-info message issue are unrelated." }, { "left": "issue:38175", "right": "issue:43295", "accept": false, - "reason": "siglip2 zero-probabilities output issue is unrelated to processor.tokenizer regression." - }, - { - "left": "issue:44361", - "right": "issue:44442", - "accept": false, - "reason": "Different tokenizer failures: MLuke task AttributeError vs FastSpeech2ConformerTokenizer loading failure." - }, - { - "left": "issue:43653", - "right": "issue:44361", - "accept": false, - "reason": "BigBird special-token registration and MLuke task AttributeError are unrelated tokenizer bugs." + "reason": "Unexpected zero probabilities in SigLIP2 and a processor.tokenizer regression are not the same bug." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and the soft-edge judgments are conservative. The lone accepted pair, issue:44466 and issue:45005, is a plausible same-bug cluster around v5 tied-weights serialization, while the other pairs are reasonably rejected based on distinct titles/symptoms." + "feedback": "The assessment is grounded in the packet, and the soft-edge rejections are conservative. The 44466/45005 pair is the only one that plausibly describes the same tied-weights/save_pretrained regression, and the summary does not overstate the weaker matches. " }, "evaluator_used": true, "retried": false, @@ -5902,7 +5780,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:28282", "issue:30064", @@ -5935,6 +5813,7 @@ "issue:43906", "issue:43927", "issue:44361", + "issue:44442", "issue:44451", "issue:44464", "issue:44479", @@ -5942,18 +5821,18 @@ "issue:44743", "issue:44821", "issue:44843", - "issue:44849", "issue:44913", "issue:44936", "issue:44977", "issue:44991", "issue:45325", "issue:45356", - "issue:45357", "issue:45440", "issue:45588" ], "soft_pairs": [ + "issue:44361|issue:44442", + "issue:43653|issue:44361", "issue:30064|issue:33666", "issue:44464|issue:44743", "issue:43577|issue:43697", @@ -5963,8 +5842,8 @@ "issue:44977|issue:45356", "issue:43688|issue:45440", "issue:43531|issue:43742", - "issue:43577|issue:45588", "issue:43653|issue:43927", + "issue:43577|issue:45588", "issue:44464|issue:44977", "issue:43550|issue:43927", "issue:43550|issue:44361", @@ -5976,237 +5855,242 @@ "issue:28282|issue:35141", "issue:43504|issue:44610", "issue:43452|issue:44451", + "issue:43688|issue:43749", + "issue:43329|issue:45325", "issue:43742|issue:44821", - "issue:43525|issue:43572", - "issue:41720|issue:44464", - "issue:43756|issue:44913", "issue:33666|issue:36010", - "issue:43329|issue:45325", - "issue:43688|issue:43749", + "issue:41720|issue:44464", "issue:43262|issue:43295", + "issue:43756|issue:44913", "issue:43493|issue:43825", - "issue:44849|issue:45357" + "issue:43525|issue:43572" ], - "prepared_review_unit_hash": "fbecede648f077eed7a0ff2919c66a30462ce755ea8218dfcc69ac88c7abae68", + "prepared_review_unit_hash": "39cf0dd8dbb710d492916152516702fb311c2ed421e9b5057d46f72555346f74", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 32, - "serialized_chars": 37462, - "estimated_input_tokens": 9366, - "estimated_eval_tokens": 18988 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37169, + "estimated_input_tokens": 9293, + "estimated_eval_tokens": 18842 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous. A few items concern tokenizer/processor/model-loading regressions, but the soft pairs generally describe different concrete failures and code paths, so I would not merge any of them as duplicates.", - "confidence": 0.88, - "canonical_issue_reason": "issue:44843 is the strongest canonical issue here: it describes a concrete regression in a core loading path, has a clear root cause (`_patch_mistral_regex` calling `model_info()` unconditionally), and has broad user impact compared with the more model-specific reports.", + "summary": "This cluster is heterogeneous: it mixes unrelated bug reports across tokenizers, config loading, multimodal processors, training, generation, and model-specific regressions. The soft pairs mostly share only broad subsystem similarity, not the same concrete defect or fix path, so none should be merged as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:44843 is also the best single representative of the set because it is a well-scoped, high-impact library regression with an explicit failure mode and a clear fix target.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Different failures in different tokenizers: MLukeTokenizer task handling vs AutoTokenizer loading FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "BigBirdTokenizer special-token registration bug is unrelated to MLukeTokenizer's task AttributeError." + }, { "left": "issue:30064", "right": "issue:33666", "accept": false, - "reason": "Different subsystems and failures: image processor segmentation-map handling vs multi-GPU Qwen2-VL training." + "reason": "Image processor void segmentation maps and Qwen2-VL multi-GPU training are unrelated subsystems." }, { "left": "issue:44464", "right": "issue:44743", "accept": false, - "reason": "Both are generation correctness bugs, but they hit different code paths: compiled chunked generation vs recurrent-state reset with cache." + "reason": "Both involve generation, but one is compiled forward inconsistency while the other is recurrent-state reset in modular_qwen3_5.py." }, { "left": "issue:43577", "right": "issue:43697", "accept": false, - "reason": "Unrelated bugs: dtype propagation in BLIP-2 loading vs RTDetrV2 output divergence in v5." + "reason": "Blip2 dtype loading and RTDetrV2 output drift are different model-specific bugs." }, { "left": "issue:44843", "right": "issue:44991", "accept": false, - "reason": "Both touch tokenizer loading, but one is an offline-mode/model_info regression and the other is a model-specific tokenizer/codec loading failure; not the same bug." + "reason": "Offline model_info patching in Mistral regex is unrelated to est-roberta tokenizer loading regression." }, { "left": "issue:44479", "right": "issue:44936", "accept": false, - "reason": "Video-input regression in Qwen VL models vs trainer.evaluate() failure after training; different code paths." + "reason": "Video-input regression in Qwen VL models is unrelated to trainer.evaluate() failing after train()." }, { "left": "issue:41720", "right": "issue:43906", "accept": false, - "reason": "One is an A800 cudaErrorAssert from auto device mapping; the other is only an isolated reproduction of a different issue." + "reason": "Qwen3 auto device mapping cuda assert is not the same as an isolated reproduction of another issue." }, { "left": "issue:44977", "right": "issue:45356", "accept": false, - "reason": "Flash-attention generation failure in Qwen3.5 vs tokenizer codec/regression warning in Kimi-K2.5; not the same underlying change." + "reason": "Qwen3.5 flash-attention generation bug and Kimi-K2.5 tokenizer codec regression are different code paths." }, { "left": "issue:43688", "right": "issue:45440", "accept": false, - "reason": "Both involve MoE models, but they are different implementations and different bugs: auxiliary-loss normalization vs native DeepseekV3MoE divergence." + "reason": "Auxiliary-loss normalization in OLMoE/GPT Oss is not the same as DeepseekV3MoE divergence from remote implementation." }, { "left": "issue:43531", "right": "issue:43742", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and MobileLLM loading KeyError are unrelated." + "reason": "Qwen3-MoE sliding_window behavior and MobileLLM-125M load KeyError do not share a concrete fix path." }, { - "left": "issue:43577", - "right": "issue:45588", + "left": "issue:43653", + "right": "issue:43927", "accept": false, - "reason": "Model dtype staying float32 during loading is unrelated to flash_attention.py crashing on `s_aux=None`." + "reason": "BigBirdTokenizer decode output bug and DiaConfig token-ID loss on reload are separate persistence/tokenization issues." }, { - "left": "issue:43653", - "right": "issue:43927", + "left": "issue:43577", + "right": "issue:45588", "accept": false, - "reason": "Tokenizer special-token registration bug vs DiaConfig losing custom token IDs on save/load; different failure modes." + "reason": "Blip2 dtype retention and flash_attention crash on s_aux=None are unrelated." }, { "left": "issue:44464", "right": "issue:44977", "accept": false, - "reason": "Both are generation-related, but one is compiled chunked generation inconsistency and the other is flash-attention generation failure." + "reason": "Compiled-forward chunked generation instability is different from Qwen3.5 flash-attention generation failure." }, { "left": "issue:43550", "right": "issue:43927", "accept": false, - "reason": "torch.compile/SDPA failure in Bamba vs config token-ID loss in Dia; unrelated." + "reason": "Bamba-9B-v2 torch.compile SDPA failure is unrelated to DiaConfig custom token ID loss." }, { "left": "issue:43550", "right": "issue:44361", "accept": false, - "reason": "Different areas: Bamba compile/SDPA vs MLukeTokenizer task AttributeError." + "reason": "Bamba torch.compile/SDPA issue and MLukeTokenizer task AttributeError are unrelated." }, { "left": "issue:32090", "right": "issue:34567", "accept": false, - "reason": "Trainer GPU broadcast NoneType error and TrainerState token-count metric not updating are unrelated." + "reason": "Trainer broadcast NoneType error and num_input_tokens_seen not updating are different trainer bugs." }, { "left": "issue:43504", "right": "issue:43550", "accept": false, - "reason": "Semantic-segmentation model loading with a legacy field is unrelated to Bamba torch.compile/SDPA failures." + "reason": "Beit legacy-field loading and Bamba SDPA compilation failure are different model/load paths." }, { "left": "issue:43504", "right": "issue:43720", "accept": false, - "reason": "Legacy preset-loading bug in BEiT vs BitNet packed-weight unpacking during accelerate loading; different bugs." + "reason": "Beit pretrained preset loading and BitNet packed-weight unpacking during accelerate loading are unrelated." }, { "left": "issue:43122", "right": "issue:43906", "accept": false, - "reason": "Tokenizer behavior change across versions is not the same as an isolated reproduction of another issue." + "reason": "Tokenization changes across versions and an isolated reproduction of issue 38071 are not the same defect." }, { "left": "issue:32090", "right": "issue:33666", "accept": false, - "reason": "Trainer broadcast error vs Qwen2-VL multi-GPU training; not the same underlying bug." + "reason": "Trainer _gpu_broadcast_one NoneType failure is not the same as Qwen2-VL multi-GPU training support." }, { "left": "issue:28282", "right": "issue:35141", "accept": false, - "reason": "Missing PyTorch import error and token-embedding reinit are completely different issues." + "reason": "Missing PyTorch ImportError and embedding reinitialization after resize are unrelated." }, { "left": "issue:43504", "right": "issue:44610", "accept": false, - "reason": "BEiT legacy-field loading bug vs OmDet processor/model input-size mismatch; unrelated." + "reason": "Beit legacy-field loading bug and OmDet-Turbo processor size mismatch are different issues." }, { "left": "issue:43452", "right": "issue:44451", "accept": false, - "reason": "gguf_file loading breakage for tokenizer/CausalLM is not the same as a specific model repository failing to load." - }, - { - "left": "issue:43742", - "right": "issue:44821", - "accept": false, - "reason": "KeyError while loading a model and failure to load AutoImageProcessor from URL are different loading paths." + "reason": "GGUF from_pretrained breakage and inability to load ScandiBERT are separate tokenizer/model-loading regressions." }, { - "left": "issue:43525", - "right": "issue:43572", + "left": "issue:43688", + "right": "issue:43749", "accept": false, - "reason": "Both mention missing pad-token-related config fields, but they affect different models and different attribute names; not enough to treat as the same bug." + "reason": "Aux-loss normalization and FSDP CPU RAM efficient loading are different problems." }, { - "left": "issue:41720", - "right": "issue:44464", + "left": "issue:43329", + "right": "issue:45325", "accept": false, - "reason": "Auto device mapping cuda assert and compiled chunked-generation inconsistency are unrelated." + "reason": "Undefined helpers in multimodal token counting and still-image rope-index scaling are distinct Qwen-VL bugs." }, { - "left": "issue:43756", - "right": "issue:44913", + "left": "issue:43742", + "right": "issue:44821", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch is unrelated to GPTNeoX rotary_pct not persisting on reload." + "reason": "MobileLLM-125M load KeyError and AutoImageProcessor URL loading failure are unrelated loading paths." }, { "left": "issue:33666", "right": "issue:36010", "accept": false, - "reason": "Multi-GPU training issue vs GenerationMixin import error; no shared code path." + "reason": "Qwen2-VL multi-GPU training and GenerationMixin import failure are unrelated." }, { - "left": "issue:43329", - "right": "issue:45325", + "left": "issue:41720", + "right": "issue:44464", "accept": false, - "reason": "Undefined helpers in video token counting are unrelated to still-image temporal position scaling in Qwen2.5-VL." + "reason": "Auto device mapping cuda assert and compiled-forward generation inconsistency are different failures." }, { - "left": "issue:43688", - "right": "issue:43749", + "left": "issue:43262", + "right": "issue:43295", "accept": false, - "reason": "Aux-loss normalization bug and FSDP CPU RAM efficient loading breakage are different issues." + "reason": "Audio chat-template sampling-rate defaulting and processor.tokenizer regression are separate processor bugs." }, { - "left": "issue:43262", - "right": "issue:43295", + "left": "issue:43756", + "right": "issue:44913", "accept": false, - "reason": "Audio chat-template sampling-rate default and processor.tokenizer/image regression are separate processor bugs." + "reason": "Smollm3 RoPE-layer mismatch and GPTNeoX rotary_pct reload drift are only loosely related to RoPE and not the same bug." }, { "left": "issue:43493", "right": "issue:43825", "accept": false, - "reason": "SigLIP2 implementation discrepancy is unrelated to a pipeline error-message wording issue." + "reason": "SigLIP2 HF/JAX numerical discrepancy and pipeline translation error-message regression are unrelated." }, { - "left": "issue:44849", - "right": "issue:45357", + "left": "issue:43525", + "right": "issue:43572", "accept": false, - "reason": "Both concern Qwen3.5, but one is an output_hidden_states bug and the other is incorrect save_pretrained visual-encoder keys; different code paths." + "reason": "Both are config padding-field regressions, but they affect different models and attributes and do not look like one shared fix." } ] }, @@ -6228,7 +6112,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:28282", "issue:32090", @@ -6245,7 +6129,6 @@ "issue:43504", "issue:43540", "issue:43572", - "issue:43575", "issue:43577", "issue:43606", "issue:43646", @@ -6268,29 +6151,31 @@ "issue:44625", "issue:44734", "issue:44843", + "issue:44849", "issue:44964", "issue:44977", "issue:45083", "issue:45200", "issue:45229", + "issue:45357", "issue:45405", "issue:45520", "issue:45701", - "issue:45704", - "issue:45759" + "issue:45704" ], "soft_pairs": [ - "issue:43540|issue:43716", + "issue:44849|issue:45357", "issue:43720|issue:43927", "issue:43606|issue:43927", "issue:28282|issue:32090", - "issue:43334|issue:43572", + "issue:43540|issue:43716", "issue:28282|issue:33357", + "issue:43334|issue:43572", "issue:43577|issue:44554", - "issue:33666|issue:36331", "issue:43688|issue:45704", - "issue:43720|issue:44361", + "issue:33666|issue:36331", "issue:39692|issue:43493", + "issue:43720|issue:44361", "issue:44964|issue:45405", "issue:44479|issue:44625", "issue:43540|issue:44977", @@ -6304,182 +6189,186 @@ "issue:43824|issue:44589", "issue:39692|issue:43295", "issue:43688|issue:45701", - "issue:44589|issue:45520", "issue:33357|issue:33666", + "issue:44589|issue:45520", "issue:43504|issue:43606", - "issue:43572|issue:44291", "issue:44423|issue:44734", + "issue:43572|issue:44291", "issue:43927|issue:44361", - "issue:43504|issue:43653", - "issue:43575|issue:45405", - "issue:44625|issue:45759" + "issue:43504|issue:43653" ], - "prepared_review_unit_hash": "7af25d1b01b6496f6fbbf3f199bf459d46c130d89f6ca3c2b5f878ca0dfa0f62", + "prepared_review_unit_hash": "e744884dba59f2f4762cec9a801a17408d0824e0c548f5b46c9580e3ace3d091", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 33, - "serialized_chars": 37931, - "estimated_input_tokens": 9483, - "estimated_eval_tokens": 19222 + "soft_pair_count": 32, + "serialized_chars": 37618, + "estimated_input_tokens": 9405, + "estimated_eval_tokens": 19066 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster looks like a set of mostly false-positive semantic matches across unrelated Transformers issues; I don\u2019t see a convincing duplicate pair to merge.", - "confidence": 0.61, - "canonical_issue_reason": "Issue 43646 is the broadest representative of the recurring v5 regression/loading theme, so it is the best single anchor for this mostly heterogeneous issue cluster.", + "summary": "The set is highly heterogeneous. A few issues share a broad \u201ctransformers version/regression around loading/custom model code\u201d theme, but none of the soft pairs look like true duplicates or the same fixable code-path bug.", + "confidence": 0.27, + "canonical_issue_reason": "issue:43646 is the broadest representative of the repeated version-regression/custom-model-loading complaints; the rest are narrower model- or subsystem-specific failures.", "canonical_pr_reason": null, - "best_issue_reason": "Same as canonical: it is the most general issue here and best captures the shared high-level theme of v5 breakages in model initialization/loading.", + "best_issue_reason": "issue:43646 is the most representative single issue for the cluster\u2019s loose theme of Transformers 5 breaking model initialization/loading behavior.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43540", - "right": "issue:43716", + "left": "issue:44849", + "right": "issue:45357", "accept": false, - "reason": "Different models and failure modes: Qwen3OmniMoe video-input processing vs Mistral-3 image preprocessor/model dtype mismatch." + "reason": "Different Qwen3.5 regressions: hidden-states output vs incorrect save_pretrained visual keys." }, { "left": "issue:43720", "right": "issue:43927", "accept": false, - "reason": "BitNet packed-weight loading bug and DiaConfig save/load token-ID loss are unrelated code paths." + "reason": "Unrelated bugs: BitNet accelerate-loading unpacking vs DiaConfig losing custom token IDs." }, { "left": "issue:43606", "right": "issue:43927", "accept": false, - "reason": "CPU offload device mismatch in bark-small is unrelated to DiaConfig custom token ID persistence." + "reason": "CPU offload device mismatch is unrelated to config token-ID persistence." }, { "left": "issue:28282", "right": "issue:32090", "accept": false, - "reason": "Missing PyTorch ImportError and Trainer _gpu_broadcast_one NoneType error are unrelated." + "reason": "Different failures: missing PyTorch import vs Trainer GPU broadcast TypeError." }, { - "left": "issue:43334", - "right": "issue:43572", + "left": "issue:43540", + "right": "issue:43716", "accept": false, - "reason": "Both involve missing token-id fields, but they affect different configs/models and are not the same bug." + "reason": "Both are multimodal issues, but on different models and different code paths/errors." }, { "left": "issue:28282", "right": "issue:33357", "accept": false, - "reason": "PyTorch import failure vs MacOS bus error on community CLIP are different issues." + "reason": "Missing PyTorch import and Mac bus error are unrelated." }, { - "left": "issue:43577", - "right": "issue:44554", + "left": "issue:43334", + "right": "issue:43572", "accept": false, - "reason": "Blip2 dtype-casting bug and MPS attention correctness issue are unrelated." + "reason": "Both mention missing pad-token fields, but they affect different configs/models and do not read as the same bug." }, { - "left": "issue:33666", - "right": "issue:36331", + "left": "issue:43577", + "right": "issue:44554", "accept": false, - "reason": "Qwen2-VL multi-GPU training request vs CustomTrainer compute_loss signature break are different problems." + "reason": "BLIP2 dtype propagation and MPS attention correctness are different issues." }, { "left": "issue:43688", "right": "issue:45704", "accept": false, - "reason": "Auxiliary-loss normalization bug and T5 apex RMSNorm memory leak are unrelated." + "reason": "MoE auxiliary-loss normalization is unrelated to a T5 RMSNorm memory leak." }, { - "left": "issue:43720", - "right": "issue:44361", + "left": "issue:33666", + "right": "issue:36331", "accept": false, - "reason": "Packed-weight accelerate loading and MLukeTokenizer task failures are different code paths." + "reason": "Multi-GPU training support request and custom compute_loss keyword regression are not the same bug." }, { "left": "issue:39692", "right": "issue:43493", "accept": false, - "reason": "Both mention SigLIP2, but one is a broken docs example and the other is an implementation discrepancy; not the same concrete bug." + "reason": "Docs/example errors and SigLIP2 HF-vs-JAX implementation discrepancy are distinct." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "BitNet packed-weight loading is unrelated to MLukeTokenizer task AttributeError." }, { "left": "issue:44964", "right": "issue:45405", "accept": false, - "reason": "Phi-4 multimodal loading failure and unreleased MIN_PEFT_VERSION bump are unrelated." + "reason": "Model-loading failure and an unreleased PEFT version bump are not the same underlying problem." }, { "left": "issue:44479", "right": "issue:44625", "accept": false, - "reason": "Video-input regression for Qwen models and Qwen3.5 num_labels propagation are different issues." + "reason": "Video-input regression across Qwen models is separate from num_labels propagation in Qwen3.5 config." }, { "left": "issue:43540", "right": "issue:44977", "accept": false, - "reason": "Qwen3OmniMoe video processing bug vs Qwen3.5 flash-attention generation bug are unrelated." + "reason": "Video-input processing crash and flash-attention generation bug are different code paths." }, { "left": "issue:43493", "right": "issue:43844", "accept": false, - "reason": "SigLIP2 HF/JAX discrepancy and ZeRO-3 gradient inflation are different failure modes." + "reason": "SigLIP2 implementation discrepancy is unrelated to ZeRO-3 gradient growth." }, { "left": "issue:44488", "right": "issue:44843", "accept": false, - "reason": "sleng-bert load failure and offline model_info patch regression are separate loading issues." + "reason": "Model loading failure and offline-mode model_info() regression are different bugs." }, { "left": "issue:43756", "right": "issue:45083", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch and qwen3_omni_moe feature-length helper behavior are unrelated." + "reason": "Smollm3 RoPE-layer dropping and qwen3_omni_moe feature-length behavior are unrelated." }, { "left": "issue:45200", "right": "issue:45229", "accept": false, - "reason": "Text-only mm_token_type_ids defaulting and Gemma4 multi-GPU OOM are not the same bug." + "reason": "Text-only mm_token_type_ids defaulting and multi-GPU inference CUDA OOM are not the same issue." }, { "left": "issue:43278", "right": "issue:43844", "accept": false, - "reason": "Embedding dtype mismatch during eval and ZeRO-3 gradient growth are different issues." + "reason": "Embedding dtype drift in eval and ZeRO-3 gradient anomalies are different problems." }, { "left": "issue:43257", "right": "issue:43299", "accept": false, - "reason": "Qwen3 MoE weight conversion bug vs Qwen3VL MoE loading regression are related only by model family, not the same code-path problem." + "reason": "Both involve Qwen MoE, but one is accelerate/deepspeed weight conversion and the other is dev loading breakage." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "Custom model initialization breakage and lm_head serialization inconsistency are separate v5 regressions." + "reason": "Custom model initialization regression and tied-weight serialization differences are separate failures." }, { "left": "issue:43824", "right": "issue:44589", "accept": false, - "reason": "Missing Qwen2.5-VL export and Float8 storage lookup failure are unrelated." + "reason": "Importing a missing model class is unrelated to the Float8 storage-object lookup error." }, { "left": "issue:39692", "right": "issue:43295", "accept": false, - "reason": "Broken SigLIP2 docs example and processor/tokenizer API regression are not the same issue." + "reason": "Docs example mistakes are not the same as the processor.tokenizer / image-to-tokenizer regression." }, { "left": "issue:43688", @@ -6487,77 +6376,47 @@ "accept": false, "reason": "Aux-loss normalization and tokenization changes are unrelated." }, - { - "left": "issue:44589", - "right": "issue:45520", - "accept": false, - "reason": "Float8 storage type lookup and flash_attn import key error are unrelated runtime failures." - }, { "left": "issue:33357", "right": "issue:33666", "accept": false, - "reason": "MacOS CLIP bus error and Qwen2-VL multi-GPU training are unrelated." + "reason": "Mac bus error on a pretrained CLIP model is unrelated to Qwen2-VL multi-GPU training." }, { - "left": "issue:43504", - "right": "issue:43606", + "left": "issue:44589", + "right": "issue:45520", "accept": false, - "reason": "Beit legacy-field loading bug and bark-small CPU offload mismatch are different code paths." + "reason": "Float8 storage lookup and flash_attn Python 3.13 import key error are different issues." }, { - "left": "issue:43572", - "right": "issue:44291", + "left": "issue:43504", + "right": "issue:43606", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig and init_empty_weights unexpected argument are different v5 regressions." + "reason": "Legacy field model-load failure and Bark CPU offload device mismatch are unrelated." }, { "left": "issue:44423", "right": "issue:44734", "accept": false, - "reason": "Both are serve bugs, but continuous-batching multimodal crash and KV-cache continuation indexing are distinct problems." - }, - { - "left": "issue:43927", - "right": "issue:44361", - "accept": false, - "reason": "DiaConfig save/load token-ID loss and MLukeTokenizer AttributeError are unrelated." - }, - { - "left": "issue:43504", - "right": "issue:43653", - "accept": false, - "reason": "Legacy field load failure in Beit and BigBird special-token registration bug are different issues." + "reason": "Continuous-batching multimodal crash and KV-cache continuation indexing bug are different server paths." }, { - "left": "issue:43575", - "right": "issue:45405", + "left": "issue:43572", + "right": "issue:44291", "accept": false, - "reason": "Qwen2-57B tensor-parallel OOM and MIN_PEFT_VERSION bump are unrelated." + "reason": "Missing pad_token_idx in StableLmConfig and unexpected _is_hf_initialized in init_empty_weights are unrelated." }, { - "left": "issue:44625", - "right": "issue:45759", + "left": "issue:43927", + "right": "issue:44361", "accept": false, - "reason": "num_labels propagation in Qwen3.5 and AutoModelForCausalLM text_config unwrapping are adjacent but distinct config-handling bugs." + "reason": "Token-ID persistence and tokenizer AttributeError on tasks are different bugs." }, { "left": "issue:43504", "right": "issue:43653", "accept": false, - "reason": "Beit legacy-field loading bug and BigBird special-token registration bug are different issues." - }, - { - "left": "issue:43575", - "right": "issue:45405", - "accept": false, - "reason": "Qwen2-57B tensor-parallel OOM and MIN_PEFT_VERSION bump are unrelated." - }, - { - "left": "issue:44625", - "right": "issue:45759", - "accept": false, - "reason": "num_labels propagation in Qwen3.5 and composite-config unwrapping in AutoModelForCausalLM are related only at a very high level." + "reason": "Legacy-field model loading and BigBirdTokenizer special-token registration are unrelated." } ] }, @@ -6579,12 +6438,11 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:15354", "issue:29942", "issue:30333", - "issue:37428", "issue:38175", "issue:39401", "issue:43232", @@ -6594,12 +6452,12 @@ "issue:43493", "issue:43525", "issue:43540", + "issue:43575", "issue:43606", "issue:43688", "issue:43701", "issue:43844", "issue:43873", - "issue:43957", "issue:43994", "issue:44112", "issue:44315", @@ -6610,6 +6468,7 @@ "issue:44554", "issue:44568", "issue:44610", + "issue:44625", "issue:44792", "issue:44849", "issue:44898", @@ -6627,9 +6486,12 @@ "issue:45446", "issue:45520", "issue:45584", - "issue:45701" + "issue:45701", + "issue:45759" ], "soft_pairs": [ + "issue:43575|issue:45405", + "issue:44625|issue:45759", "issue:44792|issue:45405", "issue:43606|issue:44112", "issue:44521|issue:45290", @@ -6637,11 +6499,11 @@ "issue:44464|issue:45584", "issue:29942|issue:30333", "issue:43844|issue:44964", - "issue:45084|issue:45520", - "issue:44554|issue:45446", "issue:43688|issue:43873", "issue:15354|issue:38175", + "issue:45084|issue:45520", "issue:44568|issue:45701", + "issue:44554|issue:45446", "issue:43540|issue:45200", "issue:45072|issue:45198", "issue:43232|issue:44849", @@ -6653,152 +6515,163 @@ "issue:45200|issue:45325", "issue:43334|issue:43525", "issue:44466|issue:45127", - "issue:44898|issue:45412", "issue:43493|issue:43701", - "issue:37428|issue:43957" + "issue:44898|issue:45412" ], - "prepared_review_unit_hash": "2d8df1576462d209d0945ae55b90ae3811c93716a546523d60f6d063d5d5606f", + "prepared_review_unit_hash": "cecfff152a77325070ccffe5a2e3723f46000170f8ce191b6577d484d6eb78fd", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 26, - "serialized_chars": 35087, - "estimated_input_tokens": 8772, - "estimated_eval_tokens": 17800 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35906, + "estimated_input_tokens": 8977, + "estimated_eval_tokens": 18210 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: most items are unrelated model-specific bugs, regressions, or CI failures that only share broad Transformers subsystems. None of the soft pairs is convincing enough to merge, and there is no single canonical issue for the whole set.", - "confidence": 0.9, - "canonical_issue_reason": null, + "summary": "This set is mostly heterogeneous model- and subsystem-specific bugs, not a tight duplicate cluster. The soft-similarity pairs are generally false positives; no PRs are present.", + "confidence": 0.93, + "canonical_issue_reason": "No single issue clearly subsumes the others; if forced to pick an umbrella-style representative, issue 45701 is the broadest, but it is still too vague to be a true canonical duplicate target.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Issue 45701 is the most general and reusable representative among the set, with the broadest scope around tokenizer/version behavior, though it does not actually match the other items closely.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "Different problems: TP OOM on model load vs unreleased PEFT version requirement." + }, + { + "left": "issue:44625", + "right": "issue:45759", + "accept": false, + "reason": "Both involve Qwen config plumbing, but one is num_labels propagation and the other is unwrapping text_config from composite configs; not the same bug." + }, { "left": "issue:44792", "right": "issue:45405", "accept": false, - "reason": "Janus image generation test failure and an unreleased PEFT version bump are unrelated." + "reason": "Janus image-generation test failure is unrelated to a PEFT version bump." }, { "left": "issue:43606", "right": "issue:44112", "accept": false, - "reason": "Both mention device/offload/CI, but they are different models and different failure modes." + "reason": "Different models and failures: CPU offload device mismatch vs a stale CI test override." }, { "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "Both involve apply_chat_template, but one is zero assistant masks for multimodal inputs and the other is a crash on tool-call messages." + "reason": "Both hit apply_chat_template, but one is zero assistant masks for multimodal inputs and the other is a crash on tool-call assistant messages." }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "Different symptoms, different models, and no shared concrete code-path bug." + "reason": "Different symptoms and code paths: bad SigLIP2 outputs vs unexpected position_id keys." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Compiled-forward chunked generation and Whisper empty-transcription handling are unrelated issues." + "reason": "Chunked generation with compiled forward is unrelated to Whisper empty-transcription generation after align_special_tokens." }, { "left": "issue:29942", "right": "issue:30333", "accept": false, - "reason": "Flash Attention 2 CI failures and MLflow reporting on failed training jobs are unrelated." + "reason": "Flash Attention CI failures and MLflow job-state reporting are unrelated." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "ZeRO-3 gradient growth and Phi-4 multimodal loading are different problems." - }, - { - "left": "issue:45084", - "right": "issue:45520", - "accept": false, - "reason": "Template compilation error and a Python 3.13 flash_attn import KeyError are not the same bug." - }, - { - "left": "issue:44554", - "right": "issue:45446", - "accept": false, - "reason": "MPS attention correctness and a flex_attention version check are separate code paths." + "reason": "DeepSpeed ZeRO-3 gradient inflation is unrelated to loading Phi-4 multimodal models." }, { "left": "issue:43688", "right": "issue:43873", "accept": false, - "reason": "Aux-loss normalization and quantization/offloading are unrelated." + "reason": "Aux-loss normalization in MoE models is unrelated to quantized offloading behavior." }, { "left": "issue:15354", "right": "issue:38175", "accept": false, - "reason": "TorchScript generator-expression export failure and SigLIP2 zero probabilities are unrelated." + "reason": "TorchScript GeneratorExp export support and SigLIP2 zero-probability outputs are different issues." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Non-template-node compile error and flash_attn import KeyError are unrelated." }, { "left": "issue:44568", "right": "issue:45701", "accept": false, - "reason": "Tokenizer special-token omission and a broad tokenization regression are too generic to be the same concrete bug." + "reason": "Specific BOS/EOS token regression for one tokenizer is not the same as a broad tokenization-version-change report." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "MPS attention correctness and a PyTorch version check for AuxRequest import are different bugs." }, { "left": "issue:43540", "right": "issue:45200", "accept": false, - "reason": "Video input processing in Qwen3OmniMoe and Gemma 4 mm_token_type_ids handling are unrelated." + "reason": "Video-input processing failure in Qwen3OmniMoe is unrelated to Gemma 4 token-type defaults." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "BF16 dtype mismatch in inference and Wav2Vec2 save/tokenization failure do not share a concrete fix." + "reason": "BFloat16 dtype mismatches in inference are unrelated to Wav2Vec2 save/tokenization failures." }, { "left": "issue:43232", "right": "issue:44849", "accept": false, - "reason": "Different generation-related bugs affecting different model behaviors and inputs." + "reason": "Generation kwargs after sync_gpus and Qwen3.5 hidden-state output bugs are unrelated." }, { "left": "issue:39401", "right": "issue:44568", "accept": false, - "reason": "Offset mapping errors and missing BOS/EOS insertion are distinct tokenizer bugs." + "reason": "Wrong offset_mapping for Qwen3 tokenizer is not the same as missing BOS/EOS insertion for mDeBERTa." }, { "left": "issue:43278", "right": "issue:43701", "accept": false, - "reason": "Evaluation dtype drift and resume_from_checkpoint key mismatch are unrelated." + "reason": "BF16-to-FP32 dtype change during eval is unrelated to resume_from_checkpoint key mismatch." }, { "left": "issue:44315", "right": "issue:44849", "accept": false, - "reason": "Liger kernel application with model_init is unrelated to the Qwen3.5 hidden-states bug." + "reason": "Liger Kernel not applied with model_init is unrelated to Qwen3.5 output_hidden_states behavior." }, { "left": "issue:43295", "right": "issue:44610", "accept": false, - "reason": "Processor.tokenizer regression and OmDet-Turbo input-size mismatch are different processor/model issues." + "reason": "Processor/tokenizer access regression is different from OmDet-Turbo\u2019s image-size mismatch." }, { "left": "issue:44554", @@ -6810,37 +6683,31 @@ "left": "issue:45200", "right": "issue:45325", "accept": false, - "reason": "Gemma 4 text-only mm_token_type_ids and Qwen2.5-VL rope index scaling are different multimodal bugs." + "reason": "Gemma 4 mm_token_type_ids defaults and Qwen2.5-VL rope-index scaling are different multimodal bugs." }, { "left": "issue:43334", "right": "issue:43525", "accept": false, - "reason": "Both mention missing pad_token_id, but they are separate model families and separate loading failures." + "reason": "Both are pad_token_id AttributeErrors, but they affect different model/config classes and likely need separate fixes." }, { "left": "issue:44466", "right": "issue:45127", "accept": false, - "reason": "Serialization of tied lm_head weights and LoRA merge collapse are not the same underlying issue." - }, - { - "left": "issue:44898", - "right": "issue:45412", - "accept": false, - "reason": "Perceiver interpolation failure and RT-DETR memory release behavior are unrelated." + "reason": "Inconsistent serialization of lm_head.weight is unrelated to LoRA merge collapse with extended vocabulary." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "SigLIP2 implementation discrepancy and checkpoint key mismatch are separate problems." + "reason": "SigLIP2 JAX-vs-HF discrepancy is unrelated to checkpoint resume key mismatch." }, { - "left": "issue:37428", - "right": "issue:43957", + "left": "issue:44898", + "right": "issue:45412", "accept": false, - "reason": "Flash-attention import failure and meta-device model loading breakage are distinct regressions." + "reason": "Perceiver resolution/interpolation failure is unrelated to RT-DETR memory not being released." } ] }, @@ -6862,7 +6729,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:37428", "issue:38175", @@ -6913,20 +6780,21 @@ "issue:45593" ], "soft_pairs": [ + "issue:37428|issue:43957", "issue:43572|issue:44987", - "issue:43572|issue:43957", "issue:43873|issue:44464", "issue:43295|issue:43450", + "issue:43572|issue:43957", "issue:43824|issue:43975", "issue:43844|issue:44936", "issue:43295|issue:43873", "issue:44514|issue:44521", - "issue:38175|issue:45412", "issue:44442|issue:45198", - "issue:37428|issue:43824", - "issue:43295|issue:43994", + "issue:38175|issue:45412", "issue:43278|issue:43856", + "issue:43295|issue:43994", "issue:38175|issue:43493", + "issue:37428|issue:43824", "issue:42915|issue:43824", "issue:43827|issue:43901", "issue:43653|issue:44112", @@ -6939,233 +6807,239 @@ "issue:43262|issue:43450", "issue:44206|issue:45584", "issue:43819|issue:44112", - "issue:37428|issue:43425", "issue:44610|issue:44857", + "issue:37428|issue:43425", "issue:44857|issue:45593", "issue:43278|issue:43825", - "issue:45290|issue:45381", - "issue:43606|issue:45083" + "issue:43606|issue:45083", + "issue:45290|issue:45381" ], - "prepared_review_unit_hash": "26d06c417393c01188287a9fd855482d04e70e28ab9f0701883a4638778e179b", + "prepared_review_unit_hash": "6552d82c24b2c529d057f7b4575a0bf446e000f4b50bf0a80c7ded158a6f2535", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 32, - "serialized_chars": 36643, - "estimated_input_tokens": 9161, - "estimated_eval_tokens": 18578 + "soft_pair_count": 33, + "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 set looks like a grab bag of unrelated regression reports and model-specific bugs. The soft pairs are mostly topic-adjacent but not duplicate-level, so I would not merge any of them into a single canonical thread.", - "confidence": 0.74, - "canonical_issue_reason": "issue:43493 is the most root-cause-like report in the set: it frames the SigLIP2 problem as HF-vs-JAX implementation parity, which is broader and more canonical than narrower symptom reports like zero probabilities.", + "summary": "The set is mostly heterogeneous and the soft edges are too broad to treat as true duplicates. There is a small chat-template/multimodal subgroup, but even there the paired issues describe different failure modes, so I would not merge any of the suggested edges.", + "confidence": 0.27, + "canonical_issue_reason": "No issue cleanly represents the whole set. If forced to pick one representative, issue 44521 is the broadest open, user-facing report in the only clearly related subgroup (multimodal `apply_chat_template` behavior), but it is not a duplicate of the others.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43493 is the best representative issue overall because it is the broadest, most general description of a concrete bug class among the provided issues.", + "best_issue_reason": "Issue 44521 is the best single representative among these because it has the broadest framing and strongest overlap with the chat-template/multimodal subset, though the overall cluster is too mixed for a strong canonical choice.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43572", - "right": "issue:44987", + "left": "issue:37428", + "right": "issue:43957", "accept": false, - "reason": "Both are v5-era load/config regressions, but they point to different failures and model paths; too broad to treat as the same bug." + "reason": "Different failures: an ImportError for `_flash_supports_window_size` versus meta-device model loading regressions. Same library, not the same bug." }, { "left": "issue:43572", - "right": "issue:43957", + "right": "issue:44987", "accept": false, - "reason": "Shared theme is model loading after v5 changes, but one is a missing config field and the other is meta-device loading; different code-path failures." + "reason": "Different model/config breakages with no clear shared code-path; one is `StableLmConfig.pad_token_idx`, the other is a load failure for a specific model on v5.1+." }, { "left": "issue:43873", "right": "issue:44464", "accept": false, - "reason": "Quantized offloading and compiled chunked generation are unrelated behaviors; no sign of one underlying defect." + "reason": "Quantization/offload behavior and compiled-forward chunked generation are unrelated symptoms and code paths." }, { "left": "issue:43295", "right": "issue:43450", "accept": false, - "reason": "Both involve processors, but one is a tokenizer/image API regression and the other is batched video shape handling; not the same bug." + "reason": "Processor/tokenizer regression is not the same bug as batched video shape handling." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Different regressions: missing config field versus meta-device loading failures; no evidence of one underlying issue." }, { "left": "issue:43824", "right": "issue:43975", "accept": false, - "reason": "ImportError for a missing exported class and incorrect detokenization are distinct problems." + "reason": "An import/export failure for Qwen2.5-VL is not the same as DeepSeek detokenization corruption." }, { "left": "issue:43844", "right": "issue:44936", "accept": false, - "reason": "Training instability with ZeRO-3 is not the same as trainer.evaluate() failing after train()." + "reason": "Training-gradient instability with ZeRO-3 is a different problem from `trainer.evaluate()` failing after training." }, { "left": "issue:43295", "right": "issue:43873", "accept": false, - "reason": "Processor API regression and quantization/offload behavior are unrelated." + "reason": "Processor API regression and offloading/quantization behavior are separate subsystems and bugs." }, { "left": "issue:44514", "right": "issue:44521", "accept": false, - "reason": "Both are about apply_chat_template, but one is a batching crash and the other is all-zero assistant masks; similar surface area, not clearly the same defect." + "reason": "Both involve `apply_chat_template`, but one is a batched-input crash with `padding=False` and the other is all-zero assistant masks; related area, different concrete bug." }, { - "left": "issue:38175", - "right": "issue:45412", + "left": "issue:44442", + "right": "issue:45198", "accept": false, - "reason": "These are unrelated model-specific issues with different subsystems and symptoms." + "reason": "Different tokenizer/model families and different failure modes (`AutoTokenizer` load vs save/tokenization failure)." }, { - "left": "issue:44442", - "right": "issue:45198", + "left": "issue:38175", + "right": "issue:45412", "accept": false, - "reason": "Different speech/tokenizer code paths and different failure modes; too weak to merge." + "reason": "SigLIP2 zero probabilities and RT-DETR memory not released are unrelated." }, { - "left": "issue:37428", - "right": "issue:43824", + "left": "issue:43278", + "right": "issue:43856", "accept": false, - "reason": "Both are ImportErrors, but for different missing symbols in different modules." + "reason": "Embedding dtype mismatch during eval is not the same as Qwen3 MoE memory usage." }, { "left": "issue:43295", "right": "issue:43994", "accept": false, - "reason": "Both mention processor/model usage, but one is a custom processor/tokenizer regression and the other is a SigLIP2 output-quality issue." + "reason": "Custom processor/tokenizer regression and SigLIP2 output quality issues are different bugs." }, { - "left": "issue:43278", - "right": "issue:43856", + "left": "issue:38175", + "right": "issue:43493", "accept": false, - "reason": "Embedding dtype drift and MoE training memory usage are distinct training issues." + "reason": "Both mention SigLIP2, but one is zero probabilities in a model run and the other is a discrepancy versus JAX implementation; not clearly the same concrete defect." }, { - "left": "issue:38175", - "right": "issue:43493", + "left": "issue:37428", + "right": "issue:43824", "accept": false, - "reason": "The titles are related by model family, but one is a specific zero-probability symptom and the other is a broader HF-vs-JAX parity complaint; not enough to call them the same bug." + "reason": "Import errors for different symbols and models; no shared underlying bug." }, { "left": "issue:42915", "right": "issue:43824", "accept": false, - "reason": "FP8 configuration failure in Qwen3MoE and a missing export ImportError are unrelated." + "reason": "FP8 config failure for Qwen3 MoE versus missing Qwen2.5-VL export are unrelated." }, { "left": "issue:43827", "right": "issue:43901", "accept": false, - "reason": "Both are documentation complaints, but they concern different APIs and different deprecations; too coarse to merge." + "reason": "Both are docs/API transition issues around v5, but they concern different pipeline behaviors and should not be merged." }, { "left": "issue:43653", "right": "issue:44112", "accept": false, - "reason": "Tokenizer special-token registration and a GraniteSpeech CI device override test are unrelated." + "reason": "Tokenizer special-token registration and GraniteSpeech device override tests are different issues." }, { "left": "issue:43450", "right": "issue:43493", "accept": false, - "reason": "Video batch-shape handling and SigLIP2 parity/output discrepancy are different bugs in different subsystems." + "reason": "Video batched-shape bug and SigLIP2 implementation discrepancy do not share the same failure mode." }, { "left": "issue:43450", "right": "issue:43575", "accept": false, - "reason": "Video processor shape handling and tensor-parallel OOM are not the same issue." + "reason": "Batched video shape handling is unrelated to Qwen2-57B tensor-parallel OOM." }, { "left": "issue:44360", "right": "issue:44561", "accept": false, - "reason": "A ReLU omission in an indexer and removal of a v5 helper function are unrelated." + "reason": "DSA indexer ReLU discussion and removal of `is_torch_fx_available` breakage are not the same bug." }, { "left": "issue:44792", "right": "issue:45563", "accept": false, - "reason": "A failing Janus image-generation test and a stale warning in paged generate are different generation issues." + "reason": "Janus image-generation test failure is unrelated to paged `generate()` warning behavior." }, { "left": "issue:43819", "right": "issue:45083", "accept": false, - "reason": "DAC latent conversion mismatch and qwen3_omni_moe feature-length helper behavior are distinct code paths." + "reason": "DAC latent reconstruction mismatch and Qwen3 Omni MoE feature-length helper behavior are different code paths." }, { "left": "issue:45127", "right": "issue:45245", "accept": false, - "reason": "LoRA merge collapse and a category-cardinality runtime error are unrelated defects." + "reason": "LoRA merge/collapsed model behavior is unrelated to category-count limits." }, { "left": "issue:43262", "right": "issue:43450", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting and video batch shape handling are unrelated processor bugs." + "reason": "Audio chat-template sample-rate defaulting and video batched shape handling are different processor bugs." }, { "left": "issue:44206", "right": "issue:45584", "accept": false, - "reason": "Unsupported feature-extractor args and Whisper empty-transcription failure are different audio issues." + "reason": "Feature-extractor argument regression and Whisper empty-transcription failure are unrelated." }, { "left": "issue:43819", "right": "issue:44112", "accept": false, - "reason": "DAC latent mismatch and GraniteSpeech test failure do not share a concrete bug." + "reason": "DAC forward/latents mismatch and GraniteSpeech CI device override are separate issues." }, { - "left": "issue:37428", - "right": "issue:43425", + "left": "issue:44610", + "right": "issue:44857", "accept": false, - "reason": "ImportError for a flash-attention helper and torch 2.10 incompatibility are separate compatibility problems." + "reason": "OmDet-Turbo preprocessing size mismatch and LwDetrImageLoss AMP crash are different bugs." }, { - "left": "issue:44610", - "right": "issue:44857", + "left": "issue:37428", + "right": "issue:43425", "accept": false, - "reason": "Processor image-size mismatch and a float16 AMP loss crash are different model bugs." + "reason": "ImportError for a flash-attention helper and Torch 2.10 incompatibility are not the same defect." }, { "left": "issue:44857", "right": "issue:45593", "accept": false, - "reason": "AMP crash in LwDetrImageLoss and missing auxiliary losses in D-FINE are unrelated training behaviors." + "reason": "AMP/CUDA loss crash versus missing auxiliary losses when denoising is off are unrelated." }, { "left": "issue:43278", "right": "issue:43825", "accept": false, - "reason": "Evaluation dtype drift and a pipeline deprecation/error-message issue are not the same bug." + "reason": "Embedding dtype change and pipeline translation error-message regression are different problems." }, { - "left": "issue:45290", - "right": "issue:45381", + "left": "issue:43606", + "right": "issue:45083", "accept": false, - "reason": "Chat-template tool-call handling and Qwen2.5-VL video position IDs are different multimodal regressions." + "reason": "CPU-offload device mismatch in Bark and a Qwen3 Omni MoE helper behavior issue are unrelated." }, { - "left": "issue:43606", - "right": "issue:45083", + "left": "issue:45290", + "right": "issue:45381", "accept": false, - "reason": "CPU offload device mismatch and qwen3_omni_moe feature-length behavior are unrelated." + "reason": "Assistant-message tool-call crash in `apply_chat_template` and Qwen2.5-VL video `vision_position_ids` mismatch are different bugs." } ] }, @@ -7187,7 +7061,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:29127", "issue:29942", @@ -7229,7 +7103,6 @@ "issue:45200", "issue:45229", "issue:45310", - "issue:45372", "issue:45459", "issue:45478", "issue:45491", @@ -7242,290 +7115,283 @@ "issue:44448|issue:44743", "issue:45200|issue:45588", "issue:43335|issue:43550", - "issue:37428|issue:44351", "issue:44462|issue:44561", "issue:44561|issue:44779", + "issue:37428|issue:44351", "issue:43611|issue:43931", "issue:43388|issue:43526", "issue:39692|issue:43526", "issue:43819|issue:44493", "issue:43335|issue:43653", - "issue:44855|issue:45083", "issue:29942|issue:32090", + "issue:44855|issue:45083", "issue:45310|issue:45478", "issue:43404|issue:45200", "issue:43550|issue:44112", - "issue:43335|issue:45083", "issue:39692|issue:44493", + "issue:43335|issue:45083", "issue:43335|issue:43504", - "issue:44292|issue:45520", "issue:44448|issue:44779", + "issue:44292|issue:45520", "issue:38175|issue:44493", "issue:43335|issue:45072", "issue:44977|issue:45229", - "issue:43720|issue:44112", "issue:43335|issue:45593", + "issue:43720|issue:44112", "issue:44448|issue:44977", "issue:45459|issue:45712", "issue:44568|issue:45491", "issue:29127|issue:32090", "issue:43606|issue:44857", "issue:40444|issue:43526", - "issue:44857|issue:45072", "issue:34689|issue:35141", - "issue:45372|issue:45588", + "issue:44857|issue:45072", "issue:43701|issue:45083" ], - "prepared_review_unit_hash": "edc94b1ef60fc758b9032e67af082c83b12b04f539e9927e038fcdb519848939", + "prepared_review_unit_hash": "9aa6aac0bb1fa2d3b540fb4296da362c82c0f006b36ee77beb66468d23891552", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 36, - "serialized_chars": 38838, - "estimated_input_tokens": 9710, - "estimated_eval_tokens": 19676 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 35, + "serialized_chars": 37901, + "estimated_input_tokens": 9476, + "estimated_eval_tokens": 19208 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a set of unrelated issue candidates that only look superficially similar. I found one clear duplicate pair: the two Qwen3.5 MoE from_pretrained reports. Everything else should stay separate.", - "confidence": 0.95, - "canonical_issue_reason": "issue:45310 is the better canonical issue because it is the earlier, clearer report for the Qwen3.5 MoE from_pretrained failure; issue:45478 is the same bug with a shorter follow-up report.", + "summary": "Mostly unrelated issue set; the only clear duplicate is the Qwen3.5 Moe from_pretrained regression, which appears as an exact title match in issues 45310 and 45478.", + "confidence": 0.94, + "canonical_issue_reason": "Issue 45310 is the earlier, better-supported report of the exact Qwen3.5 Moe from_pretrained regression and has an inbound reference, making it the best canonical issue.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the strongest issue to keep as representative because it has the same exact bug title as issue:45478 and is the most complete instance of that duplicate pair.", + "best_issue_reason": "Issue 45310 is the strongest representative issue here because it is the original of the only exact duplicate pair and is phrased as a concrete, reproducible bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44448", "right": "issue:44743", "accept": false, - "reason": "Different models and failure modes: Pegasus output drift vs Qwen3.5 recurrent-state reset." + "reason": "Different models and failure modes: Pegasus output regression vs Qwen3.5 recurrent-state reset." }, { "left": "issue:45200", "right": "issue:45588", "accept": false, - "reason": "Unrelated bugs: Gemma 4 token-type defaults vs flash-attention sink-less model crash." + "reason": "Different code paths: Gemma4 token-type handling vs flash_attention integration crash on s_aux=None." }, { "left": "issue:43335", "right": "issue:43550", "accept": false, - "reason": "Different subsystems and symptoms: SwitchTransformers config bug vs Bamba SDPA/torch.compile failure." - }, - { - "left": "issue:37428", - "right": "issue:44351", - "accept": false, - "reason": "Both are import errors, but for different missing symbols and different code paths." + "reason": "Different subsystems and bugs: SwitchTransformers config layer count vs Bamba torch.compile/SDPA failure." }, { "left": "issue:44462", "right": "issue:44561", "accept": false, - "reason": "Tokenizer loading vs removal of a utility function; not the same underlying bug." + "reason": "Unrelated regressions: tokenizer.json loading vs removal of is_torch_fx_available affecting trust_remote_code." }, { "left": "issue:44561", "right": "issue:44779", "accept": false, - "reason": "Both relate to v5 breakage, but one is a missing helper API and the other is incorrect DeepSeek tokenizer behavior." + "reason": "Both are v5 regressions, but one is a remote-code API removal issue and the other is a Deepseek tokenizer correctness bug." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and likely different fixes." }, { "left": "issue:43611", "right": "issue:43931", "accept": false, - "reason": "Both are loading issues, but one is a generic base_model_prefix regression and the other is a specific Qwen3-VL weight-shape mismatch." + "reason": "Different loading failures for different models: base_model_prefix handling vs Qwen3-VL weight-shape mismatch." }, { "left": "issue:43388", "right": "issue:43526", "accept": false, - "reason": "Different code paths: metric gathering bug vs Beit label reduction bug." + "reason": "Both mention labels, but one is metric gathering and the other is Beit image-processor label reduction." }, { "left": "issue:39692", "right": "issue:43526", "accept": false, - "reason": "Docs/example problems are not the same as the BeitImageProcessorFast label handling bug." + "reason": "Documentation/example failures are not the same bug as BeitImageProcessorFast label reduction." }, { "left": "issue:43819", "right": "issue:44493", "accept": false, - "reason": "DAC latent decoding mismatch is unrelated to the position-id key issue." + "reason": "Different symptoms and components: DAC latent reconstruction vs unexpected position-id key handling." }, { "left": "issue:43335", "right": "issue:43653", "accept": false, - "reason": "Different components and failure modes: config-layer construction vs tokenizer special-token registration." + "reason": "Different models and bugs: SwitchTransformers sparse-layer creation vs BigBirdTokenizer special-token registration." }, { - "left": "issue:44855", - "right": "issue:45083", + "left": "issue:29942", + "right": "issue:32090", "accept": false, - "reason": "Python 3.13 import/parsing error is unrelated to the qwen3_omni_moe helper behavior." + "reason": "Flash Attention test failures and Trainer _gpu_broadcast_one NoneType errors are unrelated." }, { - "left": "issue:29942", - "right": "issue:32090", + "left": "issue:44855", + "right": "issue:45083", "accept": false, - "reason": "FlashAttention test failures and NoneType GPU broadcast errors are distinct trainer/runtime issues." + "reason": "Python parsing/import issue in DebertaV2 vs qwen3_omni_moe feature-length helper behavior." }, { "left": "issue:45310", "right": "issue:45478", "accept": true, - "reason": "Exact duplicate report: same title, same Qwen3.5 MoE from_pretrained regression." + "reason": "Exact same title and same Qwen3.5 Moe from_pretrained error report; clearly the same underlying bug." }, { "left": "issue:43404", "right": "issue:45200", "accept": false, - "reason": "Both affect multimodal generation, but the bugs are different: weight tying vs missing mm_token_type_ids defaults." + "reason": "Different generation/training bugs: lm_head weight tying vs mm_token_type_ids defaulting." }, { "left": "issue:43550", "right": "issue:44112", "accept": false, - "reason": "SDPA compilation failure in Bamba is unrelated to a GraniteSpeech CI device-override test." - }, - { - "left": "issue:43335", - "right": "issue:45083", - "accept": false, - "reason": "Different failures in different models; no shared concrete bug." + "reason": "Different failures in different models: torch.compile SDPA issue vs stale CI device-override test." }, { "left": "issue:39692", "right": "issue:44493", "accept": false, - "reason": "Documentation/example mismatch is unrelated to unexpected position-id keys." + "reason": "SigLIP2 docs/example problems are unrelated to position-id key regressions." }, { "left": "issue:43335", - "right": "issue:43504", + "right": "issue:45083", "accept": false, - "reason": "SwitchTransformers sparse-layer construction bug is not the same as BEiT legacy-field loading failure." + "reason": "Different model families and code paths; no shared concrete bug." }, { - "left": "issue:44292", - "right": "issue:45520", + "left": "issue:43335", + "right": "issue:43504", "accept": false, - "reason": "Different runtime problems: Qwen-3 NVFP4 execution vs Python 3.13 flash_attn import KeyError." + "reason": "SwitchTransformers config bug and Beit legacy-field loading bug are unrelated." }, { "left": "issue:44448", "right": "issue:44779", "accept": false, - "reason": "Different models and regression types; only a broad v4/v5 theme overlaps." + "reason": "Both are transformer v5 regressions, but they affect different models and causes." + }, + { + "left": "issue:44292", + "right": "issue:45520", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime error is not the same as a Python 3.13 flash_attn import KeyError." }, { "left": "issue:38175", "right": "issue:44493", "accept": false, - "reason": "Zero-probabilities issue in SigLIP2 is not the same as position-id key warnings." + "reason": "Different issues: zero probabilities in SigLIP2 vs unexpected position-id keys." }, { "left": "issue:43335", "right": "issue:45072", "accept": false, - "reason": "Different bugs in different models; bfloat16 dtype mismatch is not the same as SwitchTransformers config behavior." + "reason": "SwitchTransformers config bug is unrelated to bfloat16 dtype mismatches in inference." }, { "left": "issue:44977", "right": "issue:45229", "accept": false, - "reason": "Flash-attention generation bug and CUDA OOM are distinct failure modes." + "reason": "Different models and failures: flash-attention generation bug vs CUDA OOM on multi-GPU inference." }, { - "left": "issue:43720", - "right": "issue:44112", + "left": "issue:43335", + "right": "issue:45593", "accept": false, - "reason": "Packed-weight loading bug is unrelated to the GraniteSpeech stale device override test." + "reason": "Sparse-layer config bug and missing auxiliary losses are unrelated." }, { - "left": "issue:43335", - "right": "issue:45593", + "left": "issue:43720", + "right": "issue:44112", "accept": false, - "reason": "Sparse-layer config bug is unrelated to missing auxiliary losses in D-FINE." + "reason": "Packed-weight loading failure is unrelated to a stale CI device-override test." }, { "left": "issue:44448", "right": "issue:44977", "accept": false, - "reason": "Both mention output changes, but they concern different models and different causes." + "reason": "Different models and symptoms; one is output drift, the other is flash-attention generation failure." }, { "left": "issue:45459", "right": "issue:45712", "accept": false, - "reason": "Tokenizer error masking is unrelated to dummy class cleanup/check_repo failures." + "reason": "Tokenizer error masking and leftover dummy classes are separate maintenance issues." }, { "left": "issue:44568", "right": "issue:45491", "accept": false, - "reason": "Tokenizer special-token addition bug is unrelated to Gemma3 NaN embeddings." + "reason": "Tokenizer special-token addition bug and Gemma3 NaN embeddings are unrelated." }, { "left": "issue:29127", "right": "issue:32090", "accept": false, - "reason": "LayoutLMv3 error messaging and trainer GPU broadcast failures are unrelated." + "reason": "Error-message clarity in layoutlmv3 is not the same as a Trainer broadcast TypeError." }, { "left": "issue:43606", "right": "issue:44857", "accept": false, - "reason": "CPU offload device mismatch is unrelated to AMP/CUDA loss crashes." + "reason": "Device-mismatch CPU offload bug and float16 AMP crash are different failures." }, { "left": "issue:40444", "right": "issue:43526", "accept": false, - "reason": "Multi-image iterable finetuning failure is unrelated to Beit image processor label reduction." - }, - { - "left": "issue:44857", - "right": "issue:45072", - "accept": false, - "reason": "AMP loss crash and bfloat16 dtype mismatch are different issues." + "reason": "Multi-image iterable-dataset finetuning bug is unrelated to Beit label reduction." }, { "left": "issue:34689", "right": "issue:35141", "accept": false, - "reason": "Both are model-loading regressions, but for different models and different causes." + "reason": "Model-loading regression for Llama 3.2 vision and post_init reinitialization are distinct bugs." }, { - "left": "issue:45372", - "right": "issue:45588", + "left": "issue:44857", + "right": "issue:45072", "accept": false, - "reason": "Gemma 4 processor import failure is unrelated to sink-less flash-attention attribute handling." + "reason": "Different model/runtime issues: LwDetr AMP crash vs SwitchTransformers dtype mismatch." }, { "left": "issue:43701", "right": "issue:45083", "accept": false, - "reason": "Checkpoint key mismatch is unrelated to the qwen3_omni_moe helper function behavior." + "reason": "Checkpoint resume key mismatch is unrelated to qwen3_omni_moe feature-length behavior." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded overall: the only clear duplicate in the packet is issue:45310 / issue:45478, which have identical titles and strongly overlapping descriptions. The other soft-edge rejections stay conservative and do not overclaim a shared bug based on the packet evidence." + "feedback": "The summary is grounded in the packet: 45310 and 45478 are the only exact title match, and 45310 is earlier with an inbound reference. The soft-edge verdicts are conservative and do not overstate evidence; the lone true duplicate call is well supported." }, "evaluator_used": true, "retried": false, @@ -7544,7 +7410,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:25251", "issue:29942", @@ -7596,236 +7462,237 @@ ], "soft_pairs": [ "issue:43927|issue:45698", - "issue:44977|issue:45563", + "issue:45372|issue:45588", "issue:43550|issue:45083", + "issue:44977|issue:45563", "issue:43412|issue:45588", "issue:43066|issue:44462", - "issue:45245|issue:45563", "issue:43873|issue:44493", + "issue:45245|issue:45563", "issue:29942|issue:35141", - "issue:36010|issue:37428", "issue:45372|issue:45459", + "issue:36010|issue:37428", "issue:44448|issue:44466", "issue:43278|issue:43381", "issue:25251|issue:32090", - "issue:43065|issue:45412", "issue:44857|issue:45588", + "issue:43065|issue:45412", "issue:43653|issue:44743", "issue:44186|issue:45198", "issue:44451|issue:45520", - "issue:43976|issue:45440", "issue:44361|issue:45198", + "issue:43976|issue:45440", "issue:44265|issue:45198", "issue:43994|issue:45412", - "issue:42915|issue:42994", "issue:44871|issue:44977", "issue:43421|issue:45491", + "issue:42915|issue:42994", "issue:29942|issue:30064", "issue:44206|issue:44479", "issue:43065|issue:43994" ], - "prepared_review_unit_hash": "f9e3c94d5ca984fefe9a93712d43ba302311add170bc45df57af97dedaeab0f4", + "prepared_review_unit_hash": "09e4521981c1d1f90387584150bc47f175cfd3533569445e7b1aa38193ab27c2", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 28, - "serialized_chars": 35521, - "estimated_input_tokens": 8881, - "estimated_eval_tokens": 18018 + "soft_pair_count": 29, + "serialized_chars": 35812, + "estimated_input_tokens": 8953, + "estimated_eval_tokens": 18162 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft edge looks like a true duplicate: the pairs mostly connect unrelated bugs across different models, tokenizer paths, export flows, and runtime features.", + "summary": "This cluster is not a true duplicate set; the items span many unrelated bugs and regressions, so all soft edges should be rejected. There are no PRs in the input, and no single issue cleanly represents the whole set.", "confidence": 0.96, - "canonical_issue_reason": "No canonical issue is justified here; the items do not form a single duplicate cluster and the soft similarities are only superficial.", + "canonical_issue_reason": "No single canonical issue fits: the cluster is heterogeneous across tokenization, generation, quantization, vision, import, and trainer/runtime bugs.", "canonical_pr_reason": null, - "best_issue_reason": "No global best issue is appropriate because there is no shared underlying bug/change to anchor the set.", + "best_issue_reason": "No global best issue is suitable because none of the issues subsumes the others or represents one underlying defect.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43927", "right": "issue:45698", "accept": false, - "reason": "Different bugs: custom token IDs lost on save/load vs loading the wrong custom module after save_pretrained." + "reason": "Both involve save/load behavior around custom tokens or modules, but one is about token IDs being lost and the other is about loading the wrong custom module; different failure modes." }, { - "left": "issue:44977", - "right": "issue:45563", + "left": "issue:45372", + "right": "issue:45588", "accept": false, - "reason": "One is a Qwen3.5 flash-attention generation bug; the other is a stale warning in paged generate for num_return_sequences." + "reason": "ImportError in Gemma 4 processor loading from a missing dependency is unrelated to a flash-attention crash on sink-less models." }, { "left": "issue:43550", "right": "issue:45083", "accept": false, - "reason": "Bamba torch.compile+SDPA crash and a qwen3_omni_moe feature-length helper issue are unrelated code paths." + "reason": "torch.compile SDPA failure in Bamba-9B-v2 and a helper output-length bug in qwen3_omni_moe are distinct code paths." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Qwen3.5 flash-attention generation regression and a stale num_return_sequences warning in paged generate() are not the same bug." }, { "left": "issue:43412", "right": "issue:45588", "accept": false, - "reason": "Gemma3n export failures and a flash_attention.py None aux crash are distinct problems." + "reason": "Executorch export failure due to missing training guard/erfinv support is unrelated to the s_aux=None flash_attention.py crash." }, { "left": "issue:43066", "right": "issue:44462", "accept": false, - "reason": "Tokenizer decoder type regression and AutoTokenizer ignoring tokenizer.json are separate loader/serialization bugs." + "reason": "Wrong tokenizer decoder type and ignoring tokenizer.json are both tokenizer issues, but they affect different loading/serialization logic." }, { - "left": "issue:45245", - "right": "issue:45563", + "left": "issue:43873", + "right": "issue:44493", "accept": false, - "reason": "Category-count limit runtime error is unrelated to a stale generate warning." + "reason": "Quantization offloading behavior and an unexpected position-id key error are different symptoms with no clear shared code-path." }, { - "left": "issue:43873", - "right": "issue:44493", + "left": "issue:45245", + "right": "issue:45563", "accept": false, - "reason": "Quantization/offloading behavior and unexpected position-id keys are different failure modes." + "reason": "Category-count runtime error and a stale generation warning are unrelated." }, { "left": "issue:29942", "right": "issue:35141", "accept": false, - "reason": "Flash Attention 2 test failures do not match the embedding reinitialization bug in post_init." + "reason": "Flash Attention 2 test failures and embedding reinitialization in post_init are unrelated regressions." }, { - "left": "issue:36010", - "right": "issue:37428", + "left": "issue:45372", + "right": "issue:45459", "accept": false, - "reason": "Both are import errors, but for different symbols and modules, so not the same underlying issue." + "reason": "Missing ReasoningEffort import and protobuf decode-error masking affect different tokenizer/processor paths." }, { - "left": "issue:45372", - "right": "issue:45459", + "left": "issue:36010", + "right": "issue:37428", "accept": false, - "reason": "ReasoningEffort import breakage and protobuf-related tokenizer error masking are unrelated." + "reason": "Both are import errors, but they are for different symbols, modules, and likely fixes." }, { "left": "issue:44448", "right": "issue:44466", "accept": false, - "reason": "Pegasus output differences and lm_head.weight serialization inconsistency are distinct v5 regressions." + "reason": "Pegasus output differences across versions and inconsistent lm_head.weight serialization are separate model-behavior issues." }, { "left": "issue:43278", "right": "issue:43381", "accept": false, - "reason": "Embedding dtype drift in eval and gradient checkpointing being disallowed in eval are separate issues." + "reason": "Embedding dtype drift in evaluate() and gradient-checkpointing incompatibility in eval mode are distinct." }, { "left": "issue:25251", "right": "issue:32090", "accept": false, - "reason": "Pipeline top_k nesting change and Trainer NoneType broadcast failure have no shared code-path." + "reason": "top_k nesting output change in pipeline and Trainer NoneType broadcast error are unrelated." }, { - "left": "issue:43065", - "right": "issue:45412", + "left": "issue:44857", + "right": "issue:45588", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder and RT-DETR memory not released are unrelated." + "reason": "AMP/CUDA crash in LwDetrImageLoss is unrelated to the sink-less flash_attention None crash." }, { - "left": "issue:44857", - "right": "issue:45588", + "left": "issue:43065", + "right": "issue:45412", "accept": false, - "reason": "LwDetrImageLoss AMP crash is not the same as the s_aux=None flash_attention AttributeError." + "reason": "Dummy Conv2d in Sam3PixelDecoder and RT-DETR memory leak on deletion are different problems." }, { "left": "issue:43653", "right": "issue:44743", "accept": false, - "reason": "BigBirdTokenizer special-token decode and Qwen3.5 recurrent-state reset are unrelated." + "reason": "BigBirdTokenizer special-token registration and qwen3_5 recurrent-state reset are unrelated." }, { "left": "issue:44186", "right": "issue:45198", "accept": false, - "reason": "LayoutLMv2Tokenizer NER/padding crash and Wav2Vec2 save_pretrained/tokenization failure are different tokenizer bugs." + "reason": "LayoutLMv2Tokenizer NER/padding crash and Wav2Vec2 save_pretrained/tokenization failure are different model/tokenizer bugs." }, { "left": "issue:44451", "right": "issue:45520", "accept": false, - "reason": "Loading ScandiBERT failing is separate from the Python 3.13 flash_attn import_utils KeyError." + "reason": "ScandiBERT loading failure and flash_attn KeyError on Python 3.13 are unrelated." }, { - "left": "issue:43976", - "right": "issue:45440", + "left": "issue:44361", + "right": "issue:45198", "accept": false, - "reason": "Python version incompatibility and DeepseekV3MoE divergence are unrelated." + "reason": "MLukeTokenizer task-time AttributeError and Wav2Vec2 save_pretrained/tokenization failure do not share the same underlying defect." }, { - "left": "issue:44361", - "right": "issue:45198", + "left": "issue:43976", + "right": "issue:45440", "accept": false, - "reason": "MLukeTokenizer task AttributeError and Wav2Vec2 save/tokenization failure do not share the same root cause." + "reason": "Python version compatibility problem and DeepseekV3MoE divergence are unrelated." }, { "left": "issue:44265", "right": "issue:45198", "accept": false, - "reason": "torch.export/export with torch_compilable_check is a different issue from Wav2Vec2 save_pretrained/tokenization failure." + "reason": "torch.export failure with torch_compilable_check and Wav2Vec2 save_pretrained/tokenization failure are different code paths." }, { "left": "issue:43994", "right": "issue:45412", "accept": false, - "reason": "Siglip2 bad outputs and RT-DETR memory retention are unrelated bugs." - }, - { - "left": "issue:42915", - "right": "issue:42994", - "accept": false, - "reason": "Quantized inference failure on Qwen3Moe and quantized model saving failure are different code paths." + "reason": "Siglip2 nonsense outputs and RT-DETR memory retention are not the same bug." }, { "left": "issue:44871", "right": "issue:44977", "accept": false, - "reason": "Gemma3 eos_token_id inconsistency and Qwen3.5 flash-attention generation are unrelated." + "reason": "Gemma-3 eos_token_id configuration mismatch and Qwen3.5 flash-attention generation regression are separate issues." }, { "left": "issue:43421", "right": "issue:45491", "accept": false, - "reason": "Runtime post-processor updates in TokenizersBackend do not match the Gemma3 NaN embedding issue." + "reason": "TokenizerBackend post-processor update behavior and Gemma3 NaN embeddings on GPU are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "Qwen3Moe FP8 failure during inference/training and quantized model saving failure are different quantization bugs." }, { "left": "issue:29942", "right": "issue:30064", "accept": false, - "reason": "Flash Attention 2 tests and void segmentation map processing are unrelated subsystems." + "reason": "Flash Attention 2 test failures and void segmentation map processing are unrelated." }, { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Unsupported center arg in LasrFeatureExtractor and Qwen video regression are distinct failures." + "reason": "Unsupported center arg in LasrFeatureExtractor and Qwen video input regression are different failures." }, { "left": "issue:43065", "right": "issue:43994", "accept": false, - "reason": "Sam3PixelDecoder dummy Conv2d and Siglip2 nonsensical outputs are not the same bug." - }, - { - "left": "issue:43873", - "right": "issue:43994", - "accept": false, - "reason": "Quantization/offloading behavior and Siglip2 model outputs are unrelated." + "reason": "Sam3PixelDecoder dummy Conv2d and Siglip2 AutoModel/pipeline behavior are unrelated." } ] }, @@ -7847,7 +7714,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:29127", "issue:29942", @@ -7863,7 +7730,6 @@ "issue:43329", "issue:43334", "issue:43381", - "issue:43441", "issue:43452", "issue:43526", "issue:43531", @@ -7879,6 +7745,7 @@ "issue:44206", "issue:44279", "issue:44442", + "issue:44464", "issue:44554", "issue:44610", "issue:44734", @@ -7894,6 +7761,7 @@ "issue:45406", "issue:45412", "issue:45440", + "issue:45491", "issue:45588", "issue:45593" ], @@ -7905,231 +7773,238 @@ "issue:45103|issue:45372", "issue:43873|issue:45229", "issue:43756|issue:45440", + "issue:43723|issue:44554", "issue:45372|issue:45406", "issue:43278|issue:43575", - "issue:43723|issue:44554", "issue:43381|issue:43701", "issue:43278|issue:43329", "issue:43653|issue:44442", - "issue:41720|issue:45229", "issue:45440|issue:45593", + "issue:41720|issue:45229", "issue:29942|issue:36010", "issue:32090|issue:33357", "issue:43334|issue:43531", - "issue:43824|issue:44279", "issue:43526|issue:44186", "issue:44977|issue:45245", + "issue:43824|issue:44279", "issue:43526|issue:44857", - "issue:44442|issue:45072", - "issue:43526|issue:44610", - "issue:43262|issue:44206", "issue:42915|issue:43931", + "issue:43526|issue:44610", "issue:44743|issue:44811", - "issue:43441|issue:45588", - "issue:43452|issue:45588" + "issue:44442|issue:45072", + "issue:43262|issue:44206", + "issue:43452|issue:45588", + "issue:43575|issue:43701", + "issue:44464|issue:45491" ], - "prepared_review_unit_hash": "41b4599dea153880ab3eed3fae1e738f37a158f2215713117ac3398e3a26c7d3", + "prepared_review_unit_hash": "06a6e03128fae22379544ad63841cff0e045ee1290318171009ea6a79bb25d93", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 29, - "serialized_chars": 35874, - "estimated_input_tokens": 8969, - "estimated_eval_tokens": 18194 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36804, + "estimated_input_tokens": 9201, + "estimated_eval_tokens": 18658 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a grab-bag of unrelated closed issues. Several pairs share a subsystem or model family name, but the underlying bugs, code paths, and fixes are different, so none of the soft edges look like true duplicates.", - "confidence": 0.87, - "canonical_issue_reason": "Issue 45588 is the best anchor: it is a concrete, reproducible crash in a shared integration code path (`integrations/flash_attention.py`) with a narrowly scoped failure mode and clear fix surface.", + "summary": "This cluster is heterogeneous: the soft pairs are mostly surface-similar but refer to different models, components, and failure modes. I found no true duplicate/merge-worthy PR-style overlap, and no single issue cleanly represents the whole set.", + "confidence": 0.96, + "canonical_issue_reason": "No clear canonical duplicate source: the items span unrelated bugs across training, tokenizers, multimodal processors, generation, and model loading.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45588 is the strongest representative because it is specific, actionable, and tied to a shared library path that other nearby flash-attention regressions could plausibly relate to.", + "best_issue_reason": "If a single anchor is needed, issue:45229 is the strongest standalone report because it has the most discussion and describes a high-impact multi-GPU inference OOM regression.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44734", "right": "issue:45406", "accept": false, - "reason": "Both involve `transformers serve`, but one is a KV-cache tensor indexing bug and the other is a missing `_tokenizer` attribute on Gemma4Processor; different failures and fixes." + "reason": "Different Gemma4-related failures: KV-cache continuation tensor indexing vs missing `_tokenizer` on processor load/serve." }, { "left": "issue:33357", "right": "issue:39692", "accept": false, - "reason": "Both mention vision-language/model loading problems, but one is a Mac bus error with CLIP and the other is a SigLIP2 doc example mismatch/quantization issue." + "reason": "Both mention CLIP/SigLIP-style usage, but one is a MacOS bus error in a community CLIP model and the other is a docs example mismatch/quantization failure." }, { "left": "issue:29127", "right": "issue:34567", "accept": false, - "reason": "LayoutLMV3 error messaging vs TrainerState token counter update are unrelated features and failure modes." + "reason": "LayoutLMv3 error messaging vs TrainerState token counting update bug; unrelated subsystems." }, { "left": "issue:44610", "right": "issue:45412", "accept": false, - "reason": "Processor image-size mismatch and model memory not releasing are separate bugs with no shared code path." + "reason": "Processor input-size mismatch for OmDet-Turbo vs RT-DETR memory not being released; different bugs." }, { "left": "issue:45103", "right": "issue:45372", "accept": false, - "reason": "Auto-docstring annotation handling and a mistral_common import regression are unrelated." + "reason": "auto_docstring annotation crash vs Gemma4 loading import failure; no shared code path." }, { "left": "issue:43873", "right": "issue:45229", "accept": false, - "reason": "Quantization/offloading behavior and Gemma4 multi-GPU OOM are different resource-management problems, not the same bug." + "reason": "Both are memory/loading-adjacent, but one is quantization/offloading behavior and the other is a model-specific multi-GPU OOM." }, { "left": "issue:43756", "right": "issue:45440", "accept": false, - "reason": "Smollm3 RoPE layer selection and DeepseekV3MoE divergence are distinct model-implementation issues." + "reason": "Different model families and different architectural issues: Smollm3 RoPE layer handling vs DeepSeekV3 MoE divergence." + }, + { + "left": "issue:43723", + "right": "issue:44554", + "accept": false, + "reason": "Tokenizer loading regression in v5 vs MPS attention correctness with mismatched value/query head dims." }, { "left": "issue:45372", "right": "issue:45406", "accept": false, - "reason": "Gemma4 processor loading via mistral_common import failure is separate from `transformers serve` missing `_tokenizer`." + "reason": "Both touch Gemma4 loading, but the root causes differ: missing `ReasoningEffort` import vs missing `_tokenizer` attribute in serve." }, { "left": "issue:43278", "right": "issue:43575", "accept": false, - "reason": "BF16/FP32 dtype drift during eval and TP loading OOM for a large Qwen2 model are not the same underlying bug." - }, - { - "left": "issue:43723", - "right": "issue:44554", - "accept": false, - "reason": "Tokenizer loading in v5 and an MPS attention correctness issue are unrelated." + "reason": "Embedding dtype drift during eval vs Qwen2-57B tensor-parallel OOM; not the same bug." }, { "left": "issue:43381", "right": "issue:43701", "accept": false, - "reason": "Gradient checkpointing being invalid in eval mode is different from a resume-from-checkpoint key mismatch." + "reason": "Eval-mode gradient checkpointing restriction vs resume-from-checkpoint key mismatch; unrelated trainer behaviors." }, { "left": "issue:43278", "right": "issue:43329", "accept": false, - "reason": "Dtype changes in evaluation and a multimodal token-counting bug in video paths are unrelated." + "reason": "Different issues entirely: dtype change in evaluation vs multimodal token counting video-branch undefined names." }, { "left": "issue:43653", "right": "issue:44442", "accept": false, - "reason": "BigBirdTokenizer special-token registration and FastSpeech2ConformerTokenizer auto-loading failures are different tokenizer-specific issues." + "reason": "Tokenizer special-token registration bug vs AutoTokenizer failing on a FastSpeech2Conformer tokenizer; different failure modes." }, { - "left": "issue:41720", - "right": "issue:45229", + "left": "issue:45440", + "right": "issue:45593", "accept": false, - "reason": "Qwen3 auto device mapping cuda assert and Gemma4 multi-GPU OOM are separate large-model inference problems." + "reason": "DeepSeekV3MoE implementation divergence vs D-FINE auxiliary-loss behavior when denoising is off." }, { - "left": "issue:45440", - "right": "issue:45593", + "left": "issue:41720", + "right": "issue:45229", "accept": false, - "reason": "DeepseekV3MoE implementation divergence and D-FINE auxiliary-loss handling are different model bugs." + "reason": "Different model-specific GPU problems: Qwen3 auto device-map assert vs Gemma4 multi-GPU OOM." }, { "left": "issue:29942", "right": "issue:36010", "accept": false, - "reason": "Flash Attention 2 test failures and a GenerationMixin import error are unrelated regressions." + "reason": "Flash Attention 2 test failures and a GenerationMixin import error are unrelated." }, { "left": "issue:32090", "right": "issue:33357", "accept": false, - "reason": "NoneType broadcast error in Trainer and a MacOS CLIP bus error are not the same failure." + "reason": "Trainer NoneType broadcast error vs MacOS CLIP bus error; no shared bug." }, { "left": "issue:43334", "right": "issue:43531", "accept": false, - "reason": "Qwen3-VL missing `pad_token_id` and Qwen3-MoE sliding-window behavior are different model/config issues." - }, - { - "left": "issue:43824", - "right": "issue:44279", - "accept": false, - "reason": "A missing `Qwen2_5_VLForConditionalGeneration` import and a generic dependency issue are too vague and do not share a concrete bug." + "reason": "Qwen3-VL missing `pad_token_id` vs Qwen3-MoE sliding-window issue; different components and symptoms." }, { "left": "issue:43526", "right": "issue:44186", "accept": false, - "reason": "BEiT `reduce_labels` returning one label and LayoutLMv2Tokenizer crashing on NER/padding are different processor/tokenizer bugs." + "reason": "BeitImageProcessorFast label reduction bug vs LayoutLMv2Tokenizer crashes on NER/padding; not the same defect." }, { "left": "issue:44977", "right": "issue:45245", "accept": false, - "reason": "Flash-attention generation problems for Qwen3.5 and a category-cardinality runtime error are unrelated." + "reason": "Qwen3.5 flash-attention generation problems vs category-count overflow; unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Specific missing `Qwen2_5_VLForConditionalGeneration` import vs generic dependency issue; too broad and not the same root cause." }, { "left": "issue:43526", "right": "issue:44857", "accept": false, - "reason": "A label-reduction bug and an AMP/CUDA crash in LwDetrImageLoss are not the same issue." + "reason": "Beit label reduction bug vs LwDetrImageLoss AMP/CUDA crash; different model paths." }, { - "left": "issue:44442", - "right": "issue:45072", + "left": "issue:42915", + "right": "issue:43931", "accept": false, - "reason": "Tokenizer loading failure and dtype mismatches in inference are different classes of bugs." + "reason": "Qwen3Moe FP8 config failure vs Qwen3-VL weight shape mismatch; distinct model-loading errors." }, { "left": "issue:43526", "right": "issue:44610", "accept": false, - "reason": "BEiT label reduction and OmDet-Turbo image-size mismatch are unrelated." + "reason": "Different processor bugs: Beit label reduction vs OmDet-Turbo image size mismatch." }, { - "left": "issue:43262", - "right": "issue:44206", + "left": "issue:44743", + "right": "issue:44811", "accept": false, - "reason": "Audio chat-template sample-rate defaulting and a feature-extractor unsupported-arg crash are separate audio-processing regressions." + "reason": "Qwen3.5 recurrent-state reset with cache vs Whisper batch_decode ignoring skip_special_tokens; unrelated." }, { - "left": "issue:42915", - "right": "issue:43931", + "left": "issue:44442", + "right": "issue:45072", "accept": false, - "reason": "Qwen3MoE FP8 config failure and Qwen3-VL weight-shape mismatch are different load-time problems." + "reason": "FastSpeech2ConformerTokenizer loading bug vs dtype mismatch in SwitchTransformers/TimmWrapperModel." }, { - "left": "issue:44743", - "right": "issue:44811", + "left": "issue:43262", + "right": "issue:44206", "accept": false, - "reason": "Recurrent-state reset with cache and Whisper batch_decode skip-special-tokens handling are unrelated." + "reason": "Audio chat-template sample-rate default bug vs LasrFeatureExtractor unsupported argument crash; different paths." }, { - "left": "issue:43441", + "left": "issue:43452", "right": "issue:45588", "accept": false, - "reason": "Both mention flash attention, but one is a Ministral-3 generation failure and the other is an `s_aux=None` crash in a different integration path." + "reason": "gguf_file loading regression vs flash_attention `s_aux=None` crash; unrelated." }, { - "left": "issue:43452", - "right": "issue:45588", + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "Qwen2 tensor-parallel OOM vs checkpoint resume key mismatch; different problems." + }, + { + "left": "issue:44464", + "right": "issue:45491", "accept": false, - "reason": "GGUF loading issues for tokenizer/model constructors are not the same as the flash_attention integration crash." + "reason": "Compiled-forward chunked generation inconsistency vs Gemma3 NaN embeddings from sliding-window/all-padding behavior." } ] }, @@ -8151,7 +8026,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:29127", "issue:33290", @@ -8165,13 +8040,13 @@ "issue:43295", "issue:43316", "issue:43381", + "issue:43441", "issue:43493", "issue:43525", "issue:43550", "issue:43575", "issue:43611", "issue:43646", - "issue:43701", "issue:43761", "issue:43827", "issue:43873", @@ -8183,8 +8058,8 @@ "issue:44206", "issue:44387", "issue:44442", - "issue:44464", "issue:44479", + "issue:44568", "issue:44743", "issue:44898", "issue:44964", @@ -8194,6 +8069,7 @@ "issue:45229", "issue:45290", "issue:45440", + "issue:45479", "issue:45491", "issue:45584", "issue:45588", @@ -8202,8 +8078,7 @@ "issue:45759" ], "soft_pairs": [ - "issue:44464|issue:45491", - "issue:43575|issue:43701", + "issue:43441|issue:45588", "issue:29127|issue:35141", "issue:43646|issue:44077", "issue:43575|issue:44387", @@ -8211,10 +8086,10 @@ "issue:44977|issue:45584", "issue:33357|issue:38175", "issue:45588|issue:45600", - "issue:43316|issue:43525", "issue:44479|issue:45290", - "issue:44964|issue:45440", "issue:43761|issue:44079", + "issue:44964|issue:45440", + "issue:43316|issue:43525", "issue:43931|issue:45216", "issue:33290|issue:34567", "issue:42907|issue:45698", @@ -8224,195 +8099,196 @@ "issue:39692|issue:45491", "issue:43381|issue:43873", "issue:43493|issue:43827", - "issue:43295|issue:45584", "issue:43550|issue:44743", + "issue:43295|issue:45584", "issue:43994|issue:44206", + "issue:44568|issue:45479", "issue:43316|issue:45759" ], - "prepared_review_unit_hash": "f253d518eca547f12da5da228d2480f761765dd03cb7c52d1d65586e47f4a9f9", + "prepared_review_unit_hash": "7cd5289733dfe4a5679224f87d2069fdd78ac97bf6ef64b2e098b69806c20de5", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 26, - "serialized_chars": 34962, - "estimated_input_tokens": 8741, - "estimated_eval_tokens": 17738 + "serialized_chars": 35673, + "estimated_input_tokens": 8919, + "estimated_eval_tokens": 18094 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These soft pairs are mostly superficial matches: they share broad themes like OOMs, loading regressions, config mismatches, or save/load behavior, but each issue targets a different model, code path, or failure mode. None looks like the same underlying bug or a mergeable duplicate pair.", - "confidence": 0.97, - "canonical_issue_reason": "No single canonical issue fits this set; the candidates are heterogeneous across subsystems and failure types, so there isn't one underlying bug to anchor the cluster.", + "summary": "The cluster is heterogeneous: most pairs share only broad themes like v5 regressions, OOMs, or model-loading failures, but they do not appear to be the same underlying bug/change. I found no safe duplicate merges among the soft edges.", + "confidence": 0.44, + "canonical_issue_reason": "issue:43646 is the broadest, most representative regression report in the set ('Transformers 5.0.0 breaks custom model initialization'), so it best serves as a cluster anchor despite the overall heterogeneity.", "canonical_pr_reason": null, - "best_issue_reason": "No best issue is truly representative because the cluster is not a real duplicate group; at most, a few issues share broad regression language, but not the same concrete defect.", + "best_issue_reason": "issue:43646 is the best single issue to represent the cluster because it is a generic v5 regression report and not tied to one narrow model or subsystem.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44464", - "right": "issue:45491", - "accept": false, - "reason": "Both are inference correctness bugs, but one is about compiled forward chunked generation and the other about NaN embeddings in Gemma3 mixed-length batching; different models and code paths." - }, - { - "left": "issue:43575", - "right": "issue:43701", + "left": "issue:43441", + "right": "issue:45588", "accept": false, - "reason": "Qwen2 tensor-parallel OOM and resume_from_checkpoint key mismatch are unrelated failures with no shared underlying mechanism." + "reason": "Both involve FlashAttention, but one is a Ministral v5 RC generation failure and the other is an AttributeError in flash_attention.py for sink-less models; different root causes." }, { "left": "issue:29127", "right": "issue:35141", "accept": false, - "reason": "LayoutLMv3 error messaging for insufficient box info is a validation/UX issue, while the other is about output embeddings being reinitialized during resize; different bugs." + "reason": "LayoutLMv3 error-message clarity and token-embedding reinitialization on resize are unrelated bugs." }, { "left": "issue:43646", "right": "issue:44077", "accept": false, - "reason": "Both mention initialization/post_init, but one is a broad Transformers 5.0 custom model init regression and the other is a patchtsmixer-specific post_init policy change; not the same defect." + "reason": "One is a broad custom model initialization regression; the other is a PatchTSTMixer-specific post_init policy issue." }, { "left": "issue:43575", "right": "issue:44387", "accept": false, - "reason": "Both are OOM reports, but one is TP load OOM for Qwen2 and the other is increased CUDA reserved memory under int4 quantization; distinct causes and paths." + "reason": "Both mention OOM, but one is tensor-parallel loading of Qwen2-57B and the other is increased reserved memory under int4 quantization; different failure modes." }, { "left": "issue:45072", "right": "issue:45229", "accept": false, - "reason": "Bfloat16 dtype mismatch in inference is unrelated to a multi-GPU CUDA OOM for Gemma4." + "reason": "dtype mismatch CI failures and Gemma4 multi-GPU CUDA OOM are not the same bug." }, { "left": "issue:44977", "right": "issue:45584", "accept": false, - "reason": "Flash-attention generation issues for Qwen3.5 and Whisper empty-transcription failure after align_special_tokens are different models and different failure conditions." + "reason": "Qwen3.5 flash-attention generation issues and Whisper empty-transcription failure after align_special_tokens are unrelated." }, { "left": "issue:33357", "right": "issue:38175", "accept": false, - "reason": "MacOS bus error on a community CLIP model and zero-probability outputs on SigLIP2 are not the same bug." + "reason": "A MacOS bus error on CLIP and zero probabilities on SigLIP2 are distinct model/runtime problems." }, { "left": "issue:45588", "right": "issue:45600", "accept": false, - "reason": "A crash in flash_attention.py with s_aux=None is unrelated to stale auto_mappings references to removed Sam3LiteText configs." + "reason": "FlashAttention crash on s_aux=None and auto_mappings references to removed configs affect different code paths." }, { - "left": "issue:43316", - "right": "issue:43525", + "left": "issue:44479", + "right": "issue:45290", "accept": false, - "reason": "Both are config/API discrepancies, but one concerns Gemma3TextConfig parity and the other a missing pad_token_id on Llama4Config; different API gaps." + "reason": "Video-input regression for Qwen multimodal models and chat-template crashes on tool-call messages are separate regressions." }, { - "left": "issue:44479", - "right": "issue:45290", + "left": "issue:43761", + "right": "issue:44079", "accept": false, - "reason": "Video-input regression across Qwen VL variants and chat template crashes on tool-call assistant messages are separate feature paths." + "reason": "Missing hidden_states in CLIPVisionModel.forward and ModelOutput key assignment when previous value is None are different defects." }, { "left": "issue:44964", "right": "issue:45440", "accept": false, - "reason": "Phi-4 multimodal loading failure and DeepseekV3MoE divergence from remote implementation are unrelated model-specific issues." + "reason": "Phi-4 multimodal loading failure and DeepseekV3MoE divergence from remote implementation are unrelated." }, { - "left": "issue:43761", - "right": "issue:44079", + "left": "issue:43316", + "right": "issue:43525", "accept": false, - "reason": "CLIPVisionModel ignoring output_hidden_states and a generic ModelOutput key assignment bug are different layers of the stack and not the same concrete bug." + "reason": "A config API discrepancy and a missing pad_token_id attribute are not the same underlying issue." }, { "left": "issue:43931", "right": "issue:45216", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch on load and Qwen3.5 save_pretrained regression affect different stages of the lifecycle and different failure modes." + "reason": "Weight-shape mismatch on load and incorrect save_pretrained checkpoint output are different lifecycle stages and bugs." }, { "left": "issue:33290", "right": "issue:34567", "accept": false, - "reason": "AdaFactor+DeepSpeed OOM is unrelated to TrainerState.num_input_tokens_seen not updating." + "reason": "Adafactor OOM in DeepSpeed and TrainerState token-count tracking are unrelated." }, { "left": "issue:42907", "right": "issue:45698", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals and loading the wrong custom module after save_pretrained are both save/load related, but they are different bugs with different causes." + "reason": "Saving dequantized models and loading the wrong custom module after save_pretrained concern different serialization paths." }, { "left": "issue:34689", "right": "issue:43611", "accept": false, - "reason": "Both are loading regressions, but one is Llama 3.2 vision model loading and the other is a base_model_prefix attribute regression in Transformers 5.0; not mergeable as one fix." + "reason": "Llama 3.2 Vision loading breakage and base_model_prefix loading regression are separate model-loading failures." }, { "left": "issue:44442", "right": "issue:44898", "accept": false, - "reason": "Tokenizer loading failure and Perceiver interpolation failure are unrelated model/component bugs." + "reason": "Tokenizer loading failure and Perceiver non-default-resolution failure are unrelated." }, { "left": "issue:43761", "right": "issue:44188", "accept": false, - "reason": "Hidden-states being None in CLIPVisionModel is unrelated to attention kernel divergence caused by torch.compile branching." + "reason": "A forward-output bug in CLIPVisionModel and attention-kernel divergence under torch.compile are different problems." }, { "left": "issue:39692", "right": "issue:45491", "accept": false, - "reason": "SigLIP2 docs/example errors and Gemma3 NaN embeddings are different problem classes; one is documentation/example correctness, the other runtime numerical instability." + "reason": "SigLIP2 docs/example issues and Gemma3 mixed-length NaN embeddings are not the same bug." }, { "left": "issue:43381", "right": "issue:43873", "accept": false, - "reason": "Gradient checkpointing in eval mode and offloading with quantization are separate runtime behaviors with no shared root cause." + "reason": "Gradient checkpointing in eval mode and quantization offloading behavior are unrelated." }, { "left": "issue:43493", "right": "issue:43827", "accept": false, - "reason": "SigLIP2 HF-vs-JAX implementation discrepancy is unrelated to docs still referencing removed pipeline() APIs." + "reason": "SigLIP2 implementation discrepancy and docs referencing removed pipeline APIs are distinct." }, { - "left": "issue:43295", - "right": "issue:45584", + "left": "issue:43550", + "right": "issue:44743", "accept": false, - "reason": "Processor.tokenizer regression and Whisper empty-transcription failure are different regressions in different code paths." + "reason": "torch.compile SDPA failure in Bamba and recurrent-state reset in Qwen3.5 are different code-path bugs." }, { - "left": "issue:43550", - "right": "issue:44743", + "left": "issue:43295", + "right": "issue:45584", "accept": false, - "reason": "torch.compile+SDPA failure in Bamba is unrelated to recurrent states being reset in modular_qwen3_5 cache handling." + "reason": "Processor/tokenizer access regression and Whisper empty-transcription failure are unrelated." }, { "left": "issue:43994", "right": "issue:44206", "accept": false, - "reason": "SigLIP2 nonsensical outputs and LasrFeatureExtractor center-arg crash do not describe the same defect." + "reason": "Bad outputs from SigLIP2 AutoModel/pipeline and a LaserFeatureExtractor argument crash are different issues." + }, + { + "left": "issue:44568", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer special-token addition regression and degenerate zero-loss classification behavior do not share the same root cause." }, { "left": "issue:43316", "right": "issue:45759", "accept": false, - "reason": "Both involve config APIs, but one is a generic Gemma3TextConfig discrepancy and the other is specifically AutoModelForCausalLM.from_config not unwrapping text_config for composite Qwen configs; different fixes." + "reason": "Gemma3TextConfig API mismatch and Qwen multimodal from_config text_config unwrapping are separate config-handling bugs." } ] }, @@ -8434,7 +8310,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:25251", "issue:28282", @@ -8443,6 +8319,7 @@ "issue:33290", "issue:33357", "issue:34689", + "issue:35141", "issue:36010", "issue:41628", "issue:41950", @@ -8472,237 +8349,251 @@ "issue:44479", "issue:44496", "issue:44514", - "issue:44568", "issue:44792", "issue:44855", "issue:44871", "issue:44877", "issue:44913", + "issue:45127", "issue:45137", + "issue:45200", "issue:45372", - "issue:45479", "issue:45584", "issue:45656" ], "soft_pairs": [ - "issue:44568|issue:45479", - "issue:41628|issue:45372", + "issue:32090|issue:33290", "issue:43720|issue:44442", "issue:43643|issue:44913", - "issue:32090|issue:33290", - "issue:44190|issue:44496", + "issue:41628|issue:45372", "issue:43873|issue:44871", + "issue:44190|issue:44496", "issue:34689|issue:36010", "issue:43381|issue:43856", "issue:44855|issue:45137", "issue:44077|issue:44479", "issue:33290|issue:45656", - "issue:43540|issue:44514", "issue:29127|issue:33357", - "issue:43316|issue:44877", + "issue:43540|issue:44514", "issue:43381|issue:43742", - "issue:42175|issue:43316", + "issue:43316|issue:44877", "issue:33290|issue:34689", + "issue:42175|issue:43316", "issue:25251|issue:28282", - "issue:44162|issue:44792", "issue:43404|issue:43441", + "issue:44162|issue:44792", "issue:43232|issue:43388", - "issue:41950|issue:43475", "issue:43295|issue:43825", + "issue:41950|issue:43475", "issue:44871|issue:45584", "issue:43550|issue:43701", - "issue:43606|issue:43701" + "issue:43606|issue:43701", + "issue:45127|issue:45200", + "issue:43295|issue:44871", + "issue:33290|issue:35141" ], - "prepared_review_unit_hash": "4c573659d4264c635c2dfa94fe187fd56e068b6276117de963e7143c594b0302", + "prepared_review_unit_hash": "b730d94e30325e4ed69e90dd8ed9b27c75c735c90bcb3b06ad8eedcb82dd9de4", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 27, - "serialized_chars": 35398, - "estimated_input_tokens": 8850, - "estimated_eval_tokens": 17956 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36563, + "estimated_input_tokens": 9141, + "estimated_eval_tokens": 18538 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are a loose collection of unrelated Transformers bugs and regressions. The soft pairs share only broad subsystem similarity; none look like the same underlying bug/change, so no soft edge should be merged.", + "summary": "The cluster is heterogeneous: the paired issues only overlap at a broad subsystem level (pipelines, model loading, deepspeed, multimodal/video), but each describes a different concrete bug or regression. No soft edge should be merged.", "confidence": 0.96, - "canonical_issue_reason": null, + "canonical_issue_reason": "No single canonical duplicate stands out; these issues span unrelated code paths in loading, config, generation, training, and pipeline behavior.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:41950 is the most central/open pipeline-related report and is the closest thing to an anchor, but it still does not duplicate the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44568", - "right": "issue:45479", - "accept": false, - "reason": "Different failures: tokenizer special-token handling vs a degenerate zero-loss classification bug." - }, - { - "left": "issue:41628", - "right": "issue:45372", + "left": "issue:32090", + "right": "issue:33290", "accept": false, - "reason": "Both are import errors, but for different symbols and different loading paths (AutoImageProcessor vs ReasoningEffort/Gemma 4)." + "reason": "Different failures: Trainer NoneType broadcast vs deepspeed/adafactor OOM." }, { "left": "issue:43720", "right": "issue:44442", "accept": false, - "reason": "Unrelated problems: unpacking packed weights during accelerate loading vs AutoTokenizer failing on a specific tokenizer class." + "reason": "Unrelated loading problems: BitNet packed-weight loading vs tokenizer class import/load failure." }, { "left": "issue:43643", "right": "issue:44913", "accept": false, - "reason": "Different config bugs: missing fields with trust_remote_code vs rotary_pct not persisting on reload." + "reason": "Both are config-related, but one is missing fields with trust_remote_code and the other is rotary_pct not persisting on reload." }, { - "left": "issue:32090", - "right": "issue:33290", + "left": "issue:41628", + "right": "issue:45372", "accept": false, - "reason": "Both involve Trainer/DeepSpeed, but one is a NoneType broadcast error and the other is OOM with Adafactor." + "reason": "Different import errors for different symbols and different model families." }, { - "left": "issue:44190", - "right": "issue:44496", + "left": "issue:43873", + "right": "issue:44871", "accept": false, - "reason": "Different areas entirely: example script dataset loading vs unrecognized model/config loading." + "reason": "Quantization/offloading behavior and Gemma eos_token_id mismatch are distinct bugs." }, { - "left": "issue:43873", - "right": "issue:44871", + "left": "issue:44190", + "right": "issue:44496", "accept": false, - "reason": "Quantization/offloading behavior is unrelated to Gemma eos_token_id config mismatch." + "reason": "Dataset loading in a training script is unrelated to model config recognition in AutoModel." }, { "left": "issue:34689", "right": "issue:36010", "accept": false, - "reason": "Different import/model-load regressions: Llama 3.2 vision loading vs GenerationMixin import failure." + "reason": "One is a model-loading regression; the other is a missing GenerationMixin import." }, { "left": "issue:43381", "right": "issue:43856", "accept": false, - "reason": "Gradient checkpointing eval-mode restriction is not the same as Qwen3 MoE memory inefficiency." + "reason": "Gradient checkpointing in eval mode and MoE memory usage are different issues." }, { "left": "issue:44855", "right": "issue:45137", "accept": false, - "reason": "A Python 3.13 parsing/IndentationError bug is unrelated to DeepSpeed ZeRO3 deque underflow." + "reason": "Python 3.13 IndentationError in torch.jit parsing is unrelated to DeepSpeed ZeRO3 deque failure." }, { "left": "issue:44077", "right": "issue:44479", "accept": false, - "reason": "Different model/library issues: patchtsmixer post_init validation vs video-input regression across Qwen VL models." + "reason": "Post-init handling for patchtsmixer is unrelated to the Qwen video-input regression." }, { "left": "issue:33290", "right": "issue:45656", "accept": false, - "reason": "Both mention DeepSpeed/optimizers, but one is OOM with Adafactor and the other is double optimizer-step behavior." - }, - { - "left": "issue:43540", - "right": "issue:44514", - "accept": false, - "reason": "Different processor/input bugs: Qwen3OmniMoe video ValueError vs batched apply_chat_template padding crash." + "reason": "Both involve deepspeed/optimizer behavior, but OOM from adafactor is not the same as double optimizer step calls." }, { "left": "issue:29127", "right": "issue:33357", "accept": false, - "reason": "LayoutLMv3 error-message clarity issue is unrelated to a MacOS bus error in CLIP model loading." + "reason": "LayoutLMv3 error-message clarity and MacOS CLIP bus error are unrelated." }, { - "left": "issue:43316", - "right": "issue:44877", + "left": "issue:43540", + "right": "issue:44514", "accept": false, - "reason": "Different config-loading issues affecting different models and validation rules." + "reason": "Both are multimodal/video processor issues, but they fail in different APIs and code paths." }, { "left": "issue:43381", "right": "issue:43742", "accept": false, - "reason": "Gradient checkpointing eval-mode behavior is unrelated to a MobileLLM key error during loading." + "reason": "Gradient checkpointing eval-mode restriction is unrelated to MobileLLM loading key errors." }, { - "left": "issue:42175", - "right": "issue:43316", + "left": "issue:43316", + "right": "issue:44877", "accept": false, - "reason": "One is packaging/backend dependency coverage; the other is a Gemma3TextConfig API mismatch." + "reason": "Gemma3TextConfig API discrepancy and granite_speech strict config loading are different config-schema problems." }, { "left": "issue:33290", "right": "issue:34689", "accept": false, - "reason": "DeepSpeed optimizer OOM and Llama 3.2 vision load regression are not the same bug." + "reason": "Deepspeed OOM and Llama 3.2 vision model loading are not the same bug." }, { - "left": "issue:25251", - "right": "issue:28282", + "left": "issue:42175", + "right": "issue:43316", "accept": false, - "reason": "Pipeline output nesting bug and missing PyTorch ImportError are completely different failure modes." + "reason": "TensorFlow backend packaging and Gemma3TextConfig API mismatch are unrelated." }, { - "left": "issue:44162", - "right": "issue:44792", + "left": "issue:25251", + "right": "issue:28282", "accept": false, - "reason": "ESM2 breakage affecting workflows is unrelated to a janus image-generation test failure." + "reason": "Pipeline top_k nesting behavior has nothing to do with PyTorch import availability." }, { "left": "issue:43404", "right": "issue:43441", "accept": false, - "reason": "One is a tied-weight bug in Mistral3; the other is a FlashAttention regression for Ministral-3." + "reason": "Mistral3 weight tying and Ministral flash-attention failure are separate model-specific regressions." }, { - "left": "issue:43232", - "right": "issue:43388", + "left": "issue:44162", + "right": "issue:44792", "accept": false, - "reason": "Both touch generation/training utilities, but the concrete bugs differ: sync_gpus model-kwargs handling vs metric gathering dropping tuple labels." + "reason": "ESM2 breakage and janus image-generation test failure are different model/test problems." }, { - "left": "issue:41950", - "right": "issue:43475", + "left": "issue:43232", + "right": "issue:43388", "accept": false, - "reason": "Video-classification processor lookup and SAM 3 Video missing attribute are distinct video-model issues." + "reason": "Generation kwargs update after sync_gpus and gather_for_metrics label dropping are different training/eval bugs." }, { "left": "issue:43295", "right": "issue:43825", "accept": false, - "reason": "Custom processor/tokenizer regression is unrelated to pipeline translation-task messaging." + "reason": "Custom processor.tokenizer regression is unrelated to the pipeline translation-task error message." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Video-classification pipeline choosing image processors is not the same as Sam3VisionEncoderOutput missing fpn_position_embeddings." }, { "left": "issue:44871", "right": "issue:45584", "accept": false, - "reason": "Gemma eos_token_id inconsistency and Whisper empty-transcription handling after align_special_tokens are different bugs." + "reason": "Gemma eos_token_id inconsistency and Whisper empty-transcription generation failure are different issues." }, { "left": "issue:43550", "right": "issue:43701", "accept": false, - "reason": "torch.compile/SDPA failure is unrelated to resume_from_checkpoint key mismatch." + "reason": "torch.compile/SDPA failure and resume_from_checkpoint key mismatch are unrelated." }, { "left": "issue:43606", "right": "issue:43701", "accept": false, - "reason": "CPU offload device mismatch is unrelated to resume_from_checkpoint key mismatch." + "reason": "CPU-offload device mismatch and resume_from_checkpoint key mismatch are unrelated." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "LoRA merge collapse with extended vocabulary and mm_token_type_ids defaulting are different model-behavior bugs." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Processor/tokenizer API regression and Gemma eos_token_id mismatch are unrelated." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "Deepspeed/adafactor OOM is not the same as embedding reinitialization after resize." } ] }, @@ -8724,18 +8615,17 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:33290", - "issue:35141", "issue:36331", "issue:41720", "issue:42617", "issue:43065", - "issue:43295", "issue:43335", "issue:43452", "issue:43575", + "issue:43644", "issue:43646", "issue:43673", "issue:43742", @@ -8761,14 +8651,15 @@ "issue:44792", "issue:44855", "issue:44871", + "issue:44877", "issue:44913", "issue:44936", "issue:44964", "issue:44977", "issue:45127", - "issue:45200", "issue:45245", "issue:45290", + "issue:45357", "issue:45381", "issue:45584", "issue:45588", @@ -8776,245 +8667,245 @@ "issue:45698" ], "soft_pairs": [ - "issue:45127|issue:45200", - "issue:43295|issue:44871", - "issue:33290|issue:35141", "issue:43761|issue:44479", "issue:33290|issue:36331", + "issue:43575|issue:43742", "issue:43065|issue:44617", "issue:43452|issue:44496", - "issue:43575|issue:43742", - "issue:43673|issue:44743", "issue:42617|issue:43575", "issue:43646|issue:44913", + "issue:43673|issue:44743", "issue:44936|issue:45698", "issue:43761|issue:44242", "issue:43575|issue:43873", "issue:44077|issue:44534", "issue:44479|issue:44734", - "issue:43931|issue:44060", - "issue:44617|issue:45584", "issue:44464|issue:44871", + "issue:44617|issue:45584", + "issue:43931|issue:44060", "issue:43065|issue:44610", "issue:43906|issue:45290", - "issue:44792|issue:44977", "issue:43827|issue:43994", + "issue:44792|issue:44977", "issue:44534|issue:44913", "issue:41720|issue:45245", "issue:44190|issue:44964", "issue:44792|issue:45600", "issue:44855|issue:45588", "issue:44521|issue:45381", - "issue:43335|issue:44855" + "issue:43335|issue:44855", + "issue:45127|issue:45357", + "issue:44077|issue:44877", + "issue:43644|issue:44913" ], - "prepared_review_unit_hash": "46dd9a3dd50d17b435b41420c31b54da3bb60c8f2099306562bf51738f474a3b", + "prepared_review_unit_hash": "b6bceb6b257993b2adcf6bc2f3710f4c9dda6e4740b429e1cba036ff94ec0d28", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36751, - "estimated_input_tokens": 9188, - "estimated_eval_tokens": 18632 + "serialized_chars": 36638, + "estimated_input_tokens": 9160, + "estimated_eval_tokens": 18576 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are separate model-specific regressions, loading errors, or docs/test failures rather than one duplicated bug. I found no soft-edge pair that clearly looks like the same underlying change or defect.", - "confidence": 0.32, - "canonical_issue_reason": "issue:43646 is the broadest, most umbrella-like report in the set (\u201cTransformers 5.0.0 breaks custom model initialization\u201d) and is the closest representative of the cluster\u2019s recurring v5/custom-model loading theme, though it is not a true duplicate of the others.", + "summary": "This cluster is heterogeneous: the issues cover unrelated bugs across training, loading, generation, multimodal inputs, configs, and docs. I do not see any true duplicate pairs among the soft-edge candidates, so all soft edges are rejected. No PRs are present in the cluster.", + "confidence": 0.91, + "canonical_issue_reason": "No single issue cleanly represents the cluster because the items are not one underlying bug. If forced to pick the broadest/most central metadata-wise, issue 43873 has the most discussion and inbound references, but it still does not match the rest of the items.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43646 is the best global representative because it captures the highest-level regression theme and is broader than the narrower model-specific or API-specific reports.", + "best_issue_reason": "Issue 43873 is the most central by activity and inbound links, so it is the best available anchor only from a triage/metadata standpoint. It is not a good semantic canonical for the cluster because the surrounding issues are unrelated.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:45127", - "right": "issue:45200", + "left": "issue:43761", + "right": "issue:44479", "accept": false, - "reason": "Different bugs: LoRA/extended-vocab collapse vs Gemma 4 multimodal token-type defaults." + "reason": "Both are model-generation regressions in broad terms, but they target different model families and different failure modes." }, { - "left": "issue:43295", - "right": "issue:44871", + "left": "issue:33290", + "right": "issue:36331", "accept": false, - "reason": "One is a processor/tokenizer regression; the other is an eos_token_id config mismatch." + "reason": "OOM in DeepSpeed/Adafactor is unrelated to the custom compute_loss signature TypeError." }, { - "left": "issue:33290", - "right": "issue:35141", - "accept": false, - "reason": "OOM with Adafactor+DeepSpeed is unrelated to embedding reinitialization after resize." - }, - { - "left": "issue:36331", - "right": "issue:33290", + "left": "issue:43575", + "right": "issue:43742", "accept": false, - "reason": "Trainer API signature break vs optimizer memory blow-up; different code paths." + "reason": "One is tensor-parallel OOM on a large Qwen model; the other is a config/key error when loading MobileLLM." }, { "left": "issue:43065", "right": "issue:44617", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder is unrelated to Sam3Video CUDA OOM." + "reason": "Same Sam3 family, but one is a dummy Conv2d concern and the other is a CUDA OOM in Sam3Video." }, { "left": "issue:43452", "right": "issue:44496", "accept": false, - "reason": "GGUF loader breakage is a different issue from unrecognized model_type in a specific checkpoint." - }, - { - "left": "issue:43575", - "right": "issue:43742", - "accept": false, - "reason": "Tensor-parallel OOM on a large model vs key error loading a small MobileLLM model." + "reason": "Both are loading failures, but one is about gguf_file handling and the other is about an unrecognized model/config mismatch." }, { "left": "issue:42617", "right": "issue:43575", "accept": false, - "reason": "3d_parallel execution failure is not the same as Qwen2-57B tensor-parallel OOM." + "reason": "These are different runtime failures with different entrypoints and symptoms; not the same bug." }, { "left": "issue:43646", "right": "issue:44913", "accept": false, - "reason": "Generic custom-model initialization regression vs GPTNeoX rotary_pct reload defaulting; different defect." + "reason": "Buffer initialization junk and GPTNeoX rotary_pct reload behavior are distinct config/runtime issues." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Both mention cache behavior, but they affect different code paths and model implementations." }, { "left": "issue:44936", "right": "issue:45698", "accept": false, - "reason": "train/evaluate lifecycle failure vs wrong custom module loading after save_pretrained." + "reason": "Trainer lifecycle failure is unrelated to loading the wrong custom module after save_pretrained." }, { "left": "issue:43761", "right": "issue:44242", "accept": false, - "reason": "CLIP hidden_states regression is unrelated to MoE router loss handling." + "reason": "A CLIPVision hidden-states regression is not the same as missing load-balancing loss in MoE routing." }, { "left": "issue:43575", "right": "issue:43873", "accept": false, - "reason": "OOM from tensor parallelism is not the same as quantized offloading behavior." + "reason": "OOM on a specific model load is not the same as quantization/offloading not working as expected." }, { "left": "issue:44077", "right": "issue:44534", "accept": false, - "reason": "PatchTSMixer post_init policy and non-persistent buffer junk are separate v5 problems." + "reason": "A patchtsmixer post_init constraint issue is unrelated to the non-persistent buffer junk regression." }, { "left": "issue:44479", "right": "issue:44734", "accept": false, - "reason": "Video-input regression for Qwen VL models is unrelated to serve KV-cache indexing crash." + "reason": "Both are generation-related regressions, but they involve different subsystems and different concrete bugs." }, { - "left": "issue:43931", - "right": "issue:44060", + "left": "issue:44464", + "right": "issue:44871", "accept": false, - "reason": "Weight-shape mismatch in Qwen3-VL loading is different from a tied-weights warning in Qwen3-Next." + "reason": "Chunked generation inconsistency and eos_token_id config mismatch are unrelated problems." }, { "left": "issue:44617", "right": "issue:45584", "accept": false, - "reason": "Sam3Video OOM is unrelated to Whisper empty-transcription generation failure." + "reason": "Sam3Video OOM and Whisper empty-transcription generation failure are unrelated." }, { - "left": "issue:44464", - "right": "issue:44871", + "left": "issue:43931", + "right": "issue:44060", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency is not the same as eos_token_id config inconsistency." + "reason": "Shape mismatch on loading Qwen3-VL is not the same as an incorrect tied-weights warning in Qwen3-Next." }, { "left": "issue:43065", "right": "issue:44610", "accept": false, - "reason": "Sam3 pixel decoder dummy layer and OmDet processor resolution mismatch are different model bugs." + "reason": "Both touch vision/model components, but one concerns a dummy Conv2d and the other an input-size mismatch in OmDet-Turbo." }, { "left": "issue:43906", "right": "issue:45290", "accept": false, - "reason": "Isolated reproduction of an earlier issue is not the same as tool-call tokenization crash." + "reason": "An isolated reproduction of an older issue is not the same as a chat template crash on tool-call assistant messages." }, { - "left": "issue:44792", - "right": "issue:44977", + "left": "issue:43827", + "right": "issue:43994", "accept": false, - "reason": "Janus image-generation test failure is unrelated to Qwen3.5 flash-attention generation behavior." + "reason": "Docs still referencing pipeline() is not the same bug as SigLIP2 producing nonsensical outputs with pipeline." }, { - "left": "issue:43827", - "right": "issue:43994", + "left": "issue:44792", + "right": "issue:44977", "accept": false, - "reason": "Docs still referencing pipeline() is not the same as a bad model output bug in SigLIP2." + "reason": "A failing Janus image-generation test and a Qwen3.5 flash-attention generation bug are different issues." }, { "left": "issue:44534", "right": "issue:44913", "accept": false, - "reason": "Non-persistent buffer initialization and GPTNeoX config reload defaulting are distinct defects." + "reason": "The non-persistent buffer regression and GPTNeoXConfig reload behavior do not share the same root cause." }, { "left": "issue:41720", "right": "issue:45245", "accept": false, - "reason": "Qwen3 auto device mapping CUDA assert is unrelated to the category-cardinality runtime error." + "reason": "CUDA assert on A800 with auto device mapping is unrelated to the category-count runtime limit." }, { "left": "issue:44190", "right": "issue:44964", "accept": false, - "reason": "Local dataset loading failure in a training script is different from loading Phi-4 multimodal model failure." + "reason": "Local dataset loading in a training script is unrelated to loading a multimodal Phi-4 model." }, { "left": "issue:44792", "right": "issue:45600", "accept": false, - "reason": "Janus generate-image test failure and removed Sam3LiteText auto_mappings references are unrelated." + "reason": "A Janus model test failure is unrelated to stale auto_mappings references to removed configs." }, { "left": "issue:44855", "right": "issue:45588", "accept": false, - "reason": "Python 3.13 torch.jit parsing error is unrelated to flash_attention s_aux=None crash." + "reason": "Python 3.13 import/parsing failure in DebertaV2 is unrelated to flash_attention crashing on s_aux=None." }, { "left": "issue:44521", "right": "issue:45381", "accept": false, - "reason": "All-zero assistant masks in multimodal chat templating is different from wrong vision_position_ids in Qwen2.5-VL video input." + "reason": "Both are multimodal/template issues, but one is assistant-mask generation and the other is video vision_position_ids." }, { - "left": "issue:43673", - "right": "issue:44743", + "left": "issue:43335", + "right": "issue:44855", "accept": false, - "reason": "Missing GenerationMixin cache in v5 chunked_prefill and Qwen3.5 recurrent-state reset are not the same code path." + "reason": "A SwitchTransformers sparse-layer bug is unrelated to the DebertaV2 import/parsing failure." }, { - "left": "issue:43873", - "right": "issue:44521", + "left": "issue:45127", + "right": "issue:45357", "accept": false, - "reason": "Quantized offloading behavior and multimodal assistant-mask generation are unrelated." + "reason": "LoRA merge collapse with extended vocabulary is unrelated to incorrect visual encoder keys saved by Qwen3.5." }, { - "left": "issue:44936", - "right": "issue:45698", + "left": "issue:44077", + "right": "issue:44877", "accept": false, - "reason": "Evaluator failure after training and wrong custom module loading after save_pretrained are separate lifecycle bugs." + "reason": "PatchTSMixer post_init policy and strict config loading for granite_speech are different problems." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer junk and rotary_pct reverting on reload are different regressions with different scopes." } ] }, @@ -9036,12 +8927,13 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:36246", "issue:38175", "issue:38617", "issue:41628", + "issue:41950", "issue:43065", "issue:43116", "issue:43295", @@ -9050,11 +8942,11 @@ "issue:43479", "issue:43540", "issue:43550", - "issue:43644", + "issue:43643", "issue:43650", "issue:43761", "issue:43976", - "issue:44077", + "issue:44016", "issue:44112", "issue:44186", "issue:44190", @@ -9062,23 +8954,23 @@ "issue:44246", "issue:44464", "issue:44466", + "issue:44496", "issue:44561", "issue:44683", "issue:44734", "issue:44743", "issue:44855", - "issue:44877", "issue:44913", "issue:44933", "issue:44938", + "issue:44964", "issue:45003", - "issue:45127", "issue:45216", "issue:45245", "issue:45276", - "issue:45357", "issue:45362", "issue:45375", + "issue:45406", "issue:45446", "issue:45542", "issue:45563", @@ -9087,231 +8979,245 @@ "issue:45632" ], "soft_pairs": [ - "issue:45127|issue:45357", "issue:44938|issue:45542", - "issue:44077|issue:44877", - "issue:43644|issue:44913", "issue:43976|issue:44855", "issue:43550|issue:44190", - "issue:43540|issue:44734", "issue:43761|issue:44743", + "issue:43540|issue:44734", "issue:45276|issue:45584", "issue:44112|issue:45600", "issue:43650|issue:44220", "issue:43295|issue:45276", "issue:43335|issue:44190", "issue:43116|issue:44190", - "issue:36246|issue:41628", "issue:45245|issue:45584", "issue:43404|issue:45216", "issue:43479|issue:45375", + "issue:36246|issue:41628", + "issue:44186|issue:44734", "issue:44933|issue:45362", "issue:44464|issue:45563", - "issue:44186|issue:44734", "issue:38175|issue:43065", "issue:44466|issue:44913", "issue:44561|issue:45632", - "issue:44734|issue:45632", "issue:44683|issue:45446", + "issue:44734|issue:45632", "issue:44246|issue:45003", - "issue:36246|issue:38617" + "issue:43643|issue:44496", + "issue:36246|issue:38617", + "issue:45406|issue:45632", + "issue:45563|issue:45584", + "issue:44964|issue:45600", + "issue:41950|issue:44016" ], - "prepared_review_unit_hash": "559a19e5c3cbe98ee568182332248037fedf67af4b37d681b86a8d18b298ee3d", + "prepared_review_unit_hash": "a27762a12c0f13f2ea06dc6fe543b38379a92261f006f7156e94a55556253557", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 28, - "serialized_chars": 35563, - "estimated_input_tokens": 8891, - "estimated_eval_tokens": 18038 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36803, + "estimated_input_tokens": 9201, + "estimated_eval_tokens": 18658 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is very heterogeneous: most items are unrelated regressions, import errors, config bugs, or generation issues. None of the suggested soft edges look like the same underlying bug/change, so they should all stay separate.", - "confidence": 0.25, - "canonical_issue_reason": "Issue 44246 is the broadest and most discussion-heavy item in the set, so it is the best single representative if one must be chosen. That said, the overall cluster does not look like a true duplicate group.", + "summary": "This cluster is a loose collection of mostly unrelated Transformers issues (imports, config loading, generation, CI, and model-specific bugs). I don\u2019t see any soft edge pair that is the same underlying bug/change, so none should be merged.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue really fits: the cluster is heterogeneous and the soft pairs span different models, code paths, and failure modes. If forced to pick one representative issue, 44561 is the broadest regression-style report, but it does not subsume the rest.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44246 has the widest scope ('import transformers takes long sometimes') and the strongest engagement, making it the most general anchor for this mostly unrelated set.", + "best_issue_reason": "issue:44561 is the most broadly scoped, systemic regression report in the set (trust_remote_code models breaking due to a removed helper) and is the closest thing to a cluster anchor, though the cluster is not actually cohesive.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45127", - "right": "issue:45357", - "accept": false, - "reason": "Both involve Qwen3.5-related behavior, but one is about LoRA/weight tying collapse and the other about save_pretrained visual encoder keys; different failure modes and code paths." - }, { "left": "issue:44938", "right": "issue:45542", "accept": false, - "reason": "One is a Python 3.14 load failure, the other is a TensorBoard/TensorFlow backend mismatch; unrelated bugs." - }, - { - "left": "issue:44077", - "right": "issue:44877", - "accept": false, - "reason": "Both touch config validation, but they affect different models and different rules; not the same bug." - }, - { - "left": "issue:43644", - "right": "issue:44913", - "accept": false, - "reason": "One is about non-persistent buffers being filled with junk in v5, the other about a config value not persisting on reload; different mechanisms." + "reason": "Different failures: Python import/load problem vs tensorboard-without-TensorFlow backend error." }, { "left": "issue:43976", "right": "issue:44855", "accept": false, - "reason": "Both are version/platform breakages, but one is a broad Python support regression and the other is a specific DebertaV2 import parsing failure on Python 3.13." + "reason": "Both involve Python versions, but one is a package compatibility/load issue and the other is a Python 3.13 parsing/IndentationError in a model import path." }, { "left": "issue:43550", "right": "issue:44190", "accept": false, - "reason": "A torch.compile/SDPA model bug and a dataset-loading example-script issue are unrelated." + "reason": "Unrelated subsystems: torch.compile/SDPA on Bamba vs local dataset loading in an image classification example." }, { - "left": "issue:43540", - "right": "issue:44734", + "left": "issue:43761", + "right": "issue:44743", "accept": false, - "reason": "Different domains and symptoms: Qwen3Omni video processing versus serve KV-cache indexing." + "reason": "Both are state-related regressions, but they affect different models and different mechanisms (hidden states vs recurrent cache state)." }, { - "left": "issue:43761", - "right": "issue:44743", + "left": "issue:43540", + "right": "issue:44734", "accept": false, - "reason": "One is CLIPVisionModel hidden-state output, the other is recurrent-state reset in Qwen3.5 code; not the same path." + "reason": "Video input validation in Qwen3OmniMoe is unrelated to KV-cache tensor indexing in transformers serve." }, { "left": "issue:45276", "right": "issue:45584", "accept": false, - "reason": "Token embedding resize behavior and Whisper empty-transcription generation are unrelated." + "reason": "Gemma4 embedding resize bug and Whisper empty-transcription generation failure are distinct code paths." }, { "left": "issue:44112", "right": "issue:45600", "accept": false, - "reason": "Both mention CI, but one is a GraniteSpeech stale test and the other is an auto_mappings import/reference cleanup issue." + "reason": "Both touch CI, but one is a stale GraniteSpeech test and the other is a broken auto_mappings reference to removed configs." }, { "left": "issue:43650", "right": "issue:44220", "accept": false, - "reason": "No common bug: one is a placeholder issue, the other is a feature-specific fbank feature extraction problem." + "reason": "No overlap beyond being issue reports; one is placeholder text, the other is an audio feature extraction bug." }, { "left": "issue:43295", "right": "issue:45276", "accept": false, - "reason": "A processor/tokenizer regression and a token-embedding resize bug are different issues." + "reason": "Custom processor/tokenizer regression is unrelated to Gemma4 token embedding resize behavior." }, { "left": "issue:43335", "right": "issue:44190", "accept": false, - "reason": "SwitchTransformers sparse-layer config creation is unrelated to the image-classification local dataset loading bug." + "reason": "SwitchTransformers sparse-layer config bug and image-classification dataset loading are different bugs." }, { "left": "issue:43116", "right": "issue:44190", "accept": false, - "reason": "Example-script multi-label classification output handling is unrelated to local dataset loading." - }, - { - "left": "issue:36246", - "right": "issue:41628", - "accept": false, - "reason": "Both are import errors, but for different symbols and different submodules; not the same underlying defect." + "reason": "Example script multi-label output issue is not the same as local dataset loading in a separate training script." }, { "left": "issue:45245", "right": "issue:45584", "accept": false, - "reason": "A category-cardinality runtime limit and Whisper empty-transcription handling are unrelated." + "reason": "Different failures: category-count runtime limit vs Whisper empty-transcription handling." }, { "left": "issue:43404", "right": "issue:45216", "accept": false, - "reason": "Both involve model save/tied-weight behavior, but they affect different models and different serialization paths." + "reason": "Both mention tied weights/serialization, but they involve different models and different failure modes." }, { "left": "issue:43479", "right": "issue:45375", "accept": false, - "reason": "Both are multimodal config bugs, but one is default init with None values and the other is a missing field dropped by strict config handling." + "reason": "Different multimodal config bugs: default init when None vs a missing deepstack_visual_indexes field being dropped." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they are for different symbols and different import surfaces." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "Tokenizer padding/truncation crash is unrelated to serve KV-cache continuation indexing." }, { "left": "issue:44933", "right": "issue:45362", "accept": false, - "reason": "An image_utils import problem and a Qwen3.5 chat crash are unrelated." + "reason": "Missing image_utils import is unrelated to Qwen3.5 chat crashes." }, { "left": "issue:44464", "right": "issue:45563", "accept": false, - "reason": "Compiled-forward generation inconsistency is not the same as a stale warning about num_return_sequences." - }, - { - "left": "issue:44186", - "right": "issue:44734", - "accept": false, - "reason": "Tokenizer padding/truncation crashes and server KV-cache indexing crashes are different code paths." + "reason": "Chunked generation inconsistency under compilation and a stale warning about num_return_sequences are not the same bug." }, { "left": "issue:38175", "right": "issue:43065", "accept": false, - "reason": "A SigLIP probability issue and a Sam3PixelDecoder dummy Conv2d issue are unrelated." + "reason": "Unexpected zero probabilities in a SigLIP model has nothing to do with Sam3PixelDecoder's dummy Conv2d." }, { "left": "issue:44466", "right": "issue:44913", "accept": false, - "reason": "Both mention serialization/persistence, but one is tied lm_head weight serialization and the other is a config field reload default." + "reason": "One is inconsistent lm_head.weight serialization by device; the other is rotary_pct silently reverting on reload." }, { "left": "issue:44561", "right": "issue:45632", "accept": false, - "reason": "Both mention trust_remote_code, but one is a removed helper API and the other is a cache-path collision." + "reason": "Both involve trust_remote_code, but one is a removed helper breaking remote models and the other is a cache-path collision for local models." }, { - "left": "issue:44734", - "right": "issue:45632", + "left": "issue:44683", + "right": "issue:45446", "accept": false, - "reason": "Server KV-cache continuation failure and trust_remote_code cache-path collision are different problems." + "reason": "Both touch flex_attention, but one is a torch>=2.9 compiled-fail issue and the other is an incorrect version check for AuxRequest import." }, { - "left": "issue:44683", - "right": "issue:45446", + "left": "issue:44734", + "right": "issue:45632", "accept": false, - "reason": "Both are flex_attention-related, but one is a compiled-kernel failure on new torch versions and the other is a version-check/import guard bug." + "reason": "Serve KV-cache tensor indexing and trust_remote_code cache-path collisions are unrelated problems." }, { "left": "issue:44246", "right": "issue:45003", "accept": false, - "reason": "Import performance and unsafe sys.modules access are not the same underlying issue, even if both relate to import-time behavior." + "reason": "Import slowdown and unsafe sys.modules access may be adjacent, but they are not the same concrete bug." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "Missing fields from trust_remote_code configs and unrecognized model_type in config.json are different config-loading failures." }, { "left": "issue:36246", "right": "issue:38617", "accept": false, - "reason": "Separate import errors for different missing names; no evidence they share a root cause." + "reason": "Both are missing-import reports, but they refer to different symbols and different fixes." + }, + { + "left": "issue:45406", + "right": "issue:45632", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer is unrelated to trust_remote_code cache-path collisions." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "A stale warning and a generation failure after align_special_tokens are different issues." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "Phi-4 multimodal model loading failure and stale auto_mappings references to removed configs do not look like the same root cause." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "Pipeline misrouting to image processors is unrelated to a notebook syntax error." } ] }, @@ -9333,7 +9239,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:41950", "issue:42175", @@ -9348,9 +9254,9 @@ "issue:43526", "issue:43531", "issue:43577", - "issue:43643", "issue:43673", "issue:43688", + "issue:43824", "issue:43827", "issue:43854", "issue:43901", @@ -9365,271 +9271,250 @@ "issue:44336", "issue:44351", "issue:44368", + "issue:44462", "issue:44496", + "issue:44933", "issue:44964", "issue:45020", "issue:45042", + "issue:45081", "issue:45200", "issue:45216", "issue:45278", "issue:45335", "issue:45357", "issue:45405", - "issue:45406", - "issue:45563", "issue:45584", "issue:45593", "issue:45600", - "issue:45632", "issue:45759" ], "soft_pairs": [ - "issue:43643|issue:44496", - "issue:45406|issue:45632", - "issue:45563|issue:45584", - "issue:41950|issue:44016", - "issue:44964|issue:45600", - "issue:43295|issue:45335", "issue:43388|issue:43688", + "issue:43295|issue:45335", "issue:44112|issue:44964", "issue:42907|issue:43441", + "issue:43854|issue:44964", "issue:44496|issue:45042", "issue:43673|issue:44336", - "issue:43854|issue:44964", - "issue:44016|issue:44220", - "issue:42175|issue:44279", - "issue:43299|issue:45216", "issue:43531|issue:45600", + "issue:43299|issue:45216", + "issue:44016|issue:44220", "issue:43526|issue:43901", + "issue:42175|issue:44279", "issue:45335|issue:45584", - "issue:44075|issue:44368", "issue:43381|issue:45593", + "issue:44075|issue:44368", "issue:43299|issue:45357", "issue:44496|issue:45759", "issue:43901|issue:44496", - "issue:41950|issue:45020", "issue:44242|issue:45200", "issue:43577|issue:45216", + "issue:41950|issue:45020", "issue:44016|issue:44062", - "issue:44351|issue:45278", "issue:43976|issue:45405", "issue:43388|issue:43827", - "issue:43479|issue:43525" + "issue:44351|issue:45278", + "issue:43479|issue:43525", + "issue:44462|issue:45081", + "issue:43824|issue:44933" ], - "prepared_review_unit_hash": "c0ae01ce820f362e3f05338193b8496641cc9bb7b31a9ab211e514f17123ae29", + "prepared_review_unit_hash": "484124b200f6079599e090c5de4e43f90388f07c04c6e06e4b84e2d92710ce74", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 31, - "serialized_chars": 36467, - "estimated_input_tokens": 9117, - "estimated_eval_tokens": 18490 + "soft_pair_count": 28, + "serialized_chars": 35575, + "estimated_input_tokens": 8894, + "estimated_eval_tokens": 18044 }, "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 bug reports, with some superficial similarity around model-loading regressions and config handling. I would not merge any of the soft-edge pairs; they point to different concrete code paths or even different artifact types (docs, tests, notebooks).", - "confidence": 0.84, - "canonical_issue_reason": "issue:45020 is the broadest umbrella here: it is open, recent, and centered on a class of `remote_code`/model-loading regressions that overlaps several other reports more than any narrow model-specific issue.", + "summary": "These items are mostly separate regressions across different models, loaders, and utilities; they share a broad Transformers-v5 regression theme but not the same underlying bug. I would not merge any of the suggested soft pairs.", + "confidence": 0.9, + "canonical_issue_reason": "issue:45020 is the broadest open umbrella for recent-version regressions breaking remote_code model loading, so it is the best representative issue for this mixed cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the best cluster anchor because it is the most general, still-active issue and can reasonably absorb related loader regressions without overfitting to one model family.", + "best_issue_reason": "issue:45020 best captures the cluster\u2019s common theme without being overly narrow to one model family or code path.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43643", - "right": "issue:44496", - "accept": false, - "reason": "`trust_remote_code` missing fields in AutoConfig vs unrecognized model/config type; different config-loading failure modes." - }, - { - "left": "issue:45406", - "right": "issue:45632", - "accept": false, - "reason": "Gemma4Processor missing `_tokenizer` vs trust_remote_code cache-path collision; unrelated symptoms and fixes." - }, - { - "left": "issue:45563", - "right": "issue:45584", - "accept": false, - "reason": "Stale `num_return_sequences` warning vs empty-transcription failure after token alignment; different generation bugs." - }, - { - "left": "issue:41950", - "right": "issue:44016", - "accept": false, - "reason": "Video-classification image-processor lookup vs notebook syntax error; not the same underlying problem." - }, - { - "left": "issue:44964", - "right": "issue:45600", + "left": "issue:43388", + "right": "issue:43688", "accept": false, - "reason": "Phi-4 multimodal load failure vs CI break from removed Sam3LiteText config references; distinct code paths." + "reason": "Different bugs: one drops label elements in distributed metric gathering; the other normalizes auxiliary loss for MoE models." }, { "left": "issue:43295", "right": "issue:45335", "accept": false, - "reason": "Processor/tokenizer regression vs `resize_token_embeddings` not updating `decoder.embed_tokens`; different model logic." - }, - { - "left": "issue:43388", - "right": "issue:43688", - "accept": false, - "reason": "Metric gathering drops tuple labels vs auxiliary-loss normalization; unrelated training/eval bugs." + "reason": "Both involve regressions, but one is a processor/tokenizer API break and the other is resize_token_embeddings not updating decoder embeddings." }, { "left": "issue:44112", "right": "issue:44964", "accept": false, - "reason": "Stale GraniteSpeech CI test vs Phi-4 multimodal loading issue; different artifacts and fixes." + "reason": "CI stale test failure is unrelated to loading Phi-4 multimodal with latest transformers." }, { "left": "issue:42907", "right": "issue:43441", "accept": false, - "reason": "Dequantized save failure vs FlashAttention load failure; different model-specific breakages." + "reason": "Different model families and failures: saving dequantized Ministral/Devstral vs FlashAttention load failure for Ministral-3." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Same general family only; one is a unit-test loading issue for GLM-4.7-Flash, the other is a Phi-4 multimodal load regression." }, { "left": "issue:44496", "right": "issue:45042", "accept": false, - "reason": "Unrecognized model/config error vs PIL backend incorrectly requiring torchvision; separate subsystems." + "reason": "Different problems: unrecognized config/model type vs PIL image processors incorrectly requiring torchvision." }, { "left": "issue:43673", "right": "issue:44336", "accept": false, - "reason": "Generation cache missing in chunked prefill vs ANSI codes in loading report; unrelated functionality." - }, - { - "left": "issue:43854", - "right": "issue:44964", - "accept": false, - "reason": "Both are model-load failures, but for different models and root causes; not mergeable as one bug." + "reason": "Generation cache missing in chunked prefill is unrelated to ANSI codes emitted by loading_report." }, { - "left": "issue:44016", - "right": "issue:44220", - "accept": false, - "reason": "Notebook syntax error vs `_torch_extract_fbank_features()` issue; different domains entirely." - }, - { - "left": "issue:42175", - "right": "issue:44279", + "left": "issue:43531", + "right": "issue:45600", "accept": false, - "reason": "TensorFlow missing from `.[torch]` install vs vague dependency issue; not the same concrete dependency bug." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to stale auto_mappings references breaking CI." }, { "left": "issue:43299", "right": "issue:45216", "accept": false, - "reason": "Qwen3VL load failure vs Qwen3.5 save_pretrained regression; same family, but different phase and code path." + "reason": "Both are Qwen-related regressions, but one is loading MoE models while the other is incorrect save_pretrained output for Qwen3.5." }, { - "left": "issue:43531", - "right": "issue:45600", + "left": "issue:44016", + "right": "issue:44220", "accept": false, - "reason": "Qwen3-MoE sliding-window bug vs Sam3LiteText auto-mapping CI break; unrelated issues." + "reason": "Notebook syntax error and _torch_extract_fbank_features() are unrelated bugs." }, { "left": "issue:43526", "right": "issue:43901", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug vs TextClassificationPipeline docs mismatch; runtime bug vs documentation." + "reason": "BEiT image processor label reduction bug is unrelated to TextClassificationPipeline docs mentioning return_all_scores." }, { - "left": "issue:45335", - "right": "issue:45584", + "left": "issue:42175", + "right": "issue:44279", "accept": false, - "reason": "Embedding resize bug vs Whisper empty transcription after align_special_tokens; unrelated model behaviors." + "reason": "Packaging/backend dependency issue for pip install '.[torch]' is not the same as a generic dependency issue with transformers." }, { - "left": "issue:44075", - "right": "issue:44368", + "left": "issue:45335", + "right": "issue:45584", "accept": false, - "reason": "SGD args ignored vs Qwen3.5 tie_word_embeddings warning; different problems." + "reason": "resize_token_embeddings affecting decoder.embed_tokens is unrelated to Whisper empty-transcription failure after align_special_tokens." }, { "left": "issue:43381", "right": "issue:45593", "accept": false, - "reason": "Gradient checkpointing eval-mode restriction vs D-FINE auxiliary-loss omission; different training behaviors." + "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary losses are different training/evaluation bugs." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "Optimizer SGD args ignored is unrelated to a tie_word_embeddings warning in Qwen3.5 LoRA fine-tuning." }, { "left": "issue:43299", "right": "issue:45357", "accept": false, - "reason": "Both are Qwen3.5-related, but one is loading and the other is save_pretrained serialization; different bugs." + "reason": "Both are Qwen3.5-related, but one concerns model loading and the other incorrect visual encoder keys in save_pretrained." }, { "left": "issue:44496", "right": "issue:45759", "accept": false, - "reason": "Unrecognized model/config error vs missing `text_config` unwrap in `from_config`; related area, but distinct failures." + "reason": "Unrecognized model config and AutoModelForCausalLM.from_config not unwrapping text_config are distinct config-resolution bugs." }, { "left": "issue:43901", "right": "issue:44496", "accept": false, - "reason": "Docs mention `return_all_scores` vs model/config loading error; unrelated docs vs runtime bug." - }, - { - "left": "issue:41950", - "right": "issue:45020", - "accept": false, - "reason": "Video-classification processor lookup bug is unrelated to the broad remote_code regression reports." + "reason": "Docs mentioning return_all_scores is unrelated to AutoModel config recognition failure." }, { "left": "issue:44242", "right": "issue:45200", "accept": false, - "reason": "MoE load-balancing loss bug vs Gemma 4 multimodal token-id defaulting; different models and code paths." + "reason": "Load-balancing loss omission and Gemma 4 mm_token_type_ids defaulting are different model-training issues." }, { "left": "issue:43577", "right": "issue:45216", "accept": false, - "reason": "BLIP-2 dtype propagation bug vs Qwen3.5 save_pretrained regression; unrelated." + "reason": "Blip2 dtype handling and Qwen3.5 save_pretrained regression are unrelated." }, { - "left": "issue:44016", - "right": "issue:44062", + "left": "issue:41950", + "right": "issue:45020", "accept": false, - "reason": "Notebook syntax error vs `AddedToken()` argument conflict; not the same bug." + "reason": "Video-classification image-processor lookup is a specific pipeline bug, not the broader remote_code breakage." }, { - "left": "issue:44351", - "right": "issue:45278", + "left": "issue:44016", + "right": "issue:44062", "accept": false, - "reason": "HybridCache import failure vs many import errors after upgrade; different specific regressions." + "reason": "Notebook syntax error and AddedToken keyword conflict are unrelated." }, { "left": "issue:43976", "right": "issue:45405", "accept": false, - "reason": "Python version support mismatch vs PEFT minimum-version bump; both dependency-related but not one bug." + "reason": "Python version compatibility and an unreleased MIN_PEFT_VERSION bump are different release issues." }, { "left": "issue:43388", "right": "issue:43827", "accept": false, - "reason": "Metric label dropping vs docs still referencing removed `pipeline()`; unrelated problem classes." + "reason": "gather_for_metrics label handling is unrelated to docs still referencing removed pipeline() APIs." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "HybridCache import failure is not the same as broad import errors after upgrading to 5.5.0." }, { "left": "issue:43479", "right": "issue:43525", "accept": false, - "reason": "Phi-4 multimodal default-config bug vs Llama4Config missing `pad_token_id`; separate config classes and failures." + "reason": "Phi4MultimodalConfig default init bug and Llama4Config missing pad_token_id are different config-class issues." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json is unrelated to _patch_mistral_regex crashing on a Tokenizer object." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Missing Qwen2_5_VL export and nonexistent image_utils import are separate import/export problems." } ] }, @@ -9651,7 +9536,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:32090", "issue:35141", @@ -9660,6 +9545,7 @@ "issue:41950", "issue:42994", "issue:43065", + "issue:43122", "issue:43295", "issue:43299", "issue:43317", @@ -9669,10 +9555,10 @@ "issue:43653", "issue:43723", "issue:43784", - "issue:43824", "issue:43854", "issue:43901", "issue:43976", + "issue:43994", "issue:44162", "issue:44261", "issue:44263", @@ -9689,9 +9575,7 @@ "issue:44749", "issue:44779", "issue:44898", - "issue:44933", "issue:44987", - "issue:45081", "issue:45276", "issue:45290", "issue:45325", @@ -9703,26 +9587,24 @@ "issue:45712" ], "soft_pairs": [ - "issue:43824|issue:44933", - "issue:44462|issue:45081", + "issue:32090|issue:35141", "issue:44683|issue:45542", "issue:44279|issue:45405", - "issue:32090|issue:35141", - "issue:43299|issue:43854", - "issue:43295|issue:44779", "issue:39692|issue:43901", + "issue:43295|issue:44779", + "issue:43299|issue:43854", "issue:44297|issue:44462", "issue:38175|issue:44617", "issue:41950|issue:43650", - "issue:43976|issue:44279", "issue:44462|issue:44779", "issue:44521|issue:45325", + "issue:43976|issue:44279", "issue:43976|issue:44987", "issue:45600|issue:45712", "issue:43450|issue:44279", - "issue:42994|issue:43317", "issue:44466|issue:45276", "issue:43065|issue:44898", + "issue:42994|issue:43317", "issue:44261|issue:44263", "issue:43650|issue:45290", "issue:43441|issue:44683", @@ -9730,225 +9612,227 @@ "issue:43723|issue:44162", "issue:44623|issue:44749", "issue:44671|issue:44779", - "issue:43784|issue:45542", "issue:43653|issue:45593", - "issue:44987|issue:45542", - "issue:44361|issue:45593" + "issue:43784|issue:45542", + "issue:44361|issue:45593", + "issue:43901|issue:43994", + "issue:43122|issue:43650", + "issue:44987|issue:45542" ], - "prepared_review_unit_hash": "b98feeb6daeb0977ad57edc85ac51d475173a3ac64c5d4b77f98a394b367762f", + "prepared_review_unit_hash": "ce2b37c2a696b79aaedbcd899788eb4a669d830b771034c2cccc8cfbfb0ec4ae", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 36931, - "estimated_input_tokens": 9233, - "estimated_eval_tokens": 18722 + "serialized_chars": 36284, + "estimated_input_tokens": 9071, + "estimated_eval_tokens": 18398 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of unrelated issues with only superficial overlap (tokenizers, multimodal models, CI, quantization, embeddings). I did not find any soft pair that clearly describes the same underlying bug or change, so all soft edges should be rejected.", + "summary": "This set is a loose grab-bag of unrelated issues across tokenization, processor/model loading, quantization, attention kernels, docs, and CI. None of the suggested soft pairs look like the same concrete bug or change.", "confidence": 0.91, - "canonical_issue_reason": "issue:44462 is the broadest tokenizer-loading regression in the set and sits closest to several tokenizer-related reports, so it is the best representative issue for the cluster.", + "canonical_issue_reason": "No true canonical duplicate stands out; the issues are heterogeneous. If one must be chosen as the representative report, issue 44521 is the clearest high-signal regression with an explicit multimodal API failure and active discussion.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44462 is the most central and reusable problem statement here: a generic AutoTokenizer/tokenizer.json loading failure in v5, which is closer to the shared tokenizer theme than the others.", + "best_issue_reason": "Issue 44521 is the strongest representative issue in the set because it has a concrete repro, clear impact, and ongoing discussion; however, it is not a duplicate hub for the other items.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43824", - "right": "issue:44933", - "accept": false, - "reason": "Both are import-related, but they concern different missing/imported symbols in different modules and do not indicate the same code-path bug." - }, - { - "left": "issue:44462", - "right": "issue:45081", + "left": "issue:32090", + "right": "issue:35141", "accept": false, - "reason": "Tokenizer loading in general vs a specific Mistral regex patch crash; different failure modes and fixes." + "reason": "Trainer broadcast TypeError and token-embedding reinitialization are different bugs in different code paths." }, { "left": "issue:44683", "right": "issue:45542", "accept": false, - "reason": "Compiled attention on newer torch versions vs TensorBoard/TensorFlow backend installation; unrelated problems." + "reason": "FlashAttention/torch compatibility and tensorboard-without-TF backend errors are unrelated." }, { "left": "issue:44279", "right": "issue:45405", "accept": false, - "reason": "A vague dependency issue is not the same as bumping MIN_PEFT_VERSION; no shared concrete bug." - }, - { - "left": "issue:32090", - "right": "issue:35141", - "accept": false, - "reason": "Trainer GPU broadcast TypeError vs embedding reinitialization after resize; different subsystems and symptoms." + "reason": "Both are dependency-related, but one is a vague dependency complaint while the other is a specific unreleased PEFT version bump." }, { - "left": "issue:43299", - "right": "issue:43854", + "left": "issue:39692", + "right": "issue:43901", "accept": false, - "reason": "Different model-loading regressions affecting different model families and code paths." + "reason": "Both mention docs, but one is a SigLIP2 example with model/processor issues and quantization failure; the other is a TextClassificationPipeline doc mismatch." }, { "left": "issue:43295", "right": "issue:44779", "accept": false, - "reason": "Processor/tokenizer API regression vs Deepseek tokenizer output regression; both tokenizer-adjacent but not the same bug." + "reason": "Different tokenizer regressions affecting different APIs/models and failure modes." }, { - "left": "issue:39692", - "right": "issue:43901", + "left": "issue:43299", + "right": "issue:43854", "accept": false, - "reason": "Documentation/example errors for SigLIP2 vs a docs mismatch for TextClassificationPipeline; unrelated issues." + "reason": "Different model-loading failures for different model families; not the same underlying bug." }, { "left": "issue:44297", "right": "issue:44462", "accept": false, - "reason": "Tokenizer class metadata written by save_pretrained vs AutoTokenizer ignoring tokenizer.json on load; related area but distinct bugs." + "reason": "Tokenizer save metadata mismatch and AutoTokenizer ignoring repository tokenizer.json are distinct serialization/load problems." }, { "left": "issue:38175", "right": "issue:44617", "accept": false, - "reason": "Unexpected probabilities in SigLIP2 vs Sam3Video OOM; unrelated model/runtime failures." + "reason": "Unexpected probabilities vs CUDA OOM are unrelated inference quality and memory issues." }, { "left": "issue:41950", "right": "issue:43650", "accept": false, - "reason": "Video-classification processor bug vs a non-informative issue title; no shared underlying defect." - }, - { - "left": "issue:43976", - "right": "issue:44279", - "accept": false, - "reason": "Python version compatibility regression vs a generic dependency issue; not the same problem." + "reason": "Video-classification processor selection and an empty title issue have no substantive overlap." }, { "left": "issue:44462", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer-related, but one is about repository tokenizer.json being ignored while the other is a model-specific correctness regression; not the same fix." + "reason": "Different tokenizer bugs: repository tokenizer.json loading versus Deepseek tokenization changes in v5." }, { "left": "issue:44521", "right": "issue:45325", "accept": false, - "reason": "Chat-template assistant masks for multimodal inputs vs Qwen2.5-VL rope position scaling; different multimodal bugs." + "reason": "Both are multimodal/VL-related, but one is assistant mask generation and the other is temporal position-id scaling; different code paths." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "Python-version incompatibility and a generic dependency issue are not the same bug." }, { "left": "issue:43976", "right": "issue:44987", "accept": false, - "reason": "Python/dependency compatibility vs loading a specific model under transformers>=5.1.0; different root causes." + "reason": "Different failures: one is Python support, the other is loading a specific model on newer transformers." }, { "left": "issue:45600", "right": "issue:45712", "accept": false, - "reason": "Both involve repository-maintenance/cleanup, but they affect different files and failure modes, so they are not the same change." + "reason": "Both are CI/repo hygiene, but they involve different files and different failure mechanisms." }, { "left": "issue:43450", "right": "issue:44279", "accept": false, - "reason": "Video processor batch-shape bug vs an unrelated dependency complaint." - }, - { - "left": "issue:42994", - "right": "issue:43317", - "accept": false, - "reason": "Quantized model saving failure vs dequantized model loading with gpu/cpu offload; both quantization-related but different code paths." + "reason": "Video processor batch shape bug and a dependency issue are unrelated." }, { "left": "issue:44466", "right": "issue:45276", "accept": false, - "reason": "Both touch embedding/weight handling, but one is inconsistent serialization while the other is resize_token_embeddings not updating Gemma4 outputs; not the same defect." + "reason": "Both touch tied/output embeddings, but one is device-dependent serialization and the other is Gemma4 resize_token_embeddings propagation." }, { "left": "issue:43065", "right": "issue:44898", "accept": false, - "reason": "Dummy Conv2d placeholder in Sam3PixelDecoder vs Perceiver non-default resolution failure; unrelated." + "reason": "Dummy Conv2d in Sam3PixelDecoder and Perceiver non-default-resolution failure are unrelated model bugs." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Quantized model saving failure and dequantized model loading with offload are different code paths." }, { "left": "issue:44261", "right": "issue:44263", "accept": false, - "reason": "Missing rms_norm_eps in MLA q_a_layernorm vs torch.split return handling in GlmMoeDsaIndexer; unrelated implementation bugs." + "reason": "Different GLM-related bugs: missing rms_norm_eps precision issue versus torch.split return handling." }, { "left": "issue:43650", "right": "issue:45290", "accept": false, - "reason": "A vague data-related issue vs chat-template crashes on tool-call assistant messages; no shared underlying bug." + "reason": "An empty-title issue and a chat-template crash on tool-call assistant messages are unrelated." }, { "left": "issue:43441", "right": "issue:44683", "accept": false, - "reason": "Ministral FlashAttention failure vs flex_attention failure on torch>=2.9; both attention-related but distinct concrete incompatibilities." + "reason": "Both involve attention kernels, but they are different kernels, models, and failure modes." }, { "left": "issue:44623", "right": "issue:45397", "accept": false, - "reason": "Missing files in processor.save_pretrained vs gemma-4 zero3 from_pretrained; different save/load paths and failure modes." + "reason": "Processor.save_pretrained missing files and Gemma-4 zero3 from_pretrained are unrelated." }, { "left": "issue:43723", "right": "issue:44162", "accept": false, - "reason": "AutoTokenizer loading in v5 vs ESM2 being broken; the titles point to different model families and problems." + "reason": "Tokenizer loading regression and ESM2 model breakage are different problems." }, { "left": "issue:44623", "right": "issue:44749", "accept": false, - "reason": "processor.save_pretrained file omission vs slow filtering after upgrading transformers; unrelated." + "reason": "Processor save files and tokenizer/filtering slowdown are unrelated." }, { "left": "issue:44671", "right": "issue:44779", "accept": false, - "reason": "CamemBERT masked-LM prediction regression vs Deepseek tokenizer regression; different models and code paths." + "reason": "CamemBERT masked-LM prediction regression and Deepseek tokenization regression are different bugs." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "BigBirdTokenizer special-token decoding bug and D-FINE auxiliary-loss behavior are unrelated." }, { "left": "issue:43784", "right": "issue:45542", "accept": false, - "reason": "Sentence-transformers import NameError vs TensorBoard/TensorFlow backend selection; unrelated." + "reason": "Sentence-transformers import NameError and tensorboard/TensorFlow backend error are different dependency/import issues." }, { - "left": "issue:43653", + "left": "issue:44361", "right": "issue:45593", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug vs D-FINE auxiliary loss behavior; not the same bug." + "reason": "MLukeTokenizer task AttributeError and D-FINE auxiliary-loss omission are unrelated." }, { - "left": "issue:44987", - "right": "issue:45542", + "left": "issue:43901", + "right": "issue:43994", "accept": false, - "reason": "Model-loading failure in transformers>=5.1.0 vs backend installation issue; unrelated." + "reason": "Docs mention of return_all_scores versus actual SigLIP2 incorrect outputs are different issue types and code paths." }, { - "left": "issue:44361", - "right": "issue:45593", + "left": "issue:43122", + "right": "issue:43650", + "accept": false, + "reason": "Tokenizer behavior change across versions and an empty-title issue are unrelated." + }, + { + "left": "issue:44987", + "right": "issue:45542", "accept": false, - "reason": "MLukeTokenizer AttributeError vs D-FINE loss behavior; unrelated code paths." + "reason": "Model-loading regression and tensorboard/TensorFlow backend error are unrelated." } ] }, @@ -9970,7 +9854,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:16998", "issue:22355", @@ -9979,22 +9863,20 @@ "issue:41950", "issue:42491", "issue:42757", - "issue:43122", "issue:43262", "issue:43299", "issue:43352", "issue:43450", "issue:43475", "issue:43525", - "issue:43650", "issue:43673", "issue:43761", "issue:43824", "issue:43881", "issue:43901", "issue:43976", - "issue:43994", "issue:44016", + "issue:44162", "issue:44188", "issue:44220", "issue:44242", @@ -10012,248 +9894,257 @@ "issue:44561", "issue:44589", "issue:44617", - "issue:44661", + "issue:44857", + "issue:44871", + "issue:44964", + "issue:45200", "issue:45250", "issue:45356", "issue:45362", "issue:45397", - "issue:45542", "issue:45593", "issue:45759" ], "soft_pairs": [ - "issue:43122|issue:43650", - "issue:43901|issue:43994", "issue:44016|issue:44291", "issue:44297|issue:45397", "issue:43262|issue:44521", - "issue:43475|issue:44016", "issue:44188|issue:44336", "issue:43673|issue:43761", + "issue:43475|issue:44016", "issue:43525|issue:45759", "issue:43525|issue:43881", "issue:43450|issue:44617", "issue:44589|issue:45362", "issue:43901|issue:44246", - "issue:41628|issue:43976", "issue:42491|issue:43299", "issue:44336|issue:44373", + "issue:41628|issue:43976", "issue:16998|issue:22355", "issue:44242|issue:45593", "issue:43352|issue:45250", "issue:44295|issue:45397", - "issue:44661|issue:45542", "issue:44246|issue:44496", - "issue:41950|issue:44220", "issue:44462|issue:45356", + "issue:41950|issue:44220", "issue:44336|issue:44485", "issue:36246|issue:42757", "issue:36246|issue:44351", "issue:43673|issue:44561", - "issue:43824|issue:43976" + "issue:43824|issue:43976", + "issue:44246|issue:44871", + "issue:43976|issue:44964", + "issue:44242|issue:44857", + "issue:44162|issue:45200" ], - "prepared_review_unit_hash": "426997d16e26b55de8a6e949561684b80a9325e83947d689e1ef2bb71b5d6d97", + "prepared_review_unit_hash": "cf84359d19b1747fc5f8bdba4df3bcfb6279ecff605a1195c32ddb261481a6d4", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36117, - "estimated_input_tokens": 9030, - "estimated_eval_tokens": 18316 + "soft_pair_count": 30, + "serialized_chars": 36485, + "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 very loose: most items are unrelated Transformers bugs/regressions that only share broad themes like imports, chat templates, or model-loading failures. I do not see any soft pair that clearly describes the same underlying bug/change, so I would reject all candidate duplicate edges. No PRs are present.", - "confidence": 0.66, - "canonical_issue_reason": "issue:43824 is the most central and well-referenced concrete bug in the set, and it sits closest to the small import-regression theme shared by a few other issues.", + "summary": "These are mostly unrelated Transformers issues that only share broad subsystem keywords. None of the soft pairs look like true duplicates of the same bug/change, so I would not merge any of them.", + "confidence": 0.73, + "canonical_issue_reason": "Issue 43824 has the strongest external signal and community impact in this set (multiple inbound references), making it the best anchor issue among these heterogeneous reports, though it is not a duplicate umbrella for the rest.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43824 is the best representative issue because it is specific, reproducible, and has the strongest discussion/inbound-reference signal among the loosely related items.", + "best_issue_reason": "Issue 43824 is the most visible and broadly referenced report here, so it is the best single issue for triage and follow-up.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43122", - "right": "issue:43650", - "accept": false, - "reason": "Different problems: tokenizer output regression vs a vague data-related issue; no shared code path." - }, - { - "left": "issue:43901", - "right": "issue:43994", - "accept": false, - "reason": "Docs mismatch for a pipeline option vs incorrect model/pipeline outputs for SigLIP; same area, different bug." - }, { "left": "issue:44016", "right": "issue:44291", "accept": false, - "reason": "Notebook syntax error and init_empty_weights TypeError are unrelated." + "reason": "Completely different bugs: notebook syntax error vs. init_empty_weights argument handling." }, { "left": "issue:44297", "right": "issue:45397", "accept": false, - "reason": "Tokenizer class metadata bug vs gemma-4 ZeRO-3 loading failure; not the same defect." + "reason": "Different failure modes and model families; tokenizer_class metadata mismatch is unrelated to gemma-4 ZeRO3 loading." }, { "left": "issue:43262", "right": "issue:44521", "accept": false, - "reason": "Both touch chat templates, but one is audio sampling-rate defaulting and the other is multimodal assistant mask generation." - }, - { - "left": "issue:43475", - "right": "issue:44016", - "accept": false, - "reason": "SAM3 video encoder attribute error is unrelated to a notebook syntax issue." + "reason": "Both involve multimodal chat/template code, but one is audio sampling-rate defaults and the other is assistant mask generation; not the same bug." }, { "left": "issue:44188", "right": "issue:44336", "accept": false, - "reason": "Attention-kernel divergence on torch.compile is unrelated to ANSI codes in loading_report." + "reason": "Attention-kernel divergence under torch.compile is unrelated to ANSI escaping in loading reports." }, { "left": "issue:43673", "right": "issue:43761", "accept": false, - "reason": "Cache handling regression and CLIP vision hidden_states regression affect different code paths." + "reason": "Generation cache handling during chunked_prefill is unrelated to CLIPVisionModel hidden_states regression." + }, + { + "left": "issue:43475", + "right": "issue:44016", + "accept": false, + "reason": "SAM3 video encoder attribute error vs. notebook syntax error; no shared code path." }, { "left": "issue:43525", "right": "issue:45759", "accept": false, - "reason": "Both are config/model-loading related, but one is missing pad_token_id on Llama4Config and the other is failure to unwrap text_config for composite Qwen configs." + "reason": "Both are config-related, but one is a missing pad_token_id on Llama4Config and the other is composite Qwen text_config unwrapping; different defects." }, { "left": "issue:43525", "right": "issue:43881", "accept": false, - "reason": "Different model-loading failures with no clear shared root cause." + "reason": "Different model-specific loading issues with different missing attributes and symptoms." }, { "left": "issue:43450", "right": "issue:44617", "accept": false, - "reason": "Video processor shape bug vs SAM3Video OOM; not the same bug." + "reason": "A batched-shape bug in video processors is not the same as a CUDA OOM in Sam3Video." }, { "left": "issue:44589", "right": "issue:45362", "accept": false, - "reason": "Float8 storage deserialization error and Qwen3.5 chat crash are distinct failures." + "reason": "Float8 storage lookup failure is a serialization/runtime issue, not the same as a Qwen3.5 chat crash." }, { "left": "issue:43901", "right": "issue:44246", "accept": false, - "reason": "Pipeline docs issue vs slow import performance; no duplicate relationship." - }, - { - "left": "issue:41628", - "right": "issue:43976", - "accept": false, - "reason": "Missing export/import error and Python-version compatibility issue are not the same defect." + "reason": "Docs mismatch for return_all_scores is unrelated to import latency." }, { "left": "issue:42491", "right": "issue:43299", "accept": false, - "reason": "Both mention Qwen3 MoE, but one is a hf4.x-to-hf5.x LoRA compatibility problem and the other is a Qwen3VL loading regression; too different to merge." + "reason": "Both mention Qwen/MoE compatibility, but they affect different model types and loading paths; not clearly the same bug." }, { "left": "issue:44336", "right": "issue:44373", "accept": false, - "reason": "ANSI escape-code bug in loading_report is unrelated to a position_ids docstring issue." + "reason": "ANSI codes in loading_report and a wrong docstring for position_ids are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "Missing AutoImageProcessor export is unrelated to a Python version compatibility regression." }, { "left": "issue:16998", "right": "issue:22355", "accept": false, - "reason": "DeBERTa model_max_length question and missing transformers.onnx module are unrelated." + "reason": "A model_max_length question and a missing transformers.onnx module are unrelated." }, { "left": "issue:44242", "right": "issue:45593", "accept": false, - "reason": "Auxiliary-loss omission in MoE routing and D-FINE denoising-off losses are different model-specific bugs." + "reason": "Both are loss-related, but they are different models and different gating conditions; not the same concrete bug." }, { "left": "issue:43352", "right": "issue:45250", "accept": false, - "reason": "The latter is too generic to prove it is the same Flash Attention 2.0 issue; no concrete shared code-path evidence." + "reason": "Generic Flash Attention 2.0 discussion is not the same as a specific model explicitly lacking FA2 support." }, { "left": "issue:44295", "right": "issue:45397", "accept": false, - "reason": "Position_ids buffer access bug and gemma-4 ZeRO-3 loading failure are unrelated." - }, - { - "left": "issue:44661", - "right": "issue:45542", - "accept": false, - "reason": "Tokenizer-mapping name bug and undefined TensorFlow backend due to tensorboard-only install are unrelated." + "reason": "Position_ids buffer access error is unrelated to gemma-4 ZeRO3 loading." }, { "left": "issue:44246", "right": "issue:44496", "accept": false, - "reason": "Import slowdown and unrecognized model/config loading failure do not describe the same underlying bug." + "reason": "Import performance issue is unrelated to model-type resolution failure." }, { - "left": "issue:41950", - "right": "issue:44220", + "left": "issue:44462", + "right": "issue:45356", "accept": false, - "reason": "Video-classification image-processor lookup and fbank feature extraction are different subsystems and failures." + "reason": "Tokenizer ignores tokenizer.json vs. codec handling regression and warning text; different tokenizer bugs." }, { - "left": "issue:44462", - "right": "issue:45356", + "left": "issue:41950", + "right": "issue:44220", "accept": false, - "reason": "Tokenizer.json selection issue and Kimi-K2.5 codec/regression warning are distinct tokenizer bugs." + "reason": "Video-classification processor selection and audio feature extraction are different pipelines with different failures." }, { "left": "issue:44336", "right": "issue:44485", "accept": false, - "reason": "Loading-report ANSI formatting and GLM-5 RoPE implementation are unrelated." + "reason": "Loading-report ANSI output is unrelated to GLM-5 RoPE implementation." }, { "left": "issue:36246", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but they concern different missing symbols from different packages/models." + "reason": "Both are ImportErrors, but they involve different missing symbols in different packages." }, { "left": "issue:36246", "right": "issue:44351", "accept": false, - "reason": "Different missing exports: Qwen2_5_VLImageProcessor vs HybridCache." + "reason": "Different missing imports/exported symbols; no evidence of the same underlying packaging bug." }, { "left": "issue:43673", "right": "issue:44561", "accept": false, - "reason": "Cache regression in generation and removal of is_torch_fx_available breaking trust_remote_code are separate compatibility issues." + "reason": "Cache handling in chunked_prefill and removal of is_torch_fx_available are separate regressions." }, { "left": "issue:43824", "right": "issue:43976", "accept": false, - "reason": "ImportError for a Qwen2.5 VL class and Python-version incompatibility are not the same bug." + "reason": "Missing Qwen2_5_VLForConditionalGeneration export is unrelated to a Python 3.9/3.10 support regression." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Import slowness is unrelated to Gemma-3 eos_token_id configuration inconsistency." + }, + { + "left": "issue:43976", + "right": "issue:44964", + "accept": false, + "reason": "Python-version compatibility and Phi-4-multimodal loading failure are different issues." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Different loss implementations and different runtime conditions; not the same bug." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "ESM2 breakage and Gemma 4 mm_token_type_ids defaults are unrelated model-specific issues." } ] }, @@ -10275,10 +10166,9 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:36246", - "issue:41628", "issue:41950", "issue:42886", "issue:43065", @@ -10293,20 +10183,18 @@ "issue:43596", "issue:43653", "issue:43901", - "issue:43976", "issue:44162", - "issue:44242", "issue:44246", "issue:44295", "issue:44336", + "issue:44368", + "issue:44451", "issue:44466", "issue:44496", "issue:44749", "issue:44779", - "issue:44857", - "issue:44871", "issue:44877", - "issue:44964", + "issue:44977", "issue:44987", "issue:44995", "issue:45003", @@ -10318,254 +10206,249 @@ "issue:45335", "issue:45362", "issue:45375", + "issue:45397", "issue:45459", "issue:45561", "issue:45563", "issue:45584", "issue:45588", + "issue:45656", "issue:45701", "issue:45759" ], "soft_pairs": [ - "issue:43976|issue:44964", - "issue:44246|issue:44871", - "issue:44242|issue:44857", - "issue:44162|issue:45200", "issue:45081|issue:45362", - "issue:44295|issue:44749", "issue:44779|issue:45292", + "issue:44295|issue:44749", "issue:43441|issue:43540", "issue:45292|issue:45584", "issue:44877|issue:45375", - "issue:36246|issue:43596", - "issue:44987|issue:45459", - "issue:45292|issue:45701", - "issue:41628|issue:41950", "issue:44779|issue:45701", + "issue:45292|issue:45701", + "issue:44987|issue:45459", "issue:43901|issue:45003", "issue:45561|issue:45563", "issue:43232|issue:44336", + "issue:41950|issue:44451", + "issue:36246|issue:43596", "issue:43065|issue:43408", - "issue:45375|issue:45759", "issue:43526|issue:45003", + "issue:45375|issue:45759", "issue:43257|issue:45137", "issue:44496|issue:45003", "issue:44995|issue:45200", "issue:43262|issue:44246", + "issue:44466|issue:45335", "issue:43475|issue:43540", "issue:42886|issue:45229", - "issue:44466|issue:45335", "issue:44995|issue:45588", - "issue:43653|issue:44162" + "issue:45397|issue:45656", + "issue:43653|issue:44162", + "issue:44368|issue:45563", + "issue:44368|issue:44977" ], - "prepared_review_unit_hash": "d06130180abecc35930c7a1febc2fb9956661662fad5f1b15a2447e3396dc147", + "prepared_review_unit_hash": "28ed0275ff8949ae0639fce186478a544963a2a9fe8079bdfd6806ba92dc43e3", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36874, - "estimated_input_tokens": 9219, - "estimated_eval_tokens": 18694 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35988, + "estimated_input_tokens": 8997, + "estimated_eval_tokens": 18250 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most items are unrelated Transformer bugs or regressions. The only clear duplicate pair is the resize_token_embeddings propagation bug: the generic output_embeddings report and the T5Gemma decoder.embed_tokens report. No PRs were provided.", - "confidence": 0.83, - "canonical_issue_reason": "issue:45292 is the broader, more generic report about resize_token_embeddings not updating output embeddings, and it likely subsumes the T5Gemma-specific manifestation.", + "summary": "This cluster is heterogeneous: it mixes unrelated tokenizer, config, generation, DeepSpeed, SAM, audio, docs, and performance issues. No soft pair looks like the same underlying bug/change, and there are no PRs here.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue stands out because the items do not share one underlying defect; choosing any single issue would be arbitrary.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45292 is the strongest canonical anchor because it is model-agnostic, clearly reproducible, and matches the one accepted duplicate pair.", + "best_issue_reason": "No single issue is a good global representative for duplicate triage; the cluster is too broad and unrelated.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43976", - "right": "issue:44964", - "accept": false, - "reason": "Different failures: Python version compatibility/regression vs a specific model loading error." - }, - { - "left": "issue:44246", - "right": "issue:44871", - "accept": false, - "reason": "Import slowdown and Gemma eos_token_id inconsistency are unrelated bugs." - }, - { - "left": "issue:44242", - "right": "issue:44857", - "accept": false, - "reason": "MoE load-balancing loss logic vs float16 CUDA crash in a different loss path." - }, - { - "left": "issue:44162", - "right": "issue:45200", - "accept": false, - "reason": "ESM2 breakage and Gemma 4 multimodal token defaults are different code paths." - }, { "left": "issue:45081", "right": "issue:45362", "accept": false, - "reason": "Tokenizer regex patch crash vs Qwen3.5 chat crash; not the same bug." + "reason": "Different models and failure modes: Mistral tokenizer regex patch crash vs Qwen3.5 chat crash." }, { - "left": "issue:44295", - "right": "issue:44749", + "left": "issue:44779", + "right": "issue:45292", "accept": false, - "reason": "Buffer access IndexError and tokenizer performance regression are unrelated." + "reason": "One is tokenization regression in DeepSeek; the other is resize_token_embeddings not updating output embeddings." }, { - "left": "issue:44779", - "right": "issue:45292", + "left": "issue:44295", + "right": "issue:44749", "accept": false, - "reason": "DeepSeek tokenization correctness vs token embedding resize propagation are different issues." + "reason": "Unrelated bugs: position_ids buffer access error vs a post-upgrade performance slowdown." }, { "left": "issue:43441", "right": "issue:43540", "accept": false, - "reason": "FlashAttention failure for Ministral and video-input handling failure for Qwen3OmniMoe are unrelated." + "reason": "Different subsystems and symptoms: FlashAttention on Ministral-3 vs video input validation in Qwen3OmniMoe." }, { "left": "issue:45292", "right": "issue:45584", "accept": false, - "reason": "Embedding resizing bug and Whisper empty transcription after align_special_tokens are different problems." + "reason": "Embedding resize bug and Whisper empty-transcription failure are separate code paths." }, { "left": "issue:44877", "right": "issue:45375", "accept": false, - "reason": "Both involve config strictness, but they describe opposite symptoms on different model families." - }, - { - "left": "issue:36246", - "right": "issue:43596", - "accept": false, - "reason": "ImportError for a Qwen image processor is unrelated to a DeepSpeed ZeRO3/BERT index error." + "reason": "Both mention config strictness, but they target different model configs and different missing-field behaviors." }, { - "left": "issue:44987", - "right": "issue:45459", + "left": "issue:44779", + "right": "issue:45701", "accept": false, - "reason": "Model loading regression and protobuf error masking are not the same underlying bug." + "reason": "Both concern tokenization changes, but one is a specific DeepSeek regression and the other is a broad version-dependent tokenization complaint." }, { "left": "issue:45292", "right": "issue:45701", "accept": false, - "reason": "Token embedding resize propagation is unrelated to version-dependent tokenization changes." - }, - { - "left": "issue:41628", - "right": "issue:41950", - "accept": false, - "reason": "Shared image-processor area, but one is an import/export issue and the other is a pipeline behavior bug." + "reason": "Embedding resize behavior is unrelated to a broad tokenization-version regression." }, { - "left": "issue:44779", - "right": "issue:45701", + "left": "issue:44987", + "right": "issue:45459", "accept": false, - "reason": "Both touch tokenization, but the symptoms and fix surfaces differ." + "reason": "Loading failure for a specific model is unrelated to protobuf exception masking in tokenizer decode errors." }, { "left": "issue:43901", "right": "issue:45003", "accept": false, - "reason": "Documentation mismatch and unsafe sys.modules access are unrelated." + "reason": "Docs mismatch for TextClassificationPipeline is unrelated to unsafe sys.modules access in modeling_utils." }, { "left": "issue:45561", "right": "issue:45563", "accept": false, - "reason": "pytest-xdist file race and stale num_return_sequences warning are unrelated." + "reason": "pytest-xdist file race and stale generate() warning are distinct issues." }, { "left": "issue:43232", "right": "issue:44336", "accept": false, - "reason": "Generation kwargs handling and ANSI output in loading_report are separate concerns." + "reason": "Generation kwargs handling after sync_gpus is unrelated to ANSI formatting in loading_report." }, { - "left": "issue:43065", - "right": "issue:43408", + "left": "issue:41950", + "right": "issue:44451", "accept": false, - "reason": "Both are SAM3-related, but one is a dummy layer concern and the other is a model-type/config mismatch." + "reason": "Video-classification pipeline processor lookup and ScandiBERT model loading are unrelated." }, { - "left": "issue:45375", - "right": "issue:45759", + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "ImportError for a missing processor class and a DeepSpeed/BertModel index error are different bugs." + }, + { + "left": "issue:43065", + "right": "issue:43408", "accept": false, - "reason": "Both involve composite Qwen configs, but one is strict-field dropping and the other is text_config unwrapping." + "reason": "Both involve SAM3, but one is a dummy layer implementation issue and the other is a model-type mismatch warning." }, { "left": "issue:43526", "right": "issue:45003", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug and sys.modules access bug are unrelated." + "reason": "BEiT label reduction bug is unrelated to unsafe sys.modules access." + }, + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Both touch composite Qwen config handling, but one drops a field under @strict while the other fails to unwrap text_config in from_config." }, { "left": "issue:43257", "right": "issue:45137", "accept": false, - "reason": "Both involve DeepSpeed, but the failure modes and code paths are different." + "reason": "Different DeepSpeed problems: missing MOE weight conversion vs ZeRO3 deque underflow." }, { "left": "issue:44496", "right": "issue:45003", "accept": false, - "reason": "Unrecognized model/config loading and sys.modules access are not the same issue." + "reason": "Unrecognized model/config loading is unrelated to sys.modules access in model utilities." }, { "left": "issue:44995", "right": "issue:45200", "accept": false, - "reason": "Stale cache crash on second forward pass is unrelated to token-id defaults." + "reason": "A stale cache crash on second forward is not the same as missing mm_token_type_ids defaults." }, { "left": "issue:43262", "right": "issue:44246", "accept": false, - "reason": "Audio sampling-rate default bug and MoE load-balancing loss bug are unrelated." + "reason": "Audio chat template sampling-rate defaulting and slow imports are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Serialization of tied weights depending on device is different from resize_token_embeddings not updating decoder embeddings." }, { "left": "issue:43475", "right": "issue:43540", "accept": false, - "reason": "SAM3 VisionEncoderOutput attribute error and Qwen3OmniMoe video processing error are different bugs." + "reason": "Sam3 vision encoder missing an attribute is unrelated to Qwen3OmniMoe video-input validation." }, { "left": "issue:42886", "right": "issue:45229", "accept": false, - "reason": "Offline cache loading and CUDA OOM are unrelated." + "reason": "Offline cache loading for tokenizers and Gemma4 multi-GPU OOM are unrelated." }, { - "left": "issue:44466", - "right": "issue:45335", + "left": "issue:44995", + "right": "issue:45588", "accept": false, - "reason": "Tied-weight serialization inconsistency is different from resize_token_embeddings not updating decoder embeddings." + "reason": "Different failure points: stale indexer cache on second forward vs flash_attention.py crashing on sink-less models." }, { - "left": "issue:44995", - "right": "issue:45588", + "left": "issue:45397", + "right": "issue:45656", "accept": false, - "reason": "GlmMoeDsa stale cache crash and flash_attention s_aux=None crash are unrelated." + "reason": "Gemma4 ZeRO3 from_pretrained and double optimizer step under DeepSpeed are separate problems." }, { "left": "issue:43653", "right": "issue:44162", "accept": false, - "reason": "Tokenizer special-token registration bug and ESM2 model breakage are unrelated." + "reason": "BigBirdTokenizer special-token registration bug is unrelated to ESM2 being broken." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "A tie_word_embeddings warning in Qwen3.5 fine-tuning is unrelated to a stale num_return_sequences warning." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "Both are Qwen3.5-related, but one is a config warning and the other is a flash-attention generation failure." } ] }, @@ -10587,11 +10470,12 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:28282", "issue:30990", "issue:41628", + "issue:42222", "issue:42371", "issue:42673", "issue:42994", @@ -10619,12 +10503,12 @@ "issue:44749", "issue:44869", "issue:44933", - "issue:44977", "issue:45003", "issue:45081", "issue:45092", "issue:45103", "issue:45137", + "issue:45200", "issue:45230", "issue:45310", "issue:45356", @@ -10634,14 +10518,10 @@ "issue:45459", "issue:45478", "issue:45561", - "issue:45563", "issue:45600", "issue:45656" ], "soft_pairs": [ - "issue:45397|issue:45656", - "issue:44368|issue:45563", - "issue:44368|issue:44977", "issue:44060|issue:44077", "issue:45137|issue:45600", "issue:43673|issue:44734", @@ -10650,10 +10530,10 @@ "issue:44933|issue:45459", "issue:44869|issue:45356", "issue:44297|issue:44623", - "issue:43646|issue:44162", "issue:44361|issue:44749", - "issue:44623|issue:45656", + "issue:43646|issue:44162", "issue:43452|issue:45003", + "issue:44623|issue:45656", "issue:43673|issue:43827", "issue:43232|issue:43673", "issue:28282|issue:30990", @@ -10667,250 +10547,233 @@ "issue:43475|issue:45381", "issue:45081|issue:45103", "issue:41628|issue:44933", + "issue:42673|issue:43704", "issue:42994|issue:43525", "issue:43475|issue:44661", - "issue:42673|issue:43704", "issue:45375|issue:45478", - "issue:45310|issue:45375" + "issue:45310|issue:45375", + "issue:42222|issue:45200" ], - "prepared_review_unit_hash": "1bfede4eaf4018be3002735786c091ec91342ef0a49ba46a9177155662d5b492", + "prepared_review_unit_hash": "b6bcce7cba1c5a25f0fee41eeedba27ad2f49b634e73084333853a14a123a9df", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 33, - "serialized_chars": 37517, - "estimated_input_tokens": 9380, - "estimated_eval_tokens": 19016 + "soft_pair_count": 31, + "serialized_chars": 36943, + "estimated_input_tokens": 9236, + "estimated_eval_tokens": 18728 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most pairs are only loosely related by model family or subsystem and should stay separate. Two clear duplicate groups stand out: Qwen3ForCausalLM VRAM leak (42673/43704) and Qwen3.5 MoE from_pretrained error (45310/45478).", + "summary": "Most items are distinct model- or subsystem-specific regressions. Only two soft edges look like true duplicates: the repeated Qwen3.5-Moe from_pretrained error and the repeated Qwen3ForCausalLM VRAM leak. The rest share only broad themes (generation, tokenizer, save/load, v5 regressions) and should stay separate.", "confidence": 0.93, - "canonical_issue_reason": "Issue 42673 is the best canonical anchor: it is a concrete, well-scoped bug report with strong exact-duplicate evidence from 43704, and it has the richer discussion/activity of the pair.", + "canonical_issue_reason": "issue:43646 is the best umbrella issue: it describes a broad Transformers v5 custom-model initialization regression, which is the closest cluster-wide anchor among the issues here.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 42673 is the strongest overall representative in this set because it is a clear, actionable bug with an exact duplicate and minimal ambiguity.", + "best_issue_reason": "issue:43646 is the most general and reusable issue to represent the cluster; the other candidates are narrower, model-specific fallout or unrelated bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45397", - "right": "issue:45656", - "accept": false, - "reason": "Different problems: Zero3 from_pretrained failure vs optimizer stepping twice in DeepSpeed." - }, - { - "left": "issue:44368", - "right": "issue:45563", - "accept": false, - "reason": "Both mention warnings in generation/fine-tuning, but the symptoms and code paths are unrelated." - }, - { - "left": "issue:44368", - "right": "issue:44977", - "accept": false, - "reason": "Qwen3.5 tie_word_embeddings warning is not the same as flash-attention generation failure." - }, { "left": "issue:44060", "right": "issue:44077", "accept": false, - "reason": "Unrelated bugs: tied-weights warning in Qwen3-Next vs patchtsmixer post_init behavior." + "reason": "Both are model-specific warnings/errors, but they are different bugs in different code paths (tie-word-embeddings warning vs optional post_init handling)." }, { "left": "issue:45137", "right": "issue:45600", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error is unrelated to removed Sam3LiteText auto-mapping references." + "reason": "DeepSpeed ZeRO3 deque handling and auto_mappings cleanup are unrelated bugs despite both affecting training/tooling." }, { "left": "issue:43673", "right": "issue:44734", "accept": false, - "reason": "Both touch generation/cache, but they describe different failures in different paths." + "reason": "Both concern generation/KV-cache code, but one is a missing cache field in chunked_prefill and the other is a tensor-indexing crash in /v1/responses continuation; not the same defect." }, { "left": "issue:45003", "right": "issue:45561", "accept": false, - "reason": "sys.modules access safety and pytest-xdist file races are different bugs." + "reason": "Unsafe sys.modules access and the xdist captured_info.txt race are different failures in different subsystems." }, { "left": "issue:43931", "right": "issue:44368", "accept": false, - "reason": "Weight-shape mismatch in Qwen3-VL is unrelated to a tie-word-embeddings warning." + "reason": "Qwen3-VL weight-shape mismatch and a Qwen3.5 LoRA tie-word-embeddings warning are not the same underlying bug." }, { "left": "issue:44933", "right": "issue:45459", "accept": false, - "reason": "Import error from image_utils and protobuf-error masking are different tokenizer/import issues." + "reason": "A missing import from image_utils is unrelated to tokenizer error masking when protobuf is absent." }, { "left": "issue:44869", "right": "issue:45356", "accept": false, - "reason": "Whisper timestamp crash and Kimi-K2.5 tokenizer regression are different code paths." + "reason": "Whisper timestamp decode crash and Kimi-K2.5 tokenizer codec/regression handling are different concrete tokenizer bugs." }, { "left": "issue:44297", "right": "issue:44623", "accept": false, - "reason": "Both concern save_pretrained, but tokenizer metadata mismatch and processor missing files are distinct failures." - }, - { - "left": "issue:43646", - "right": "issue:44162", - "accept": false, - "reason": "Custom model initialization breakage and ESM2 breakage are separate issues." + "reason": "Both are save_pretrained-related, but one is tokenizer metadata mismatch and the other is processor file omission; too different to merge." }, { "left": "issue:44361", "right": "issue:44749", "accept": false, - "reason": "Tokenizer AttributeError on tasks is unrelated to the performance regression after upgrade." + "reason": "MLukeTokenizer task AttributeError and the reported performance slowdown after upgrade are unrelated." }, { - "left": "issue:44623", - "right": "issue:45656", + "left": "issue:43646", + "right": "issue:44162", "accept": false, - "reason": "Missing processor save files is unrelated to DeepSpeed optimizer double-step behavior." + "reason": "44162 looks like a specific model breakage, but there is not enough evidence that it is the same concrete failure as the broad custom-model initialization regression." }, { "left": "issue:43452", "right": "issue:45003", "accept": false, - "reason": "GGUF/AutoTokenizer loading failure is unrelated to unsafe sys.modules access." + "reason": "GGUF from_pretrained breakage and sys.modules access safety are separate code-path problems." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "Processor save_pretrained file output and a double optimizer-step bug under deepspeed are unrelated." }, { "left": "issue:43673", "right": "issue:43827", "accept": false, - "reason": "Cache missing in chunked_prefill is not the same as stale docs referencing removed pipeline()." + "reason": "Generation cache handling and docs still referencing removed pipeline() APIs are not the same bug." }, { "left": "issue:43232", "right": "issue:43673", "accept": false, - "reason": "Generation kwargs update bug and missing cache in chunked_prefill are different generation issues." + "reason": "These are both generation-related, but one is a sync_gpus kwargs update issue and the other is a chunked_prefill cache regression." }, { "left": "issue:28282", "right": "issue:30990", "accept": false, - "reason": "PyTorch ImportError and Sentence Transformers loading stall are unrelated." + "reason": "ImportError about missing PyTorch and Sentence Transformers hanging on load are different user-facing failures." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Quantized model saving failure is not the same as Qwen2.5-GGUF loading failure." + "reason": "Quantized saving failure and Qwen2.5-GGUF loading failure in v5 are different directions and different bugs." }, { "left": "issue:44749", "right": "issue:45656", "accept": false, - "reason": "Filtering slowdown after upgrade and optimizer double-step under DeepSpeed are unrelated." + "reason": "A performance regression after upgrading versions is not the same as deepspeed causing optimizer.step() to run twice." }, { "left": "issue:44749", "right": "issue:45310", "accept": false, - "reason": "Performance regression in filtering is unrelated to Qwen3.5 MoE from_pretrained errors." + "reason": "Slow filtering after upgrade and Qwen3.5-Moe from_pretrained error are unrelated regressions." }, { "left": "issue:43475", "right": "issue:43643", "accept": false, - "reason": "SAM 3 video missing attribute and trust_remote_code config field loss are different bugs." + "reason": "SAM3 vision output attribute loss and trust_remote_code returning missing config fields are different problems." }, { "left": "issue:45230", "right": "issue:45397", "accept": false, - "reason": "Generic bug report and gemma-4 Zero3 issue are not enough to identify the same bug." + "reason": "A placeholder bug report and Gemma-4 ZeRO3 from_pretrained failure are not the same underlying issue." }, { "left": "issue:44038", "right": "issue:44623", "accept": false, - "reason": "Qwen3-VL-Moe v5 bug and processor.save_pretrained missing files are unrelated." + "reason": "Qwen3-VL-Moe v5 issues and missing processor save artifacts are different bugs." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF32 API settings warning is unrelated to InternVL2 meta-initialization incompatibility." + "reason": "TF32 settings deprecation and InternVL2 meta-init incompatibility are unrelated." }, { "left": "issue:43475", "right": "issue:45381", "accept": false, - "reason": "SAM3 encoder attribute error and qwen2.5-vl video position ids are different failures." + "reason": "SAM3 encoder output missing fields and Qwen2.5-VL video position IDs are different model-specific defects." }, { "left": "issue:45081", "right": "issue:45103", "accept": false, - "reason": "Mistral regex crash and auto_docstring future-annotations crash are unrelated." + "reason": "Tokenizer regex patching and auto_docstring future-annotations handling are unrelated crash paths." }, { "left": "issue:41628", "right": "issue:44933", "accept": false, - "reason": "Two different import errors involving different symbols and modules." + "reason": "Both are import-related, but one is AutoImageProcessor import availability and the other is a missing image_utils import; not the same concrete bug." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Same title and same reported behavior: Qwen3ForCausalLM leaks VRAM in multiple dataloader threads." }, { "left": "issue:42994", "right": "issue:43525", "accept": false, - "reason": "Quantized saving failure and missing pad_token_id in Llama4Config are unrelated." + "reason": "Quantized model saving failure and missing pad_token_id on Llama4Config are different issues." }, { "left": "issue:43475", "right": "issue:44661", "accept": false, - "reason": "SAM3 video encoder bug and tokenizer-mapping add-new-model-like failure are different concerns." - }, - { - "left": "issue:42673", - "right": "issue:43704", - "accept": true, - "reason": "Same exact bug title and same VRAM leak symptom in multi-dataloader-thread use; clear duplicate." + "reason": "A SAM3 output attribute bug and add-new-model-like failing on TOKENIZER_MAPPING_NAMES are unrelated." }, { "left": "issue:45375", "right": "issue:45478", "accept": false, - "reason": "Both involve Qwen3.5 MoE, but one is a missing config field and the other is a from_pretrained error; not the same concrete bug." + "reason": "Both are Qwen3.5-Moe-related, but one is a specific config field being dropped and the other is a from_pretrained failure; not enough to treat as the same bug." }, { "left": "issue:45310", "right": "issue:45375", "accept": false, - "reason": "Qwen3.5 MoE from_pretrained error and missing deepstack_visual_indexes are different symptoms and likely different code paths." + "reason": "The from_pretrained error and the missing deepstack_visual_indexes config field are related by model family only, not by the same concrete failure." }, { - "left": "issue:45310", - "right": "issue:45478", - "accept": true, - "reason": "Same title-level bug report about Qwen3.5 MoE from_pretrained failing on transformers>=5.4.0; clear duplicate." + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "VitPose breakage and Gemma-4 token-type defaults are unrelated model-specific bugs." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: both cited duplicate groups have identical titles, and the rest of the pairwise rejections stay conservative by not overclaiming equivalence from shared model families or subsystems alone. No obvious evidence inflation." + "feedback": "The cluster choice is grounded in the packet, and the soft-edge decisions are conservative overall. The only accepted duplicate pair is well-supported by identical titles/behavior, and the false rejections avoid over-claiming sameness from broad thematic overlap. Minor note: the summary mentions the repeated Qwen3.5-Moe from_pretrained error as a true duplicate, but that exact pair is not one of the listed soft edges; still, this does not materially undermine the evaluation." }, "evaluator_used": true, "retried": false, @@ -10929,7 +10792,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:30333", "issue:30990", @@ -10938,7 +10801,6 @@ "issue:36010", "issue:41628", "issue:42175", - "issue:42222", "issue:42831", "issue:42898", "issue:42994", @@ -10958,6 +10820,7 @@ "issue:44188", "issue:44222", "issue:44242", + "issue:44295", "issue:44351", "issue:44373", "issue:44448", @@ -10972,26 +10835,25 @@ "issue:45103", "issue:45125", "issue:45137", - "issue:45200", "issue:45310", "issue:45356", "issue:45405", "issue:45561", "issue:45600", + "issue:45656", "issue:45712" ], "soft_pairs": [ - "issue:42222|issue:45200", - "issue:41628|issue:44351", "issue:42831|issue:44222", - "issue:42994|issue:44164", + "issue:41628|issue:44351", "issue:44222|issue:45310", + "issue:42994|issue:44164", + "issue:43377|issue:44242", "issue:36010|issue:41628", "issue:42175|issue:45712", - "issue:43377|issue:44242", "issue:44038|issue:45310", - "issue:44496|issue:45561", "issue:44964|issue:45137", + "issue:44496|issue:45561", "issue:45561|issue:45600", "issue:42898|issue:44568", "issue:30990|issue:35141", @@ -11000,8 +10862,8 @@ "issue:43232|issue:44188", "issue:44188|issue:44373", "issue:43278|issue:45137", - "issue:43784|issue:43824", "issue:44623|issue:45310", + "issue:43784|issue:43824", "issue:30333|issue:33290", "issue:44734|issue:45103", "issue:43723|issue:44568", @@ -11012,228 +10874,229 @@ "issue:43874|issue:45092", "issue:42898|issue:44448", "issue:43479|issue:44079", - "issue:44188|issue:44964" + "issue:44188|issue:44964", + "issue:44295|issue:45656" ], - "prepared_review_unit_hash": "f0f30ed9a54ba338123bd1b7990d5f53625b47d58b2035b7851c53ac3d290b9b", + "prepared_review_unit_hash": "40571a51d021bb4a9facc6fe18684887c1c4e12bcc574d86b76966c1cfaf0cfa", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37295, - "estimated_input_tokens": 9324, - "estimated_eval_tokens": 18904 + "serialized_chars": 37269, + "estimated_input_tokens": 9318, + "estimated_eval_tokens": 18892 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: it mixes unrelated bugs across imports, tokenizer behavior, multimodal model loading, quantization/save-pretrained, distributed training, and test/CI infrastructure. None of the soft pairs look like the same underlying bug or change.", - "confidence": 0.95, - "canonical_issue_reason": null, + "summary": "This is a heterogeneous collection of unrelated Transformers issues (tokenizer regressions, multimodal import/load failures, quantization, DeepSpeed, docs, and packaging). None of the soft-edge pairs look like the same underlying bug/change, so they should all remain separate.", + "confidence": 0.79, + "canonical_issue_reason": "No single true canonical fits this mixed cluster; if one issue must be chosen, #43723 is the broadest user-facing regression here (AutoTokenizer.from_pretrained loading in v5), but it does not actually subsume the rest.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "#43723 is the best representative only because it is the most general, well-scoped tokenizer-loading regression among the set; the other issues are too diverse to serve as duplicates of it.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:42222", - "right": "issue:45200", + "left": "issue:42831", + "right": "issue:44222", "accept": false, - "reason": "Different model families and failures: vitpose model breakage vs Gemma 4 mm_token_type_ids handling." + "reason": "Both mention FP8, but one is an accuracy regression and the other is a save_pretrained failure for MoE; different bug classes." }, { "left": "issue:41628", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but for different symbols and modules; no shared code path is evident." + "reason": "Both are import errors, but they concern different symbols (`AutoImageProcessor` vs `HybridCache`) and different code paths." }, { - "left": "issue:42831", - "right": "issue:44222", + "left": "issue:44222", + "right": "issue:45310", "accept": false, - "reason": "One is an accuracy regression with FineGrainedFP8, the other is save_pretrained failure for FP8 MoE models." + "reason": "Both involve MoE/FP8 and loading/saving, but one is a save_pretrained bug and the other is a from_pretrained error; not the same defect." }, { "left": "issue:42994", "right": "issue:44164", "accept": false, - "reason": "Both involve serialization, but one is quantized model saving and the other is extra_state handling in save/from_pretrained." + "reason": "Both are save/load problems, but one is quantized-model saving and the other is handling `extra_state`; too different to merge." }, { - "left": "issue:44222", - "right": "issue:45310", + "left": "issue:43377", + "right": "issue:44242", "accept": false, - "reason": "Different phases and symptoms: FP8 save_pretrained bug vs Qwen3.5 MoE from_pretrained loading error." + "reason": "One is a missing padding-mask bug in MIMI encoder outputs; the other is load-balancing loss logic when `output_router_logits=False`." }, { "left": "issue:36010", "right": "issue:41628", "accept": false, - "reason": "Separate import breakages for different APIs: GenerationMixin vs AutoImageProcessor." + "reason": "Different missing imports (`GenerationMixin` vs `AutoImageProcessor`) and different failure surfaces." }, { "left": "issue:42175", "right": "issue:45712", "accept": false, - "reason": "Packaging/backend dependency issue vs dummy classes leaking into dir/check_repo; unrelated." - }, - { - "left": "issue:43377", - "right": "issue:44242", - "accept": false, - "reason": "MIMI encoder padding-mask behavior is unrelated to MoE load-balancing loss computation." + "reason": "Both touch packaging/dependency behavior, but one is about missing TensorFlow in a torch install and the other is dummy classes leaking into the namespace." }, { "left": "issue:44038", "right": "issue:45310", "accept": false, - "reason": "Different Qwen model/version problems with different failure modes; not the same concrete bug." + "reason": "Same model family, but one is a Transformers 5.0 Qwen3-VL-Moe bug and the other is a Qwen3.5 MoE from_pretrained failure." }, { - "left": "issue:44496", - "right": "issue:45561", + "left": "issue:44964", + "right": "issue:45137", "accept": false, - "reason": "Unrecognized model config loading vs pytest-xdist file race; entirely different subsystems." + "reason": "Completely different areas: multimodal model loading versus DeepSpeed ZeRO3 deque underflow." }, { - "left": "issue:44964", - "right": "issue:45137", + "left": "issue:44496", + "right": "issue:45561", "accept": false, - "reason": "Phi-4 multimodal loading issue and DeepSpeed ZeRO3 deque error are unrelated." + "reason": "Unrecognized model/config loading versus a pytest-xdist file race; unrelated bugs." }, { "left": "issue:45561", "right": "issue:45600", "accept": false, - "reason": "Test-worker race on captured_info.txt vs stale auto_mappings references; different CI/codegen problems." + "reason": "Both are CI/test-related, but one is a parallel test race and the other is stale auto_mappings references." }, { "left": "issue:42898", "right": "issue:44568", "accept": false, - "reason": "Both are tokenizer regressions, but one concerns clean_up_tokenization_spaces and the other add_special_tokens/BOS-EOS behavior." + "reason": "Both are tokenizer regressions in v5, but they affect different behaviors (`clean_up_tokenization_spaces` vs `add_special_tokens`) and different models." }, { "left": "issue:30990", "right": "issue:35141", "accept": false, - "reason": "Sentence Transformers loading hang and embedding reinitialization after resize are unrelated bugs." + "reason": "Sentence-transformers loading hang is unrelated to token embedding resizing/reinitialization." }, { "left": "issue:43901", "right": "issue:45561", "accept": false, - "reason": "Docs mismatch for return_all_scores vs xdist file-race in test utilities; no overlap." + "reason": "Docs mismatch for `return_all_scores` versus a test utility race; not the same underlying issue." }, { "left": "issue:43232", "right": "issue:44792", "accept": false, - "reason": "Generation KV-cache sync bug vs janus image generation test failure; different code paths and symptoms." + "reason": "Generation kwarg update after `sync_gpus` is unrelated to janus image-generation test failure." }, { "left": "issue:43232", "right": "issue:44188", "accept": false, - "reason": "Generation sync_gpus bug and attention-kernel divergence under torch.compile are separate issues." + "reason": "Both are generation/attention code, but one is a `sync_gpus` generation bug and the other is torch.compile branching in attention kernels." }, { "left": "issue:44188", "right": "issue:44373", "accept": false, - "reason": "Kernel branching bug versus a docstring issue; not the same underlying change." + "reason": "Attention-kernel divergence is unrelated to a wrong docstring for `position_ids`." }, { "left": "issue:43278", "right": "issue:45137", "accept": false, - "reason": "Embedding dtype change in evaluation is unrelated to the DeepSpeed ZeRO3 deque crash." + "reason": "dtype change during evaluate versus a ZeRO3 deque pop error; unrelated." }, { - "left": "issue:43784", - "right": "issue:43824", + "left": "issue:44623", + "right": "issue:45310", "accept": false, - "reason": "Different missing-import breakages for different packages/models; no shared root cause." + "reason": "Processor save_pretrained missing files and Qwen3.5 MoE from_pretrained failure are different save/load issues." }, { - "left": "issue:44623", - "right": "issue:45310", + "left": "issue:43784", + "right": "issue:43824", "accept": false, - "reason": "Processor save_pretrained missing files is a save-path issue, while Qwen3.5 MoE from_pretrained is a load-path/model issue." + "reason": "Both are import errors, but one is a third-party sentence-transformers compatibility issue and the other is a missing model class export." }, { "left": "issue:30333", "right": "issue:33290", "accept": false, - "reason": "MLflow reporting on failed training jobs and DeepSpeed OOM with Adafactor are unrelated." + "reason": "MLFlow reporting on training failure is unrelated to DeepSpeed OOM with Adafactor." }, { "left": "issue:44734", "right": "issue:45103", "accept": false, - "reason": "Serve KV-cache indexing crash and auto_docstring future-annotations crash are unrelated." + "reason": "Serving KV-cache indexing crash and auto_docstring annotation handling are unrelated subsystems." }, { "left": "issue:43723", "right": "issue:44568", "accept": false, - "reason": "AutoTokenizer.from_pretrained loading issue is distinct from add_special_tokens tokenization behavior." + "reason": "Tokenizer loading in v5 and missing BOS/EOS insertion are distinct tokenizer bugs." }, { "left": "issue:42898", "right": "issue:45356", "accept": false, - "reason": "Tokenizer cleanup flag behavior and Kimi-K2.5 codec/fix_mistral_regex regression are different bugs." + "reason": "Both are tokenizer regressions, but they concern different models and different codec/regex behavior." }, { "left": "issue:42898", "right": "issue:44779", "accept": false, - "reason": "Both affect tokenizers, but the concrete failure modes and model-specific paths differ." + "reason": "General tokenizer behavior change versus DeepSeek-specific incorrect tokenization; not the same bug." }, { "left": "issue:44060", "right": "issue:45125", "accept": false, - "reason": "Qwen3-Next tied-weights warning bug is unrelated to missing _tp_plan for tensor parallelism." + "reason": "Both are Qwen3.5 MoE-related, but one is a tied-weights warning and the other is missing tensor-parallel plan support." }, { "left": "issue:45405", "right": "issue:45712", "accept": false, - "reason": "Dependency version bump and leftover dummy classes are unrelated repository-maintenance issues." + "reason": "A version pin to unreleased PEFT is unrelated to leftover dummy classes in repo checks." }, { "left": "issue:43874", "right": "issue:45092", "accept": false, - "reason": "Different multimodal-model incompatibilities: GLM46V image-patch accounting vs InternVL2 meta-init failure." + "reason": "Both involve multimodal model loading, but one is a missing image-patch method and the other is remote-code/meta-init incompatibility." }, { "left": "issue:42898", "right": "issue:44448", "accept": false, - "reason": "Both mention version regressions, but one is tokenization cleanup and the other is Pegasus output differences." + "reason": "Tokenizer cleanup behavior in v5 versus Pegasus output differences between v4 and v5; related theme, different concrete bug." }, { "left": "issue:43479", "right": "issue:44079", "accept": false, - "reason": "Config default initialization bug and ModelOutput key assignment bug are different abstractions." + "reason": "Config default initialization and ModelOutput key assignment are different code paths and failure modes." }, { "left": "issue:44188", "right": "issue:44964", "accept": false, - "reason": "Attention-kernel divergence under compile and Phi-4 multimodal loading are unrelated." + "reason": "Attention-kernel divergence under compile is unrelated to Phi-4 multimodal model loading." + }, + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "Position_ids buffer access and DeepSpeed double optimizer step are different runtime bugs with no shared code path." } ] }, @@ -11255,7 +11118,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:22355", "issue:28282", @@ -11269,12 +11132,13 @@ "issue:42175", "issue:42757", "issue:42831", + "issue:42886", "issue:42898", + "issue:42994", "issue:43295", "issue:43479", "issue:43644", "issue:43723", - "issue:43824", "issue:43873", "issue:43976", "issue:44206", @@ -11306,14 +11170,13 @@ "issue:45712" ], "soft_pairs": [ - "issue:44295|issue:45656", "issue:42898|issue:43295", "issue:44933|issue:45712", "issue:42175|issue:44933", "issue:41628|issue:45459", "issue:44222|issue:44297", - "issue:43976|issue:44933", "issue:44297|issue:45656", + "issue:43976|issue:44933", "issue:43479|issue:44869", "issue:44206|issue:45103", "issue:42831|issue:44623", @@ -11335,251 +11198,245 @@ "issue:44514|issue:45103", "issue:44295|issue:44623", "issue:43723|issue:44871", - "issue:44261|issue:45092", "issue:45092|issue:45357", + "issue:44261|issue:45092", "issue:43479|issue:44913", "issue:30990|issue:38175", - "issue:41628|issue:43824", - "issue:22355|issue:28282" + "issue:22355|issue:28282", + "issue:42886|issue:42994" ], - "prepared_review_unit_hash": "987fbbd1e2e49d14772c2ff8e01506702d43c26595030c4dccead3baab22d800", + "prepared_review_unit_hash": "a5aba265cf0e7fbb56eb585d9afb44d3040b518553c02cf74815689adb2f9763", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 35, - "serialized_chars": 37442, - "estimated_input_tokens": 9361, - "estimated_eval_tokens": 18978 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37655, + "estimated_input_tokens": 9414, + "estimated_eval_tokens": 19084 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is not a duplicate cluster: it mixes many unrelated import, packaging, tokenizer, save/load, model-config, serving, and training bugs. The soft links are mostly broad lexical similarity, not the same underlying issue.", - "confidence": 0.96, - "canonical_issue_reason": "No single issue is a good canonical anchor; the cluster spans unrelated regressions across imports, saving/loading, training, and model behavior.", + "summary": "The cluster is heterogeneous: it mixes unrelated import errors, save/load regressions, training/runtime bugs, and a few API/serialization issues. None of the soft pairs look like the same underlying defect, so I would not merge any of them as duplicates.", + "confidence": 0.69, + "canonical_issue_reason": "If one issue must anchor the cluster, issue 44933 is the cleanest narrow import/export failure and the closest representative of the small API-surface breakage subset. That said, the overall set is too diverse for a strong duplicate canonical.", "canonical_pr_reason": null, - "best_issue_reason": "No global best issue: none of these items clearly represents the others, so selecting one would be arbitrary.", + "best_issue_reason": "Issue 44933 is the best single representative only because it is concrete, scoped, and closest to several other import/export-themed reports; it is not a true umbrella for the whole cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44295", - "right": "issue:45656", - "accept": false, - "reason": "Position_ids buffer read error and double optimizer step under deepspeed are unrelated bugs." - }, { "left": "issue:42898", "right": "issue:43295", "accept": false, - "reason": "Tokenization cleanup behavior and processor.tokenizer access are different regression surfaces." + "reason": "Different regressions: tokenization cleanup behavior vs processor/tokenizer access in a multimodal custom model path." }, { "left": "issue:44933", "right": "issue:45712", "accept": false, - "reason": "Both are import-related, but one is a missing image_utils symbol and the other is dummy class leakage." + "reason": "Both touch import/export surfaces, but one is a missing import in image_utils and the other is dummy class leakage in a different file; not the same bug." }, { "left": "issue:42175", "right": "issue:44933", "accept": false, - "reason": "Backend extra packaging issue and missing image_utils import are different problems." + "reason": "Packaging/backend dependency omission is unrelated to a missing import from image_utils." }, { "left": "issue:41628", "right": "issue:45459", "accept": false, - "reason": "AutoImageProcessor import failure and tokenizer protobuf error masking are unrelated." + "reason": "AutoImageProcessor import failure and protobuf-related tokenizer error handling are different code paths." }, { "left": "issue:44222", "right": "issue:44297", "accept": false, - "reason": "FP8 save_pretrained/moe bug and tokenizer_class mismatch in tokenizer_config are different save-related issues." + "reason": "Both involve save_pretrained, but one is FP8/MoE save failure and the other is tokenizer metadata mismatch." }, { - "left": "issue:43976", - "right": "issue:44933", + "left": "issue:44297", + "right": "issue:45656", "accept": false, - "reason": "Python version support problem and missing image_utils import are unrelated." + "reason": "Tokenizer save metadata bug is unrelated to a deepspeed optimizer step-count issue." }, { - "left": "issue:44297", - "right": "issue:45656", + "left": "issue:43976", + "right": "issue:44933", "accept": false, - "reason": "Tokenizer config mismatch and optimizer step duplication are unrelated." + "reason": "Python-version compatibility failure does not match a missing internal import." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Phi4 multimodal config defaults and Whisper timestamp decode crash are different code paths." + "reason": "A multimodal config default-init bug is unrelated to Whisper timestamp decoding crashing on a trailing replacement character." }, { "left": "issue:44206", "right": "issue:45103", "accept": false, - "reason": "Unsupported center arg in a feature extractor and auto_docstring annotation crash are unrelated." + "reason": "Feature-extractor argument regression and auto_docstring annotation parsing crash are different failures." }, { "left": "issue:42831", "right": "issue:44623", "accept": false, - "reason": "FP8 accuracy issue and missing processor.save_pretrained files are different bugs." + "reason": "FP8 accuracy problem is not the same as processor.save_pretrained missing files." }, { "left": "issue:44222", "right": "issue:44623", "accept": false, - "reason": "Both mention saving, but one is FP8 MoE serialization and the other is processor artifact omission." + "reason": "Different save_pretrained bugs: FP8/MoE serialization versus missing processor assets." }, { "left": "issue:30333", "right": "issue:36331", "accept": false, - "reason": "MLflow job-state reporting and CustomTrainer.compute_loss signature mismatch are different training issues." + "reason": "MLflow reporting on failed training jobs is unrelated to a CustomTrainer compute_loss signature break." }, { "left": "issue:41628", "right": "issue:42757", "accept": false, - "reason": "AutoImageProcessor import failure and huggingface_hub is_offline_mode import error are different missing-import problems." + "reason": "Different import failures from different dependencies; one is transformers API exposure, the other is huggingface_hub symbol availability." }, { "left": "issue:30333", "right": "issue:34689", "accept": false, - "reason": "MLflow reporting and Llama 3.2 model loading regression are unrelated." + "reason": "Training-job tracking and Llama 3.2 vision model loading are unrelated regressions." }, { "left": "issue:43723", "right": "issue:44977", "accept": false, - "reason": "AutoTokenizer loading in v5 and Qwen3.5 flash-attention generation are different model/tokenizer issues." + "reason": "Tokenizer loading in v5 and Qwen3.5 flash-attention generation failures are distinct model/runtime issues." }, { "left": "issue:44297", "right": "issue:45310", "accept": false, - "reason": "Tokenizer_class save mismatch and Qwen3.5 MoE from_pretrained failure are unrelated." + "reason": "Tokenizer save metadata mismatch is not the same as Qwen3.5 MoE from_pretrained failing on newer transformers." }, { "left": "issue:28282", "right": "issue:30333", "accept": false, - "reason": "Missing torch for AutoModel and MLflow reporting of failed jobs are completely different bugs." + "reason": "Missing PyTorch dependency for AutoModel is unrelated to MLflow job reporting on failure." }, { "left": "issue:30333", "right": "issue:35141", "accept": false, - "reason": "MLflow reporting and output embedding reinitialization are unrelated." + "reason": "Training-job reporting and output-embedding reinitialization during token resize are different bugs." }, { "left": "issue:44263", "right": "issue:44360", "accept": false, - "reason": "torch.split return-value issue and missing ReLU in the DSA indexer are different implementation bugs." + "reason": "Both mention the DSA indexer, but one is about torch.split return handling and the other about a missing ReLU; not the same defect." }, { "left": "issue:44222", "right": "issue:44484", "accept": false, - "reason": "FP8 save_pretrained MoE bug and max_shard_size default discussion are unrelated." + "reason": "FP8 MoE save failure and max_shard_size default question are not the same issue." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Quantization offloading behavior and pipeline-parallelism output errors are different problems." + "reason": "Quantization offloading behavior and pipeline-parallelism output correctness are separate runtime problems." }, { "left": "issue:44483", "right": "issue:45406", "accept": false, - "reason": "Chat/completions request rejection and Gemma4Processor missing _tokenizer are unrelated serving bugs." + "reason": "OpenAI chat endpoint request rejection and Gemma4Processor missing _tokenizer in serve are different failures." }, { "left": "issue:44263", "right": "issue:45092", "accept": false, - "reason": "torch.split behavior in an indexer and old InternVL2 meta-init incompatibility are unrelated." + "reason": "DSA indexer return-value issue is unrelated to remote-code checkpoints failing meta initialization in v5." }, { "left": "issue:44373", "right": "issue:44485", "accept": false, - "reason": "Wrong docstring for position_ids and GLM-5 RoPE implementation are not the same issue." + "reason": "Docstring correctness and GLM-5 RoPE implementation are unrelated." }, { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Phi4 config initialization and junk-filled non-persistent buffers are different bugs." + "reason": "Config initialization for Phi4Multimodal is not the same as junk-filled non-persistent buffers in Transformers 5.0.0." }, { "left": "issue:44514", "right": "issue:45103", "accept": false, - "reason": "Batched apply_chat_template crash and auto_docstring future-annotations crash are unrelated." + "reason": "Batched apply_chat_template crash and auto_docstring annotation crash affect different subsystems." }, { "left": "issue:44295", "right": "issue:44623", "accept": false, - "reason": "Position_ids buffer access error and processor.save_pretrained missing files are different save/buffer bugs." + "reason": "Reading position_ids after buffer registration is unrelated to processor save_pretrained missing files." }, { "left": "issue:43723", "right": "issue:44871", "accept": false, - "reason": "AutoTokenizer loading regression and Gemma eos_token_id inconsistency are different tokenizer/config problems." + "reason": "Tokenizer loading in v5 and Gemma eos_token_id inconsistency are different model/config issues." }, { - "left": "issue:44261", - "right": "issue:45092", + "left": "issue:45092", + "right": "issue:45357", "accept": false, - "reason": "Missing rms_norm_eps in q_a_layernorm and InternVL2 meta-init incompatibility are unrelated." + "reason": "Old InternVL2 meta-init incompatibility and Qwen3.5 visual encoder key saving are separate checkpoint/serialization bugs." }, { - "left": "issue:45092", - "right": "issue:45357", + "left": "issue:44261", + "right": "issue:45092", "accept": false, - "reason": "Old InternVL2 meta-init incompatibility and Qwen3.5 visual-key save bug are different multimodal regressions." + "reason": "Missing rms_norm_eps precision issue in MLA q_a_layernorm is unrelated to remote-code meta initialization compatibility." }, { "left": "issue:43479", "right": "issue:44913", "accept": false, - "reason": "Phi4 multimodal defaults and GPTNeoX rotary_pct reload behavior are unrelated config bugs." + "reason": "Phi4Multimodal config initialization and GPTNeoXConfig rotary_pct reloading are different config serialization bugs." }, { "left": "issue:30990", "right": "issue:38175", "accept": false, - "reason": "Sentence Transformers loading hang and SigLIP2 zero probabilities are different model issues." + "reason": "SentenceTransformers loading stall and zero probabilities in siglip2-base are different model-runtime symptoms." }, { - "left": "issue:41628", - "right": "issue:43824", + "left": "issue:22355", + "right": "issue:28282", "accept": false, - "reason": "AutoImageProcessor import failure and missing Qwen2_5_VLForConditionalGeneration import are different missing-symbol issues." + "reason": "Missing transformers.onnx module and AutoModel requiring PyTorch are separate missing-dependency/import issues." }, { - "left": "issue:22355", - "right": "issue:28282", + "left": "issue:42886", + "right": "issue:42994", "accept": false, - "reason": "Missing transformers.onnx module and AutoModel requiring torch are unrelated import errors." + "reason": "Offline tokenizer cache loading and quantized model saving failure are different behaviors." } ] }, @@ -11601,7 +11458,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:29127", "issue:30990", @@ -11612,7 +11469,6 @@ "issue:42831", "issue:42886", "issue:42907", - "issue:42994", "issue:43066", "issue:43116", "issue:43122", @@ -11623,6 +11479,7 @@ "issue:43723", "issue:43756", "issue:43784", + "issue:43824", "issue:43937", "issue:44038", "issue:44060", @@ -11653,7 +11510,7 @@ "issue:45753" ], "soft_pairs": [ - "issue:42886|issue:42994", + "issue:41628|issue:43824", "issue:42831|issue:45310", "issue:44263|issue:44373", "issue:44945|issue:45753", @@ -11665,8 +11522,8 @@ "issue:43232|issue:45593", "issue:42371|issue:44263", "issue:30990|issue:33357", - "issue:44704|issue:45290", "issue:43756|issue:44060", + "issue:44704|issue:45290", "issue:42907|issue:44945", "issue:43122|issue:44514", "issue:44295|issue:44297", @@ -11676,249 +11533,243 @@ "issue:43408|issue:43475", "issue:43784|issue:45600", "issue:44261|issue:44373", - "issue:41628|issue:45278", "issue:43122|issue:44521", "issue:42907|issue:43937", + "issue:41628|issue:45278", "issue:44360|issue:44485", "issue:43232|issue:44743", - "issue:43116|issue:45200", "issue:44625|issue:44829", + "issue:43116|issue:45200", "issue:43122|issue:44779", "issue:29127|issue:30990", "issue:44038|issue:44297" ], - "prepared_review_unit_hash": "81fc8fbd7ec89e2520aef8d4e649b2adceb2098f1158e07c899d3201d6876f4d", + "prepared_review_unit_hash": "d78fba7e157338a690a75b506207bee0924de36e3c1adf4df45be14dc97b3cd5", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37185, - "estimated_input_tokens": 9297, - "estimated_eval_tokens": 18850 + "serialized_chars": 37250, + "estimated_input_tokens": 9313, + "estimated_eval_tokens": 18882 }, "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. There are a few loose v5/import/tokenizer regressions, but none of the soft-edge pairs look like the same underlying bug or change.", - "confidence": 0.9, - "canonical_issue_reason": "issue:45278 is the broadest umbrella for the shared v5 import-regression theme; most other items are narrower or unrelated model/config bugs.", + "summary": "This cluster is mostly a grab-bag of unrelated Transformers v5 regression reports. The soft-similarity pairs are mostly thematic matches only; I would not merge any of them as true duplicates.", + "confidence": 0.74, + "canonical_issue_reason": "issue:45278 is the broadest umbrella report in the set (many import errors after upgrading to v5.5.0), so it is the best single issue to represent the cluster\u2019s main regression theme.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most representative standalone issue because it summarizes a general post-upgrade import failure pattern rather than a single narrow symptom.", + "best_issue_reason": "issue:45278 is the most representative issue because it spans the cluster\u2019s dominant failure mode\u2014package import regressions after a v5 upgrade\u2014and is broader than the narrower, model-specific or API-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:42886", - "right": "issue:42994", + "left": "issue:41628", + "right": "issue:43824", "accept": false, - "reason": "Offline tokenizer cache loading and quantized model saving are different code paths and different bugs." + "reason": "Both are import errors, but one is root-level AutoImageProcessor export and the other is Qwen2.5-VL class export; different missing symbols and likely different fixes." }, { "left": "issue:42831", "right": "issue:45310", "accept": false, - "reason": "FP8 accuracy regression and Qwen3.5 from_pretrained import failure are unrelated symptoms." + "reason": "FineGrainedFP8 accuracy drift and Qwen3.5-MoE from_pretrained failure are unrelated bug classes." }, { "left": "issue:44263", "right": "issue:44373", "accept": false, - "reason": "A torch.split bug and a wrong docstring are not the same underlying issue." + "reason": "One is a torch.split return-value issue in a DSA indexer; the other is a docstring problem for position_ids." }, { "left": "issue:44945", "right": "issue:45753", "accept": false, - "reason": "Pipeline-parallel output corruption and an image-triggered infinite loop are distinct failures." + "reason": "Pipeline-parallelism output corruption and a Qwen3_5 image-specific infinite loop are different runtime failures." }, { "left": "issue:42831", "right": "issue:44038", "accept": false, - "reason": "An FP8 accuracy issue and a Qwen3-VL-Moe bug do not appear to share the same concrete defect." + "reason": "These are different model-family bugs: FP8 accuracy regression vs Qwen3-VL-Moe behavior in v5." }, { "left": "issue:42757", "right": "issue:45278", "accept": false, - "reason": "Both involve imports, but one is a specific huggingface_hub symbol error while the other is a broad v5 import regression." + "reason": "The second is a broad import-regression report, but the first is a specific missing is_offline_mode symbol; not enough evidence of the same underlying bug." }, { "left": "issue:43576", "right": "issue:43784", "accept": false, - "reason": "Broken env command and a sentence-transformers import NameError are different failures." + "reason": "Broken env command and a NameError while importing sentence-transformers are unrelated." }, { "left": "issue:42886", "right": "issue:43066", "accept": false, - "reason": "Offline cache loading and wrong tokenizer decoder type are separate tokenizer issues." + "reason": "Offline cache loading failure and wrong tokenizer decoder type are tokenizer bugs, but not the same failure or code path." }, { "left": "issue:44117", "right": "issue:44704", "accept": false, - "reason": "TOKENIZER_MAPPING_NAMES handling and AutoProcessor kwarg forwarding are different from_pretrained paths." + "reason": "TOKENIZER_MAPPING_NAMES returning None and AutoProcessor kwargs forwarding are different loader issues." }, { "left": "issue:43232", "right": "issue:45593", "accept": false, - "reason": "Generation kwarg update logic and D-FINE auxiliary-loss behavior are unrelated." + "reason": "Generation kwargs state handling and D-FINE auxiliary-loss behavior are unrelated." }, { "left": "issue:42371", "right": "issue:44263", "accept": false, - "reason": "TF32 settings guidance and a GlmMoeDsaIndexer split issue are unrelated." + "reason": "TF32 settings guidance and a torch.split/indexer bug do not match." }, { "left": "issue:30990", "right": "issue:33357", "accept": false, - "reason": "Both are loading problems, but they affect different models and different failure modes." + "reason": "Sentence-Transformers loading hang and a MacOS bus error with a CLIP model are different problems." }, { - "left": "issue:44704", - "right": "issue:45290", + "left": "issue:43756", + "right": "issue:44060", "accept": false, - "reason": "AutoProcessor cached_file kwargs and chat-template tool-call crashes are different code paths." + "reason": "Smollm3 RoPE-layer dropping and Qwen3-Next tied-weights warnings are unrelated model-implementation issues." }, { - "left": "issue:43756", - "right": "issue:44060", + "left": "issue:44704", + "right": "issue:45290", "accept": false, - "reason": "Smollm3 RoPE layer handling and Qwen3-Next tied-weights warnings are unrelated model implementation bugs." + "reason": "AutoProcessor cached_file kwargs forwarding and apply_chat_template crashing on tool-call messages are separate API bugs." }, { "left": "issue:42907", "right": "issue:44945", "accept": false, - "reason": "Quantized model save failures and pipeline-parallel output errors are not the same bug." + "reason": "Saving dequantized Ministrals/Devstrals and pipeline-parallelism output errors are not the same code path." }, { "left": "issue:43122", "right": "issue:44514", "accept": false, - "reason": "Tokenizer output drift and multimodal chat-template padding crashes are different issues." + "reason": "Different tokenization across versions and a batched multimodal apply_chat_template crash are distinct regressions." }, { "left": "issue:44295", "right": "issue:44297", "accept": false, - "reason": "Buffer reading for position_ids and tokenizer_config save mismatch are unrelated." + "reason": "Reading position_ids after buffer registration and tokenizer_class mismatch on save are unrelated serialization issues." }, { "left": "issue:43232", "right": "issue:45245", "accept": false, - "reason": "Generation state update logic and a category-cardinality runtime error do not match." + "reason": "Generation sync_gpus handling and a category-cardinality runtime error are unrelated." }, { "left": "issue:43723", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions in v5, but one is load failure and the other is incorrect tokenization output." + "reason": "Both involve tokenizer loading in v5, but one is AutoTokenizer load failure and the other is incorrect Deepseek tokenization output; not the same bug." }, { "left": "issue:43723", "right": "issue:44442", "accept": false, - "reason": "Different tokenizer classes failing to load is too broad to be the same concrete bug." + "reason": "AutoTokenizer load failure and FastSpeech2ConformerTokenizer load failure are related only at a very high level." }, { "left": "issue:43408", "right": "issue:43475", "accept": false, - "reason": "Model-type warning and missing attribute in Sam3VisionEncoderOutput are different SAM3 problems." + "reason": "SAM3 model-type warning and missing fpn_position_embeddings attribute are different SAM 3 issues." }, { "left": "issue:43784", "right": "issue:45600", "accept": false, - "reason": "Sentence-transformers import failure and Sam3LiteText auto_mapping CI breakage are unrelated." + "reason": "Sentence-transformers import NameError and stale auto_mappings references to removed Sam3LiteText configs are unrelated." }, { "left": "issue:44261", "right": "issue:44373", "accept": false, - "reason": "A precision/config bug and a docstring issue are not duplicates." - }, - { - "left": "issue:41628", - "right": "issue:45278", - "accept": false, - "reason": "Missing AutoImageProcessor is one specific import break; the other is a broader set of import errors." + "reason": "A precision issue in MLA q_a_layernorm and a docstring issue for position_ids are not duplicates." }, { "left": "issue:43122", "right": "issue:44521", "accept": false, - "reason": "Tokenizer behavior drift and zeroed assistant masks for multimodal chat templates are different bugs." + "reason": "Both involve apply_chat_template, but one is a version-to-version tokenization change and the other is all-zero assistant masks for multimodal inputs." }, { "left": "issue:42907", "right": "issue:43937", "accept": false, - "reason": "Saving dequantized models and invalid GenerationConfig are unrelated." + "reason": "Saving dequantized models and invalid GenerationConfig for GLM-5 are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "The latter is a broad import-regression report, but the former is a specific missing AutoImageProcessor export; not enough to call them the same issue." }, { "left": "issue:44360", "right": "issue:44485", "accept": false, - "reason": "A missing ReLU in the DSA indexer and GLM-5 RoPE implementation discussion are not the same defect." + "reason": "A missing ReLU in the DSA indexer and GLM-5 RoPE implementation issues are different." }, { - "left": "issue:44261", - "right": "issue:44373", + "left": "issue:43232", + "right": "issue:44743", "accept": false, - "reason": "MLA q_a_layernorm precision/config handling and a wrong docstring do not match." + "reason": "Generation state update logic and Qwen3.5 recurrent-state reset under cache are separate bugs." }, { - "left": "issue:41628", - "right": "issue:45278", + "left": "issue:44625", + "right": "issue:44829", "accept": false, - "reason": "AutoImageProcessor import removal and general upgrade import errors are related only at a high level, not as one concrete bug." + "reason": "num_labels propagation in Qwen3.5 and flash_attention_3 training degeneration are unrelated." }, { "left": "issue:43116", "right": "issue:45200", "accept": false, - "reason": "Empty multi-label classification results in an example script and Gemma 4 token-type defaults are different issues." - }, - { - "left": "issue:44625", - "right": "issue:44829", - "accept": false, - "reason": "num_labels propagation and flash_attention_3 training degeneration are unrelated model behaviors." + "reason": "Multi-label example-script behavior and Gemma 4 mm_token_type_ids defaults are different concerns." }, { - "left": "issue:43723", + "left": "issue:43122", "right": "issue:44779", "accept": false, - "reason": "Tokenizer load failure and tokenizer output regression may both involve v5, but not the same specific bug." + "reason": "Tokenization drift across versions and Deepseek tokenizer correctness regressions are not the same underlying bug." }, { "left": "issue:29127", "right": "issue:30990", "accept": false, - "reason": "LayoutLMv3 box-validation messaging and Sentence Transformers loading are unrelated." + "reason": "A LayoutLMv3 error message clarity issue is unrelated to Sentence Transformers loading hang." }, { "left": "issue:44038", "right": "issue:44297", "accept": false, - "reason": "Qwen3-VL-Moe bug reports and tokenizer save metadata mismatch are different subsystems." + "reason": "Qwen3-VL-Moe bug report and tokenizer save_pretrained class mismatch are different." } ] }, @@ -11940,8 +11791,10 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ + "issue:22355", + "issue:29127", "issue:33453", "issue:36296", "issue:38617", @@ -11969,7 +11822,6 @@ "issue:43992", "issue:44038", "issue:44075", - "issue:44117", "issue:44222", "issue:44295", "issue:44351", @@ -11988,7 +11840,6 @@ "issue:45446", "issue:45478", "issue:45563", - "issue:45588", "issue:45753" ], "soft_pairs": [ @@ -11996,17 +11847,17 @@ "issue:42907|issue:43452", "issue:44038|issue:44295", "issue:39401|issue:44488", - "issue:43824|issue:44351", "issue:36296|issue:43116", + "issue:43824|issue:44351", "issue:44038|issue:44222", "issue:43066|issue:44779", "issue:39401|issue:44521", - "issue:42757|issue:44351", - "issue:45216|issue:45478", "issue:43122|issue:44977", + "issue:45216|issue:45478", + "issue:42757|issue:44351", "issue:43992|issue:44704", - "issue:43784|issue:45446", "issue:42907|issue:44222", + "issue:43784|issue:45446", "issue:43937|issue:44485", "issue:42222|issue:43441", "issue:43976|issue:45446", @@ -12022,243 +11873,243 @@ "issue:45083|issue:45446", "issue:43937|issue:45563", "issue:42175|issue:44938", - "issue:38617|issue:45278", "issue:33453|issue:39401", "issue:39401|issue:44779", - "issue:44117|issue:45588" + "issue:38617|issue:45278", + "issue:22355|issue:29127" ], - "prepared_review_unit_hash": "1adc32386fb9b3000880303e666b4a119860e804276c0dc53131931bcfe21d97", + "prepared_review_unit_hash": "6b631e4eb0e48ffa2e7e10566df912707e6da446a7c1f2ee9fc0ab5e5282b49e", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37344, - "estimated_input_tokens": 9336, - "estimated_eval_tokens": 18928 + "serialized_chars": 37298, + "estimated_input_tokens": 9325, + "estimated_eval_tokens": 18906 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a loose cluster of Transformers v5-era regressions spanning tokenizer behavior, model loading, import errors, and a few training/generation issues. The soft pairs mostly share only broad subsystem similarity, not the same concrete bug or change, so none should be merged as duplicates.", - "confidence": 0.74, - "canonical_issue_reason": "Issue 33453 is the broadest representative of the repeated loading/tokenizer regression theme, though the overall cluster is still too heterogeneous to be a true duplicate set.", + "summary": "This cluster is heterogeneous: the soft pairs span unrelated import errors, tokenizer regressions, multimodal/model-specific bugs, training issues, and warning/message issues. I reject all soft edges because none clearly share the same underlying bug or code change.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue stands out because the items do not form one duplicate family; they cover many distinct bugs across different APIs, models, and subsystems.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 33453 is the most suitable global anchor because it is generic and closest to the recurring load/tokenizer failure pattern across the set.", + "best_issue_reason": "No best issue for the cluster overall: it is not coherent enough to justify choosing one issue as the representative duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42907", "right": "issue:44964", "accept": false, - "reason": "Dequantized save failure vs multimodal model load failure; different code paths and likely different root causes." + "reason": "Different failures: saving/dequantizing a model versus loading a multimodal model." }, { "left": "issue:42907", "right": "issue:43452", "accept": false, - "reason": "Saving dequantized models and gguf AutoTokenizer/AutoModel loading are unrelated symptoms." + "reason": "Unrelated code paths: dequantized save failure vs gguf/AutoTokenizer loading." }, { "left": "issue:44038", "right": "issue:44295", "accept": false, - "reason": "Qwen3-VL-Moe FlashAttention issue vs position_ids buffer access error; not the same bug." + "reason": "Different problems: Qwen3-VL-Moe behavior versus position_ids buffer access." }, { "left": "issue:39401", "right": "issue:44488", "accept": false, - "reason": "Different models and different failures: offset_mapping bug vs failing to load a specific tokenizer model." + "reason": "Tokenizer offset-mapping bug is unrelated to loading a specific model checkpoint." }, { - "left": "issue:43824", - "right": "issue:44351", + "left": "issue:36296", + "right": "issue:43116", "accept": false, - "reason": "Both are import errors, but for different missing symbols from different dependency surfaces." + "reason": "Tensor-parallel training bug and empty multi-label classification results are unrelated." }, { - "left": "issue:36296", - "right": "issue:43116", + "left": "issue:43824", + "right": "issue:44351", "accept": false, - "reason": "Tensor parallel training bug and multi-label classification empty outputs are unrelated." + "reason": "Both are import errors, but for different missing symbols and different underlying fixes." }, { "left": "issue:44038", "right": "issue:44222", "accept": false, - "reason": "Same broad Qwen3-family area, but one is a FlashAttention path issue and the other is FP8 save_pretrained behavior." + "reason": "Same broad model family, but one is a model loading bug and the other is an FP8 save_pretrained issue." }, { "left": "issue:43066", "right": "issue:44779", "accept": false, - "reason": "Wrong decoder type in v5 vs Deepseek tokenizer incorrect results; different tokenizer bugs." + "reason": "Different tokenizer regressions affecting different models and behaviors." }, { "left": "issue:39401", "right": "issue:44521", "accept": false, - "reason": "Tokenizer offset mapping regression vs multimodal assistant_masks issue; distinct functionality." + "reason": "Offset mapping regression is unrelated to assistant mask generation for multimodal inputs." }, { - "left": "issue:42757", - "right": "issue:44351", + "left": "issue:43122", + "right": "issue:44977", "accept": false, - "reason": "Different missing imports from different packages/modules; not one underlying defect." + "reason": "Different symptoms: tokenization change across versions versus flash-attention generation failure." }, { "left": "issue:45216", "right": "issue:45478", "accept": false, - "reason": "Both involve Qwen3.5, but save_pretrained checkpoint corruption and from_pretrained loading errors are different code paths." + "reason": "Both involve Qwen3.5, but one is about save_pretrained checkpoint contents and the other about from_pretrained loading." }, { - "left": "issue:43122", - "right": "issue:44977", + "left": "issue:42757", + "right": "issue:44351", "accept": false, - "reason": "General tokenization difference across versions vs Qwen3.5 flash-attention generation failure; not the same bug." + "reason": "Different missing imports from different packages; no shared code-path bug is evident." }, { "left": "issue:43992", "right": "issue:44704", "accept": false, - "reason": "Missing embed_tokens weight and not forwarding kwargs to cached_file are separate loading issues." + "reason": "Model weight loading bug and AutoProcessor kwargs forwarding are separate issues." }, { - "left": "issue:43784", - "right": "issue:45446", + "left": "issue:42907", + "right": "issue:44222", "accept": false, - "reason": "Sentence-transformers import NameError vs PyTorch version check bug for AuxRequest import; unrelated." + "reason": "Both mention saving, but one is dequantized model saving and the other is FP8 moe save_pretrained." }, { - "left": "issue:42907", - "right": "issue:44222", + "left": "issue:43784", + "right": "issue:45446", "accept": false, - "reason": "Saving dequantized models vs FP8 save_pretrained moe are both save-related, but not the same concrete failure." + "reason": "Sentence-transformers import failure and a PyTorch version check bug are unrelated." }, { "left": "issue:43937", "right": "issue:44485", "accept": false, - "reason": "GenerationConfig validation error and GLM-5 RoPE discussion are different problems." + "reason": "Invalid GenerationConfig and RoPE implementation discussion are not the same bug." }, { "left": "issue:42222", "right": "issue:43441", "accept": false, - "reason": "Vitpose model breakage and Ministral-3 FlashAttention failure affect different models and paths." + "reason": "Different models and failures: vitpose breakage versus Ministral flash-attention support." }, { "left": "issue:43976", "right": "issue:45446", "accept": false, - "reason": "Python version compatibility issue vs AuxRequest import version check bug; different scopes." + "reason": "Python support/version constraint issue is separate from the flex_attention import check." }, { "left": "issue:42371", "right": "issue:43874", "accept": false, - "reason": "TF32 settings warning and missing image patch count method are unrelated." + "reason": "TF32 settings guidance and a missing image-patch-count method are unrelated." }, { "left": "issue:43493", "right": "issue:45753", "accept": false, - "reason": "SigLIP2 HF/JAX discrepancy vs Qwen3_5 infinite loop on a specific image; different model behavior bugs." + "reason": "SigLIP2 implementation discrepancy and Qwen3_5 image looping are different bugs." }, { "left": "issue:44075", "right": "issue:44936", "accept": false, - "reason": "Optimizer SGD args ignored and trainer.evaluate() failing after train() are separate training issues." + "reason": "Optimizer-argument handling is different from train/evaluate state corruption." }, { "left": "issue:42831", "right": "issue:45753", "accept": false, - "reason": "FineGrainedFP8 accuracy issue and Qwen3_5 infinite loop are not the same defect." + "reason": "FP8 accuracy issue and an infinite-loop image case are unrelated." }, { "left": "issue:43116", "right": "issue:43825", "accept": false, - "reason": "Multi-label classification output bug vs pipeline translation support error; different APIs and causes." + "reason": "Example-script classification behavior and pipeline translation error messaging are separate issues." }, { "left": "issue:44038", "right": "issue:45478", "accept": false, - "reason": "Same model family, but FlashAttention generation failure and from_pretrained error are different concrete code paths." + "reason": "Both are Qwen3.5/Moe-related, but one is a Qwen3-VL-Moe bug and the other is a from_pretrained load failure." }, { "left": "issue:44222", "right": "issue:45478", "accept": false, - "reason": "FP8 save_pretrained moe vs Qwen3.5 Moe from_pretrained error; distinct operations and failures." + "reason": "Both are Moe-related save/load issues, but they affect different APIs and likely different fixes." }, { "left": "issue:43116", "right": "issue:43493", "accept": false, - "reason": "Example-script classification bug and SigLIP2 implementation discrepancy are unrelated." + "reason": "Classification example output bug and SigLIP2 model discrepancy are unrelated." }, { "left": "issue:43064", "right": "issue:43673", "accept": false, - "reason": "Wrong optimizer states under FSDP2/PEFT and missing GenerationMixin cache are separate bugs." + "reason": "FSDP2 optimizer-state bug and missing GenerationMixin cache are separate training/generation issues." }, { "left": "issue:45083", "right": "issue:45446", "accept": false, - "reason": "Helper output-length behavior in qwen3_omni_moe vs PyTorch version check bug; not the same issue." + "reason": "Helper function behavior in qwen3_omni_moe is unrelated to the flex_attention import check." }, { "left": "issue:43937", "right": "issue:45563", "accept": false, - "reason": "GenerationConfig validation failure and stale paged-generate warning are different generation-path problems." + "reason": "Invalid generation config and a stale warning about num_return_sequences are different problems." }, { "left": "issue:42175", "right": "issue:44938", "accept": false, - "reason": "Missing TensorFlow backend in torch extra vs Python 3.14 load failure; unrelated packaging/runtime issues." - }, - { - "left": "issue:38617", - "right": "issue:45278", - "accept": false, - "reason": "Specific ImportError for layer_type_validation vs broad many-import-errors report; too broad to treat as one duplicate." + "reason": "Backend dependency packaging issue versus Python 3.14 load failure; no shared bug." }, { "left": "issue:33453", "right": "issue:39401", "accept": false, - "reason": "Generic tokenizer loading regression vs Qwen3 tokenizer offset_mapping bug; related area but not the same defect." + "reason": "Tokenizer loading regression and Qwen3 offset mapping issue are not the same defect." }, { "left": "issue:39401", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer-related, but they affect different models and different tokenization failures." + "reason": "Both are tokenization bugs, but they affect different models and different behaviors." }, { - "left": "issue:44117", - "right": "issue:45588", + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "Specific missing import and broad many-import-errors report are too coarse to treat as one bug." + }, + { + "left": "issue:22355", + "right": "issue:29127", "accept": false, - "reason": "TOKENIZER_MAPPING_NAMES None handling and flash_attention s_aux=None crash are unrelated." + "reason": "No shared underlying bug: missing transformers.onnx module vs an error-message clarity issue." } ] }, @@ -12280,10 +12131,8 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ - "issue:22355", - "issue:29127", "issue:38617", "issue:39401", "issue:41628", @@ -12302,6 +12151,7 @@ "issue:43577", "issue:43742", "issue:43827", + "issue:43874", "issue:43937", "issue:43986", "issue:44075", @@ -12329,227 +12179,235 @@ "issue:45459", "issue:45464", "issue:45542", + "issue:45588", "issue:45656" ], "soft_pairs": [ - "issue:22355|issue:29127", + "issue:44117|issue:45588", "issue:43827|issue:45446", "issue:43381|issue:44945", "issue:43937|issue:44977", - "issue:43450|issue:43986", "issue:45405|issue:45446", + "issue:43450|issue:43986", "issue:39401|issue:43421", "issue:44117|issue:45081", "issue:42673|issue:45072", "issue:38617|issue:41628", "issue:43408|issue:43742", - "issue:42222|issue:45542", "issue:44829|issue:45216", "issue:44291|issue:45071", - "issue:42175|issue:42222", + "issue:42222|issue:45542", "issue:44410|issue:45310", "issue:44484|issue:45310", + "issue:42175|issue:42222", "issue:43937|issue:44496", "issue:44483|issue:45464", "issue:44075|issue:45656", - "issue:44964|issue:45278", "issue:43066|issue:43122", + "issue:44964|issue:45278", "issue:43012|issue:43577", - "issue:38617|issue:44220", "issue:42371|issue:44261", - "issue:42222|issue:45459", - "issue:43097|issue:44493" + "issue:38617|issue:44220", + "issue:43097|issue:44493", + "issue:43874|issue:44261", + "issue:42222|issue:45459" ], - "prepared_review_unit_hash": "25b94212b1c185eaa79f1c38226a8cce40e5252ea783b52a97dcbf339d77b70e", + "prepared_review_unit_hash": "9831ece046567207fefb48eacc2ee6ad9ad40f42d1939da2a9ae306515f671e3", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 27, - "serialized_chars": 35710, - "estimated_input_tokens": 8928, - "estimated_eval_tokens": 18112 + "soft_pair_count": 28, + "serialized_chars": 36085, + "estimated_input_tokens": 9022, + "estimated_eval_tokens": 18300 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No strong duplicates here; the soft pairs are mostly superficial matches on broad themes like imports, tokenizers, or chat APIs, but they describe different failures and code paths.", - "confidence": 0.69, - "canonical_issue_reason": "issue:43097 is the clearest broadly applicable regression report in this set: concrete, closed, and tied to a v5 API break with downstream references.", + "summary": "These items are mostly unrelated transformer bugs/regressions across tokenizers, model loading, precision, and optimizer behavior. The soft pairs are all false-positive duplicates; none look like the same concrete bug or mergeable change.", + "confidence": 0.9, + "canonical_issue_reason": "No single canonical issue is a safe choice here: the set is heterogeneous, and the titles point to distinct failure modes rather than one underlying bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43097 is the best representative issue overall because it is specific enough to be actionable yet broad enough to stand in for nearby v5 compatibility complaints.", + "best_issue_reason": "If one issue must be kept as the broadest representative, issue:45278 is the most umbrella-like report ('many import errors after update'), but it still does not actually unify the other issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:22355", - "right": "issue:29127", + "left": "issue:44117", + "right": "issue:45588", "accept": false, - "reason": "Unrelated import failure vs. LayoutLMv3 error-message clarity issue." + "reason": "Different failures: tokenizer mapping fallback vs flash-attention crash on missing s_aux. No shared code-path bug is evident." }, { "left": "issue:43827", "right": "issue:45446", "accept": false, - "reason": "Docs/pipeline deprecation mismatch is unrelated to a PyTorch version check bug in flex_attention." + "reason": "Docs mentioning pipeline() and a flex_attention import-version check are unrelated." }, { "left": "issue:43381", "right": "issue:44945", "accept": false, - "reason": "Different problems: eval-mode gradient checkpointing vs. incorrect outputs under pipeline parallelism." + "reason": "Gradient checkpointing in eval mode and pipeline-parallelism output errors are different runtime problems." }, { "left": "issue:43937", "right": "issue:44977", "accept": false, - "reason": "Different model-specific failures; GenerationConfig validation is not the same as flash-attention generation breakage." + "reason": "GLM-5 GenerationConfig validation is not the same as Qwen3.5 flash-attention generation failure." }, { - "left": "issue:43450", - "right": "issue:43986", + "left": "issue:45405", + "right": "issue:45446", "accept": false, - "reason": "Both touch video processing, but one is a batched-shape bug and the other is a missing-torchvision import crash." + "reason": "A released-version pin issue and a PyTorch import/version check bug are separate dependency problems." }, { - "left": "issue:45405", - "right": "issue:45446", + "left": "issue:43450", + "right": "issue:43986", "accept": false, - "reason": "PEFT version pinning issue is unrelated to the flex_attention import-version check." + "reason": "Batched video shape handling and AutoProcessor crash without torchvision are distinct video-processing issues." }, { "left": "issue:39401", "right": "issue:43421", "accept": false, - "reason": "Tokenizer offset-mapping bug and runtime special-token post-processing bug are different code paths." + "reason": "Wrong offset mapping for Qwen3 tokenizer is not the same as runtime post-processor updates for special tokens." }, { "left": "issue:44117", "right": "issue:45081", "accept": false, - "reason": "TOKENIZER_MAPPING_NAMES fallback bug is unrelated to the Mistral regex patch AttributeError." + "reason": "Tokenizer mapping None handling and Mistral regex patch crashes are different tokenizer code paths." }, { "left": "issue:42673", "right": "issue:45072", "accept": false, - "reason": "VRAM leak in Qwen3ForCausalLM and dtype mismatches in other models are distinct issues." + "reason": "VRAM leak in Qwen3ForCausalLM and bfloat16 dtype mismatches are unrelated model/runtime bugs." }, { "left": "issue:38617", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but for different missing symbols and likely different regressions." + "reason": "Both are import errors, but for different symbols and likely different causes; not the same bug." }, { "left": "issue:43408", "right": "issue:43742", "accept": false, - "reason": "Model-type warning for sam3_tracker vs. key error loading MobileLLM are unrelated." - }, - { - "left": "issue:42222", - "right": "issue:45542", - "accept": false, - "reason": "VitPose breakage is unrelated to the TensorBoard-without-TensorFlow backend error." + "reason": "SAM3 model-type warning and MobileLLM loading KeyError are unrelated model-loading issues." }, { "left": "issue:44829", "right": "issue:45216", "accept": false, - "reason": "Training degradation with flash_attention_3 is not the same as a bad save_pretrained checkpoint regression." + "reason": "FlashAttention-3 training degeneration and Qwen3.5 save_pretrained regression do not share a concrete fix path." }, { "left": "issue:44291", "right": "issue:45071", "accept": false, - "reason": "Different compatibility regressions: unexpected init_empty_weights arg vs. PretrainedConfig type-checking breakage." + "reason": "init_empty_weights/_is_hf_initialized and PretrainedConfig type-checking breakage are separate v5 regressions." }, { - "left": "issue:42175", - "right": "issue:42222", + "left": "issue:42222", + "right": "issue:45542", "accept": false, - "reason": "Packaging/backend dependency issue is unrelated to VitPose model breakage." + "reason": "VitPose model breakage and tensorboard-without-TF backend errors are different backend/model issues." }, { "left": "issue:44410", "right": "issue:45310", "accept": false, - "reason": "Missing qwen3next projections and Qwen3.5 MoE from_pretrained errors are different model-loading failures." + "reason": "Missing qkv/gate projections in qwen3next is not the same as Qwen3.5 MoE from_pretrained failure." }, { "left": "issue:44484", "right": "issue:45310", "accept": false, - "reason": "max_shard_size behavior is unrelated to Qwen3.5 MoE loading failure." + "reason": "save_pretrained max_shard_size behavior is unrelated to Qwen3.5 MoE loading." + }, + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "TF not installed in a torch-only extra and VitPose breakage are separate dependency/model problems." }, { "left": "issue:43937", "right": "issue:44496", "accept": false, - "reason": "GenerationConfig validation and unrecognized-model config.json errors are different issues." + "reason": "Invalid GenerationConfig for GLM-5 and unrecognized model/config.json loading are different errors." }, { "left": "issue:44483", "right": "issue:45464", "accept": false, - "reason": "Same API surface, but one is a generic request rejection and the other is a streaming inference failure on a specific model." + "reason": "Both involve chat/completions, but one is an API request-format regression and the other is a streaming inference failure on a specific model." }, { "left": "issue:44075", "right": "issue:45656", "accept": false, - "reason": "Optimizer argument handling and double optimizer-step behavior are distinct bugs." + "reason": "Unused SGD arguments and double optimizer.step() under deepspeed are different optimizer issues." }, { - "left": "issue:44964", - "right": "issue:45278", + "left": "issue:43066", + "right": "issue:43122", "accept": false, - "reason": "Model loading failure for Phi-4 multimodal is not the same as a broad import-error regression after upgrade." + "reason": "Wrong tokenizer decoder type and changed tokenization output across versions are related thematically but not shown to be the same concrete bug." }, { - "left": "issue:43066", - "right": "issue:43122", + "left": "issue:44964", + "right": "issue:45278", "accept": false, - "reason": "Both are tokenizer regressions, but one is decoder-type mismatch and the other is changed tokenization output." + "reason": "A specific model-loading failure and a broad import-errors report are not the same underlying issue." }, { "left": "issue:43012", "right": "issue:43577", "accept": false, - "reason": "PyTorch precision warning on compile is different from BLIP2 dtype fields staying float32." + "reason": "PyTorch warning on compiling bfloat16 models and BLIP2 dtype retention are distinct precision/dtype bugs." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 settings guidance and MLA rms_norm_eps precision mismatch are unrelated." }, { "left": "issue:38617", "right": "issue:44220", "accept": false, - "reason": "ImportError for layer_type_validation is unrelated to the _torch_extract_fbank_features bug." + "reason": "ImportError for layer_type_validation and _torch_extract_fbank_features() are unrelated failures." }, { - "left": "issue:42371", - "right": "issue:44261", + "left": "issue:43097", + "right": "issue:44493", "accept": false, - "reason": "TF32 behavior-setting change is unrelated to the MLA q_a_layernorm precision/config issue." + "reason": "Removed tie_embeddings_and_encoder_decoder and unexpected position-id keys are different v5 regressions." }, { - "left": "issue:42222", - "right": "issue:45459", + "left": "issue:43874", + "right": "issue:44261", "accept": false, - "reason": "VitPose model breakage is unrelated to tokenizer error masking when protobuf is missing." + "reason": "Missing get_number_of_image_patches and missing rms_norm_eps are separate multimodal/model-config issues." }, { - "left": "issue:43097", - "right": "issue:44493", + "left": "issue:42222", + "right": "issue:45459", "accept": false, - "reason": "Removed tie_embeddings_and_encoder_decoder API and unexpected position-id keys are different regressions." + "reason": "VitPose model breakage and protobuf-hidden tokenizer errors are not the same bug." } ] }, @@ -12571,7 +12429,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:38617", "issue:41628", @@ -12623,20 +12481,19 @@ "issue:45588" ], "soft_pairs": [ - "issue:43874|issue:44261", "issue:44484|issue:45478", - "issue:44410|issue:45478", "issue:44263|issue:44485", - "issue:43012|issue:43408", + "issue:44410|issue:45478", "issue:43441|issue:44829", + "issue:43012|issue:43408", "issue:41628|issue:45071", "issue:45071|issue:45542", "issue:42673|issue:44977", "issue:42371|issue:42831", "issue:42673|issue:43906", - "issue:43986|issue:44484", "issue:44117|issue:44987", "issue:43097|issue:43452", + "issue:43986|issue:44484", "issue:43335|issue:45071", "issue:43931|issue:44038", "issue:43976|issue:44945", @@ -12649,255 +12506,255 @@ "issue:44062|issue:44117", "issue:43097|issue:45588", "issue:44410|issue:45375", - "issue:42673|issue:43577", "issue:38617|issue:44351", "issue:44704|issue:45103", + "issue:42673|issue:43577", "issue:44038|issue:44410", "issue:44261|issue:44485", - "issue:43986|issue:44117", "issue:44964|issue:45341", + "issue:43986|issue:44117", "issue:43931|issue:44410", "issue:43502|issue:43992", "issue:43097|issue:43873" ], - "prepared_review_unit_hash": "20f48114169e22b442ef10cf1ae973690da9eaee8f3f67bcfbb4967a1a9aa204", + "prepared_review_unit_hash": "3b488282b8f625eb1b6fead3af826e79dfd51b01d54279301e730a3c7992ce81", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 36, - "serialized_chars": 38467, - "estimated_input_tokens": 9617, - "estimated_eval_tokens": 19490 + "soft_pair_count": 35, + "serialized_chars": 38166, + "estimated_input_tokens": 9542, + "estimated_eval_tokens": 19340 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of unrelated Transformers regressions and import/loading issues. The only visible common thread is a loose v5-breaking-change / API-regression theme, but none of the soft pairs look like true duplicates of the same underlying bug.", - "confidence": 0.72, - "canonical_issue_reason": "issue:43097 is the best representative of the cluster\u2019s broad v5-regression theme: a concrete API removal with user-visible breakage, and it has the clearest cross-links to other nearby compatibility complaints.", - "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the strongest standalone triage target: it is open, actively discussed, has multiple inbound references, and describes a concrete high-impact behavior bug.", - "best_pr_reason": null, + "summary": "This cluster is a loose grab bag of unrelated Transformers issues across imports, model loading, training, quantization, and config bugs. None of the soft pairs look like the same underlying defect, so I would not collapse any of them.", + "confidence": 0.95, + "canonical_issue_reason": "No clear canonical issue: the items do not form one duplicate family and instead cover many unrelated regressions and feature bugs.", + "canonical_pr_reason": "No pull requests are present in the cluster.", + "best_issue_reason": "No single issue is a good global representative; the cluster is too heterogeneous to safely use one issue as the canonical duplicate target.", + "best_pr_reason": "No pull requests are present, so there is no best PR candidate.", "soft_edge_verdicts": [ { - "left": "issue:43874", - "right": "issue:44261", + "left": "issue:44484", + "right": "issue:45478", "accept": false, - "reason": "Different subsystems and failures: image-patch counting vs missing RMS norm epsilon in MLA." + "reason": "Different bugs: save_pretrained shard sizing vs Qwen3.5 MoE loading failure." }, { - "left": "issue:44484", - "right": "issue:45478", + "left": "issue:44263", + "right": "issue:44485", "accept": false, - "reason": "A save_pretrained shard-size question is unrelated to a Qwen3.5 MoE from_pretrained error." + "reason": "Unrelated failures: torch.split return handling vs GLM-5 RoPE implementation." }, { "left": "issue:44410", "right": "issue:45478", "accept": false, - "reason": "Different Qwen3 model bugs: missing projections vs loading failure." + "reason": "Different model/config problems: missing qwen3next projections vs Qwen3.5 MoE from_pretrained error." }, { - "left": "issue:44263", - "right": "issue:44485", + "left": "issue:43441", + "right": "issue:44829", "accept": false, - "reason": "torch.split/indexing behavior is unrelated to a RoPE implementation issue." + "reason": "Both touch flash attention, but one is a model load failure and the other is degenerate training; not the same bug." }, { "left": "issue:43012", "right": "issue:43408", "accept": false, - "reason": "A BF16 precision warning is not the same bug as a model-type mismatch warning." - }, - { - "left": "issue:43441", - "right": "issue:44829", - "accept": false, - "reason": "Both mention attention, but these are different concrete bugs in different code paths and symptoms." + "reason": "A PyTorch precision warning and a model-type warning are different issues." }, { "left": "issue:41628", "right": "issue:45071", "accept": false, - "reason": "ImportError for AutoImageProcessor is unrelated to PretrainedConfig type-checking regressions." + "reason": "ImportError for AutoImageProcessor is unrelated to PretrainedConfig type checking." }, { "left": "issue:45071", "right": "issue:45542", "accept": false, - "reason": "Type checking breakage and a TensorFlow backend detection error are different issues." + "reason": "Type-checking regression and undefined TensorFlow backend error are unrelated." }, { "left": "issue:42673", "right": "issue:44977", "accept": false, - "reason": "VRAM leak in multithreaded loading is not the same as flash-attention generation corruption." + "reason": "VRAM leak in dataloader threads is not the same as Qwen3.5 flash-attention generation failure." }, { "left": "issue:42371", "right": "issue:42831", "accept": false, - "reason": "TF32 API guidance and FineGrainedFP8 accuracy drift are different problems." + "reason": "TF32 API settings guidance and FineGrainedFP8 accuracy loss are different problems." }, { "left": "issue:42673", "right": "issue:43906", "accept": false, - "reason": "Insufficient evidence of the same bug; only a vague similarity in model/runtime troubleshooting." - }, - { - "left": "issue:43986", - "right": "issue:44484", - "accept": false, - "reason": "AutoProcessor crash without torchvision and max_shard_size behavior are unrelated." + "reason": "The VRAM leak report does not match the isolated reproduction of issue 38071." }, { "left": "issue:44117", "right": "issue:44987", "accept": false, - "reason": "Tokenizer mapping None and a model load failure do not look like one underlying defect." + "reason": "Tokenizer mapping None handling and loading physical-intelligence/fast are different failure modes." }, { "left": "issue:43097", "right": "issue:43452", "accept": false, - "reason": "Different failures: removed config flag vs gguf_file loading breakage." + "reason": "Removed tie_embeddings API and gguf_file loading breakage are unrelated." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Missing torchvision crash in AutoProcessor is unrelated to save_pretrained shard size." }, { "left": "issue:43335", "right": "issue:45071", "accept": false, - "reason": "SwitchTransformers sparse-layer construction and config type-checking are unrelated." + "reason": "SwitchTransformers sparse-layer config bug is unrelated to type-checking regressions." }, { "left": "issue:43931", "right": "issue:44038", "accept": false, - "reason": "Different Qwen3-VL-related loading issues with no clear shared code-path." + "reason": "Different Qwen-related load issues with different symptoms and likely different code paths." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python version compatibility and pipeline-parallelism output corruption are unrelated." }, { "left": "issue:43408", "right": "issue:44368", "accept": false, - "reason": "A model-type warning and a tie_word_embeddings warning are distinct config messages, not the same bug." + "reason": "A SAM model-type mismatch warning is unrelated to a Qwen fine-tuning warning about tie_word_embeddings." }, { "left": "issue:43874", "right": "issue:44263", "accept": false, - "reason": "Different GLM-related bugs: missing image-patch helper vs torch.split return handling." + "reason": "Different GLM bugs in different components: missing image-patch method vs torch.split return values." }, { "left": "issue:44829", "right": "issue:45245", "accept": false, - "reason": "flash_attention_3 training collapse is unrelated to the 2^24 category limit error." + "reason": "FlashAttention-3 training degeneration and category-cardinality limits are unrelated." }, { "left": "issue:38617", "right": "issue:42757", "accept": false, - "reason": "Two different missing-symbol import errors from different packages/modules." + "reason": "Both are import errors, but for different symbols and different underlying causes." }, { "left": "issue:43064", "right": "issue:44315", "accept": false, - "reason": "FSDP/optimizer-state corruption is not the same as Liger kernel application in model_init." + "reason": "FSDP2/PEFT optimizer-state corruption is not the same as Liger Kernel not being applied with model_init." }, { "left": "issue:44464", "right": "issue:45071", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency is unrelated to PretrainedConfig type checking." + "reason": "Compiled-forward generation inconsistency is unrelated to PretrainedConfig type checking." }, { "left": "issue:44062", "right": "issue:44117", "accept": false, - "reason": "Tokenizer.AddedToken keyword conflict and TOKENIZER_MAPPING_NAMES returning None are different failure modes." + "reason": "AddedToken keyword duplication and tokenizer mapping None handling are separate tokenizer bugs." }, { "left": "issue:43097", "right": "issue:45588", "accept": false, - "reason": "Config-flag removal is unrelated to a flash_attention.py crash on sink-less models." + "reason": "Removed tie_embeddings API and flash_attention s_aux=None crash are unrelated." }, { "left": "issue:44410", "right": "issue:45375", "accept": false, - "reason": "Missing attention projections and missing config field are different model-definition bugs." - }, - { - "left": "issue:42673", - "right": "issue:43577", - "accept": false, - "reason": "VRAM leakage across dataloader threads is unrelated to dtype propagation in Blip2 loading." + "reason": "Missing qwen3next projections and missing Qwen3_5MoeVisionConfig field are different model/config issues." }, { "left": "issue:38617", "right": "issue:44351", "accept": false, - "reason": "Different missing imports: layer_type_validation vs HybridCache." + "reason": "Different missing imports; same general theme, but not the same concrete bug." }, { "left": "issue:44704", "right": "issue:45103", "accept": false, - "reason": "cached_file kwargs forwarding and auto_docstring annotation handling are unrelated." + "reason": "Passing kwargs to cached_file and auto_docstring future-annotations crash are unrelated." + }, + { + "left": "issue:42673", + "right": "issue:43577", + "accept": false, + "reason": "VRAM leak in multi-thread loading and BLIP2 dtype retention are unrelated." }, { "left": "issue:44038", "right": "issue:44410", "accept": false, - "reason": "Two different Qwen3-family issues, but not the same concrete bug or code path." + "reason": "Both mention Qwen models, but one is a version/load compatibility issue and the other is a missing-projection architecture bug." }, { "left": "issue:44261", "right": "issue:44485", "accept": false, - "reason": "MLA layernorm epsilon precision issue is unrelated to the GLM-5 RoPE implementation." + "reason": "Missing rms_norm_eps precision issue and GLM-5 RoPE implementation are different bugs." + }, + { + "left": "issue:44964", + "right": "issue:45341", + "accept": false, + "reason": "Phi-4 multimodal load failure and a testing_utils bug are unrelated." }, { "left": "issue:43986", "right": "issue:44117", "accept": false, - "reason": "Video AutoProcessor/torchvision crash and tokenizer mapping None are unrelated." + "reason": "Video-model AutoProcessor crash without torchvision is unrelated to tokenizer mapping None handling." }, { - "left": "issue:44964", - "right": "issue:45341", + "left": "issue:43931", + "right": "issue:44410", "accept": false, - "reason": "Phi-4 multimodal loading failure and a testing_utils bug are unrelated." + "reason": "Different Qwen model families and different load-path failures." }, { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "local_files_only still causing requests is unrelated to missing embed_tokens.weight in UMT5Encoder." + "reason": "local_files_only network leak and missing embed_tokens.weight are unrelated." }, { "left": "issue:43097", - "right": "issue:43452", - "accept": false, - "reason": "Different failures: removed config flag vs gguf_file loading breakage." - }, - { - "left": "issue:43441", - "right": "issue:45588", + "right": "issue:43873", "accept": false, - "reason": "Both involve attention internals, but the model-specific FlashAttention failure and sink-less-model AttributeError are different code-path bugs." + "reason": "Removed tie_embeddings API and quantization offloading behavior are unrelated." } ] }, @@ -12919,9 +12776,8 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ - "issue:36010", "issue:36683", "issue:38617", "issue:42617", @@ -12972,18 +12828,18 @@ ], "soft_pairs": [ "issue:43066|issue:43906", - "issue:43986|issue:45372", "issue:43986|issue:45290", + "issue:43986|issue:45372", "issue:45341|issue:45561", "issue:43986|issue:45341", "issue:43097|issue:43901", "issue:43825|issue:44655", "issue:42617|issue:45341", - "issue:43992|issue:44484", "issue:45341|issue:45588", + "issue:43992|issue:44484", + "issue:42886|issue:43381", "issue:43208|issue:43550", "issue:43012|issue:44368", - "issue:42886|issue:43381", "issue:43012|issue:43550", "issue:42907|issue:45753", "issue:43531|issue:45341", @@ -12994,267 +12850,254 @@ "issue:43012|issue:44683", "issue:43618|issue:43761", "issue:44483|issue:45290", - "issue:44038|issue:44230", "issue:43064|issue:44380", + "issue:44038|issue:44230", "issue:38617|issue:43502", - "issue:43299|issue:43867", "issue:44483|issue:45381", + "issue:43299|issue:43867", "issue:43901|issue:44655", "issue:36683|issue:43475", "issue:43867|issue:45216", "issue:42994|issue:44655", "issue:43352|issue:43937", - "issue:43066|issue:45381", - "issue:36010|issue:42757" + "issue:43066|issue:45381" ], - "prepared_review_unit_hash": "412b66d56d1d2b144839a106c323b3ab231a8f8c76a62057f0e9c9f9c40a8970", + "prepared_review_unit_hash": "756a42d4a40647e424a91863b083395f8df588ec0ab78bceb33b82b2092bdb68", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 35, - "serialized_chars": 37930, - "estimated_input_tokens": 9483, - "estimated_eval_tokens": 19222 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 34, + "serialized_chars": 37049, + "estimated_input_tokens": 9263, + "estimated_eval_tokens": 18782 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous mix of v5 regressions, import errors, and unrelated runtime bugs. I would keep the cluster conservative: 43097 is the best anchor, and no soft edges should be accepted; in particular, 43618 vs 43761 is only a topical CLIP/v5 similarity without a shared concrete fix path.", - "confidence": 0.74, - "canonical_issue_reason": "issue:43097 is the strongest anchor: it is a clear user-facing v5 API regression with inbound references and a concrete symptom tied to a specific removed/config change.", + "summary": "This cluster is mostly heterogeneous v5-era bug reports and import/runtime regressions; only the CLIPVisionModel output regressions look like the same underlying change. No PRs are present.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 43097 is the broadest, clearest representative of the cluster\u2019s dominant theme: an explicit Transformers v5 breaking-change regression with user-visible API fallout.", "canonical_pr_reason": null, - "best_issue_reason": "Same reason as the canonical pick: it is the most self-contained, high-signal regression in an otherwise mixed cluster.", + "best_issue_reason": "Issue 43097 is the best overall issue representative because it captures a concrete, high-impact v5 compatibility break rather than a narrow model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "Both are v5-related, but one is a tokenizer decoder-type bug and the other is a reproduction of a separate earlier issue; not the same defect." + "reason": "Both involve v5 behavior, but one is a tokenizer decoder-type bug and the other is an unrelated isolated reproduction; no same code-path evidence." }, { "left": "issue:43986", - "right": "issue:45372", + "right": "issue:45290", "accept": false, - "reason": "Different failure surfaces: missing torchvision handling versus a mistral_common import error during Gemma 4 processor loading." + "reason": "Different failures: missing torchvision for AutoProcessor vs chat template crashing on tool-call assistant messages." }, { "left": "issue:43986", - "right": "issue:45290", + "right": "issue:45372", "accept": false, - "reason": "AutoProcessor/video-model loading without torchvision is unrelated to chat-template crashes on tool-call messages." + "reason": "Both are import/dependency-related, but they affect different components and different missing symbols; not the same bug." }, { "left": "issue:45341", "right": "issue:45561", "accept": false, - "reason": "Both mention testing utilities, but one is a generic bug report and the other is a specific xdist race on captured_info.txt." + "reason": "Both mention testing utilities, but one is a vague bug report and the other is a specific xdist file race; insufficient evidence of identity." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "Processor-loading failure and a testing_utils bug are unrelated code paths." + "reason": "A processor-loading dependency error is unrelated to a generic testing_utils bug." }, { "left": "issue:43097", "right": "issue:43901", "accept": false, - "reason": "API removal regression and documentation mismatch are different artifacts and likely different fixes." + "reason": "One is an API removal regression; the other is documentation lag about return_all_scores." }, { "left": "issue:43825", "right": "issue:44655", "accept": false, - "reason": "Pipeline error messaging and Pipeline save_pretrained persistence are separate problems." + "reason": "Pipeline task support messaging and saving Pipeline objects are different problems." }, { "left": "issue:42617", "right": "issue:45341", "accept": false, - "reason": "3d_parallel.py execution failure is unrelated to the testing_utils race report." + "reason": "3d_parallel.py execution failure is unrelated to a testing_utils issue." + }, + { + "left": "issue:45341", + "right": "issue:45588", + "accept": false, + "reason": "A testing_utils bug and a flash-attention sink-less-model crash are unrelated." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "Missing model weights in from_pretrained and /v1/chat/completions request rejection are distinct bugs." + "reason": "Missing embed_tokens weights on load and max_shard_size default behavior in save_pretrained are different issues." }, { - "left": "issue:45341", - "right": "issue:45588", + "left": "issue:42886", + "right": "issue:43381", "accept": false, - "reason": "A testing_utils race and a flash_attention AttributeError on sink-less models are unrelated." + "reason": "Offline cache loading and eval-mode gradient checkpointing are unrelated." }, { "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "Different models and different backend/training failure modes; not the same bug." + "reason": "Different model families and different failure modes; not the same underlying bug." }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "A PyTorch precision warning and a Qwen3.5 tie-word-embeddings warning are only loosely similar." - }, - { - "left": "issue:42886", - "right": "issue:43381", - "accept": false, - "reason": "Offline cache loading and gradient checkpointing in eval mode are separate issues." + "reason": "Both are warnings, but one is a PyTorch precision warning and the other is a model config warning about tie_word_embeddings." }, { "left": "issue:43012", "right": "issue:43550", "accept": false, - "reason": "Compile-time precision warning and Bamba SDPA compile failure are not the same defect." + "reason": "A compile/precision warning is unrelated to a Bamba SDPA compile failure." }, { "left": "issue:42907", "right": "issue:45753", "accept": false, - "reason": "Dequantized model saving and an infinite generation loop are unrelated." + "reason": "Quantized model saving and Qwen image generation looping are unrelated." }, { "left": "issue:43531", "right": "issue:45341", "accept": false, - "reason": "Qwen3-MoE sliding-window behavior is unrelated to the testing_utils report." + "reason": "A Qwen3-MoE sliding_window bug is unrelated to testing_utils." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Quantized/dequantized save-pretrained behavior and Pipeline save_pretrained are different save paths." + "reason": "Dequantized model saving and saving Pipeline objects are different save_pretrained problems." }, { "left": "issue:43208", "right": "issue:43526", "accept": false, - "reason": "xLSTM training bugs and BeitImageProcessorFast label reduction are unrelated." + "reason": "xLSTM training bugs and a BeitImageProcessorFast label reduction bug are unrelated." }, { "left": "issue:44483", "right": "issue:44514", "accept": false, - "reason": "An HTTP chat-completions API rejection is not the same as a chat_template preprocessing crash." + "reason": "OpenAI-style chat completion request rejection is not the same as a Qwen2_5_VLProcessor batching crash." }, { "left": "issue:42757", "right": "issue:43824", "accept": false, - "reason": "Both are import errors, but for different symbols in different packages and different code paths." + "reason": "Both are import errors, but for different packages and different missing names; no shared underlying change is evident." }, { "left": "issue:43012", "right": "issue:44683", "accept": false, - "reason": "A compile-time warning is unrelated to a flex_attention failure on newer torch." + "reason": "A bfloat16 compile warning and a flex_attention failure on torch>=2.9 are different code paths." }, { "left": "issue:43618", "right": "issue:43761", - "accept": false, - "reason": "These are only loosely related CLIP/v5 vision-output regressions; the packet does not show a shared concrete fix path or linked evidence." + "accept": true, + "reason": "Both report a CLIPVisionModel v5 regression where optional outputs are not populated correctly; attentions/hidden_states are missing from the same forward output path." }, { "left": "issue:44483", "right": "issue:45290", "accept": false, - "reason": "Request handling in chat-completions and apply_chat_template tool-call crashes occur at different layers." + "reason": "Different APIs and symptoms: chat completion request validation vs chat template handling of tool-call assistant messages." }, { - "left": "issue:44038", - "right": "issue:44230", + "left": "issue:43064", + "right": "issue:44380", "accept": false, - "reason": "They both mention Qwen3-VL-MoE/fp8, but one is a loading regression and the other is fp8 inference support." + "reason": "FSDP2/PEFT optimizer-state corruption and GPT2 attention-scaling under SDPA/FlashAttention are unrelated." }, { - "left": "issue:43064", - "right": "issue:44380", + "left": "issue:44038", + "right": "issue:44230", "accept": false, - "reason": "FSDP2+PEFT optimizer-state corruption and GPT2 attention scaling under SDPA/FlashAttention are unrelated." + "reason": "Both mention Qwen3 VL/MoE, but one is a Transformers loading bug and the other is fp8 inference support; not the same bug." }, { "left": "issue:38617", "right": "issue:43502", "accept": false, - "reason": "Importing layer_type_validation and ignoring local_files_only are unrelated bugs." + "reason": "A missing configuration_utils import and API calls despite local_files_only=True are unrelated." }, { - "left": "issue:43299", - "right": "issue:43867", + "left": "issue:44483", + "right": "issue:45381", "accept": false, - "reason": "Qwen3-VL-MoE loading breakage and a sorted-state_dict load error are not the same regression." + "reason": "Chat-completions request handling and Qwen2.5-VL video vision_position_ids are different bugs." }, { - "left": "issue:44483", - "right": "issue:45381", + "left": "issue:43299", + "right": "issue:43867", "accept": false, - "reason": "An HTTP API request rejection is unrelated to a Qwen2.5-VL video position-id bug." + "reason": "Qwen3VL MoE loading breakage and a sorted state_dict load error are not clearly the same underlying issue." }, { - "left": "issue:43097", + "left": "issue:43901", "right": "issue:44655", "accept": false, - "reason": "A removed config flag and inability to save Pipeline objects are different problems." + "reason": "Documentation mismatch for return_all_scores is unrelated to Pipeline serialization." }, { "left": "issue:36683", "right": "issue:43475", "accept": false, - "reason": "Different missing-attribute errors on different objects/models; not the same bug." + "reason": "Gemma3Config missing vocab_size and SAM3 vision output missing fpn_position_embeddings are different model attribute regressions." }, { "left": "issue:43867", "right": "issue:45216", "accept": false, - "reason": "A sorted-state_dict load failure and a bad saved checkpoint are related in theme but not clearly the same underlying regression." + "reason": "Both involve loading/saving state_dicts, but one is a sorted-load error and the other is incorrect saved checkpoints; too little to treat as the same bug." }, { "left": "issue:42994", "right": "issue:44655", "accept": false, - "reason": "Quantized model save failure and Pipeline save_pretrained failure are distinct save-path issues." + "reason": "Quantized model saving failure and saving Pipeline objects are different save_pretrained paths." }, { "left": "issue:43352", "right": "issue:43937", "accept": false, - "reason": "Flash-Attention-2 support gating and invalid GenerationConfig are unrelated." - }, - { - "left": "issue:43066", - "right": "issue:45381", - "accept": false, - "reason": "Tokenizer decoder-type bug and Qwen2.5-VL video vision_position_ids bug are unrelated." - }, - { - "left": "issue:36010", - "right": "issue:42757", - "accept": false, - "reason": "Different import errors from different modules with no shared code-path bug." + "reason": "FA2 support not implemented for NemotronHForCausalLM is unrelated to an invalid GenerationConfig in GLM-5." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The cluster summary is mostly grounded, but one soft-edge verdict is too strong for the evidence in the packet: issue:43618 vs issue:43761 was accepted as the same regression, yet the packet only shows a loose topical similarity (both CLIP/v5 vision-output regressions) and no concrete shared fix path or linked evidence. That should stay conservative/false. Also, the summary says two soft edges were accepted, but only one accept=true appears in the verdict list." + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative about the soft edges. The only accepted pair, issue:43618 and issue:43761, is plausibly the same CLIPVisionModel v5 output regression, and the other pairwise rejections do not overclaim identity. No PRs are present, so the mergeability constraint is not an issue." }, "evaluator_used": true, - "retried": true, - "accepted_nontrivial_soft_edge": false, + "retried": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -13269,13 +13112,15 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:28282", "issue:33453", + "issue:36010", "issue:36296", "issue:36683", "issue:41669", + "issue:42757", "issue:42890", "issue:43064", "issue:43097", @@ -13288,7 +13133,6 @@ "issue:43756", "issue:43867", "issue:43881", - "issue:43901", "issue:43906", "issue:43976", "issue:43992", @@ -13308,7 +13152,6 @@ "issue:44855", "issue:44857", "issue:44938", - "issue:44977", "issue:45216", "issue:45310", "issue:45357", @@ -13322,6 +13165,7 @@ ], "soft_pairs": [ "issue:44263|issue:45440", + "issue:36010|issue:42757", "issue:44220|issue:44655", "issue:43756|issue:44485", "issue:36683|issue:43525", @@ -13332,15 +13176,15 @@ "issue:43299|issue:45753", "issue:43352|issue:43976", "issue:43441|issue:44380", - "issue:43992|issue:45310", "issue:43352|issue:44380", + "issue:43992|issue:45310", "issue:44380|issue:45479", "issue:41669|issue:44246", - "issue:43867|issue:45375", "issue:43064|issue:44492", + "issue:43867|issue:45375", "issue:43618|issue:44380", - "issue:44355|issue:44855", "issue:43352|issue:45478", + "issue:44355|issue:44855", "issue:36683|issue:44117", "issue:42890|issue:45561", "issue:43097|issue:43352", @@ -13353,266 +13197,257 @@ "issue:44077|issue:44380", "issue:43618|issue:44857", "issue:43618|issue:45588", - "issue:43867|issue:44484", "issue:28282|issue:33453", - "issue:43618|issue:44977", - "issue:43901|issue:44938", - "issue:44655|issue:44938" + "issue:43867|issue:44484" ], - "prepared_review_unit_hash": "496451ab2b43b9d56baa3bb83507f82bab8bb9045aa896e530c4488701819588", + "prepared_review_unit_hash": "3db5c8667f7f0ded7b1bb17242445473c19aad809106aedab920236f327562e0", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 37, - "serialized_chars": 38376, - "estimated_input_tokens": 9594, - "estimated_eval_tokens": 19444 + "soft_pair_count": 35, + "serialized_chars": 37790, + "estimated_input_tokens": 9448, + "estimated_eval_tokens": 19152 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated single-issue reports. Keep the Qwen3.5 MoE from_pretrained regression as the strongest issue candidate, but the import-slowdown pair (41669/44246) is only thematically similar and should not be treated as a confirmed duplicate from the packet alone.", - "confidence": 0.74, - "canonical_issue_reason": "issue:45310 is the clearest report of the Qwen3.5 MoE from_pretrained regression, and issue:45478 appears to be an identical duplicate by title and symptom.", + "summary": "Most proposed edges are false positives across unrelated model/config/version bugs. The only convincing duplicate-like pair is the import-time slowdown report versus the generic slow-import report; the rest differ in concrete model, code path, or failure mode.", + "confidence": 0.63, + "canonical_issue_reason": "issue:41669 is the most specific and actionable issue in the only convincing duplicate pair: it names the concrete import-time regression cause ('import *' in models) and the measurable slowdown, while issue:44246 is the broader symptom report.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the best overall issue candidate because it is a concrete, well-scoped regression with an identical duplicate in the cluster.", + "best_issue_reason": "issue:41669 is the best global issue representative because it is specific, root-cause oriented, and can subsume the more generic slow-import report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44263", "right": "issue:45440", "accept": false, - "reason": "Different bugs: DSA indexer split behavior vs DeepSeekV3 MoE implementation divergence." + "reason": "Different bugs in different model families; one is a torch.split / indexer issue, the other is a DeepSeekV3MoE implementation divergence." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols and different dependency breakages; not the same underlying bug." }, { "left": "issue:44220", "right": "issue:44655", "accept": false, - "reason": "Different code paths: audio fbank feature extraction vs saving Pipeline objects." + "reason": "Unrelated failures: audio feature extraction versus saving Pipeline objects." }, { "left": "issue:43756", "right": "issue:44485", "accept": false, - "reason": "Both mention RoPE, but they are model-specific bugs in different models." + "reason": "Both mention RoPE, but they are model-specific implementation bugs with different symptoms and targets." }, { "left": "issue:36683", "right": "issue:43525", "accept": false, - "reason": "Different missing config attributes on different models." + "reason": "Both are missing config attributes, but for different models and different fields; too generic to treat as one bug." }, { "left": "issue:43618", "right": "issue:45216", "accept": false, - "reason": "CLIP attentions regression is unrelated to the Qwen3.5 save_pretrained checkpoint issue." + "reason": "Different problems: CLIPOutput attentions assignment versus Qwen3.5 checkpoint save regression." }, { "left": "issue:43906", "right": "issue:45753", "accept": false, - "reason": "A reproduction of an earlier issue and an image-specific infinite loop are not the same underlying bug." + "reason": "The first is a reproduction of another issue, while the second is a specific Qwen3_5 image infinite-loop report; not enough evidence they are the same bug." }, { "left": "issue:36683", "right": "issue:44242", "accept": false, - "reason": "Different failures: config attribute error vs missing load-balancing loss." + "reason": "Different model behavior: missing Gemma3Config field versus MoE load-balancing loss not being added." }, { "left": "issue:36683", "right": "issue:43881", "accept": false, - "reason": "Different model-loading failures with different root symptoms." + "reason": "Different loading failures affecting different models and different config/runtime paths." }, { "left": "issue:43299", "right": "issue:45753", "accept": false, - "reason": "Both involve Qwen, but one is Qwen3VL MoE loading and the other is a generation loop on a specific image." + "reason": "Both involve Qwen-related multimodal/model loading behavior, but one is a v5.0.0 loading regression and the other is an image-specific infinite loop." }, { "left": "issue:43352", "right": "issue:43976", "accept": false, - "reason": "FlashAttention support error is unrelated to Python-version compatibility." + "reason": "Unrelated issues: Flash Attention support for a model versus Python version compatibility for the package." }, { "left": "issue:43441", "right": "issue:44380", "accept": false, - "reason": "Model-specific FlashAttention failure vs GPT2 attention scaling backend bug." + "reason": "Both touch attention backends, but they are different concrete failures in different models." }, { - "left": "issue:43992", - "right": "issue:45310", + "left": "issue:43352", + "right": "issue:44380", "accept": false, - "reason": "UMT5 missing weights and Qwen3.5 MoE from_pretrained errors are different model-specific loading bugs." + "reason": "Different model/code-path problems; one is model support for FA2, the other is GPT2 attention scaling being ignored." }, { - "left": "issue:43352", - "right": "issue:44380", + "left": "issue:43992", + "right": "issue:45310", "accept": false, - "reason": "Same broad backend area, but different concrete bugs and models." + "reason": "Different models and different failure modes: missing embed_tokens.weight versus Qwen3.5 MoE from_pretrained error." }, { "left": "issue:44380", "right": "issue:45479", "accept": false, - "reason": "Attention scaling backend bug vs classification loss degeneracy are unrelated." + "reason": "No shared underlying bug; one is attention scaling with backends, the other is degenerate zero loss in sequence classification." }, { "left": "issue:41669", "right": "issue:44246", - "accept": false, - "reason": "Both are about slow imports, but the packet only shows similar titles and not enough evidence for the same root cause or code path." + "accept": true, + "reason": "Both report the same underlying import-time performance regression in transformers; one is the concrete cause/root-cause report and the other is the generic symptom report." }, { - "left": "issue:43867", - "right": "issue:45375", + "left": "issue:43064", + "right": "issue:44492", "accept": false, - "reason": "State_dict ordering load failure and dropped config fields are different problems." + "reason": "Completely different topics: distributed optimizer state corruption versus a cache strategy typo." }, { - "left": "issue:43064", - "right": "issue:44492", + "left": "issue:43867", + "right": "issue:45375", "accept": false, - "reason": "Trainer/FSDP optimizer-state corruption is unrelated to a cache-strategy typo." + "reason": "Different loading/serialization bugs affecting different Qwen-related configs and fields." }, { "left": "issue:43618", "right": "issue:44380", "accept": false, - "reason": "Different regressions: CLIP attentions assignment vs GPT2 attention scaling under SDPA/FlashAttention." + "reason": "Different issues despite both involving model internals; no shared concrete code-path failure." }, { - "left": "issue:44355", - "right": "issue:44855", + "left": "issue:43352", + "right": "issue:45478", "accept": false, - "reason": "Compiled-file import errors and a torch.jit parsing IndentationError are not the same bug." + "reason": "Same broad area as another FlashAttention issue, but different model and no clear shared concrete bug." }, { - "left": "issue:43352", - "right": "issue:45478", + "left": "issue:44355", + "right": "issue:44855", "accept": false, - "reason": "FlashAttention support issue vs Qwen3.5 MoE from_pretrained regression." + "reason": "One is about compiled Python files, the other about a Python 3.13 torch.jit parsing/indentation problem." }, { "left": "issue:36683", "right": "issue:44117", "accept": false, - "reason": "Missing model config attribute vs tokenizer mapping returning None." + "reason": "Different loader assumptions in different parts of the stack; not the same bug." }, { "left": "issue:42890", "right": "issue:45561", "accept": false, - "reason": "Seed-related flaky test failures and an xdist file race are different test issues." + "reason": "Both are test-related, but one is missing set_seed and the other is an xdist race on a captured-info file." }, { "left": "issue:43097", "right": "issue:43352", "accept": false, - "reason": "Tie-embeddings API removal and FlashAttention support are unrelated." + "reason": "Unrelated: an API removal/regression versus a Flash Attention support limitation." }, { "left": "issue:43867", "right": "issue:45357", "accept": false, - "reason": "Different regressions: sorted state_dict load error vs incorrect visual encoder keys on save." + "reason": "Different save/load regressions involving different Qwen3.5 components and different key handling." }, { "left": "issue:43976", "right": "issue:44938", "accept": false, - "reason": "Both are Python compatibility complaints, but they target different versions and failure modes." + "reason": "Both are Python-version issues, but they affect different versions and appear to be separate compatibility problems." }, { "left": "issue:43992", "right": "issue:45478", "accept": false, - "reason": "Different model-specific load failures with no shared concrete code path." + "reason": "Different model families and different from_pretrained failures." }, { "left": "issue:44360", "right": "issue:45440", "accept": false, - "reason": "DSA indexer behavior and DeepSeekV3 MoE divergence are unrelated." + "reason": "Different code paths: DSA indexer math versus DeepSeekV3 implementation divergence." }, { "left": "issue:43064", "right": "issue:44655", "accept": false, - "reason": "Saving Pipeline objects and training optimizer-state corruption are different issues." + "reason": "Unrelated distributed-training state bug versus pipeline serialization issue." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Tensor-parallel training bug vs TP OOM on model loading are not the same underlying issue." + "reason": "Both mention tensor parallelism, but one is a training bug and the other is an OOM on model loading." }, { "left": "issue:44077", "right": "issue:44380", "accept": false, - "reason": "PatchTSMixer post_init policy and GPT2 attention scaling are different problems." + "reason": "Different bugs in different model implementations; no evidence they share the same failure mechanism." }, { "left": "issue:43618", "right": "issue:44857", "accept": false, - "reason": "Different subsystems: CLIP attentions vs LwDetr AMP/CUDA crash." + "reason": "Different failures: missing attentions assignment versus AMP/CUDA crash in a loss function." }, { "left": "issue:43618", - "right": "issue:44977", - "accept": false, - "reason": "Both mention attention/FlashAttention, but the model and failure mode differ." - }, - { - "left": "issue:43867", - "right": "issue:44484", + "right": "issue:45588", "accept": false, - "reason": "A load error with sorted state_dict is unrelated to the shard-size question." + "reason": "Both touch attention-related code, but one is a CLIPOutput regression and the other is a FlashAttention crash on sink-less models." }, { "left": "issue:28282", "right": "issue:33453", "accept": false, - "reason": "AutoModel missing PyTorch is unrelated to tokenizer loading regression." - }, - { - "left": "issue:43901", - "right": "issue:44938", - "accept": false, - "reason": "Docs issue vs runtime Python-version incompatibility." + "reason": "ImportError for missing PyTorch is unrelated to tokenizer loading regression." }, { - "left": "issue:44655", - "right": "issue:44938", + "left": "issue:43867", + "right": "issue:44484", "accept": false, - "reason": "Pipeline save_pretrained failure is unrelated to Python 3.14 load failure." + "reason": "Model loading error versus a save_pretrained shard-size question; not the same bug." } ] }, "evaluator_result": { - "accept": false, - "feedback": "Most of the conservative false/duplicate judgments are grounded in the packet, and the canonical Qwen3.5 issue choice is reasonable. However, the one accepted soft-edge pair (issue:41669 / issue:44246) is stated too strongly: the packet only shows two similar import-slowdown titles, not enough to conclude they are the same underlying bug or that one identifies the exact cause. The summary should stay at the level of thematic similarity unless there is more direct evidence." + "accept": true, + "feedback": "Grounded overall: the only accepted duplicate-like pair is well supported by the issue titles and descriptions, and the other soft-edge rejections stay conservative rather than overclaiming equivalence. The summary does not overstate the evidence." }, "evaluator_used": true, - "retried": true, - "accepted_nontrivial_soft_edge": false, + "retried": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -13627,7 +13462,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:30333", "issue:30990", @@ -13635,8 +13470,6 @@ "issue:33453", "issue:34689", "issue:36296", - "issue:38617", - "issue:41628", "issue:42548", "issue:42757", "issue:42831", @@ -13645,11 +13478,12 @@ "issue:43299", "issue:43502", "issue:43576", + "issue:43618", "issue:43723", "issue:43824", "issue:43827", "issue:43874", - "issue:43976", + "issue:43901", "issue:44062", "issue:44075", "issue:44112", @@ -13665,9 +13499,10 @@ "issue:44556", "issue:44655", "issue:44749", - "issue:44908", + "issue:44938", "issue:44961", "issue:44964", + "issue:44977", "issue:45071", "issue:45125", "issue:45278", @@ -13679,13 +13514,16 @@ "issue:45656" ], "soft_pairs": [ + "issue:43618|issue:44977", + "issue:43901|issue:44938", + "issue:44655|issue:44938", "issue:42890|issue:44964", "issue:42890|issue:44112", "issue:42890|issue:45600", "issue:43502|issue:44062", + "issue:42890|issue:45341", "issue:43010|issue:44075", "issue:42548|issue:44351", - "issue:42890|issue:45341", "issue:44222|issue:45125", "issue:43502|issue:44485", "issue:44355|issue:44749", @@ -13694,8 +13532,8 @@ "issue:44387|issue:44556", "issue:45125|issue:45375", "issue:43010|issue:45278", - "issue:30333|issue:33453", "issue:43723|issue:44556", + "issue:30333|issue:33453", "issue:43010|issue:44368", "issue:43502|issue:44222", "issue:43502|issue:44961", @@ -13706,295 +13544,265 @@ "issue:44222|issue:44230", "issue:44355|issue:45071", "issue:43827|issue:44961", - "issue:43010|issue:45656", "issue:33453|issue:34689", + "issue:43010|issue:45656", "issue:43299|issue:44556", "issue:42757|issue:43576", "issue:33453|issue:36296", "issue:43576|issue:45071", "issue:43576|issue:43824", - "issue:44961|issue:45468", - "issue:41628|issue:42548", - "issue:43010|issue:43976", - "issue:31795|issue:38617", - "issue:38617|issue:43576", - "issue:44908|issue:44961", - "issue:43576|issue:44351" + "issue:44961|issue:45468" ], - "prepared_review_unit_hash": "3673390da3eed75a4143dd83c5d737d4ec5adfbdbf9100d6609f40f49328fcbe", + "prepared_review_unit_hash": "08b28b7e20f8d7590a7225874567e8ee850f8545e6c8ea155d83a739a076fa97", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 41, - "serialized_chars": 39463, - "estimated_input_tokens": 9866, - "estimated_eval_tokens": 19988 + "soft_pair_count": 38, + "serialized_chars": 38498, + "estimated_input_tokens": 9625, + "estimated_eval_tokens": 19506 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a mix of separate Transformers v5 regression/import/loading reports rather than one duplicate bug. Issue 45278 is the broadest representative; no PRs were provided.", - "confidence": 0.84, - "canonical_issue_reason": "issue:45278 is the broad umbrella report for multiple import failures after upgrading to 5.5.0.", + "summary": "This cluster is mostly a false-positive grouping: the issues span unrelated areas like loading/import regressions, documentation cleanup, FP8, tensor parallelism, testing, and optimizers. I do not see a single underlying bug/change that would justify merging any pair, and there is no PR evidence here.", + "confidence": 0.88, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 best represents the cluster because it captures the shared v5 import-regression theme at the highest level.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43618", + "right": "issue:44977", + "accept": false, + "reason": "Different model/output bug classes: CLIPOutput attentions assignment vs Qwen3.5 flash-attention generation behavior." + }, + { + "left": "issue:43901", + "right": "issue:44938", + "accept": false, + "reason": "One is a docs mismatch about return_all_scores; the other is a Python 3.14 import/load failure." + }, + { + "left": "issue:44655", + "right": "issue:44938", + "accept": false, + "reason": "Saving Pipeline objects and Python 3.14 import/load failure are unrelated failure modes." + }, { "left": "issue:42890", "right": "issue:44964", "accept": false, - "reason": "Test flakiness in SamHQ vs a Phi-4 multimodal model-loading failure; unrelated bugs." + "reason": "Different test/model areas: SamHQ integration test seeding flakiness vs Phi-4 multimodal model loading." }, { "left": "issue:42890", "right": "issue:44112", "accept": false, - "reason": "Different CI/test issues: missing set_seed coverage vs a stale device override test." + "reason": "A missing set_seed in one model test is not the same as a stale GraniteSpeech device-override CI failure." }, { "left": "issue:42890", "right": "issue:45600", "accept": false, - "reason": "SamHQ test seeding flake vs removed Sam3LiteText config references; different code paths." + "reason": "Unrelated: SamHQ test randomness vs auto_mappings references to removed Sam3LiteText configs." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "local_files_only network leak vs AddedToken keyword duplication; different failure modes." + "reason": "Local-files-only network leakage and AddedToken argument duplication are different code paths." + }, + { + "left": "issue:42890", + "right": "issue:45341", + "accept": false, + "reason": "Both mention testing, but the specific SamHQ seeding issue is not clearly the same as an unspecified testing_utils bug." }, { "left": "issue:43010", "right": "issue:44075", "accept": false, - "reason": "Cache update no_grad concern vs SGD argument handling; not the same bug." + "reason": "Cache/layer no_grad decoration is unrelated to SGD optimizer arguments not being used." }, { "left": "issue:42548", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but different missing symbols with different root causes." - }, - { - "left": "issue:42890", - "right": "issue:45341", - "accept": false, - "reason": "SamHQ test seeding flake vs a testing_utils bug; unrelated." + "reason": "Different missing exports: PreTrainedModel vs HybridCache." }, { "left": "issue:44222", "right": "issue:45125", "accept": false, - "reason": "FP8 save_pretrained issue vs missing tensor-parallel plan; different model features." + "reason": "FP8 save_pretrained for MoE models is not the same bug as missing _tp_plan for tensor parallelism." }, { "left": "issue:43502", "right": "issue:44485", "accept": false, - "reason": "Offline loading API leak vs GLM-5 RoPE implementation discussion; unrelated." + "reason": "Network access despite local_files_only and GLM-5 RoPE implementation are unrelated." }, { "left": "issue:44355", "right": "issue:44749", "accept": false, - "reason": "Compiled-file error vs a slow-filtering performance regression; unrelated." + "reason": "Compiled Python file errors and tokenizer slowdown after upgrade are different problems." }, { "left": "issue:30990", "right": "issue:31795", "accept": false, - "reason": "Loading hang in Sentence Transformers vs documentation confusion; not the same bug." + "reason": "Loading hang and documentation confusion around model.forward arguments do not share a concrete root cause." }, { "left": "issue:45125", "right": "issue:45478", "accept": false, - "reason": "Related Qwen3.5 MoE area, but tensor-parallel support and from_pretrained failure are distinct." + "reason": "Both involve Qwen3.5 MoE, but one is tensor-parallel metadata and the other is a from_pretrained load error; too broad to merge." }, { "left": "issue:44387", "right": "issue:44556", "accept": false, - "reason": "Quantization OOM vs checkpoint reload regression after upgrade; different problems." + "reason": "Int4 quantization memory/OOM regression and checkpoint reload failure after v5 upgrades are different regressions." }, { "left": "issue:45125", "right": "issue:45375", "accept": false, - "reason": "Different Qwen3.5 MoE config gaps: tensor-parallel plan vs deepstack_visual_indexes." + "reason": "Missing _tp_plan and missing deepstack_visual_indexes are separate config issues." }, { - "left": "issue:43010", + "left": "issue:43502", "right": "issue:45278", "accept": false, - "reason": "Cache no_grad/update bug vs broad import-error regression; unrelated." + "reason": "Local file handling and broad import failures after 4.57->5.5 are not the same bug." }, { - "left": "issue:30333", - "right": "issue:33453", + "left": "issue:43723", + "right": "issue:44556", "accept": false, - "reason": "MLflow training-job status vs tokenizer loading regression; unrelated." + "reason": "Tokenizer loading in v5 and generic checkpoint reload after upgrade are only loosely related." }, { - "left": "issue:43723", - "right": "issue:44556", + "left": "issue:30333", + "right": "issue:33453", "accept": false, - "reason": "Tokenizer loading regression vs checkpoint reload failure; different concrete load paths." + "reason": "MLflow reporting after training failure and tokenizer loading regression are unrelated." }, { "left": "issue:43010", "right": "issue:44368", "accept": false, - "reason": "Cache update semantics vs a tied-embeddings warning; not a duplicate." + "reason": "no_grad decoration for cache updates is unrelated to a tie_word_embeddings warning during LoRA fine-tuning." }, { "left": "issue:43502", "right": "issue:44222", "accept": false, - "reason": "Offline API requests vs FP8 save_pretrained bug; unrelated." + "reason": "Network requests with local_files_only and FP8 save_pretrained for MoE are different bugs." }, { "left": "issue:43502", "right": "issue:44961", "accept": false, - "reason": "Real loading bug vs placeholder issue title; no duplicate evidence." + "reason": "The 'racoon' issue has no visible relation to local_files_only network behavior." }, { "left": "issue:44315", "right": "issue:44360", "accept": false, - "reason": "Liger Kernel model_init behavior vs DSA indexer ReLU discussion; unrelated." + "reason": "Liger Kernel application in model_init is unrelated to a DSA indexer/ReLU discussion." }, { "left": "issue:42831", "right": "issue:43874", "accept": false, - "reason": "FineGrainedFP8 accuracy issue vs GLM46V image patch API error; different bugs." + "reason": "FineGrainedFP8 accuracy and GLM46V image patch counting are different code paths." }, { "left": "issue:30333", "right": "issue:36296", "accept": false, - "reason": "MLflow reporting failure vs tensor-parallel training bug; unrelated." + "reason": "MLflow reporting after failed training and tensor-parallel training bugs are not the same underlying issue." }, { "left": "issue:44355", "right": "issue:44655", "accept": false, - "reason": "Compiled file execution errors vs Pipeline save_pretrained support; unrelated." + "reason": "Compiled Python file errors and Pipeline save_pretrained failures are different concerns." }, { "left": "issue:44222", "right": "issue:44230", "accept": false, - "reason": "Both touch FP8/MoE, but save_pretrained failure and inference support are different." + "reason": "Both mention FP8/MoE, but one is save_pretrained and the other is inference support; not the same concrete bug." }, { "left": "issue:44355", "right": "issue:45071", "accept": false, - "reason": "Compiled-file errors vs PretrainedConfig type-checking breakage; unrelated." + "reason": "Compiled file execution errors and PretrainedConfig type-checking breaks are unrelated." }, { "left": "issue:43827", "right": "issue:44961", "accept": false, - "reason": "Docs stale after pipeline removals vs unrelated placeholder issue." + "reason": "Pipeline-docs cleanup and the 'racoon' issue are unrelated." }, { - "left": "issue:43010", - "right": "issue:45656", + "left": "issue:33453", + "right": "issue:34689", "accept": false, - "reason": "Cache update semantics vs deepspeed optimizer-step duplication; different code paths." + "reason": "Tokenizer loading regression and Llama 3.2 vision model loading breakage are different failures." }, { - "left": "issue:33453", - "right": "issue:34689", + "left": "issue:43010", + "right": "issue:45656", "accept": false, - "reason": "Tokenizer-loading regression vs Llama 3.2 model-loading breakage; related theme, different bug." + "reason": "Cache update no_grad decoration is unrelated to deepspeed causing double optimizer steps." }, { "left": "issue:43299", "right": "issue:44556", "accept": false, - "reason": "Different model/version-specific loading regressions; not the same underlying issue." + "reason": "Qwen3VL MoE loading regression and checkpoint reload regression after v5.2/v5.3 are not clearly the same root cause." }, { "left": "issue:42757", "right": "issue:43576", "accept": false, - "reason": "Missing huggingface_hub symbol vs broken v5 env command; unrelated." - }, - { - "left": "issue:33453", - "right": "issue:36296", - "accept": false, - "reason": "Tokenizer loading regression vs tensor-parallel training bug; unrelated." + "reason": "Missing is_offline_mode import and a broken transformers env command are different packaging/CLI issues." }, { "left": "issue:43576", "right": "issue:45071", "accept": false, - "reason": "Broken env command vs PretrainedConfig typing breakage; unrelated." + "reason": "CLI breakage and PretrainedConfig type-checking are unrelated." }, { "left": "issue:43576", "right": "issue:43824", "accept": false, - "reason": "Broken env command vs Qwen2_5_VL import failure; different symptoms." + "reason": "Broken env command and missing Qwen2_5_VLForConditionalGeneration export are different import/CLI failures." }, { "left": "issue:44961", "right": "issue:45468", "accept": false, - "reason": "Placeholder title vs Gemma-4 audio positional encoding bug; unrelated." - }, - { - "left": "issue:41628", - "right": "issue:42548", - "accept": false, - "reason": "Both are import errors, but for different exported classes and likely different causes." - }, - { - "left": "issue:43010", - "right": "issue:43976", - "accept": false, - "reason": "Cache update semantics vs Python-version compatibility problem; unrelated." - }, - { - "left": "issue:31795", - "right": "issue:38617", - "accept": false, - "reason": "Documentation confusion vs missing import symbol; unrelated." - }, - { - "left": "issue:38617", - "right": "issue:43576", - "accept": false, - "reason": "Missing configuration_utils symbol vs broken env CLI command; unrelated." - }, - { - "left": "issue:44908", - "right": "issue:44961", - "accept": false, - "reason": "Scheduler kwargs bug vs placeholder issue title; unrelated." - }, - { - "left": "issue:43576", - "right": "issue:44351", - "accept": false, - "reason": "Broken env command vs HybridCache import failure; unrelated." + "reason": "No meaningful relation between the 'racoon' issue and Gemma-4 audio positional encoding." } ] }, @@ -14016,14 +13824,14 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:31795", - "issue:33453", "issue:36296", + "issue:38617", "issue:41628", "issue:41669", - "issue:42371", + "issue:42548", "issue:42890", "issue:43010", "issue:43232", @@ -14051,11 +13859,11 @@ "issue:44492", "issue:44556", "issue:44623", - "issue:44655", "issue:44704", "issue:44829", "issue:44861", "issue:44908", + "issue:44938", "issue:44961", "issue:45003", "issue:45071", @@ -14068,6 +13876,12 @@ "issue:45753" ], "soft_pairs": [ + "issue:31795|issue:38617", + "issue:43010|issue:43976", + "issue:41628|issue:42548", + "issue:38617|issue:43576", + "issue:44908|issue:44961", + "issue:43576|issue:44938", "issue:43010|issue:43232", "issue:36296|issue:43901", "issue:41628|issue:44355", @@ -14079,8 +13893,8 @@ "issue:44393|issue:45753", "issue:43576|issue:44556", "issue:44393|issue:45071", - "issue:44038|issue:45125", "issue:41669|issue:43673", + "issue:44038|issue:45125", "issue:44556|issue:45092", "issue:44373|issue:44961", "issue:44704|issue:44908", @@ -14097,272 +13911,273 @@ "issue:41669|issue:44351", "issue:41669|issue:45341", "issue:41669|issue:44492", - "issue:44393|issue:45478", "issue:44315|issue:44829", + "issue:44393|issue:45478", "issue:44077|issue:44861", - "issue:43827|issue:44908", - "issue:31795|issue:33453", - "issue:41669|issue:44908", - "issue:42371|issue:44492", - "issue:44655|issue:44861", - "issue:43576|issue:44273" + "issue:43827|issue:44908" ], - "prepared_review_unit_hash": "41e727272fc23e3dd25dc67d9f36ff92d3193edee76e792495e03210b0669c1a", + "prepared_review_unit_hash": "07f242530ebf0041cef551fa15be47c0e2a66a21d43f6367c1a58f3c4b632f08", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 38, - "serialized_chars": 38419, - "estimated_input_tokens": 9605, - "estimated_eval_tokens": 19466 + "soft_pair_count": 39, + "serialized_chars": 38754, + "estimated_input_tokens": 9689, + "estimated_eval_tokens": 19634 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is not a true duplicate set: most items are loosely related Transformers issues (v5 migration, lazy loading/imports, generation/cache behavior, tokenizer/processor save-load, and model-specific bugs), but the soft pairs only show superficial keyword overlap rather than the same underlying defect. No pair looks safe to merge as duplicates.", - "confidence": 0.91, - "canonical_issue_reason": "No single canonical issue cleanly covers the cluster; the issues span several unrelated bug classes. If one representative must be chosen, issue:41669 is the most concrete and triageable, but it only reflects the import/lazy-loading performance subgroup.", + "summary": "This cluster is mostly a loose collection of unrelated transformer issues with a few superficial overlaps around imports, caching, and v5 regressions. None of the soft pairs look like the same concrete bug/change, so I reject all soft edges. The broadest representative issue is the import/performance regression around `import *` usage.", + "confidence": 0.84, + "canonical_issue_reason": "issue:41669 is the broadest and most central representative of the cluster\u2019s recurring import/loading regressions, with concrete impact and multiple nearby import-related complaints.", "canonical_pr_reason": null, - "best_issue_reason": "Issue:41669 is the strongest anchor because it is specific, well-described, and has observable impact and references. It is still not a good global canonical for the whole cluster, just the best representative issue among these heterogeneous reports.", + "best_issue_reason": "issue:41669 best captures the cluster\u2019s most common underlying theme (import-time/loading regressions) and is more representative than the narrower model-specific or doc-only issues.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Docs wording for forward args vs an ImportError for a missing symbol; different problem types and code paths." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Cache mutation/no_grad concern vs Python version compatibility failure; unrelated." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Different missing imports (`AutoImageProcessor` vs `PreTrainedModel`) with separate causes." + }, + { + "left": "issue:38617", + "right": "issue:43576", + "accept": false, + "reason": "Missing import symbol vs broken `transformers env` CLI; not the same bug." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs bug vs unrelated placeholder issue title; no shared underlying change." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "CLI env command regression vs checkpoint reload incompatibility; different functionality." + }, { "left": "issue:43010", "right": "issue:43232", "accept": false, - "reason": "Both involve generation/cache update methods, but one is about no_grad decoration on cache/layer updates and the other is about _update_model_kwargs_for_generation after sync_gpus. Different code paths and fixes." + "reason": "Both are generation/cache-related, but one is about `@torch.no_grad` on updates and the other about `_update_model_kwargs_for_generation` after GPU sync; not the same fix." }, { "left": "issue:36296", "right": "issue:43901", "accept": false, - "reason": "Tensor parallel training bug vs. text classification pipeline docs mismatch; unrelated symptom and subsystem, only broad v5-era similarity." + "reason": "Tensor parallel training bug vs documentation mismatch for pipeline args; unrelated." }, { "left": "issue:41628", "right": "issue:44355", "accept": false, - "reason": "Importing AutoImageProcessor failure is not the same as errors with compiled Python files; different import/runtime failure modes." + "reason": "Import error for `AutoImageProcessor` vs errors running compiled Python files; different failure modes." }, { "left": "issue:36296", "right": "issue:43526", "accept": false, - "reason": "Tensor parallel training bug and BeitImageProcessorFast reduce_labels bug are unrelated model-specific issues." + "reason": "Tensor parallel training bug vs `reduce_labels` returning one label; no concrete overlap." }, { "left": "issue:44261", "right": "issue:44360", "accept": false, - "reason": "MLA q_a_layernorm precision/config issue is unrelated to the DSA indexer ReLU discussion; different numerical component and fix scope." + "reason": "MLA layernorm epsilon precision issue vs DSA indexer/ReLU comment; different subsystems and symptoms." }, { "left": "issue:44273", "right": "issue:45003", "accept": false, - "reason": "Lazy loading problem vs. unsafe sys.modules access in modeling_utils; both touch import behavior, but they are not the same defect or code path." + "reason": "Lazy loading bug vs unsafe `sys.modules` access; related area, but not the same underlying defect." }, { "left": "issue:44623", "right": "issue:45230", "accept": false, - "reason": "Missing processor save_pretrained files is a concrete serialization bug; the second is a generic bug report with no evidence it is the same issue." + "reason": "Processor save/load missing files vs a generic bug report title; insufficient evidence of sameness." }, { "left": "issue:44273", "right": "issue:45542", "accept": false, - "reason": "Lazy loading not working and tensorboard-without-TensorFlow backend error are separate environment/import issues." + "reason": "Lazy loading regression vs undefined TF backend when tensorboard is installed; unrelated." }, { "left": "issue:44393", "right": "issue:45753", "accept": false, - "reason": "Qwen3-VL 2D bounding box hallucination differs from Qwen3_5 infinite loop on a specific image; same model family, but different failure modes and likely different fixes." + "reason": "Qwen3-VL bbox hallucination/error vs infinite loop on a specific image; different concrete bugs." }, { "left": "issue:43576", "right": "issue:44556", "accept": false, - "reason": "Broken 'transformers env' command in v5 is not the same as checkpoint reload incompatibility after upgrade; one is CLI/config output, the other is model loading compatibility." + "reason": "Broken env command vs checkpoint reload compatibility after upgrade; different code paths." }, { "left": "issue:44393", "right": "issue:45071", "accept": false, - "reason": "Vision model bounding-box output bug is unrelated to PretrainedConfig type-checking regressions." + "reason": "Vision model bbox issue vs `PretrainedConfig` type-checking regression; unrelated." }, { - "left": "issue:44038", - "right": "issue:45125", + "left": "issue:41669", + "right": "issue:43673", "accept": false, - "reason": "Qwen3-VL-Moe bug and missing _tp_plan for Qwen3_5Moe are both model-family issues but concern different breakages and code paths." + "reason": "Import-star performance problem vs missing GenerationMixin cache in chunked prefill; adjacent theme, but not the same bug." }, { - "left": "issue:41669", - "right": "issue:43673", + "left": "issue:44038", + "right": "issue:45125", "accept": false, - "reason": "Import * performance/lazy-loading problem and GenerationMixin cache missing during chunked_prefill are distinct regressions with different mechanisms." + "reason": "Qwen3-VL-Moe bug vs missing `_tp_plan` for tensor parallelism; different model features and fixes." }, { "left": "issue:44556", "right": "issue:45092", "accept": false, - "reason": "Both mention v5 migration incompatibilities, but one is general checkpoint reload failure and the other is remote-code meta initialization incompatibility for InternVL2; not the same bug." + "reason": "Checkpoint reload breakage across versions vs remote-code/meta-init incompatibility for InternVL2; not the same issue." }, { "left": "issue:44373", "right": "issue:44961", "accept": false, - "reason": "Wrong docstring for position_ids is documentation-only; 'racoon' provides no evidence of the same issue." + "reason": "Wrong docstring for `position_ids` vs unrelated placeholder title; no evidence of a shared bug." }, { "left": "issue:44704", "right": "issue:44908", "accept": false, - "reason": "AutoProcessor.from_pretrained kwargs forwarding and inverse_sqrt scheduler kwargs handling are unrelated API plumbing bugs." + "reason": "`cached_file` kwargs forwarding bug vs scheduler `timescale` forwarding bug; both forwarding-related but different APIs and fixes." }, { "left": "issue:42890", "right": "issue:44908", "accept": false, - "reason": "SAM HQ test flakiness from missing set_seed and inverse_sqrt scheduler kwargs are unrelated; one is test determinism, the other is scheduler configuration." + "reason": "Test flakiness from missing seeds vs scheduler kwargs bug; unrelated." }, { "left": "issue:44246", "right": "issue:44273", "accept": false, - "reason": "Import taking long sometimes and lazy loading broken are both import-related, but one is performance regression and the other is functional breakage." + "reason": "Import slowness vs lazy loading failure; same broad area but not the same concrete defect." }, { "left": "issue:43408", "right": "issue:44315", "accept": false, - "reason": "sam3_video vs sam3_tracker warning is a config/model-type mismatch; Liger Kernel not applied with model_init is a training integration issue." + "reason": "Model-type warning on instantiation vs Liger Kernel not applied with `model_init`; different behaviors." }, { "left": "issue:43976", "right": "issue:44393", "accept": false, - "reason": "Python version compatibility for Transformers 5.1.0 is unrelated to Qwen3-VL bounding box hallucinations." + "reason": "Python version incompatibility vs Qwen3-VL visual hallucination/error; unrelated." }, { "left": "issue:43352", "right": "issue:44393", "accept": false, - "reason": "Nemotron Flash Attention 2 unsupported and Qwen3-VL hallucination are different model-specific failures." + "reason": "Flash Attention 2 support error vs Qwen3-VL bbox hallucination; different model/runtime issues." }, { "left": "issue:44297", "right": "issue:45230", "accept": false, - "reason": "Tokenizer save_pretrained class mismatch and generic bug report do not share a concrete underlying defect." + "reason": "Tokenizer class mismatch in save_pretrained vs generic bug report title; no match." }, { "left": "issue:43576", "right": "issue:44908", "accept": false, - "reason": "Broken env CLI and inverse_sqrt scheduler kwargs are unrelated; same release window is not enough to merge." + "reason": "CLI env regression vs scheduler kwargs bug; unrelated." }, { "left": "issue:44908", "right": "issue:45071", "accept": false, - "reason": "Scheduler kwargs bug and PretrainedConfig type-checking regression are different API layers with different fixes." + "reason": "Scheduler kwargs forwarding vs config type-checking regression; different code paths." }, { "left": "issue:44295", "right": "issue:45230", "accept": false, - "reason": "position_ids buffer/read bug is unrelated to an unspecified generic bug report." + "reason": "`position_ids` buffer access error vs generic bug report title; no concrete overlap." }, { "left": "issue:43673", "right": "issue:44492", "accept": false, - "reason": "Generation cache missing during chunked_prefill and cache strategies typo are not the same issue; one is functional, the other is wording/labeling." + "reason": "Missing generation cache vs typo in cache strategies; different fixes." }, { "left": "issue:41669", "right": "issue:44351", "accept": false, - "reason": "Import * performance/lazy loading and missing HybridCache import are both import-related but distinct symbols and failures." + "reason": "Import-star performance regression vs missing `HybridCache` import; related import area, but not the same bug." }, { "left": "issue:41669", "right": "issue:45341", "accept": false, - "reason": "Model import performance issue and a little bug in testing_utils.py are unrelated." + "reason": "Import-time regression vs a little bug in `testing_utils.py`; no shared underlying issue." }, { "left": "issue:41669", "right": "issue:44492", "accept": false, - "reason": "Import * cleanup/performance and typo in cache strategies are not the same underlying bug." - }, - { - "left": "issue:44393", - "right": "issue:45478", - "accept": false, - "reason": "Qwen3-VL 2D box output error differs from Qwen3.5 Moe from_pretrained error; separate model families and failure points." + "reason": "Import-star slowdown vs cache-strategy typo; different problems." }, { "left": "issue:44315", "right": "issue:44829", "accept": false, - "reason": "Liger Kernel not applied with model_init and flash_attention_3 degenerate training are both training-related, but different mechanisms and code paths." - }, - { - "left": "issue:44077", - "right": "issue:44861", - "accept": false, - "reason": "Optional post_init policy for patchtsmixer is unrelated to _get_tied_weight_keys crashing on list-valued _tied_weights_keys." - }, - { - "left": "issue:43827", - "right": "issue:44908", - "accept": false, - "reason": "Docs still referencing pipeline() after removals and inverse_sqrt scheduler kwargs are different issues in different areas." - }, - { - "left": "issue:31795", - "right": "issue:33453", - "accept": false, - "reason": "Confusing forward docs for input_ids/past_key_values and tokenizer loading regression share only a general docs/loading theme; no same bug or fix path." + "reason": "Liger Kernel application with `model_init` vs degenerate training under flash_attention_3; different model-training bugs." }, - { - "left": "issue:41669", - "right": "issue:44908", - "accept": false, - "reason": "Model import performance and scheduler kwargs handling are unrelated regression types." - }, - { - "left": "issue:42371", - "right": "issue:44492", + { + "left": "issue:44393", + "right": "issue:45478", "accept": false, - "reason": "TF32 API settings guidance and cache strategies typo are clearly unrelated." + "reason": "Qwen3-VL image output bug vs Qwen3.5 Moe `from_pretrained` error; different models and failure modes." }, { - "left": "issue:44655", + "left": "issue:44077", "right": "issue:44861", "accept": false, - "reason": "Pipeline save_pretrained failure and tied weight keys AttributeError are separate serialization/model-introspection bugs." + "reason": "`post_init` allowance issue vs tied-weight-key AttributeError; unrelated." }, { - "left": "issue:43576", - "right": "issue:44273", + "left": "issue:43827", + "right": "issue:44908", "accept": false, - "reason": "Broken env command and lazy loading not working are both v5-related, but they do not describe the same code path or symptom." + "reason": "Docs still referencing removed pipeline APIs vs scheduler kwargs handling; different scopes." } ] }, @@ -14384,7 +14199,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-15354-384", + "cluster_id": "cluster-15354-385", "nodes": [ "issue:30333", "issue:31515", @@ -14403,6 +14218,7 @@ "issue:43519", "issue:43525", "issue:43572", + "issue:43576", "issue:43606", "issue:43704", "issue:43827", @@ -14415,6 +14231,7 @@ "issue:44485", "issue:44492", "issue:44623", + "issue:44655", "issue:44861", "issue:44908", "issue:44998", @@ -14426,7 +14243,12 @@ "issue:45468" ], "soft_pairs": [ + "issue:31795|issue:33453", + "issue:41669|issue:44908", + "issue:42371|issue:44492", + "issue:44655|issue:44861", "issue:44360|issue:45468", + "issue:43576|issue:44273", "issue:45230|issue:45310", "issue:43827|issue:44492", "issue:44273|issue:44908", @@ -14451,8 +14273,8 @@ "issue:44861|issue:45290", "issue:43525|issue:44861", "issue:41669|issue:44998", - "issue:43066|issue:44998", "issue:43704|issue:44998", + "issue:43066|issue:44998", "issue:43572|issue:44861", "issue:31515|issue:38617", "issue:44360|issue:44998", @@ -14464,256 +14286,286 @@ "issue:31515|issue:33453", "issue:43298|issue:44230" ], - "prepared_review_unit_hash": "8cf95ad57232c36f9449edf2272e4a91e4b54219534c3d5971f4e647f5771a65", + "prepared_review_unit_hash": "50881226b04dcaf01a66949e534287454847ffa2ba9f6c2bd46607047cf6d53f", "component_budget": { - "node_count": 384, - "item_count": 384, - "soft_pair_count": 1441, - "serialized_chars": 742743, - "estimated_input_tokens": 185686, - "estimated_eval_tokens": 371628 + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 }, "budget": { - "node_count": 38, - "item_count": 38, - "soft_pair_count": 37, - "serialized_chars": 32376, - "estimated_input_tokens": 8094, - "estimated_eval_tokens": 16444 + "node_count": 40, + "item_count": 40, + "soft_pair_count": 42, + "serialized_chars": 34967, + "estimated_input_tokens": 8742, + "estimated_eval_tokens": 17740 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is highly heterogeneous: docs, tokenizer/loading regressions, scheduler issues, model-specific bugs, and a few unrelated one-off reports. No soft edge looks like the same underlying bug/change, so none should be merged as duplicates.", - "confidence": 0.9, - "canonical_issue_reason": "If one issue must anchor the cluster, issue 43519 is the most concrete user-facing runtime bug; however, the overall set is not a true duplicate cluster and the rest do not cohere around one bug.", + "summary": "The set is mostly a loose collection of unrelated Transformers bugs and docs issues. The soft pairs share broad themes like loading, tokenizers, or v5 regressions, but not the same concrete failure mode, so none should be merged as duplicates.", + "confidence": 0.74, + "canonical_issue_reason": "issue:45092 is the most representative and actionable loading-regression report here: it names a specific v5 meta-initialization incompatibility, a clear affected class of checkpoints, and a plausible shared code path with nearby loading bugs.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43519 is the best representative only in the sense that it has a specific reproducible symptom and clear code path; it is not actually duplicate-similar to most of the others.", + "best_issue_reason": "issue:45092 is the best single issue to anchor the cluster because it is specific, technically grounded, and broad enough to represent the main loading-compatibility regression theme without being overly narrow to one model.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "One is a documentation confusion report; the other is a tokenizer-loading regression. Same broad area, different problem." + }, + { + "left": "issue:41669", + "right": "issue:44908", + "accept": false, + "reason": "Import-star performance cleanup and scheduler kwargs handling are unrelated code paths." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 behavior settings and a cache-strategies typo are not the same bug." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Saving Pipeline objects and tied-weight-key handling fail in different parts of serialization." + }, { "left": "issue:44360", "right": "issue:45468", "accept": false, - "reason": "DSA indexer ReLU omission vs Gemma4AudioRelPositionalEncoding; different models and different code paths." + "reason": "A missing ReLU in one DSA indexer and Gemma4 audio positional encoding are different model-implementation issues." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Broken env command and lazy-loading problems are unrelated features." }, { "left": "issue:45230", "right": "issue:45310", "accept": false, - "reason": "Generic bug report vs a specific Qwen3.5 Moe from_pretrained failure; not the same concrete issue." + "reason": "Both mention general bugs, but only one is a specific from_pretrained failure; the other is too vague to match." }, { "left": "issue:43827", "right": "issue:44492", "accept": false, - "reason": "Docs still referencing pipeline() vs a typo in cache strategy text; unrelated." + "reason": "Outdated pipeline docs and a cache-strategy typo are different scopes and symptoms." }, { "left": "issue:44273", "right": "issue:44908", "accept": false, - "reason": "Lazy loading malfunction vs inverse_sqrt scheduler kwargs handling; different subsystems." + "reason": "Lazy loading behavior and inverse_sqrt scheduler kwargs are unrelated." }, { "left": "issue:43519", "right": "issue:44485", "accept": false, - "reason": "Qwen3VL timestamp calculation bug vs GLM-5 RoPE implementation; different bugs." + "reason": "Timestamp calculation in a processor and GLM-5 RoPE math are not the same defect." }, { "left": "issue:41669", "right": "issue:43704", "accept": false, - "reason": "import * cleanup/perf issue vs VRAM leak in multi-threaded dataloading; unrelated." + "reason": "Import-star usage/performance and VRAM leakage in multi-threaded dataloaders are different problems." }, { "left": "issue:43519", "right": "issue:44908", "accept": false, - "reason": "Timestamp calculation bug vs scheduler kwargs bug; no shared code path." + "reason": "Processor timestamp math and scheduler parameter passing do not share a concrete bug." }, { "left": "issue:31795", "right": "issue:34689", "accept": false, - "reason": "Forward-doc wording confusion vs Llama 3.2 model loading regression; different problem classes." + "reason": "Documentation confusion about forward args does not duplicate a model-loading regression." }, { "left": "issue:31795", "right": "issue:37428", "accept": false, - "reason": "Documentation clarification vs missing flash_attention utility import; unrelated." + "reason": "A doc issue about model.forward is unrelated to a flash-attention import error." }, { "left": "issue:44315", "right": "issue:45092", "accept": false, - "reason": "model_init not applying Liger Kernel vs old InternVL2 checkpoints failing meta init; different failures." + "reason": "Liger-kernel application via model_init and meta-initialization checkpoint loading are different mechanisms." }, { "left": "issue:43704", "right": "issue:44485", "accept": false, - "reason": "VRAM leak under threaded dataloading vs RoPE implementation bug; different concrete bugs." + "reason": "VRAM leakage and RoPE implementation bugs are distinct code-path failures." }, { "left": "issue:44485", "right": "issue:45468", "accept": false, - "reason": "GLM-5 RoPE implementation vs Gemma4AudioRelPositionalEncoding; both positional-encoding-ish but not the same bug." + "reason": "GLM-5 RoPE math and Gemma4 audio positional encoding are separate architecture-specific issues." }, { "left": "issue:42371", "right": "issue:43704", "accept": false, - "reason": "TF32 API deprecation warning vs VRAM leak; unrelated." + "reason": "TF32 configuration warning and VRAM leakage are unrelated." }, { "left": "issue:43502", "right": "issue:43519", "accept": false, - "reason": "local_files_only still making requests vs Qwen3VL timestamp calculation; different code paths." + "reason": "Unexpected network requests with local_files_only and a timestamp bug in a processor are not duplicates." }, { "left": "issue:43519", "right": "issue:44623", "accept": false, - "reason": "Timestamp bug in processor logic vs processor.save_pretrained missing files; not the same issue." + "reason": "Processor timestamp calculations and missing processor save files are different save/load bugs." }, { "left": "issue:43519", "right": "issue:43704", "accept": false, - "reason": "Timestamp calculation bug vs multi-thread VRAM leak; unrelated." + "reason": "Timestamp miscalculation and dataloader-thread VRAM leaks are unrelated." }, { "left": "issue:44263", "right": "issue:45468", "accept": false, - "reason": "torch.split return-value handling in DSA indexer vs Gemma4AudioRelPositionalEncoding; different bugs." + "reason": "torch.split return handling in one indexer does not match Gemma4 audio encoding behavior." }, { "left": "issue:43606", "right": "issue:45468", "accept": false, - "reason": "CPU offload device mismatch in CI vs Gemma4AudioRelPositionalEncoding; unrelated." + "reason": "CPU offload device mismatch and Gemma4 audio positional encoding are different model-runtime issues." }, { "left": "issue:44261", "right": "issue:45468", "accept": false, - "reason": "MLA q_a_layernorm precision/config issue vs Gemma4AudioRelPositionalEncoding; different model families and failures." + "reason": "A missing rms_norm_eps precision issue is not the same as an audio positional encoding bug." }, { "left": "issue:43704", "right": "issue:45341", "accept": false, - "reason": "VRAM leak in dataloader threads vs testing_utils bug; unrelated." + "reason": "VRAM leakage in dataloader threads and a testing_utils bug do not match." }, { "left": "issue:42913", "right": "issue:43066", "accept": false, - "reason": "Both are tokenizer-v5 regressions, but one is a broad behavior difference and the other is a decoder-type bug; not clearly the same concrete bug." + "reason": "Both concern tokenizer behavior across v4/v5, but one is a broad behavior change report and the other is a specific decoder-type error." }, { "left": "issue:44908", "right": "issue:44998", "accept": false, - "reason": "Scheduler kwargs bug vs an unrelated 'Unemployment' issue; no relation." + "reason": "Scheduler kwargs handling and an unrelated unemployment issue are clearly not duplicates." }, { "left": "issue:44861", "right": "issue:45290", "accept": false, - "reason": "_get_tied_weight_keys AttributeError vs chat template crashes with tool calls; different functions and symptoms." + "reason": "Tied-weight key attribute handling and chat-template tool-call crashes are different areas of the codebase." }, { "left": "issue:43525", "right": "issue:44861", "accept": false, - "reason": "Llama4Config missing pad_token_id vs tied-weights list handling; unrelated." + "reason": "Missing pad_token_id and tied-weight-key crashes are separate config/serialization defects." }, { "left": "issue:41669", "right": "issue:44998", "accept": false, - "reason": "import * removal/perf topic vs unrelated issue title; no duplicate signal." + "reason": "Import-star cleanup and an unrelated unemployment issue have no overlap." }, { - "left": "issue:43066", + "left": "issue:43704", "right": "issue:44998", "accept": false, - "reason": "Tokenizer decoder-type regression vs unrelated issue title; no duplicate." + "reason": "VRAM leaks in training and an unrelated unemployment issue are not duplicates." }, { - "left": "issue:43704", + "left": "issue:43066", "right": "issue:44998", "accept": false, - "reason": "VRAM leak vs unrelated issue title; no duplicate." + "reason": "Tokenizer decoder type in v5 and an unrelated unemployment issue are not the same bug." }, { "left": "issue:43572", "right": "issue:44861", "accept": false, - "reason": "StableLmConfig pad_token_idx missing vs tied-weight key extraction crash; different bugs." + "reason": "Missing pad_token_idx in a config and tied-weight-key crashes affect different subsystems." }, { "left": "issue:31515", "right": "issue:38617", "accept": false, - "reason": "Slow from_pretrained checkpoint loading vs import error for layer_type_validation; different failure modes." + "reason": "A slow checkpoint load and an import error for layer_type_validation are unrelated." }, { "left": "issue:44360", "right": "issue:44998", "accept": false, - "reason": "DSA indexer ReLU bug vs unrelated title; no duplicate." + "reason": "DSA indexer implementation issues are unrelated to the unemployment placeholder issue." }, { "left": "issue:44230", "right": "issue:44623", "accept": false, - "reason": "fp8 inference support request vs processor.save_pretrained missing files; different concerns." + "reason": "FP8 support for a model family and missing save_pretrained files are different topics." }, { "left": "issue:30333", "right": "issue:31515", "accept": false, - "reason": "MLFlow job-reporting failure vs slow checkpoint loading; unrelated." + "reason": "MLflow reporting on failed training jobs and slow checkpoint loading are unrelated." }, { "left": "issue:31515", "right": "issue:36296", "accept": false, - "reason": "Checkpoint loading slowdown vs tensor parallel training bug; different issues." + "reason": "Slow checkpoint loading and tensor-parallel training bugs are different failures." }, { "left": "issue:31515", "right": "issue:31795", "accept": false, - "reason": "Loading slowdown vs documentation confusion in model.forward; not the same bug." + "reason": "Checkpoint-loading performance and forward-API documentation confusion are not the same issue." }, { "left": "issue:44230", "right": "issue:45310", "accept": false, - "reason": "fp8 support request vs Qwen3.5 Moe from_pretrained error; different change requests/issues." + "reason": "FP8 support work and a Qwen3.5 MoE from_pretrained error are distinct tasks." }, { "left": "issue:31515", "right": "issue:33453", "accept": false, - "reason": "Slow checkpoint loading vs tokenizer loading regression; both about loading, but not the same underlying bug." + "reason": "Slow checkpoint loading is not the same as a tokenizer-loading regression." }, { "left": "issue:43298", "right": "issue:44230", "accept": false, - "reason": "Continuous batching lacking audio-model support vs fp8 infer support for Qwen3-VL/Qwen3.5; different feature gaps." + "reason": "Continuous batching audio-model support and FP8 inference support are different feature gaps." } ] }, @@ -14743,12 +14595,12 @@ "soft_pairs": [ "issue:26350|issue:33260" ], - "prepared_review_unit_hash": "d3be0b0b8b33b09f5745306d0d052ca48bed77528ac32be91d701350bc4b264b", + "prepared_review_unit_hash": "b5fea8e9a0d6664c58b60317903a64d8a05aeae154b894602d0b8a793b2a723b", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1589, + "serialized_chars": 1590, "estimated_input_tokens": 398, "estimated_eval_tokens": 1052 }, @@ -14756,7 +14608,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1589, + "serialized_chars": 1590, "estimated_input_tokens": 398, "estimated_eval_tokens": 1052 }, @@ -14765,7 +14617,7 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: one tracks Flash Attention 2 support across more architectures, the other tracks GGUF support across more architectures. They are related only at a very broad \u201cmore architectures\u201d level, but they target different features and code paths.", + "summary": "These are not duplicates: one requests Flash Attention 2 support for more architectures, the other requests GGUF support for more architectures. They target different feature areas and underlying code paths.", "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, @@ -14776,7 +14628,7 @@ "left": "issue:26350", "right": "issue:33260", "accept": false, - "reason": "Different underlying requests: Flash Attention 2 vs GGUF. Same broad theme, but not the same bug/change and not something that would be merged into one issue." + "reason": "Different enhancement requests (Flash Attention 2 vs GGUF support) with no shared underlying bug or change." } ] }, @@ -14813,12 +14665,12 @@ "issue:43834|issue:44365", "issue:44365|issue:44370" ], - "prepared_review_unit_hash": "e777ade57937c408d5fb32356e17f1dd6a6339321078f85639b33f4209833d54", + "prepared_review_unit_hash": "8476dcdfa0ed61b3f136c2774154fe309c9b9308b91d0bf003a7e86c95c71be4", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4455, + "serialized_chars": 4453, "estimated_input_tokens": 1114, "estimated_eval_tokens": 2484 }, @@ -14826,7 +14678,7 @@ "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4455, + "serialized_chars": 4453, "estimated_input_tokens": 1114, "estimated_eval_tokens": 2484 }, @@ -14835,48 +14687,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are all highly similar i18n docs-translation issue templates, but they appear to target different languages/locales rather than the same underlying request. The Turkish issue is the only concrete, language-specific one.", - "confidence": 0.89, - "canonical_issue_reason": "Issue 27088 is the only explicitly Turkish request; the others are generic template issues for other languages, so there is no single duplicate issue to canonicalize across all four.", + "summary": "These are all docs-translation issues, but they are not the same underlying request: one is a concrete Turkish translation issue, while the others are generic template-style placeholders for arbitrary languages. The similarity is mostly boilerplate, so I would not merge them as duplicates.", + "confidence": 0.92, + "canonical_issue_reason": "Issue #27088 is the only concrete, language-specific request ([i18n-TR] Turkish). The others are generic placeholders and do not describe the same exact translation target.", "canonical_pr_reason": null, - "best_issue_reason": "27088 is the best representative for an actual concrete user request because it names the target language, whereas the others are generic placeholders and likely distinct locale-specific tasks.", + "best_issue_reason": "#27088 is the most actionable representative of the cluster because it names a specific language and a specific docs translation task, unlike the templated generic entries.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27088", "right": "issue:44370", "accept": false, - "reason": "Same translation-template pattern, but no evidence they refer to the same target language; likely separate locale requests." + "reason": "Same broad docs-translation theme, but one is Turkish-specific and the other is a generic language placeholder; not the same underlying issue." }, { "left": "issue:27088", "right": "issue:44365", "accept": false, - "reason": "They share the same generic i18n template, but the concrete language target is not the same underlying request." + "reason": "The titles/body structure match only as a template; the actual target language differs, so this is not a duplicate." }, { "left": "issue:27088", "right": "issue:43834", "accept": false, - "reason": "Looks like another locale-specific translation issue rather than a duplicate of the Turkish request." + "reason": "A concrete Turkish translation request is not the same as a generic placeholder issue for an unspecified language." }, { "left": "issue:43834", "right": "issue:44370", "accept": false, - "reason": "Near-identical template text, but these appear to be separate language-localization issues, not one bug/change." + "reason": "Both are template-like translation requests, but they are placeholders rather than the same concrete translation task." }, { "left": "issue:43834", "right": "issue:44365", "accept": false, - "reason": "Both are generic translation template issues, but there is no sign they represent the same language or task." + "reason": "High textual similarity comes from shared boilerplate, not from the same underlying language-specific request." }, { "left": "issue:44365", "right": "issue:44370", "accept": false, - "reason": "Highly similar wording, but likely different locale targets; not enough to treat as the same issue." + "reason": "These are nearly identical templates, but without a shared concrete target language they should not be treated as duplicates of the same issue." } ] }, @@ -14945,12 +14797,12 @@ "issue:43837|issue:45027", "issue:43837|issue:45177" ], - "prepared_review_unit_hash": "c00ad3ced43872c5e961fba684a8ceca1d138bf888d8aff544a94d5c6b380593", + "prepared_review_unit_hash": "7d57b4f4777b51a6759b26c325bce7dc0b49113a630a4287cefbea30fa11818c", "component_budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 28, - "serialized_chars": 17651, + "serialized_chars": 17650, "estimated_input_tokens": 4413, "estimated_eval_tokens": 9082 }, @@ -14958,7 +14810,7 @@ "node_count": 14, "item_count": 14, "soft_pair_count": 28, - "serialized_chars": 17651, + "serialized_chars": 17650, "estimated_input_tokens": 4413, "estimated_eval_tokens": 9082 }, @@ -14967,180 +14819,180 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are mostly independent model-request issues across unrelated models and modalities. The apparent similarity comes from repeated request phrasing (\"Add/Support/Proposal\"), not the same underlying bug or change, so none of the soft edges should be merged.", - "confidence": 0.96, - "canonical_issue_reason": "issue:42738 is the broadest umbrella-style request in the set ('BERT-like models with RoPE'), so it is the closest representative issue, but the overall cluster is still heterogeneous and not truly duplicate-heavy.", + "summary": "These are mostly one-off model addition requests across different model families and modalities. They are loosely similar at the \u201cadd a model/config support\u201d level, but not duplicates of the same underlying bug or change.", + "confidence": 0.94, + "canonical_issue_reason": "issue:37015 ('Add NeoBERT') is a clear, generic model-add request and sits near the center of the similarity set, but it is still distinct from the other model-specific requests.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42738 is the most general, reusable framing among these issues and best serves as a representative anchor, though it does not subsume the other model-specific requests.", + "best_issue_reason": "issue:37015 is the best representative issue for this cluster because it is open, straightforward, and broadly matches the repeated pattern of requesting new model support without being tied to a narrow edge case.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36817", "right": "issue:41211", "accept": false, - "reason": "Different model requests: EuroBert config support vs DEIMv2. Shared request language only, no same code-path change." + "reason": "Different model requests ('EuroBert' vs 'DEIMv2'); same broad category, but not the same underlying change." }, { "left": "issue:36817", "right": "issue:43671", "accept": false, - "reason": "EuroBert config support and Qwen3-TTS support are unrelated models/modalities." + "reason": "'EuroBert' config support is unrelated to Qwen3-TTS support; different model and modality." }, { "left": "issue:37015", "right": "issue:43837", "accept": false, - "reason": "NeoBERT and Qwen3-ASR are different models and different task domains." + "reason": "'NeoBERT' vs 'Qwen3-ASR' are different model families and different tasks." }, { "left": "issue:37015", "right": "issue:44121", "accept": false, - "reason": "NeoBERT vs OpenAI weight-sparse transformer are distinct architecture requests." + "reason": "A NeoBERT addition is not the same change as adding OpenAI weight-sparse transformer support." }, { "left": "issue:37015", "right": "issue:43909", "accept": false, - "reason": "NeoBERT and LFM2.5 Audio are different model families and modalities." + "reason": "'NeoBERT' and 'LFM2.5 Audio 1.5B' are different model requests, with no shared code-path problem." }, { "left": "issue:37015", "right": "issue:42503", "accept": false, - "reason": "Both are model-addition requests, but NeoBERT and ModernVBERT are different concrete models." + "reason": "'NeoBERT' and 'ModernVBERT' are different model additions; similarity is only at the generic config-support level." }, { "left": "issue:37015", "right": "issue:45027", "accept": false, - "reason": "NeoBERT and Voxtral-4B-TTS-2603 are unrelated models; one is not a duplicate of the other." + "reason": "'NeoBERT' is unrelated to 'Voxtral-4B-TTS-2603'; different model and audio/TTS path." }, { "left": "issue:41115", "right": "issue:43837", "accept": false, - "reason": "MiniCPM3 architecture support and Qwen3-ASR support are different model integrations." + "reason": "MiniCPM3 architecture support and Qwen3-ASR support are separate model integrations." }, { "left": "issue:41115", "right": "issue:44121", "accept": false, - "reason": "MiniCPM3 vs the weight-sparse transformer request are different architectures and code paths." + "reason": "MiniCPM3 architecture support is not the same as weight-sparse transformer support." }, { "left": "issue:41115", "right": "issue:42503", "accept": false, - "reason": "MiniCPM3 architecture work and ModernVBERT support are distinct model requests." + "reason": "'MiniCPM3' and 'ModernVBERT' are distinct model requests." }, { "left": "issue:41115", "right": "issue:45027", "accept": false, - "reason": "MiniCPM3 and Voxtral TTS target different models and modalities." + "reason": "MiniCPM3 architecture work is unrelated to Voxtral TTS support." }, { "left": "issue:41211", "right": "issue:43671", "accept": false, - "reason": "DEIMv2 and Qwen3-TTS are unrelated model requests (vision vs speech)." + "reason": "'DEIMv2' and 'Qwen3-TTS' are different model additions and likely different code paths." }, { "left": "issue:42503", "right": "issue:43837", "accept": false, - "reason": "ModernVBERT models and Qwen3-ASR support are different model families/modality targets." + "reason": "ModernVBERT model support is separate from Qwen3-ASR support." }, { "left": "issue:42503", "right": "issue:43671", "accept": false, - "reason": "ModernVBERT and Qwen3-TTS are distinct model integrations." + "reason": "ModernVBERT and Qwen3-TTS are different models/tasks; no same underlying bug." }, { "left": "issue:42503", "right": "issue:42738", "accept": false, - "reason": "ModernVBERT is a specific model request; 'BERT-like models with RoPE' is a broad umbrella feature, not the same change." + "reason": "'ModernVBERT models' and 'BERT-like models with RoPE' overlap conceptually, but they are not the same concrete change." }, { "left": "issue:42738", "right": "issue:45027", "accept": false, - "reason": "BERT-like RoPE support and Voxtral TTS are unrelated feature requests." + "reason": "A generic RoPE request for BERT-like models is not the same as adding Voxtral TTS support." }, { "left": "issue:43671", "right": "issue:43837", "accept": false, - "reason": "Qwen3-TTS and Qwen3-ASR are related model family requests but different capabilities and integration paths; not the same underlying change." + "reason": "Qwen3-TTS and Qwen3-ASR share a family name, but they are different modalities and separate support requests." }, { "left": "issue:43671", "right": "issue:45027", "accept": false, - "reason": "Both are TTS-related, but Qwen3-TTS and Voxtral-4B-TTS-2603 are different models with separate support work." + "reason": "Both are audio-related, but TTS support for Qwen3 and Voxtral TTS are distinct model integrations." }, { "left": "issue:43671", "right": "issue:45177", "accept": false, - "reason": "Qwen3-TTS support and DeepSeek-OCR2 are different models and different modalities." + "reason": "Qwen3-TTS and DeepSeek-OCR2 are unrelated model requests." }, { "left": "issue:43671", "right": "issue:44121", "accept": false, - "reason": "Qwen3-TTS support and the weight-sparse transformer request are unrelated model changes." + "reason": "Qwen3-TTS support is not the same as adding a weight-sparse transformer architecture." }, { "left": "issue:43837", "right": "issue:45322", "accept": false, - "reason": "Qwen3-ASR support and EUPE are different model families and different problem types." + "reason": "Qwen3-ASR support and EUPE encoder support are different model changes." }, { "left": "issue:43837", "right": "issue:43909", "accept": false, - "reason": "Qwen3-ASR and LFM2.5 Audio are both audio-related, but they are different models and not one mergeable PR." + "reason": "ASR support and LFM2.5 Audio support are related only by being audio models, not the same change." }, { "left": "issue:43909", "right": "issue:44205", "accept": false, - "reason": "LFM2.5 Audio and SAM3-LiteText are different models/modalities." + "reason": "LFM2.5 Audio support and SAM3-LiteText are different model families." }, { "left": "issue:43671", "right": "issue:43909", "accept": false, - "reason": "Qwen3-TTS and LFM2.5 Audio are separate model support requests." + "reason": "Qwen3-TTS and LFM2.5 Audio 1.5B are separate audio model requests, not duplicates." }, { "left": "issue:44121", "right": "issue:45322", "accept": false, - "reason": "OpenAI weight-sparse transformer and EUPE are unrelated model requests." + "reason": "Weight-sparse transformer support and EUPE encoder support target different architectures." }, { "left": "issue:43837", "right": "issue:44121", "accept": false, - "reason": "Qwen3-ASR support and the weight-sparse transformer request do not share the same concrete code change." + "reason": "Qwen3-ASR support is unrelated to adding OpenAI weight-sparse transformer support." }, { "left": "issue:43837", "right": "issue:45027", "accept": false, - "reason": "Qwen3-ASR and Voxtral TTS are different speech models with different integration work." + "reason": "Qwen3-ASR and Voxtral-4B-TTS-2603 are different audio model requests." }, { "left": "issue:43837", "right": "issue:45177", "accept": false, - "reason": "Qwen3-ASR support and DeepSeek-OCR2 are unrelated model additions." + "reason": "Qwen3-ASR and DeepSeek-OCR2 are unrelated model additions." } ] }, @@ -15181,12 +15033,12 @@ "issue:43668|issue:44031", "issue:43668|issue:43813" ], - "prepared_review_unit_hash": "faa81c45faf325b62b86ea8c95d5257156fe2f46532c78e11ee0bdda563178bd", + "prepared_review_unit_hash": "73bf6ad4adb85a211f3e1ccc2beab8fe300da25fd0bb9687de6ece9ee1b13d87", "component_budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 8, - "serialized_chars": 6294, + "serialized_chars": 6293, "estimated_input_tokens": 1574, "estimated_eval_tokens": 3404 }, @@ -15194,7 +15046,7 @@ "node_count": 6, "item_count": 6, "soft_pair_count": 8, - "serialized_chars": 6294, + "serialized_chars": 6293, "estimated_input_tokens": 1574, "estimated_eval_tokens": 3404 }, @@ -15203,60 +15055,60 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These issues are only superficially similar; they cover unrelated bugs in different models/components (normalization, loss kwargs, config typing, a typo, and tokenizer warnings). None should be merged as duplicates.", - "confidence": 0.96, - "canonical_issue_reason": "No canonical issue: the cluster is heterogeneous and the titles point to distinct code paths and bug classes.", + "summary": "The cluster is only loosely similar on wording, but the issues describe unrelated bugs across different models/configs/utilities. I would not merge any of the soft pairs as duplicates.", + "confidence": 0.18, + "canonical_issue_reason": "issue:44031 is the most concrete and broadly actionable bug report in the set, with the most discussion and a clear regression-style description.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43668 is the clearest standalone report, but it is not a true duplicate representative for the rest of the cluster.", + "best_issue_reason": "issue:44031 is the best representative issue because it is specific, reproducible in intent, and not just a typo or type-hint cleanup.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "Different models and different problems: Llama4TextL2Norm/RMS norm semantics vs ModernBERTConfig type-hint correctness." + "reason": "One is about Llama4TextL2Norm possibly using RMS norm, the other is a ModernBERTConfig type-hint problem; different code paths and bug types." }, { "left": "issue:42936", "right": "issue:43240", "accept": false, - "reason": "Mask2former ignore_value not being used and kwargs not forwarded to loss are separate bug paths with different fixes." + "reason": "Mask2Former ignore_value not being used is unrelated to kwargs not being forwarded to a loss function; only broad training-loss context overlaps." }, { "left": "issue:42936", "right": "issue:44031", "accept": false, - "reason": "Mask2former loss handling and tokenizer regex warning regression are unrelated subsystems." + "reason": "Mask2Former loss handling and tokenizer regex warnings are unrelated subsystems with different failure modes." }, { "left": "issue:42936", "right": "issue:43813", "accept": false, - "reason": "A Mask2former ignore_value bug and a quantization-operation typo are not the same underlying issue." + "reason": "A missing ignore_value usage in Mask2Former is not the same as a typo in a quantization attribute name." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "Different components and symptoms: ignored loss parameter vs config type-hint error." + "reason": "Mask2Former loss behavior and a ModernBERT config type annotation issue are unrelated." }, { "left": "issue:43240", "right": "issue:43668", "accept": false, - "reason": "Loss kwargs forwarding and ModernBERTConfig typing are unrelated; no shared code path." + "reason": "Forwarding kwargs to loss calculation is a runtime behavior bug; incorrect norm_eps typing is a config typing issue." }, { "left": "issue:43668", "right": "issue:44031", "accept": false, - "reason": "A config annotation bug and tokenizer regex warnings are different issues entirely." + "reason": "ModernBERTConfig norm_eps typing and tokenizer regex warnings are unrelated concerns." }, { "left": "issue:43668", "right": "issue:43813", "accept": false, - "reason": "ModernBERTConfig type hinting and a typo in quantization_operation are distinct defects." + "reason": "These are separate small issues in different areas: config typing versus a string typo in quantization code." } ] }, @@ -15286,7 +15138,7 @@ "soft_pairs": [ "issue:41084|issue:43037" ], - "prepared_review_unit_hash": "36dd49708a945de715129816c04e6980dea1ed05662e1770ecd4f6b31f27a076", + "prepared_review_unit_hash": "6713cdc4c094c1720e2741cce0d098f986096d7a6ae74c24a8a80db4dfc521ab", "component_budget": { "node_count": 2, "item_count": 2, @@ -15308,18 +15160,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are not duplicates: one asks for Set Block Decoding, while the other requests DeepSeek v3.2 support. They describe different features/problem areas with no shared targets or evidence of the same underlying bug.", + "summary": "The two issues are not duplicates: one is about block decoding behavior, while the other is a feature/support request for DeepSeek v3.2. They do not describe the same underlying bug or change.", "confidence": 0.96, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "No single issue stands out as a canonical duplicate target; the issues are unrelated and should remain separate.", + "best_issue_reason": "If one had to be kept as the more actionable standalone item, #43037 is clearer and more specific; however, these issues should not be clustered as duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41084", "right": "issue:43037", "accept": false, - "reason": "Different requests in different areas: block decoding behavior versus model support. No indication they refer to the same underlying change or bug." + "reason": "The titles and implied scopes are unrelated: block decoding is an implementation/behavior topic, while DeepSeek v3.2 support is a model compatibility feature request. They do not appear to be the same bug or change." } ] }, @@ -15349,12 +15201,12 @@ "soft_pairs": [ "issue:43598|issue:43600" ], - "prepared_review_unit_hash": "6d3aeb85e76bdd4df7c3ca6013e0fd0f064ccadb95e39a1458ffa06d9445497c", + "prepared_review_unit_hash": "6c1ea85627d6344b0ff1098379f46bf8680483470be4944f3bb9938bd2993306", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1584, + "serialized_chars": 1583, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -15362,7 +15214,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1584, + "serialized_chars": 1583, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -15371,18 +15223,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues both mention custom trainers, but they describe different Trainer behaviors: one about `remove_unused_column` customizability, the other about FLOPS estimation. They do not look like the same underlying bug.", - "confidence": 0.91, - "canonical_issue_reason": "Issue 43598 is the more concrete and self-contained Trainer behavior change; issue 43600 is a separate FLOPS-estimation concern rather than the same bug.", + "summary": "These are related only at a high level: both mention custom trainers and Trainer behavior, but they target different problems. One is about `remove_unused_column` configurability, the other about FLOPS estimation for custom trainers.", + "confidence": 0.94, + "canonical_issue_reason": "Issue #43598 is the better anchor because it is the more general Trainer-customizability concern and is phrased as a behavior change rather than a narrow metric-specific fix.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43598 is the better representative for this cluster because it is a precise API/behavior request, while 43600 targets a different code path entirely.", + "best_issue_reason": "Issue #43598 is the best standalone issue to keep for this cluster because it captures the broader custom-trainer customization theme more cleanly than the FLOPS-specific report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43598", "right": "issue:43600", "accept": false, - "reason": "Both involve custom trainers, but one is about `remove_unused_column` behavior and the other is about FLOPS accounting. They do not describe the same underlying defect or a single fixable change." + "reason": "Do not treat as duplicates: they share the same Trainer/custom-trainer context, but one concerns `remove_unused_column` behavior and the other concerns FLOPS counting. Different underlying bugs, different code paths, and not a single merged fix." } ] }, @@ -15412,7 +15264,7 @@ "soft_pairs": [ "issue:43684|issue:44405" ], - "prepared_review_unit_hash": "c23531b7d6fb5bc04c0b0dcce1b4ed2ce330162cca429112f413c86c6d19714f", + "prepared_review_unit_hash": "ff52469a3c880ce0a151dbbe97de0e742168deae07d23d198aa63ed0591c0ab9", "component_budget": { "node_count": 2, "item_count": 2, @@ -15434,18 +15286,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related feature requests for Qwen model integration, but they target different model families and different AutoModel entry points, so they should not be treated as duplicates.", - "confidence": 0.92, - "canonical_issue_reason": "Issue 43684 is the broader request: it asks for Qwen3-Omni registration in AutoModel and AutoModelForConditionalGeneration, making it the better umbrella issue for Qwen registration work.", + "summary": "These two issues are about different model families and different AutoModel entry points, so they are not duplicates.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue: one request is for Qwen3-Omni registration in AutoModel/AutoModelForConditionalGeneration, the other is for AutoModelForSequenceClassification support for Qwen3.5.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43684 is the best issue candidate because it covers a concrete registration gap and is phrased as a general model-registration fix rather than a narrower, model-specific extension.", + "best_issue_reason": "Issue 43684 is the broader, more central model-registration request, but it should not be treated as a duplicate of 44405.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43684", "right": "issue:44405", "accept": false, - "reason": "Reject: 43684 is about Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration, while 44405 asks for AutoModelForSequenceClassification support for Qwen3.5 (Qwen3_5Config). Different model variants and different API targets mean different underlying changes." + "reason": "Different underlying changes: Qwen3-Omni auto registration vs Qwen3.5 sequence classification support. They target different models and auto classes, so they should not be merged as duplicates." } ] }, @@ -15482,12 +15334,12 @@ "issue:43845|issue:43859", "issue:43846|issue:43859" ], - "prepared_review_unit_hash": "42a7c2536660c58fd44f84cf31b350f6abb9ec58d90c1134396c82faa2d93e4a", + "prepared_review_unit_hash": "be4346cea43a9ab2f8f119c822518ec9aff176f0c9f700001d6e20a054c2a4b6", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4265, + "serialized_chars": 4266, "estimated_input_tokens": 1067, "estimated_eval_tokens": 2390 }, @@ -15495,7 +15347,7 @@ "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4265, + "serialized_chars": 4266, "estimated_input_tokens": 1067, "estimated_eval_tokens": 2390 }, @@ -15504,58 +15356,58 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Three closed issues titled \"huggingface\" appear to be duplicates of the same report; the \"Fett\" issue looks unrelated and should not be merged with them.", - "confidence": 0.61, - "canonical_issue_reason": "Issue 43845 is the earliest of the near-identical \"huggingface\" reports and is the best representative for that duplicate set.", + "summary": "This cluster is mostly low-signal noise: three near-identical closed issues titled \"huggingface \" and one unrelated issue titled \"Fett\". The three huggingface issues look similar or likely related, but the packet only provides soft similarity evidence, so I would not confirm them as duplicates; \"Fett\" is unrelated.", + "confidence": 0.78, + "canonical_issue_reason": "43845 is the earliest of the repeated \"huggingface \" issues and is a reasonable representative for the similar low-signal reports, though the evidence is not strong enough to confirm exact duplication.", "canonical_pr_reason": null, - "best_issue_reason": "43845 is the strongest issue cluster anchor because it is earliest and matches the other two nearly verbatim-title reports.", + "best_issue_reason": "43845 is the best issue representative because it is the earliest of the three nearly identical \"huggingface \" reports and has no distinguishing evidence against it.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43835", "right": "issue:43845", "accept": false, - "reason": "Different title and no shared concrete details; insufficient evidence that they are the same underlying issue." + "reason": "Different titles and no supporting body content or shared targets; only weak similarity is present." }, { "left": "issue:43835", "right": "issue:43846", "accept": false, - "reason": "Different title and no shared concrete details; this does not look like the same report." + "reason": "Different titles and no supporting body content or shared targets; not enough evidence for the same underlying issue." }, { "left": "issue:43835", "right": "issue:43859", "accept": false, - "reason": "Different title and no shared concrete details; no basis to treat these as duplicates." + "reason": "\"Fett\" is unrelated to the repeated \"huggingface \" issues and lacks evidence of the same underlying problem." }, { "left": "issue:43845", "right": "issue:43846", - "accept": true, - "reason": "Both are closed issues with the same \"huggingface\" title and nearly identical sparse metadata, suggesting the same underlying report." + "accept": false, + "reason": "The titles are the same, but the packet only shows soft similarity with no deterministic acceptance or concrete shared evidence, so this is not enough to confirm a duplicate." }, { "left": "issue:43845", "right": "issue:43859", - "accept": true, - "reason": "Same title and closely related sparse issue metadata indicate duplicate reports of the same underlying issue." + "accept": false, + "reason": "They look similar at a very high level, but there is no concrete evidence that they describe the same issue, so the edge should not be accepted." }, { "left": "issue:43846", "right": "issue:43859", - "accept": true, - "reason": "Same title and similar minimal content strongly suggest these are duplicate issue reports." + "accept": false, + "reason": "Only soft similarity is present; that is insufficient to treat these as the same issue." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary stays within the packet: 43845/43846/43859 share the same title, similar sparse metadata, and strong soft-similarity signals, while 43835 has a different title and only weak similarity. The duplicate claim is phrased conservatively, and the anchor choice (earliest of the similar set) is grounded." + "accept": false, + "feedback": "The cluster summary is mostly grounded, but the soft-edge handling is too strong. In the packet, every pair has only `soft_similarity`, and `deterministic_accept` is false for all of them. That makes the statements that 43845/43846/43859 are duplicates too confident. It is fine to say the three `huggingface ` issues are similar or likely related, but not enough evidence is shown to accept those edges as same-issue duplicates. The `Fett` issue being unrelated is reasonable." }, "evaluator_used": true, - "retried": false, - "accepted_nontrivial_soft_edge": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -15589,88 +15441,88 @@ "issue:44572|issue:44937", "issue:44573|issue:44937" ], - "prepared_review_unit_hash": "0ff7c16a0b4d78916dedca0dfbeaf0cd210f00cffbdff85e4e1360016a6dd6a8", + "prepared_review_unit_hash": "b7defc7b9b413d1537f8cfcc020d2889c8eb923f17c28a08bd665def2d3050eb", "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 looks like a mix of unrelated low-information spam/doc-placeholder issues rather than one real bug report. The only clear near-duplicate subgroup is the repeated \"\" titles, but that is not enough to treat them as the same underlying issue.", - "confidence": 0.89, - "canonical_issue_reason": "issue:44371 is the earliest of the repeated \"\" issues and is the closest representative of that spam subgroup.", + "summary": "These are mostly unrelated low-information spam issues; there is no real duplicate cluster. The only reasonable grouping signal is superficial title similarity, not shared bug content.", + "confidence": 0.93, + "canonical_issue_reason": "Issue 44371 is the earliest generic \"\" report and is as representative as any other for the spam-only subset.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44371 is the best representative only by being the earliest repeated spam placeholder; there is no meaningful technical signal in the cluster.", + "best_issue_reason": "44371 is the least-specific, earliest item and best fits as a placeholder representative for this noisy cluster, though it is not a true duplicate canonical.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44371", "right": "issue:44572", "accept": false, - "reason": "Same placeholder title, but no body detail or technical signal to show the same underlying issue; treat as separate spam reports." + "reason": "Both are spam-labeled, but there is no shared underlying problem or concrete issue description to indicate the same bug/change." }, { "left": "issue:44371", "right": "issue:44573", "accept": false, - "reason": "Same placeholder title, but nothing ties them to one concrete bug/change beyond generic spam similarity." + "reason": "Same as above: identical-looking spam titles do not establish a duplicate underlying issue." }, { "left": "issue:44371", "right": "issue:44418", "accept": false, - "reason": "Different titles and no shared issue context; the documentation placeholder does not look like the same report as the spam issue." + "reason": "One is generic spam and the other is a documentation enhancement suggestion; they are clearly different in intent and content." }, { "left": "issue:44371", "right": "issue:44937", "accept": false, - "reason": "Different titles and no evidence of the same underlying problem; this looks like a separate low-effort issue." + "reason": "Generic spam vs a named quote-like title; no evidence they describe the same problem." }, { "left": "issue:44418", "right": "issue:44572", "accept": false, - "reason": "Generic placeholder/documentation wording versus spam placeholder; no concrete overlap indicating the same issue." + "reason": "Documentation suggestion and spam are not the same underlying issue." }, { "left": "issue:44418", "right": "issue:44573", "accept": false, - "reason": "No shared bug description or target; these are just low-information issues with different titles." + "reason": "Documentation suggestion and spam are unrelated despite superficial low-information similarity." }, { "left": "issue:44572", "right": "issue:44573", "accept": false, - "reason": "Both are spam placeholders, but there is no evidence they refer to one identical underlying problem rather than two separate spam submissions." + "reason": "Both are titled \"\", but that is only a spam pattern; no concrete shared bug/change is visible." }, { "left": "issue:44572", "right": "issue:44937", "accept": false, - "reason": "The titles are unrelated and there is no technical evidence for a duplicate relationship." + "reason": "Spam placeholder vs a specific named title; no shared issue substance." }, { "left": "issue:44573", "right": "issue:44937", "accept": false, - "reason": "Different titles with no shared concrete bug/change; reject as duplicate." + "reason": "Spam placeholder and a specific named title are not enough to treat these as duplicates." } ] }, @@ -15700,12 +15552,12 @@ "soft_pairs": [ "issue:45145|issue:45151" ], - "prepared_review_unit_hash": "f838d56bdd6d9620ade23c1eb38322013a5bc4752eb00857380b3ea68ae05849", + "prepared_review_unit_hash": "5630e0b17106fee10a55a0f7e1dc2949c15354aef95e66e0c0911b79030c4a30", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1646, + "serialized_chars": 1645, "estimated_input_tokens": 412, "estimated_eval_tokens": 1080 }, @@ -15713,7 +15565,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1646, + "serialized_chars": 1645, "estimated_input_tokens": 412, "estimated_eval_tokens": 1080 }, @@ -15722,27 +15574,27 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Likely duplicate issue pair with identical titles and similar-sized bodies, but the evidence only supports a moderate-confidence deduplication call. #45145 is the best anchor because it was created first and already has one inbound reference.", - "confidence": 0.78, - "canonical_issue_reason": "#45145 is the better canonical issue because it was created earlier and has one inbound reference, making it the most natural dedupe anchor among two otherwise similarly titled issues.", + "summary": "The two issues are effectively duplicates: they have the same title, near-identical body size, and no distinguishing metadata beyond one having an inbound reference. Issue #45145 is the better canonical candidate.", + "confidence": 0.98, + "canonical_issue_reason": "#45145 is the stronger canonical issue because it has an inbound reference, suggesting other discussion already points to it.", "canonical_pr_reason": null, - "best_issue_reason": "#45145 is the best issue candidate for the same practical reasons: earliest creation time and an inbound reference give it a small but real edge as the representative issue.", + "best_issue_reason": "#45145 best represents the cluster since it appears to be the referenced/central copy of the duplicated report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45145", "right": "issue:45151", "accept": true, - "reason": "The issues have the same exact title and a soft_similarity signal, with very similar body lengths, which is enough to treat them as likely duplicates, though the evidence is not strong enough to claim near-identical content." + "reason": "Same title and highly similar content; no evidence they describe different underlying problems." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The pair is plausibly duplicate, but the writeup overstates the evidence. The packet only shows matching titles, similar body lengths, and a soft_similarity signal; it does not demonstrate near-identical body content or a near-exact duplicate with high confidence. The canonical/best-issue reasoning about creation time and inbound reference is grounded, but the summary and confidence are too strong for the available evidence." + "accept": true, + "feedback": "The summary is grounded in the packet: both issues have the same title, very similar body lengths, and no file/target links. The inbound reference on #45145 is a reasonable basis for treating it as the canonical copy. The duplicate verdict is appropriately conservative for a soft-similarity pair." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -15820,7 +15672,7 @@ "soft_pairs": [ "issue:45672|issue:45674" ], - "prepared_review_unit_hash": "dd044ff436f70335edbe05901c9cbf3f89aa017eed378f2420c82270520c29ef", + "prepared_review_unit_hash": "37bf3d74887375b3ef196de2e72d8fd7f6bdcc56ec0af2a0374a5c84b5501db6", "component_budget": { "node_count": 2, "item_count": 2, @@ -15842,18 +15694,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues both concern 4-bit inference with BitsAndBytes, but they describe different failure modes and likely different fixes: one is a torch.finfo TypeError on uint8 weights in Gemma4 audio modules, the other is an lm_head exclusion regression when llm_int8_skip_modules is set. Not duplicates.", - "confidence": 0.92, - "canonical_issue_reason": "Issue #45674 is the better canonical issue because it states a concrete, reproducible regression in BitsAndBytesConfig with a clear trigger and failure mode, making it easier to route and track.", + "summary": "These are both 4-bit inference issues, but they describe different failure modes and likely different code paths: one is a Gemma4/audio-module uint8 `torch.finfo()` TypeError, the other is a BitsAndBytesConfig regression where `llm_int8_skip_modules` removes the default `lm_head` exclusion and triggers an AssertionError. I would not deduplicate them.", + "confidence": 0.96, + "canonical_issue_reason": "None \u2014 the two issues do not appear to be the same underlying bug.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #45674 is the strongest representative of the cluster: specific configuration, explicit symptom (AssertionError), and a well-scoped behavioral regression.", + "best_issue_reason": "Issue 45674 is the better representative because it describes a clearer, more general BitsAndBytesConfig regression affecting 4-bit inference.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45672", "right": "issue:45674", "accept": false, - "reason": "Both are about 4-bit inference, but #45672 is a Gemma4/audio-module torch.finfo TypeError while #45674 is an lm_head exclusion bug from llm_int8_skip_modules. Different code paths and fixes, so not a duplicate." + "reason": "Different bugs: 45672 is a Gemma4/audio-module `torch.finfo()` TypeError on uint8 weights, while 45674 is a BitsAndBytesConfig `llm_int8_skip_modules`/`lm_head` exclusion regression causing an AssertionError in 4-bit inference. Shared quantized inference context is too broad to merge." } ] }, @@ -15883,40 +15735,40 @@ "soft_pairs": [ "issue:45762|issue:45763" ], - "prepared_review_unit_hash": "eb65dfc43adad1a71e31ea0bc7e94b664eb0c7e876130387013355cb3c846991", + "prepared_review_unit_hash": "45797cdc7e1dca785be5e88e9c0551fd59918545ff104d9c0ed22294787222ed", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1477, - "estimated_input_tokens": 370, - "estimated_eval_tokens": 996 + "serialized_chars": 1476, + "estimated_input_tokens": 369, + "estimated_eval_tokens": 994 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1477, - "estimated_input_tokens": 370, - "estimated_eval_tokens": 996 + "serialized_chars": 1476, + "estimated_input_tokens": 369, + "estimated_eval_tokens": 994 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two closed issues appear unrelated and low-signal: both are just personal-name/date titles with no body content, so they do not look like duplicate reports of the same underlying problem.", - "confidence": 0.98, + "summary": "The two closed issues do not look like duplicates of a software problem; they appear to be unrelated name/date placeholders with no substantive issue content.", + "confidence": 0.97, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Neither issue is a strong canonical representative, but if forced, issue 45762 is slightly earlier and equally sparse; however, there is no meaningful bug/report content to canonicalize.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45762", "right": "issue:45763", "accept": false, - "reason": "Unrelated titles with different names/dates and no body or shared context; not the same bug or request." + "reason": "Different personal-name titles with different dates and no body or shared problem description; not the same underlying issue." } ] }, @@ -15946,7 +15798,7 @@ "soft_pairs": [ "issue:45764|issue:45765" ], - "prepared_review_unit_hash": "190146158f3105cf31a06e7a4f5c3a473cbae9dd8b12ec718406ed1871325aef", + "prepared_review_unit_hash": "0d42dd80081cf6f91ed964b5146cded90f5a5a33a687c3d0203f68f72876e131", "component_budget": { "node_count": 2, "item_count": 2, @@ -15968,8 +15820,8 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues have unrelated person-name/date titles and no body or linked evidence, so there is no credible duplicate relationship.", - "confidence": 0.96, + "summary": "These are two closed issues with unrelated, low-information personal-name/date titles and no shared content, targets, or filenames. They do not look like duplicates or the same underlying report.", + "confidence": 0.97, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, @@ -15979,7 +15831,7 @@ "left": "issue:45764", "right": "issue:45765", "accept": false, - "reason": "Titles are different names with different dates and there is no supporting discussion, linkage, or shared technical context; they do not look like the same underlying issue." + "reason": "The titles reference different people and different dates, with no body text, targets, filenames, or other evidence of a shared bug or request. This does not appear to be the same issue." } ] }, @@ -16038,154 +15890,160 @@ "pull_request:43823|pull_request:45795", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "0a1a3975c76d68eec07b631c0ab52c7e64576407c4849919b8eac3683d36b3d8", + "prepared_review_unit_hash": "3460822e8fd9721d900972f61ee57dfb39a73e56ed8b9e8a72ad255bfb3f8101", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 19, - "serialized_chars": 35202, - "estimated_input_tokens": 8801, - "estimated_eval_tokens": 17858 + "serialized_chars": 35200, + "estimated_input_tokens": 8800, + "estimated_eval_tokens": 17856 }, "budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 19, - "serialized_chars": 35202, - "estimated_input_tokens": 8801, - "estimated_eval_tokens": 17858 + "serialized_chars": 35200, + "estimated_input_tokens": 8800, + "estimated_eval_tokens": 17856 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Most items are unrelated PRs across different models/features. The only true duplicate-looking pair is the two cross-entropy kwargs fixes (PR 43251 and PR 43254), which patch the same function, same file, and the same issue target.", - "confidence": 0.97, + "summary": "This cluster is mostly a collection of unrelated pull requests. The only likely duplicate/change overlap is the pair of PRs fixing `fixed_cross_entropy` (`43251` and `43254`), which address the same underlying loss-function kwargs gap; the rest appear to be separate model/docs/test work.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43251 is the stronger canonical PR for the duplicate pair: it directly targets issue 43240, has the fuller implementation and test coverage, and appears to subsume PR 43254\u2019s smaller variant.", + "canonical_pr_reason": "PR 43251 is the best representative for the cross-entropy kwargs fix: it directly targets issue 43240 and provides the more explicit implementation. PR 43254 is a narrower alternate patch for the same bug.", "best_issue_reason": null, - "best_pr_reason": "PR 43251 is the best representative of the only genuine duplicate cluster because it addresses the concrete loss-utils bug with the more complete patch; the rest are not duplicates of each other.", + "best_pr_reason": "PR 43251 is the strongest global representative because it has an explicit issue link and the clearest statement of the intended behavior change in `fixed_cross_entropy`.", "soft_edge_verdicts": [ { "left": "pull_request:39895", "right": "pull_request:43424", "accept": false, - "reason": "Different features: new VideoPrism model vs an Executorch dynamic-shape test. No shared code-path bug." + "reason": "Unrelated PRs: one adds VideoPrism, the other adds an Executorch dynamic-shape test." }, { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Unrelated tokenizer helper methods vs adding a new MobileLLM model. Same broad area only at the library level, not the same change." + "reason": "Different areas: Mistral tokenizer method additions vs adding a new MobileLLM model." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Tokenizer method additions and Executorch export tests are unrelated." + "reason": "Tokenizer API patch vs Executorch test; no shared code-path bug." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor auto-loading robustness changes are not the same as introducing a new model implementation." + "reason": "Processor robustness changes are unrelated to adding MobileLLM." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Cross-entropy kwargs fix and MobileLLM addition are entirely different code paths." + "reason": "Loss-function kwargs fix vs new MobileLLM model addition." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "Loss utility fix vs Mistral4 test/fix work; different bugs and different files." + "reason": "Cross-entropy helper fix and Mistral4 test/model work are different changes." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "Executorch dynamic-shape test and Mistral4 test fix are not the same underlying change." + "reason": "Executorch dynamic-shape testing is unrelated to Mistral4 fixes." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "Different domains: exportability testing vs adding MobileLLM." + "reason": "Different artifacts and code paths: export test vs new model implementation." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "New model addition vs class annotation resolution order fix; no overlap in bug or implementation." + "reason": "Adding MobileLLM and fixing config resolution order are unrelated." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "New model addition and Falcon output-interface refactor are unrelated." + "reason": "New model addition vs Falcon output-interface refactor; not the same bug/change." }, { "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "Falcon refactor vs model-creation skill docs; documentation-only PR is not a duplicate of code refactoring." + "reason": "Falcon refactor and model-creation skill docs are unrelated." }, { "left": "pull_request:43995", "right": "pull_request:44827", "accept": false, - "reason": "Falcon interface refactor and Mistral4 test fixes are different concrete changes." + "reason": "Different model families and different changes; no same underlying issue." }, { "left": "pull_request:44827", "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 tests/fixes are unrelated to the layrnorm->layernorm renaming sweep." + "reason": "Mistral4 test fixes and CLIP-like `layrnorm` rename mapping are unrelated." }, { "left": "pull_request:44827", "right": "pull_request:45795", "accept": false, - "reason": "Mistral4 test fixes and audio/video processor docs are unrelated." + "reason": "Mistral4 test/model changes do not match docs for audio/video processors." }, { "left": "pull_request:44827", "right": "pull_request:45655", "accept": false, - "reason": "Different problems: Mistral4 tests vs config resolution order in modeling_utils." + "reason": "Mistral4 fixes vs config-resolution order fix are separate issues." }, { "left": "pull_request:44827", "right": "pull_request:45213", "accept": false, - "reason": "Model tests/fixes are not the same as the model-creation skill draft documentation." + "reason": "Model fixes vs a skill/documentation PR are not the same code-path change." }, { "left": "pull_request:43823", "right": "pull_request:45213", "accept": false, - "reason": "New model implementation vs documentation skill content; not the same code-path." + "reason": "New model implementation and model-creation skill docs are not duplicates." }, { "left": "pull_request:43823", "right": "pull_request:45795", "accept": false, - "reason": "New model implementation vs docs for audio/video processors; unrelated." + "reason": "MobileLLM model addition and docs for audio/video processors are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 test/model changes vs model-creation skill docs are unrelated." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "Both PRs patch the same function in the same file for the same issue target (43240), with nearly identical cross-entropy kwargs plumbing." + "reason": "Both PRs fix the same underlying `fixed_cross_entropy` kwargs support gap in `src/transformers/loss/loss_utils.py`, and they could plausibly be merged into one patch." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded overall. The only accepted soft-edge pair is well-supported: PR 43251 and PR 43254 modify the same function in the same file and share the same issue target, so they look mergeable into one bugfix PR. The other false verdicts are conservative and consistent with the packet." + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted pair, PR 43251 and PR 43254, clearly targets the same `fixed_cross_entropy` kwargs issue in the same file and could plausibly be merged into one bugfix PR. The other pairwise rejections are supported by the diffs and do not overstate relatedness." }, "evaluator_used": true, "retried": false, @@ -16212,12 +16070,12 @@ "soft_pairs": [ "pull_request:40546|pull_request:42781" ], - "prepared_review_unit_hash": "3d77cf12bf8535465e65b14a6774d62543ed1923f5ccafc2c66f0c4d64b7addf", + "prepared_review_unit_hash": "e9931d629496dfe3663839c7b619327c155122eac0533f0fa8eeec1dd59cbdfe", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7225, + "serialized_chars": 7226, "estimated_input_tokens": 1807, "estimated_eval_tokens": 3870 }, @@ -16225,7 +16083,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7225, + "serialized_chars": 7226, "estimated_input_tokens": 1807, "estimated_eval_tokens": 3870 }, @@ -16234,18 +16092,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two related VibeVoice PRs, but they are not the same change set: one implements the base model, the other adds a realtime variant plus acoustic tokenizer support. The overlap is mostly shared model scaffolding, not a single fix/feature.", - "confidence": 0.91, + "summary": "Two related VibeVoice PRs, but they target different scopes: the first implements VibeVoice, while the second adds VibeVoice Realtime plus an acoustic tokenizer. They share substantial model scaffolding, but not the same concrete change set, so they should not be merged as duplicates.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #40546 is the broader base implementation and appears to be the primary foundational PR in this cluster; #42781 is a later, draft realtime extension built on top of the same model family.", + "canonical_pr_reason": "PR #42781 is the broader and newer VibeVoice effort, adding the realtime path and acoustic tokenizer on top of the base model work, so it is the better representative of this cluster.", "best_issue_reason": null, - "best_pr_reason": "PR #40546 is the better representative of the cluster because it is the core VibeVoice implementation, while #42781 is a more specialized follow-up that depends on the same codebase but targets a different capability.", + "best_pr_reason": "PR #42781 is the strongest cluster representative because it covers the expanded VibeVoice feature set and supersets much of the base integration work.", "soft_edge_verdicts": [ { - "left": "pull_request:40546", - "right": "pull_request:42781", + "left": "Implement VibeVoice", + "right": "Add VibeVoice Realtime", "accept": false, - "reason": "Shared filenames are mostly VibeVoice plumbing and docs, but #40546 implements VibeVoice generally while #42781 adds VibeVoice Realtime and an acoustic tokenizer. They do not appear to fix the same concrete code-path problem and would not plausibly be merged as one PR." + "reason": "Shared filenames show the same VibeVoice scaffolding, but the PRs implement different scopes: base VibeVoice integration versus a realtime variant with an acoustic tokenizer. That is related work, not the same underlying change." } ] }, @@ -16275,7 +16133,7 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "634c2e716e6379f1b84285851c0d88e5f93500bf0d550fb4035022a0c6c521b2", + "prepared_review_unit_hash": "15c624f9e2f9edc0a7e510d30a0ec401bb75d413813b7a28239299498a83bce4", "component_budget": { "node_count": 2, "item_count": 2, @@ -16297,8 +16155,8 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two 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 they should not be clustered as duplicates.", - "confidence": 0.99, + "summary": "The two pull requests are unrelated: one adds TimesFM covariate forecasting support, while the other adds DINOv3 image classification support. They do not appear to be duplicate changes or the same underlying bug fix.", + "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, @@ -16308,7 +16166,7 @@ "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different models and different feature areas; no shared underlying code-path problem or mergeable single PR." + "reason": "Reject: these target different model families, different user-facing features, and different code paths; they are not plausibly mergeable into one PR as the same change." } ] }, @@ -16403,7 +16261,7 @@ "pull_request:41797|pull_request:45186", "pull_request:45186|pull_request:45630" ], - "prepared_review_unit_hash": "6e10151ca53d8f203244d5baee8d313443a77102744aa26aaa7309f2c5ae16cc", + "prepared_review_unit_hash": "352464f829c5c7682c8fa58b9f4172bea57022050fe283ab00c256a47dbd8a6e", "component_budget": { "node_count": 3, "item_count": 3, @@ -16425,24 +16283,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate new-model addition PRs, not duplicates. They share Hugging Face model-registration scaffolding, but each targets a different model family and code path: DeepSeek OCR, Isaac, and Kimi2-6.", + "summary": "These are three unrelated model-addition PRs with similar scaffolding, not duplicates of the same change. The shared files are mostly registry/docs boilerplate; the actual model code, processors, and tests target different models.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: the cluster contains distinct model additions rather than multiple patches for the same underlying change.", + "canonical_pr_reason": "No canonical PR: the cluster is not a duplicate set. Each PR adds a different model (DeepSeek OCR, Isaac, Kimi2-6) with distinct code paths and tests, so none should represent the others.", "best_issue_reason": null, - "best_pr_reason": "41797 is the strongest standalone representative: it has broad, self-contained DeepSeek OCR implementation coverage and the most review activity, but it is still not a duplicate of the others.", + "best_pr_reason": "No best PR for deduplication. If forced to pick a representative, it would still be arbitrary because the PRs are semantically unrelated beyond shared template files.", "soft_edge_verdicts": [ { "left": "pull_request:41797", "right": "pull_request:45186", "accept": false, - "reason": "Different model additions (DeepSeek OCR vs Isaac). The overlap is only in shared auto-registration/docs plumbing, not the same code-path fix or change." + "reason": "Different models and implementations: DeepSeek OCR vs Isaac. Shared auto-registry/docs files are boilerplate and do not indicate the same underlying change." }, { "left": "pull_request:45186", "right": "pull_request:45630", "accept": false, - "reason": "Different model additions (Isaac vs Kimi2-6) with different processors/tokenizers and model code; they are not plausibly mergeable into one PR as the same underlying change." + "reason": "Different model additions: Isaac vs Kimi2-6. They share only generic framework wiring, not the same concrete bug or feature." } ] }, @@ -16472,12 +16330,12 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "93c54ffdb86339e7cdbe34b4386d5b866f01668d943f5930b33838549bcb807f", + "prepared_review_unit_hash": "7451ae42f4d3255edfe904ba4c5c295f0af070ac527a0ba41c359339ef94e59d", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5935, + "serialized_chars": 5934, "estimated_input_tokens": 1484, "estimated_eval_tokens": 3224 }, @@ -16485,7 +16343,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5935, + "serialized_chars": 5934, "estimated_input_tokens": 1484, "estimated_eval_tokens": 3224 }, @@ -16494,18 +16352,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related GitHub Actions dependency bump PRs, but they address different actions and different failure modes. PR 42975 is a broad Node 24 compatibility upgrade across many workflows; PR 42976 is a narrower docker/build-push-action upgrade for docker image workflows. Not duplicates.", - "confidence": 0.93, + "summary": "Two open PRs both update GitHub Actions workflows, but they target different action families and different upgrade reasons, so they look related rather than duplicates.", + "confidence": 0.78, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 42975 is the broader umbrella change, touching many workflows and multiple core actions for Node 24 compatibility, so it is the better representative PR for the cluster.", + "canonical_pr_reason": "PR 42975 is the better canonical representative: it has the clearer, more specific goal (Node 24 compatibility) and updates core workflow actions across many files.", "best_issue_reason": null, - "best_pr_reason": "PR 42975 is the most central and comprehensive of the two; it spans more workflows and a wider compatibility upgrade than the docker-only action bump in 42976.", + "best_pr_reason": "PR 42975 is the best overall representative of the cluster because its scope is broader and more clearly framed than PR 42976, which is mainly a docker/build-push-action version bump.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "Both are workflow dependency upgrades, but they change different actions in different code paths: 42975 upgrades checkout/cache/upload-artifact and related workflow actions, while 42976 upgrades docker/build-push-action in docker build workflows. They are related maintenance, not the same underlying change and would not be merged as one PR." + "reason": "Both are workflow dependency upgrades, but they change different concrete actions and do not appear to fix the same underlying bug or be safely mergeable into one PR." } ] }, @@ -16542,12 +16400,12 @@ "pull_request:43448|pull_request:45490", "pull_request:43451|pull_request:45490" ], - "prepared_review_unit_hash": "933a099f96ae07090ee4a6f4222902e9069727651c29f977fa4a03faf88a2931", + "prepared_review_unit_hash": "ee02c42ce084d67ffc982db7f8b8ef6d7313d26326dc6c750bfaa94308aaa48a", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 14079, + "serialized_chars": 14078, "estimated_input_tokens": 3520, "estimated_eval_tokens": 7296 }, @@ -16555,7 +16413,7 @@ "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 14079, + "serialized_chars": 14078, "estimated_input_tokens": 3520, "estimated_eval_tokens": 7296 }, @@ -16564,48 +16422,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are separate model-addition efforts, not duplicates: ViT NEPA, Molmo, Molmo2, and CTSM each introduce a different architecture and supporting files/tests. The shared docs/auto-registration changes are boilerplate for adding new models, not evidence of the same underlying change.", + "summary": "These PRs are only superficially similar: they are separate model-addition changes for ViT NEPA, Molmo, Molmo2, and CTSM. They share standard Transformers scaffolding (docs, auto-mapping, tests), but each targets a different model and code path, so this is not a duplicate cluster.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "No duplicate canonical PR: the four PRs target different models and code paths, so there is no single underlying change to canonicalize.", + "canonical_pr_reason": "If a representative is needed, #43451 is the strongest anchor because it is the most extensive and heavily reviewed PR, but it is not a duplicate of the others.", "best_issue_reason": null, - "best_pr_reason": "PR 43451 (Molmo2) is the strongest standalone representative of the cluster because it is the most comprehensive and review-heavy model-addition PR, with broader support files (image/video processing, auto mappings, docs, tests).", + "best_pr_reason": "#43451 is the best cluster representative by scope and review activity, but only as an exemplar of a model-addition PR; it should not be treated as the same change as the others.", "soft_edge_verdicts": [ { "left": "pull_request:42978", "right": "pull_request:43448", "accept": false, - "reason": "Different models (ViT NEPA vs Molmo). Shared auto/docs plumbing is expected for model onboarding, not a shared bug or change." + "reason": "Different model additions (ViT NEPA vs Molmo). Shared auto-mapping/docs/test scaffolding is not enough to treat them as the same fix or merge into one PR." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "Different models (ViT NEPA vs Molmo2) and different implementation stacks; not plausibly mergeable as one fix." + "reason": "ViT NEPA and Molmo2 are distinct models with different configs, processors, and implementation files; they do not address the same underlying change." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "Different models (ViT NEPA vs CTSM). Overlap is limited to generic registration/docs updates." + "reason": "Different model families and code paths (ViT NEPA vs CTSM); only generic repository integration files overlap." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are distinct model families with separate configs, processors, and tests; they are not the same underlying change." + "reason": "Molmo and Molmo2 are related by naming, but these are separate model introductions with different model/processing stacks and cannot be merged as one duplicate PR." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Molmo vs CTSM are unrelated model additions. Similar touched files reflect shared framework integration, not duplication." + "reason": "Molmo and CTSM are unrelated model additions; overlapping docs/auto-registry changes are incidental." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 and CTSM are different model implementations with different processing paths and test suites; not a duplicate pair." + "reason": "Molmo2 and CTSM are distinct features touching different model implementations; they are not the same bug or change." } ] }, @@ -16638,52 +16496,52 @@ "pull_request:43291|pull_request:44815", "pull_request:43665|pull_request:44815" ], - "prepared_review_unit_hash": "999f067dd138cf2f3f7b0417c8f551c9d4432a4b252b6d04b8dd4632e1a9062d", + "prepared_review_unit_hash": "63cf15d43b6a40f0f894f6c482c8f3ebb559a5b542a8e18246eeb833479af5af", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 7091, - "estimated_input_tokens": 1773, - "estimated_eval_tokens": 3802 + "serialized_chars": 7093, + "estimated_input_tokens": 1774, + "estimated_eval_tokens": 3804 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 7091, - "estimated_input_tokens": 1773, - "estimated_eval_tokens": 3802 + "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": "No pair appears to be the same underlying bug or change. PR 44815 is the strongest standalone fix; the other two are unrelated test/model-specific changes, with 43665 looking especially test-only and not mergeable with the others.", - "confidence": 0.93, + "summary": "These three pull requests do not look like duplicates of one another: one is a Whisper tokenizer/test fix, one is a clearly broken test-only change, and one is a separate FP8 dequantization fix. The only similarity is superficial model/test overlap.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44815 is the most concrete, code-level fix with a clear cross-file dequantization path and supporting loader/kernel changes; the other PRs are narrower and less clearly substantive.", + "canonical_pr_reason": "pull_request:44815 is the best representative of the set because it is the most coherent, substantive bug fix with a clear code-path change and accompanying tests; the other two are either unrelated or test-only noise.", "best_issue_reason": null, - "best_pr_reason": "PR 44815 is the best representative of the cluster because it addresses a specific runtime/code-path problem in model loading and FP8 dequantization, rather than just tests or an ambiguous Whisper test tweak.", + "best_pr_reason": "pull_request:44815 is the strongest standalone PR here: it targets a concrete dequantization/loading issue, spans the relevant implementation files, and looks like a real fix rather than a test tweak.", "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "Different scopes and models: Whisper tokenizer/decode behavior versus CLIP/ViT test edits. They do not look like the same bug or a single mergeable PR." + "reason": "Whisper tokenizer decode fix vs unrelated CLIP/ViT test edits; no shared bug or code path." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper decode/timestamp handling is unrelated to FP8 dequantization and model loading. No shared concrete code path." + "reason": "Whisper batch/timestamp decoding bug is unrelated to FP8 dequantization/loading logic." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "Test-only CLIP/ViT edits are unrelated to FP8 quantization/dequant loading fixes. They would not plausibly be merged as one PR." + "reason": "Test-only changes in vision models are not the same underlying problem as the FP8 dequantization fix." } ] }, @@ -16713,12 +16571,12 @@ "soft_pairs": [ "pull_request:43488|pull_request:43532" ], - "prepared_review_unit_hash": "edef2ca87041b8e1c6fd6ffa0b5252c67b2ece5b43d221b8c19d3586eea34b41", + "prepared_review_unit_hash": "cee35efe8962a9e421767ef7de6d7e7683237372175210fec9d8d04674e9db79", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4563, + "serialized_chars": 4562, "estimated_input_tokens": 1141, "estimated_eval_tokens": 2538 }, @@ -16726,7 +16584,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4563, + "serialized_chars": 4562, "estimated_input_tokens": 1141, "estimated_eval_tokens": 2538 }, @@ -16735,18 +16593,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs are not duplicates: they share one test file, but one is a setup/version bot-check change while the other rewires CI/workflow and testing utilities. The overlap looks incidental rather than the same underlying fix.", + "summary": "These two PRs are both intentionally non-mergeable test/demonstration changes, but they target different areas: one tweaks packaging/version metadata, while the other rewires CI/workflow and testing utilities. The shared Vit test file is incidental and not evidence of the same underlying fix.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical duplicate PR: the PRs target different areas and goals, with only a shared Vit test file in common.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR #43532 is the better representative of the cluster because it is the larger, more substantive infrastructure/testing change; PR #43488 is mostly a narrow setup/version/test-dependency tweak.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "Reject: they do not fix the same concrete problem. #43488 changes setup/versioning and adds pytest-random-order, while #43532 changes GitHub workflow behavior plus testing/notification utilities; the shared Vit test file is incidental." + "reason": "Not the same concrete change: 43488 is about setup/version/dependency metadata, while 43532 changes GitHub workflows plus testing utilities. The overlap in a Vit test file is incidental, and these would not plausibly merge into one PR." } ] }, @@ -16776,7 +16634,7 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "0e828c5ff42329c59bafc88105762da5ea7b4524ecec4801e3ada73eb534de8b", + "prepared_review_unit_hash": "a464f95e8fd489e4706cb6d5b349e58d7fadb8470381c9de593504b69d86d3dd", "component_budget": { "node_count": 2, "item_count": 2, @@ -16798,18 +16656,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two distinct feature PRs for different pipelines: one adds Promptable Concept Segmentation, the other adds Promptable Visual Segmentation. They share scaffolding files and pipeline docs, but they target different tasks, models, tests, and code paths, so they should not be merged as duplicates.", + "summary": "These are related but distinct PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They touch overlapping docs and registration files, but the model support, tests, and filenames show different concrete features, so they should not be deduplicated or merged as one change.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "No single canonical PR for a duplicate set: these are separate pipeline additions with different targets and code paths, not the same underlying change.", "best_issue_reason": null, - "best_pr_reason": null, + "best_pr_reason": "PR #43613 is the broader representative of the cluster because it adds more model integrations and tests, but it is still not a duplicate of #43612.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Different concrete features and implementations: concept segmentation vs visual segmentation. Shared docs/registry/test scaffolding is expected for new pipeline additions, but they do not fix the same bug or represent the same change set." + "reason": "Same broad theme and shared docs/registration updates, but they implement different pipelines with different task/model/test files and different concrete functionality." } ] }, @@ -16839,7 +16697,7 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "238107644ce2b6499f93b4c021f8be2e298a08c656dd30a5ea2d137b4bff177c", + "prepared_review_unit_hash": "38e9015a7dbccab547382d5732f4cb6cbd9bc2f8520457364fdccbfda8199583", "component_budget": { "node_count": 2, "item_count": 2, @@ -16861,18 +16719,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both items are separate Trainer enhancements in the same file: one changes gradient-accumulation loss scaling, the other refactors signature-column filtering. They do not appear to fix the same bug or implement the same change.", - "confidence": 0.96, + "summary": "These are both Trainer.py refactors, but they address different behaviors: one changes gradient-accumulation loss scaling, the other factors out signature-column filtering for custom trainers. They are not the same underlying fix, so the soft duplicate link should be rejected.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43651 is the stronger canonical representative because it introduces a concrete trainer behavior change with an explicit API hook for loss scaling, whereas 43663 is a separate internal refactor for column filtering.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR 43651 is the best overall fit because it addresses a clear runtime training behavior and is more directly user-facing than the signature-column refactor in 43663.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Same broad subsystem and same file, but they address different code paths: gradient-accumulation loss normalization vs. dataset column filtering. Not the same underlying bug/change." + "reason": "Different code paths and goals: 43651 changes loss normalization during training_step, while 43663 changes data-column filtering based on model signatures. Same file, but not the same bug or mergeable change set." } ] }, @@ -16904,7 +16762,7 @@ "pull_request:43656|pull_request:43842", "pull_request:43836|pull_request:43842" ], - "prepared_review_unit_hash": "e50b75e5665b49d420822d838f7ce00dfb7970921160934da8c96fe4b6ab3cb1", + "prepared_review_unit_hash": "73ed8cd93ea341612b011e0e4f2bdae8c54695002b624c56aadb19432c1ba459", "component_budget": { "node_count": 3, "item_count": 3, @@ -16926,30 +16784,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three PRs target import-time NameError problems in `src/transformers/cli/serve.py` caused by type annotations/TypeAdapter handling. The two later PRs are very close variants of the same fix, and the earliest PR is in the same code path but broader.", + "summary": "All three PRs target the same serve CLI TypeAdapter/annotation NameError family in src/transformers/cli/serve.py. PR 43842 is the most focused fix; the others are broader or earlier variants of the same underlying problem.", "confidence": 0.84, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43842 is the cleanest, most focused representation of the bug: it directly guards the `TypeAdapter` reference for the pydantic-absent path without the extra annotation cleanup seen in the other PRs.", + "canonical_pr_reason": "PR 43842 is the most directly targeted and minimal fix for the TypeAdapter NameError when pydantic is absent, with a clear runtime-safety change and less unrelated churn than the others.", "best_issue_reason": null, - "best_pr_reason": "PR 43842 best captures the concrete failure mode and the minimal fix in `serve.py`, making it the strongest representative of the cluster.", + "best_pr_reason": "PR 43842 best represents the cluster because it addresses the concrete failure mode explicitly and cleanly in the same code path, without the broader annotation refactor noise present in the other PRs.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", "accept": true, - "reason": "Both are fixes for the same import-time NameError in the CLI serve module caused by runtime-evaluated annotations; the patches are alternative ways to prevent the same failure path." + "reason": "Both are fixing the same serve CLI TypeAdapter NameError in the same file; 43656 is a broader annotations-related variant, but it plausibly overlaps with and could be merged with the targeted pydantic-safety fix in 43842." }, { "left": "pull_request:43836", "right": "pull_request:43842", "accept": true, - "reason": "These address the same `TypeAdapter`/annotation NameError in the same file and are effectively alternative implementations of the same fix." + "reason": "Both PRs aim at the same concrete TypeAdapter annotation/NameError issue in cli/serve.py. 43836 is the broader wrapper-style patch, while 43842 is a smaller, more explicit version of the same fix and could plausibly be combined." } ] }, "evaluator_result": { "accept": true, - "feedback": "The cluster summary is conservative and supported by the packet: all three PRs touch `src/transformers/cli/serve.py` and concern import-time `NameError` issues around annotations/`TypeAdapter`. The distinction between the two later PRs as close variants of the same fix is well grounded. The soft-edge accept for 43656 vs 43842 is a bit weaker than the other pair, but it is still plausible and not overstated given the shared file, titles, and bug pattern." + "feedback": "The summary is grounded in the packet: all three PRs are in src/transformers/cli/serve.py, all are tied to the same TypeAdapter/annotation NameError family, and 43842 is indeed the most narrowly targeted fix. The soft-edge acceptances are conservative enough because each pair plausibly merges into one bugfix PR without clear contradiction or unrelated scope." }, "evaluator_used": true, "retried": false, @@ -17093,8 +16951,8 @@ "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:44098|pull_request:44101", "pull_request:44018|pull_request:44071", "pull_request:44017|pull_request:44722", "pull_request:44017|pull_request:44161", @@ -17139,352 +16997,376 @@ "pull_request:44044|pull_request:44154", "pull_request:44024|pull_request:44076" ], - "prepared_review_unit_hash": "e12447c3a2ef77cf738e4c2ec42c43b59783ee3f62706f2d75bd75228fa3d995", + "prepared_review_unit_hash": "f71f2370d0c8b506074c8d5b2f21a825a99d3bd8706feb80868693c287d3b25a", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84503, - "estimated_input_tokens": 21126, - "estimated_eval_tokens": 42508 + "serialized_chars": 84497, + "estimated_input_tokens": 21125, + "estimated_eval_tokens": 42506 }, "budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 72, - "serialized_chars": 76939, - "estimated_input_tokens": 19235, - "estimated_eval_tokens": 38726 + "serialized_chars": 76932, + "estimated_input_tokens": 19233, + "estimated_eval_tokens": 38722 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly model-specific PRs that all migrate different architectures to the new standardized output-tracing/decorator pattern. They are not duplicates across models; only the GPT-J and GPT-Neo pairs look like true overlap within the same code path.", - "confidence": 0.83, + "summary": "This cluster is mostly a set of independent, model-specific output-tracing refactors. The only clear duplicate groups are the GPT-Neo pair (44018/44068) and the GPT-J/CodeGen family (44066/44085/44722).", + "confidence": 0.82, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #44007 is the strongest anchor: it is explicitly issue-linked, clearly implements the standardized output-tracing refactor, and spans multiple related model files in one coherent change set.", + "canonical_pr_reason": "PR 44722 is the strongest representative of the GPT-J/CodeGen tracing refactor: it covers the same concrete code path as the earlier GPT-J PRs, includes both related files, and is the most complete version tied to the umbrella issue #43979.", "best_issue_reason": null, - "best_pr_reason": "PR #44007 is the best overall representative for this cluster because it is issue-linked and the clearest, most complete example of the standardized output-tracing migration pattern.", + "best_pr_reason": "44722 is the best single PR to represent the duplicate GPT-J/CodeGen change set because it is the most comprehensive and directly subsumes the earlier GPT-J-focused attempts.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different model families; similar tracing refactor wording is not enough to treat them as the same change." + "reason": "Different models and code paths (CVT/FNet vs GPT-J); same umbrella topic is too broad to merge." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "Different architectures and code paths; both are output-tracing refactors, but not the same underlying bug/change." + "reason": "Different model families and different refactors; only share a broad standardized-output theme." }, { "left": "pull_request:44066", "right": "pull_request:44085", "accept": true, - "reason": "Both touch GPT-J output tracing in the same code path; 44066 also includes CodeGen, but the GPT-J refactor substantially overlaps and could plausibly be merged into one PR." + "reason": "Both edit GPT-J output tracing in the same file and target the same underlying change; one is just a broader version with CodeGen included." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "Different models and implementations; same broad refactor theme, but not the same concrete change." + "reason": "RegNet/ResNet-style vision backbones are separate concrete code paths from EfficientNet; not the same bug/change." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "EfficientNet vs GPT-J/CodeGen; same migration style, but unrelated code paths." + "reason": "EfficientNet output tracing and GPT-J/CodeGen output tracing are unrelated model-specific changes." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen versus EfficientNet; not the same underlying bug or change." + "reason": "GPT-J/CodeGen and EfficientNet are different implementations; no shared concrete fix." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 and DeBERTaV2 are separate model implementations with different output-tracing migrations." + "reason": "MobileNetV2 and DeBERTa are unrelated model stacks; same refactor pattern is not enough." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "Different model families; both refactor output collection, but they are unrelated changes." + "reason": "GPT-J/CodeGen output tracing vs MGP-STR output tracing are different model-specific changes." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "GPT-J/CodeGen and MPT are distinct code paths; this is only thematic similarity." + "reason": "GPT-J/CodeGen and MPT touch different code paths and should not be merged as duplicates." }, { "left": "pull_request:44018", "right": "pull_request:44068", "accept": true, - "reason": "Both are GPT-Neo output-tracing refactors on the same model code path, so they look like overlapping implementations of the same change." + "reason": "Same GPT-Neo file and same output-tracing refactor; these are duplicate fixes for the same code path." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPT-J/CodeGen versus GPT-Neo; similar standardized-decorator work, but not the same change." + "reason": "GPT-J/CodeGen versus GPT-Neo are different model implementations." }, { "left": "pull_request:44018", "right": "pull_request:44066", "accept": false, - "reason": "GPT-Neo and GPT-J are different model families; shared refactor style does not make them duplicates." + "reason": "Different models (GPT-Neo vs GPT-J/CodeGen); only the refactor style is similar." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet versus GPT-J/CodeGen; unrelated concrete code paths." + "reason": "ResNet and GPT-J/CodeGen are unrelated concrete changes." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "MGP-STR and GPT-J/CodeGen are different model implementations, so this is not the same change." + "reason": "MGP-STR and GPT-J/CodeGen are different code paths; not mergeable as one PR." }, { "left": "pull_request:44019", "right": "pull_request:44085", "accept": false, - "reason": "ResNet and GPT-J refactors are separate migrations, not one underlying fix." + "reason": "ResNet output tracing and GPT-J output tracing are separate model fixes." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "GPT-Neo versus ResNet; same pattern, different model-specific change sets." + "reason": "GPT-Neo and ResNet are unrelated model implementations." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet and MPT are separate output-tracing migrations." + "reason": "ResNet and MPT are distinct model-specific refactors." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo and Flaubert/XLM are unrelated model code paths." + "reason": "GPT-Neo and Flaubert/XLM are unrelated code paths." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo versus GPT-J/CodeGen; same refactor theme, but different concrete model migrations." + "reason": "GPT-Neo is not the same underlying change as GPT-J/CodeGen, despite similar output-tracing mechanics." }, { "left": "pull_request:44071", "right": "pull_request:44722", "accept": false, - "reason": "MPT and GPT-J/CodeGen are unrelated implementations." + "reason": "MPT and GPT-J/CodeGen touch different model internals." }, { - "left": "pull_request:44059", + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert/XLM are different model families and fixes." + }, + { + "left": "pull_request:44018", "right": "pull_request:44086", "accept": false, - "reason": "GPT-2 and MGP-STR are different model code paths; only the refactor style matches." + "reason": "GPT-Neo and MGP-STR are different concrete changes." }, { - "left": "pull_request:44085", + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo and GPT-J are separate code paths." + }, + { + "left": "pull_request:44068", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J and MGP-STR are separate model migrations, not the same bug/change." + "reason": "GPT-Neo and MGP-STR are unrelated model-specific refactors." }, { - "left": "pull_request:44098", - "right": "pull_request:44101", + "left": "pull_request:44071", + "right": "pull_request:44086", + "accept": false, + "reason": "MPT and MGP-STR are different models; not the same change." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT2 output tracing and MGP-STR output tracing are separate fixes." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44086", "accept": false, - "reason": "ViLT and Flaubert/XLM are different model families with different output-handling changes." + "reason": "GPT-J and MGP-STR are different model code paths." }, { "left": "pull_request:44059", "right": "pull_request:44722", "accept": false, - "reason": "GPT-2 versus GPT-J/CodeGen; same standardization theme, but not the same concrete change." + "reason": "GPT2 refactor is unrelated to GPT-J/CodeGen." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT and Flaubert/XLM are different model implementations." }, { "left": "pull_request:44017", "right": "pull_request:44722", "accept": false, - "reason": "SegFormer and GPT-J/CodeGen are unrelated model-specific migrations." + "reason": "Segformer and GPT-J/CodeGen are unrelated model-specific refactors." }, { "left": "pull_request:44017", "right": "pull_request:44161", "accept": false, - "reason": "SegFormer and LongT5 touch different code paths; same umbrella refactor only." + "reason": "Segformer and LongT5 are different code paths." }, { "left": "pull_request:44017", "right": "pull_request:44073", "accept": false, - "reason": "SegFormer and VisualBert are separate model refactors." + "reason": "Segformer and VisualBert are separate model fixes." }, { "left": "pull_request:44068", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo and MPT are different architectures; not the same change." - }, - { - "left": "pull_request:44071", - "right": "pull_request:44076", - "accept": false, - "reason": "MPT and ImageGPT are unrelated model-specific migrations." + "reason": "GPT-Neo and MPT are different model families." }, { - "left": "pull_request:44074", - "right": "pull_request:44101", + "left": "pull_request:43997", + "right": "pull_request:44073", "accept": false, - "reason": "TextNet and Flaubert/XLM are different code paths and model families." + "reason": "RegNet and VisualBert are unrelated concrete changes." }, { "left": "pull_request:43997", "right": "pull_request:44017", "accept": false, - "reason": "RegNet and SegFormer are different vision models; only the refactor style overlaps." + "reason": "RegNet and Segformer are separate model-specific refactors." }, { "left": "pull_request:43997", "right": "pull_request:44056", "accept": false, - "reason": "RegNet and MPNet are separate model migrations." + "reason": "RegNet and MPNet are different model code paths." }, { - "left": "pull_request:44071", - "right": "pull_request:44129", + "left": "pull_request:43997", + "right": "pull_request:44018", "accept": false, - "reason": "MPT and SpeechT5 are unrelated concrete changes." + "reason": "RegNet and GPT-Neo are unrelated fixes." }, { - "left": "pull_request:43996", + "left": "pull_request:44073", "right": "pull_request:44161", "accept": false, - "reason": "CVT/FNet and LongT5 are different model code paths." + "reason": "VisualBert and LongT5 are different model implementations." }, { - "left": "pull_request:43996", - "right": "pull_request:44017", + "left": "pull_request:44074", + "right": "pull_request:44101", "accept": false, - "reason": "CVT/FNet and SegFormer are separate migrations." + "reason": "TextNet and Flaubert/XLM are unrelated model-specific refactors." }, { - "left": "pull_request:44010", - "right": "pull_request:44071", + "left": "pull_request:44056", + "right": "pull_request:44059", "accept": false, - "reason": "SqueezeBERT and MPT are unrelated model implementations." + "reason": "MPNet and GPT2 are different code paths." }, { - "left": "pull_request:44010", - "right": "pull_request:44073", + "left": "pull_request:44071", + "right": "pull_request:44129", "accept": false, - "reason": "SqueezeBERT and VisualBert are different model-specific refactors." + "reason": "MPT and SpeechT5 are different concrete changes." }, { - "left": "pull_request:44018", - "right": "pull_request:44129", + "left": "pull_request:44056", + "right": "pull_request:44161", "accept": false, - "reason": "GPT-Neo and SpeechT5 are unrelated concrete changes." + "reason": "MPNet and LongT5 are unrelated model fixes." }, { - "left": "pull_request:43997", - "right": "pull_request:44018", + "left": "pull_request:43996", + "right": "pull_request:44161", "accept": false, - "reason": "RegNet and GPT-Neo are different models; same refactor theme only." + "reason": "FNet/CVT and LongT5 are different model families." }, { - "left": "pull_request:43997", - "right": "pull_request:44161", + "left": "pull_request:43996", + "right": "pull_request:44017", "accept": false, - "reason": "RegNet and LongT5 are unrelated migrations." + "reason": "FNet/CVT and Segformer are separate concrete code paths." }, { - "left": "pull_request:44074", - "right": "pull_request:44161", + "left": "pull_request:44010", + "right": "pull_request:44044", "accept": false, - "reason": "TextNet and LongT5 are different model code paths." + "reason": "SqueezeBert and DeBERTa are unrelated model-specific changes." }, { - "left": "pull_request:43997", + "left": "pull_request:44044", "right": "pull_request:44074", "accept": false, - "reason": "RegNet and TextNet are separate model-specific output-tracing changes." + "reason": "DeBERTa and TextNet are different model implementations." }, { "left": "pull_request:44010", - "right": "pull_request:44044", + "right": "pull_request:44073", "accept": false, - "reason": "SqueezeBERT and DeBERTaV2 are unrelated model refactors." + "reason": "SqueezeBert and VisualBert are unrelated refactors." }, { - "left": "pull_request:44044", - "right": "pull_request:44074", + "left": "pull_request:44018", + "right": "pull_request:44059", "accept": false, - "reason": "DeBERTaV2 and TextNet are different model families." + "reason": "GPT-Neo and GPT2 are different model code paths." }, { - "left": "pull_request:44066", - "right": "pull_request:44722", - "accept": true, - "reason": "Both are GPT-J output-tracing refactors over the same code path, and the overlap is close enough that they could plausibly be merged into one PR." + "left": "pull_request:44074", + "right": "pull_request:44076", + "accept": false, + "reason": "TextNet and ImageGPT are separate model-specific changes." }, { - "left": "pull_request:44085", - "right": "pull_request:44722", - "accept": true, - "reason": "Both target GPT-J output tracing in the same file, with highly overlapping standardized-decorator changes." + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet and TextNet are unrelated concrete fixes." }, { - "left": "pull_request:44085", - "right": "pull_request:44129", + "left": "pull_request:44056", + "right": "pull_request:44101", "accept": false, - "reason": "GPT-J and SpeechT5 are unrelated model migrations." + "reason": "MPNet and Flaubert/XLM are different models." }, { - "left": "pull_request:43996", - "right": "pull_request:44007", + "left": "pull_request:44074", + "right": "pull_request:44071", "accept": false, - "reason": "CVT/FNet and ResNet/RegNet/RT-DETR are different model families and code paths." + "reason": "TextNet and MPT are unrelated model-specific refactors." }, { - "left": "pull_request:43996", - "right": "pull_request:44129", + "left": "pull_request:44010", + "right": "pull_request:44018", "accept": false, - "reason": "CVT/FNet and SpeechT5 are unrelated changes." + "reason": "SqueezeBert and GPT-Neo are different code paths." }, { - "left": "pull_request:44044", - "right": "pull_request:44129", + "left": "pull_request:44019", + "right": "pull_request:44071", "accept": false, - "reason": "DeBERTaV2 and SpeechT5 are separate model-specific refactors." + "reason": "ResNet and MPT are unrelated model fixes." }, { - "left": "pull_request:44044", - "right": "pull_request:44154", + "left": "pull_request:44010", + "right": "pull_request:44044", "accept": false, - "reason": "DeBERTaV2 and ViT-S are different code paths; same refactor pattern is not enough." + "reason": "SqueezeBert and DeBERTa are different model families." }, { - "left": "pull_request:44024", - "right": "pull_request:44076", + "left": "pull_request:44044", + "right": "pull_request:44074", "accept": false, - "reason": "FocalNet and ImageGPT are different models; these are parallel migrations, not duplicates." + "reason": "DeBERTa and TextNet are separate concrete changes." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet, and the soft-edge calls stay conservative overall. The accepted overlaps are limited to the same model code paths (GPT-J and GPT-Neo), while the rejects correctly treat other model-family matches as only thematic similarity. The anchor choice of PR #44007 is also supported by its explicit issue link and coherent multi-file change set." + "feedback": "The summary is grounded in the packet: it sticks to the model-specific output-tracing refactors and only treats the GPT-Neo and GPT-J/CodeGen sequences as duplicate families. The accepted soft edges are conservative because they involve the same model/file and look like broader-vs-subset versions of the same bugfix. The rejected pairs are also appropriately limited to different model families or unrelated code paths." }, "evaluator_used": true, "retried": false, @@ -17519,64 +17401,64 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "21ad35761be65b328f074753a2646c2320dcd45542da946315ab03e0920af650", + "prepared_review_unit_hash": "3671a9ca317cef380c6c3a4f0136839b04a078538852c315c32c18929129d0be", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84503, - "estimated_input_tokens": 21126, - "estimated_eval_tokens": 42508 + "serialized_chars": 84497, + "estimated_input_tokens": 21125, + "estimated_eval_tokens": 42506 }, "budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 5, - "serialized_chars": 13064, - "estimated_input_tokens": 3266, - "estimated_eval_tokens": 6788 + "serialized_chars": 13065, + "estimated_input_tokens": 3267, + "estimated_eval_tokens": 6790 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These PRs are all model-specific refactors toward standardized output capturing, but they touch different code paths and models. They look related by pattern, not duplicates of the same change, so none of the soft edges should be merged.", - "confidence": 0.93, + "summary": "These PRs all follow the same broad theme\u2014migrating model forwards to standardized output tracing\u2014but they are model-specific refactors in different code paths, not duplicate fixes for one underlying bug. No soft edge should be merged as the same change.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44073 is the cleanest representative of the shared refactor pattern: it introduces the capture_outputs/can_return_tuple plumbing in a single model without extra scope beyond the output-tracing migration.", + "canonical_pr_reason": "44129 is the best representative PR because it has an explicit issue target and a concrete, self-contained SpeechT5 refactor with clear scope.", "best_issue_reason": null, - "best_pr_reason": "PR 44073 is the best single representative of this cluster\u2019s theme, but only as an exemplar; it is not a duplicate of the others because the underlying model code paths differ.", + "best_pr_reason": "44129 is the strongest standalone PR: explicit issue linkage, clear problem statement, and the most complete/grounded implementation among the candidates.", "soft_edge_verdicts": [ { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "Both are output-tracing refactors, but they target different models (VisualBert vs TextNet) and different files/tests; same pattern, not the same concrete change." + "reason": "Same refactor pattern, but different models and different implementation details/tests; not the same underlying bug or mergeable into one PR." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "Different model implementations (MPNet vs VisualBert) with separate tracing plumbing; related refactor style, but not one mergeable fix." + "reason": "Both are output-tracing migrations, but MPNet and VisualBert are separate model code paths with distinct changes." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "MPNet and TextNet are separate model ports with distinct output-capture integration details, so these are not duplicates." + "reason": "Different model families and separate refactors; similarity is only at the decorator/migration pattern level." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are unrelated model-specific refactors; they share a migration theme but not the same bug or code path." + "reason": "RegNet vs SqueezeBert are unrelated model implementations; these are parallel cleanup refactors, not one shared fix." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "SqueezeBert and SpeechT5 each modify different model internals; similar output-tracing cleanup, but not a single underlying change." + "reason": "Different models and code paths; they share a migration style but not the same concrete change." } ] }, @@ -17656,12 +17538,12 @@ "pull_request:44028|pull_request:44030", "pull_request:44029|pull_request:44030" ], - "prepared_review_unit_hash": "a28c0fa4c7710e1353130f74d48df0674e468c8ef58740613074ea3359b4beb0", + "prepared_review_unit_hash": "360f58a728175ce26e42ad2615ca3efc6f1db7e33516338a9518351f5d115e45", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40171, + "serialized_chars": 40169, "estimated_input_tokens": 10043, "estimated_eval_tokens": 20342 }, @@ -17669,7 +17551,7 @@ "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40171, + "serialized_chars": 40169, "estimated_input_tokens": 10043, "estimated_eval_tokens": 20342 }, @@ -17678,252 +17560,258 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is a set of parallel, model-specific output-tracing refactors. They share a theme and some utility imports, but they target different architectures and code paths, so they are not duplicates of the same underlying change. The broadest representative is PR 44003.", - "confidence": 0.81, + "summary": "All items are distinct model-specific refactors around output tracing/can_return_tuple. They share a common pattern, but they do not target the same concrete bug or change and should not be deduplicated.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44003 is the broadest exemplar here: it updates two related model files and introduces the core output-capturing refactor machinery, making it the best representative anchor for the cluster.", + "canonical_pr_reason": "PR 44003 is the strongest representative: it covers two closely related Mamba-family implementations and includes the core tracing/capture plumbing, making it the most substantial exemplar of this refactor cluster.", "best_issue_reason": null, - "best_pr_reason": "PR 44003 is the strongest single representative of the pattern because it is the most substantial refactor in the set and spans multiple files in one related model family.", + "best_pr_reason": "PR 44003 is the best cluster representative because it is broader than the single-model PRs and shows the main output-tracing pattern without being just a narrow variant.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Different model families (`timm_backbone` vs `upernet`) and distinct forward paths; same refactor theme, not the same change." + "reason": "Same broad refactor theme, but timm_backbone and upernet are different code paths with different output logic." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "`timm_backbone` and `vision_text_dual_encoder` are unrelated code paths; similar output-handling cleanup is not enough to merge them." + "reason": "Different model families and different forward signatures; only the refactor pattern is shared." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "These are separate model implementations (`timm_backbone` vs `univnet`) with no shared bug or concrete code-path overlap." + "reason": "Both are output-tracing refactors, but timm_backbone and univnet are unrelated implementations." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "Different subsystems (`timm_backbone` vs `mamba/falcon_mamba`); both touch tracing utilities, but they fix separate model-specific forwards." + "reason": "Mismatched model families; 43998 is a single backbone wrapper, while 44003 changes Mamba/FalconMamba internals." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "`timm_backbone` and `rwkv` are unrelated model code paths; this is a shared refactor pattern, not one duplicate fix." + "reason": "timm_backbone and rwkv use different tracing mechanisms and code paths; not the same change." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "`mobilenet_v1` and `univnet` are distinct architectures; each PR changes its own output contract separately." + "reason": "Both are refactors, but MobileNetV1 and UnivNet are separate models with unrelated forward contracts." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "Different models and different forward implementations; no evidence of one underlying bug." + "reason": "MobileNetV1 and UperNet are different model targets; shared refactor wording is not enough." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "`mobilenet_v1` and `depth_anything` are separate model families; similar tracing refactor only." + "reason": "Different model families and different files; one is MobileNetV1, the other depth-anything variants." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "`mobilenet_v1` and `vision_encoder_decoder` do not share the same code path or concrete fix." + "reason": "MobileNetV1 output tracing is unrelated to VisionEncoderDecoder wrapper changes." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "`mobilenet_v1` and `speech_encoder_decoder` are unrelated wrappers/models; same theme, different changes." + "reason": "MobileNetV1 and SpeechEncoderDecoder are separate encoder-decoder/model wrapper code paths." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "`vision_text_dual_encoder` and `univnet` are unrelated; output-tracing cleanup alone does not make them duplicates." + "reason": "Vision-Text Dual Encoder and UnivNet are unrelated models; only the refactor motif overlaps." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Different model classes and different forward semantics; not the same concrete bug or change." + "reason": "Different architectures and forward semantics; not a single bug fix." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "These touch different subsystems (`vision_text_dual_encoder` vs `mamba/falcon_mamba`), so they should stay separate." + "reason": "Vision-text dual encoder changes do not overlap with Mamba/FalconMamba tracing changes." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "`vision_text_dual_encoder` and `depth_anything` are different architectures; no shared duplicate fix is evident." + "reason": "Different model families and different output-capture implementations." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "`univnet` and `upernet` are separate model paths; both are output-refactor PRs but not the same change." + "reason": "UnivNet and UperNet are unrelated model code paths; same pattern, different change." }, { "left": "pull_request:44001", "right": "pull_request:44003", "accept": false, - "reason": "Different model families and different implementation details; no reason to merge these as duplicates." + "reason": "Mamba/FalconMamba and UnivNet are different implementations with different tracing plumbing." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "`univnet` and `depth_anything` are unrelated model code paths; same refactor pattern is insufficient." + "reason": "UnivNet is not the same bug/change as the depth-anything refactor." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "UnivNet output tracing is unrelated to CodeGen cache/output handling." }, { "left": "pull_request:44002", "right": "pull_request:44029", "accept": false, - "reason": "`upernet` and `rwkv` are different models with different tracing behavior; not one underlying bug/change." + "reason": "UperNet segmentation wrapper and RWKV recurrent core are different code paths." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "`upernet` and `mamba/falcon_mamba` are distinct code paths; these are separate model-specific refactors." + "reason": "UperNet and Mamba/FalconMamba are separate model families; not mergeable as one PR." }, { "left": "pull_request:44003", "right": "pull_request:44028", "accept": false, - "reason": "`mamba/falcon_mamba` and `superpoint` do not share the same concrete forward path or bug." + "reason": "Mamba-family tracing changes do not match SuperPoint's feature extraction/output refactor." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "`mamba/falcon_mamba` and `codegen` are unrelated model implementations; the refactor intent is similar but the fixes are separate." + "reason": "Different model families and different refactor mechanics; CodeGen cache changes are not the same change." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "`timm_backbone` and `codegen` are different code paths and differ in both purpose and implementation." + "reason": "Timm backbone output tracing and CodeGen cache/output typing are unrelated." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "`univnet` and `codegen` are distinct models; not the same underlying issue." + "reason": "UnivNet and CodeGen touch unrelated forward logic and output shapes." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "`upernet` and `codegen` are separate model-specific refactors, not one mergeable fix." + "reason": "UperNet segmentation outputs are not the same code path as CodeGen's attention/cache logic." }, { "left": "pull_request:44000", "right": "pull_request:44004", "accept": false, - "reason": "`vision_text_dual_encoder` and `codegen` do not share a concrete bug or code path." + "reason": "Vision-text dual encoder and CodeGen are unrelated; shared refactor theme is too broad." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "`depth_anything` and `vision_encoder_decoder` are unrelated model families; similar output-tracing cleanup is not enough." + "reason": "Both are refactors, but they target different wrapper/model families: depth-anything vs vision/speech encoder-decoder." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "`depth_anything` and `speech_encoder_decoder` are separate wrappers/models with distinct forward logic." + "reason": "Depth Anything and Speech Encoder-Decoder are unrelated changes, not one bug fix." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "`depth_anything` and `superpoint` are different vision models and different code paths." + "reason": "Depth-estimation models and SuperPoint do not share the same concrete output-tracing change." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "`depth_anything` and `rwkv` are unrelated architectures; same refactor theme only." + "reason": "Depth Anything output tracing is separate from RWKV's recurrent state/output handling." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "`depth_anything` and `dpr` are distinct models with separate output contracts." + "reason": "Depth-estimation refactor and DPR encoder pooling refactor are distinct changes." }, { "left": "pull_request:44026", "right": "pull_request:44029", "accept": false, - "reason": "`vision_encoder_decoder` and `rwkv` are unrelated code paths; no shared duplicate fix." + "reason": "Vision/Speech encoder-decoder wrappers and RWKV internals are different code paths." }, { "left": "pull_request:44026", "right": "pull_request:44027", "accept": false, - "reason": "`vision_encoder_decoder` and `speech_encoder_decoder` are similar wrappers but still separate implementations and changes." + "reason": "VisionEncoderDecoder and SpeechEncoderDecoder are parallel wrappers but still separate PR targets and changes." }, { "left": "pull_request:44026", "right": "pull_request:44028", "accept": false, - "reason": "`vision_encoder_decoder` and `superpoint` are different model families and do not look mergeable into one PR." + "reason": "Encoder-decoder wrappers are not the same change as SuperPoint feature tracing." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "`vision_encoder_decoder` and `dpr` are distinct model code paths; similar refactor pattern only." + "reason": "VisionEncoderDecoder wrapper changes and DPR encoder changes are unrelated." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "`speech_encoder_decoder` and `superpoint` are unrelated models; not the same underlying bug/change." + "reason": "SpeechEncoderDecoder and SuperPoint are unrelated model implementations." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "`speech_encoder_decoder` and `rwkv` are different architectures and forward paths." + "reason": "Speech encoder-decoder wrapper logic is not the same as RWKV state/output refactoring." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "`speech_encoder_decoder` and `dpr` are separate model implementations; no duplicate evidence." + "reason": "SpeechEncoderDecoder and DPR touch different architectures and output contracts." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "`superpoint` and `rwkv` do not share the same concrete change, only a general refactor style." + "reason": "SuperPoint and RWKV are different code paths with different output semantics." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "`superpoint` and `dpr` are unrelated models with different output behavior." + "reason": "SuperPoint feature tracing is unrelated to DPR encoder pooling/output handling." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "`rwkv` and `dpr` are separate model families; these should not be merged as duplicates." + "reason": "RWKV recurrent tracing and DPR encoder output refactors are distinct model changes." } ] }, @@ -17956,7 +17844,7 @@ "pull_request:44662|pull_request:45075", "pull_request:45075|pull_request:45401" ], - "prepared_review_unit_hash": "1272ab810d881b814803a9eee80da6492700d722b5176e7355be3aed56f63639", + "prepared_review_unit_hash": "5b0970c67d84b73184eb185c6296518accef95aa2b257ddf82ef81e8028fb20b", "component_budget": { "node_count": 3, "item_count": 3, @@ -17978,30 +17866,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three pull requests are separate model-integration additions (PenguinVL, DeepSeek-OCR-2, Voxtral TTS). They overlap on shared Transformers registration/docs plumbing, but they do not fix the same concrete code-path problem and should not be merged as duplicates.", + "summary": "Three open PRs, but they are separate model-addition implementations rather than duplicates. They share some auto/docs plumbing, yet each targets a different model family and code path.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44662 is the best canonical representative because it is the most complete, non-draft model implementation with the largest implementation/test surface and strong review activity; the other PRs target different models entirely.", + "canonical_pr_reason": "PR 45075 is the best representative of the cluster because it is a fuller, non-draft implementation with substantial tests and conversion support, but it is not a duplicate of the others.", "best_issue_reason": null, - "best_pr_reason": "PR 44662 is the strongest standalone PR in the cluster: it is non-draft, appears substantially implemented, and is more review-mature than the other two. However, it is not a duplicate of them.", + "best_pr_reason": "PR 45075 is the strongest standalone PR in this set: it looks the most complete and mature, with broader implementation coverage and high review activity.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different models and different implementations: PenguinVL vs Voxtral TTS. Shared auto-mapping/docs files are generic integration plumbing, not evidence of the same underlying change." + "reason": "PenguinVL vs Voxtral TTS are different model families with different functionality and code paths; shared auto/docs files are just framework plumbing." }, { "left": "pull_request:44662", "right": "pull_request:45075", "accept": false, - "reason": "Different model additions (PenguinVL vs DeepSeek-OCR-2). The overlap is only in common registration and documentation files, so they are not the same change." + "reason": "PenguinVL and DeepSeek-OCR-2 are unrelated model additions; they touch overlapping registration files but implement different concrete models." }, { "left": "pull_request:45075", "right": "pull_request:45401", "accept": false, - "reason": "These add different model families with different code paths and tests. Similar repo-wide wiring does not make them duplicates." + "reason": "DeepSeek-OCR-2 and Voxtral TTS solve different tasks and modify different model code, so they are not the same underlying change." } ] }, @@ -18031,12 +17919,12 @@ "soft_pairs": [ "pull_request:44979|pull_request:45363" ], - "prepared_review_unit_hash": "6cfab1da37a7ed08131f004523a142ac768f2fb01d153da38707c4b699db578e", + "prepared_review_unit_hash": "dbe44d063893dc357fbc04a83dfc9f85ceabd52fe063fe3581db146b42bca775", "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 }, @@ -18044,7 +17932,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4408, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -18053,18 +17941,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs concern \u201cfusion,\u201d but they target different abstractions and code paths: #44979 introduces a generic Module Fusion API, while #45363 adds kernel-based n-to-1 fusion through KernelConfig and integration points. They are related in theme, not the same change.", - "confidence": 0.89, + "summary": "These PRs are related in the broad area of model/kernel fusion, but they do not look like the same concrete change. One introduces a generic Module Fusion API and tests; the other wires in n-to-1 kernel fusion through KernelConfig and integration points. I would not merge them as duplicates.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "#45363 is the more concrete, end-to-end implementation PR: it wires KernelConfig through model loading/integrations and includes an example. #44979 is a separate generic API layer for module fusion.", + "canonical_pr_reason": "PR #45363 is the more complete, user-facing feature: it adds KernelConfig-based fusion plumbing across model loading and integrations, with a concrete example. PR #44979 is a lower-level API addition for module fusion and appears to be a separate foundational piece rather than the same change.", "best_issue_reason": null, - "best_pr_reason": "#45363 is the strongest representative of the cluster because it contains the most explicit user-facing fusion workflow and touches the actual runtime path. #44979 is a distinct API proposal rather than the same code-path change.", + "best_pr_reason": "PR #45363 is the best representative of the cluster because it is the broader end-to-end implementation and most clearly defines the actual fused-kernel workflow. PR #44979 is adjacent but not the same change.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "Different underlying changes: generic module-fusion API vs kernel-config-driven n-to-1 fusion. They share a broad topic but not the same concrete bug or mergeable PR." + "reason": "Both concern fusion, but #44979 adds a generic Module Fusion API while #45363 implements KernelConfig-driven n-to-1 kernel fusion and integration wiring. Different abstractions, different code paths, and not plausibly one merged PR." } ] }, @@ -18094,12 +17982,12 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "31e08d33267ea82ff71dbc4bd4fff888d806d2303c2f8284c18665df4b97f5b9", + "prepared_review_unit_hash": "8e1683248f8f91e0e7e8a9846ef70332eb5cc454db8a2e536855ddd992998558", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4414, + "serialized_chars": 4413, "estimated_input_tokens": 1104, "estimated_eval_tokens": 2464 }, @@ -18107,7 +17995,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4414, + "serialized_chars": 4413, "estimated_input_tokens": 1104, "estimated_eval_tokens": 2464 }, @@ -18116,18 +18004,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs patch the same Mistral regex bug in the same code path; 45317 is the more complete version because it includes the same fix plus tests.", - "confidence": 0.96, - "canonical_issue_reason": null, - "canonical_pr_reason": "45317 is the stronger canonical PR: it implements the same AttributeError fix as 45086 and also adds test coverage, making it the more complete representative.", - "best_issue_reason": null, - "best_pr_reason": "45317 is the best representative overall because it targets the same concrete bug as 45086 but is more polished and verified with tests.", + "summary": "Both PRs fix the same Mistral tokenizer AttributeError by switching from backend_tokenizer.pre_tokenizer to tokenizer.pre_tokenizer in the same file. PR 45317 is the more complete implementation because it includes an added test, while 45086 is the smaller precursor.", + "confidence": 0.97, + "canonical_issue_reason": "Issue 45081 is the shared bug report both PRs explicitly target, so it is the natural canonical issue for this cluster.", + "canonical_pr_reason": "PR 45317 is the better canonical PR because it addresses the same bug fix and also adds test coverage, making it the more complete representative change.", + "best_issue_reason": "If selecting one issue node, 45081 best represents the underlying bug since both PRs are direct fixes for that exact reported problem.", + "best_pr_reason": "45317 is the best PR overall because it fixes the same concrete AttributeError path as 45086 but is more complete due to the accompanying test file.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Accept: both PRs change the same file and the same _patch_mistral_regex logic to fix the same AttributeError; 45317 is effectively an expanded version of 45086 with tests." + "reason": "Same underlying bug fix in the same code path: both replace backend_tokenizer.pre_tokenizer with tokenizer.pre_tokenizer in _patch_mistral_regex for issue 45081. 45317 is essentially the extended version with tests, so they are merge-equivalent." } ] }, @@ -18157,7 +18045,7 @@ "soft_pairs": [ "pull_request:45115|pull_request:45569" ], - "prepared_review_unit_hash": "c0b4f6145aacbe1a0350c65719a034ae66089af50b36c221d200bddd466cf911", + "prepared_review_unit_hash": "277c3d8c80453e28a761c90e13983981c88b623862e2769bd0c3a0896550d9c5", "component_budget": { "node_count": 2, "item_count": 2, @@ -18179,18 +18067,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are about Nemotron H, but they target different changes: one is a conversion/refactor step for NemotronH \u2192 GraniteMoeHybrid-compatible weights, while the other splits Nemotron H into dense/sparse model variants and updates docs/auto mappings. They are related, but not the same concrete fix or mergeable into one PR without conflating separate work.", - "confidence": 0.9, + "summary": "Two NemotronH pull requests overlap on model files, but they are not the same change: one is a narrow conversion/mapping refactor, the other is a broader model split into NemotronH/NemotronHDense/NemotronHSparse with docs and tests.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45569 is the better canonical PR for this cluster because it is the broader, more complete architectural change: it introduces the new NemotronHDense/NemotronHSparse paths, updates auto-mapping and docs, and reworks NemotronH into a dispatcher.", + "canonical_pr_reason": "PR #45569 is the better canonical representative because it is the broader, more complete architectural change and includes the user-facing model split, docs, and tests. PR #45115 is a narrower internal refactor focused on conversion mappings and inheritance wiring.", "best_issue_reason": null, - "best_pr_reason": "PR #45569 is the strongest representative overall since it covers the end-state model layout and user-facing integration, whereas PR #45115 is a narrower intermediate refactor focused on conversion mapping and inheritance cleanup.", + "best_pr_reason": "PR #45569 is the strongest overall PR for the cluster since it covers the more substantial end-state of the NemotronH work. PR #45115 appears to be a partial, implementation-focused step rather than the main deliverable.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "Shared NemotronH files and subsystem are not enough: #45115 is a weight-conversion/refactor PR, while #45569 is a broader model-splitting and documentation/mapping PR. They do not appear to fix the same concrete code-path problem." + "reason": "Do not merge as duplicates: #45115 targets conversion-key renames and inheritance adjustments, while #45569 introduces a broader NemotronH reorganization into separate dense/sparse models plus docs and tests. They share subsystem files but not the same concrete bug or mergeable code-path fix." } ] }, @@ -18220,7 +18108,7 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "d7a99a2686e8d012bd357599325d895ce7e77794cd1c3a3b05461379aecd05c6", + "prepared_review_unit_hash": "34184b235bab7babd767dfbd1c42543bfc2f90d2122b9c13d01b72661b5c429c", "component_budget": { "node_count": 2, "item_count": 2, @@ -18242,18 +18130,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related heterogeneous-config/modeling PRs, but they are not duplicates: one adds per-layer config plumbing and the other adds downstream modeling/cache support that builds on it.", - "confidence": 0.9, + "summary": "Two related but distinct PRs: one adds the heterogeneous config plumbing, the other adds heterogeneous model/runtime support built on top of it. They share helper files, but they are not the same underlying change and should not be merged as duplicates.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45332 is the broader umbrella change and likely the better representative of the cluster because it includes the modeling/cache side plus the heterogeneity work.", + "canonical_pr_reason": "PR #45332 is the broader end-to-end feature PR: it includes runtime/model changes, cache/masking updates, and tests, so it is the most representative single artifact for the heterogeneous-model work.", "best_issue_reason": null, - "best_pr_reason": "PR #45332 is the most comprehensive artifact in this set and best represents the overall heterogeneous-model support effort.", + "best_pr_reason": "PR #45332 is the best standalone representative because it exercises the full heterogeneous-model code path rather than only the configuration layer; #45333 is a narrower prerequisite-style config PR.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "They share heterogeneous configuration files, but the scope differs: #45333 is config plumbing only, while #45332 adds model/cache behavior on top. This looks like dependent work, not the same concrete change or bug fix." + "reason": "Related feature work, but not the same concrete change: #45333 adds per-layer configuration support, while #45332 adds model/cache/runtime support that consumes that config. They touch different code paths and would not be merged as one PR." } ] }, @@ -18283,12 +18171,12 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "466254382f8f1ec018b2da9a5f6d73a5682d0c88d1dc2973870c570f6fb1402c", + "prepared_review_unit_hash": "f348b1de3926d44f0e704422f8a49f8bd39fd503c4e44d95b0819bbbbb65ce2b", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5513, + "serialized_chars": 5514, "estimated_input_tokens": 1379, "estimated_eval_tokens": 3014 }, @@ -18296,7 +18184,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5513, + "serialized_chars": 5514, "estimated_input_tokens": 1379, "estimated_eval_tokens": 3014 }, @@ -18305,18 +18193,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related by theme (typing/type-checking work) but they are not the same concrete change. PR #45415 is a broad typing sweep across many files, while PR #45425 adds a more focused set of typing helpers and model-related annotations around `modeling_utils` and PEFT integration.", - "confidence": 0.92, + "summary": "Both PRs are typing-related, but they are not the same change: one is a broad repo-wide type-checking cleanup, the other is a focused typing/Protocol update for modeling-related code. Shared files alone are not enough to merge them as duplicates.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45425 is the better anchor because it is the more focused, concrete typing change; PR #45415 is a larger umbrella sweep touching many unrelated files and concerns.", + "canonical_pr_reason": "PR #45425 is the more focused, self-contained change: it introduces typing aliases/Protocols and updates specific transformer/modeling code paths, whereas #45415 is a broad type-checking sweep across many files plus config changes.", "best_issue_reason": null, - "best_pr_reason": "PR #45425 is the best representative of the cluster because it centers on one specific typing fix area, whereas PR #45415 is a broad refactor-style sweep and is less suitable as a duplicate target.", + "best_pr_reason": "PR #45425 is the better representative for the cluster because it has a clearer, concrete implementation scope and stronger review activity; #45415 is more of an umbrella typing cleanup than a single bug/change.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Both are typing-related, but they address different scopes and likely different problems: #45415 is a wide type-checking pass across many modules, while #45425 adds specific typing protocols/helpers and related model integration changes. They would not plausibly merge into one single PR without becoming an umbrella sweep." + "reason": "Reject: both are typing-themed and overlap on a few files, but they target different concrete changes. #45415 is a broad mypy/type-checking pass across many modules; #45425 adds specific typing protocols/aliases for modeling-utils integration. They could be merged independently and are not the same underlying fix." } ] }, @@ -18346,7 +18234,7 @@ "soft_pairs": [ "pull_request:45487|pull_request:45615" ], - "prepared_review_unit_hash": "cd7f9869eccfcb945432c331720824e63dfa1c5a55797445b4d46f95ad9b92ac", + "prepared_review_unit_hash": "31e2ed1a51e054fee5ec2c579256755731d2d38862818f9a3750a9dca92daebc", "component_budget": { "node_count": 2, "item_count": 2, @@ -18368,18 +18256,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs are not duplicates: one fixes a real model-parallel device placement bug in AltCLIP/ChineseCLIP, while the other only updates qianfan_ocr test expectations for XPU.", - "confidence": 0.98, + "summary": "These two PRs are unrelated: one fixes a model-parallel/device placement bug in AltCLIP/ChineseCLIP and similar text models, while the other only updates Qianfan OCR XPU test expectations. They do not appear to be the same underlying change.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45487 is the substantive code fix: it changes model forward behavior to move buffered token type IDs onto the correct device in AltCLIP/ChineseCLIP, addressing an actual runtime bug.", + "canonical_pr_reason": "PR #45487 is the more canonical PR because it addresses a concrete runtime bug in model code across affected models, rather than only adjusting test expectations.", "best_issue_reason": null, - "best_pr_reason": "PR #45487 is the best representative because it targets the concrete code-path defect; PR #45615 is just test expectation updates and is not a duplicate of the device-placement fix.", + "best_pr_reason": "PR #45487 is the best representative of the cluster since it is the substantive code fix and has broader impact; PR #45615 is a narrow test-only update for a different model.", "soft_edge_verdicts": [ { "left": "pull_request:45487", "right": "pull_request:45615", "accept": false, - "reason": "Different artifacts and different problems: #45487 fixes AltCLIP/ChineseCLIP model-parallel tensor placement in modeling code, while #45615 only adds XPU golden outputs for qianfan_ocr tests. They do not appear to fix the same underlying bug and could not plausibly be merged into one PR." + "reason": "Different models and different kinds of changes: #45487 fixes device placement in model forward code, while #45615 only adds XPU test outputs for Qianfan OCR. They do not look like the same underlying bug or mergeable change." } ] }, @@ -18409,7 +18297,7 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "f19667f80befd8b30e4e268a20869b1e0fee18001081c49dfd22677745590118", + "prepared_review_unit_hash": "aef7968e10581de1f1f998011ea9ddd976c4762a3f1cbd67ddfdd464dc4b398d", "component_budget": { "node_count": 2, "item_count": 2, @@ -18431,18 +18319,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs both concern expert-parallel distributed loading, but they fix different bugs in different code paths: one for accelerate/DeepSpeed ZeRO-3 config handling, the other for FSDP2/rank-0 broadcast overwriting experts.", - "confidence": 0.87, + "summary": "Both PRs concern expert parallelism, but they fix different failure modes: one is about DeepSpeed ZeRO-3 loading/serialization via accelerate, the other about FSDP2 rank-0 broadcast overwriting experts. They are related but not duplicates.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45662 is the more specific and substantial fix: it addresses expert parameters being overwritten during distributed loading, with changes in core loading plus MoE/tensor-parallel handling. PR 45548 is a separate EP/ZeRO-3 loading fix centered on config serialization and an EP helper property.", + "canonical_pr_reason": "PR #45662 is the better representative of the cluster because it targets a concrete EP correctness bug in distributed loading/parameter materialization and includes the broader fix path across loading and MoE integration.", "best_issue_reason": null, - "best_pr_reason": "PR 45662 is the best standalone PR in the cluster because it targets a concrete failure mode in the actual parameter-loading path and has broader, more directly related code changes than 45548.", + "best_pr_reason": "PR #45662 is the strongest standalone change: it addresses a specific, reproducible corruption path for experts under FSDP2 and has the more substantive code-path fix.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "They share the broad EP/distributed-loading theme, but the bugs and fixes differ: 45548 removes distributed_config from serialization and adds has_ep for accelerate/ZeRO-3, while 45662 prevents experts from being overwritten during FSDP2 loading via post_shard_wrap. Not the same underlying change." + "reason": "Same general EP area, but different bugs and code paths: ZeRO-3/accelerate config handling vs FSDP2 rank-0 broadcast overwriting experts. They would not plausibly be merged as one PR." } ] }, @@ -18472,12 +18360,12 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "0cb8d3934d9076a1cdf068ac020bea4df45ac2c244c9710b84affc56fac1332b", + "prepared_review_unit_hash": "b8daa5471c31ab4185860429ad22fea64323e30161e3640608cdb3a1ddb82f79", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4590, + "serialized_chars": 4589, "estimated_input_tokens": 1148, "estimated_eval_tokens": 2552 }, @@ -18485,7 +18373,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4590, + "serialized_chars": 4589, "estimated_input_tokens": 1148, "estimated_eval_tokens": 2552 }, @@ -18494,18 +18382,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests target the same underlying xdist-safe captured_info logging problem in testing_utils; PR 45645 is the broader, more complete fix and includes the workflow/log-collection follow-up.", - "confidence": 0.92, + "summary": "Both PRs address the same xdist/captured_info logging collision around issue #45561, with 45645 effectively subsuming 45639 and adding CI-side log collection changes.", + "confidence": 0.87, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45645 is the best canonical PR because it directly fixes the xdist collision in testing_utils and also updates CI/workflow handling to read multiple captured_info files, making it the more complete end-to-end change.", + "canonical_pr_reason": "PR 45645 is the better canonical representative because it contains the core xdist-safe artifact fix from 45639 and extends it to the CI workflow/notification path that consumes those logs.", "best_issue_reason": null, - "best_pr_reason": "PR 45645 is the strongest representative of the cluster: it covers the core bug fix plus the downstream CI artifact handling, whereas PR 45639 only addresses the library-side portion.", + "best_pr_reason": "PR 45645 is the strongest overall fit for the cluster: it resolves the underlying captured_info collision and also updates the log aggregation path so the fix is complete end-to-end.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", "accept": true, - "reason": "Both PRs fix the same concrete xdist collision in captured_info logging from testing_utils. 45645 extends the same fix with workflow support for multiple captured_info files, so they look like the same underlying change and could plausibly be merged into one PR." + "reason": "Same underlying bug: xdist-safe captured_info handling for patched testing logs. 45645 includes the exact testing_utils change from 45639 plus related CI consumption updates, so the two are mergeable as one change set." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index 0690bbd69cd71497b7c262d789537df2559f1d6f..fe1848d700ae1a5141f1ac7db2f09ba76b4ec964 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260506t000041z", + "analysis_id": "hybrid-model-20260506t060033z", "archived_artifacts": { - "hybrid": "snapshots/20260506T000041Z/analysis-runs/hybrid-model-20260506t000041z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260506T000041Z/analysis-runs/hybrid-model-20260506t000041z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-05-06T00:10:51Z", + "published_at": "2026-05-06T06:08:43Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260506T000041Z", + "snapshot_id": "20260506T060033Z", "variant": "hybrid" } diff --git a/snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.json b/snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..3e15036e2c68c7f50ee341d3e4aa61055a972d7a --- /dev/null +++ b/snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.json @@ -0,0 +1,2243 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260506T060033Z", + "generated_at": "2026-05-06T06:08:42Z", + "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-43656-4", + "summary": "Cluster of 1 issues and 3 PRs centered on issue #43824.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 43824, + "canonical_pr_number": 43656, + "issue_numbers": [ + 43824 + ], + "pr_numbers": [ + 43656, + 43836, + 43842 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 43656, + "right_pr_number": 43836, + "code_similarity": 0.36, + "size_similarity": 0.176, + "file_overlap": 1.0, + "area_overlap": 0.051, + "patch_similarity": 0.048, + "shared_filenames": [ + "src/transformers/cli/serve.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/cli/serve.py", + "left_ranges": [ + [ + 11, + 18 + ], + [ + 30, + 36 + ], + [ + 315, + 323 + ], + [ + 665, + 671 + ], + [ + 931, + 937 + ], + [ + 1843, + 1849 + ], + [ + 1868, + 1874 + ] + ], + "right_ranges": [ + [ + 11, + 18 + ], + [ + 359, + 429 + ], + [ + 584, + 590 + ], + [ + 1892, + 1910 + ], + [ + 1917, + 1923 + ] + ] + } + ] + }, + { + "left_pr_number": 43656, + "right_pr_number": 43842, + "code_similarity": 0.405, + "size_similarity": 0.5, + "file_overlap": 1.0, + "area_overlap": 0.0, + "patch_similarity": 0.036, + "shared_filenames": [ + "src/transformers/cli/serve.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/cli/serve.py", + "left_ranges": [ + [ + 11, + 18 + ], + [ + 30, + 36 + ], + [ + 315, + 323 + ], + [ + 665, + 671 + ], + [ + 931, + 937 + ], + [ + 1843, + 1849 + ], + [ + 1868, + 1874 + ] + ], + "right_ranges": [ + [ + 54, + 61 + ], + [ + 587, + 593 + ] + ] + } + ] + }, + { + "left_pr_number": 43836, + "right_pr_number": 43842, + "code_similarity": 0.332, + "size_similarity": 0.088, + "file_overlap": 1.0, + "area_overlap": 0.033, + "patch_similarity": 0.017, + "shared_filenames": [ + "src/transformers/cli/serve.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/cli/serve.py", + "left_ranges": [ + [ + 11, + 18 + ], + [ + 359, + 429 + ], + [ + 584, + 590 + ], + [ + 1892, + 1910 + ], + [ + 1917, + 1923 + ] + ], + "right_ranges": [ + [ + 54, + 61 + ], + [ + 587, + 593 + ] + ] + } + ] + } + ] + }, + { + "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-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 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-44024-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44076, + "issue_numbers": [], + "pr_numbers": [ + 44024, + 44076 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44024, + "right_pr_number": 44076, + "code_similarity": 0.269, + "size_similarity": 0.976, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.492, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + } + ], + "duplicate_issues": [], + "duplicate_prs": [ + { + "cluster_id": "cluster-41115-3", + "canonical_pr_number": 45613, + "duplicate_pr_numbers": [ + 41116 + ], + "target_issue_number": 41115, + "reason": "PRs in cluster-41115-3 are treated as duplicates because they converge on issue #41115 with closing_reference evidence." + }, + { + "cluster_id": "cluster-41211-3", + "canonical_pr_number": 44339, + "duplicate_pr_numbers": [ + 41356 + ], + "target_issue_number": 41211, + "reason": "PRs in cluster-41211-3 are treated as duplicates because they converge on issue #41211 with closing_reference evidence." + }, + { + "cluster_id": "cluster-43240-3", + "canonical_pr_number": 43251, + "duplicate_pr_numbers": [ + 43254 + ], + "target_issue_number": 43240, + "reason": "PRs in cluster-43240-3 are treated as duplicates because they converge on issue #43240 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43656-4", + "canonical_pr_number": 43656, + "duplicate_pr_numbers": [ + 43836, + 43842 + ], + "target_issue_number": 43824, + "reason": "PRs in cluster-43656-4 are treated as duplicates because they converge on issue #43824 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-43698-3", + "canonical_pr_number": 43779, + "duplicate_pr_numbers": [ + 43816 + ], + "target_issue_number": 43698, + "reason": "PRs in cluster-43698-3 are treated as duplicates because they converge on issue #43698 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43979-11", + "canonical_pr_number": 44007, + "duplicate_pr_numbers": [ + 43996, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "target_issue_number": 43979, + "reason": "PRs in cluster-43979-11 are treated as duplicates because they converge on issue #43979 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-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-44024-2", + "canonical_pr_number": 44076, + "duplicate_pr_numbers": [ + 44024 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44024-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-45081-3", + "canonical_pr_number": 45317, + "duplicate_pr_numbers": [ + 45086 + ], + "target_issue_number": 45081, + "reason": "PRs in cluster-45081-3 are treated as duplicates because they converge on issue #45081 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-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." + } + ], + "best_issue": { + "cluster_id": "cluster-45647-3", + "issue_number": 45647, + "reason": "Issue #45647 is the strongest global issue candidate because it is open, belongs to a cluster with 3 artifacts, and carries 12 discussion comments plus 8 inbound references.", + "score": 162.0 + }, + "best_pr": { + "cluster_id": "cluster-43979-11", + "pr_number": 44007, + "reason": "PR #44007 is the strongest global PR candidate because it is open, belongs to a cluster with 11 artifacts, links to 1 issues, and carries 2 review/discussion events.", + "score": 238.886 + } +} diff --git a/snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..b2a480868e3bcef2699f240bca7bfc9785ce4b23 --- /dev/null +++ b/snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,18537 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260506T060033Z", + "generated_at": "2026-05-06T06:08:42Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 16, + "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": "09c84118ccf2dde47d3770ebc7e6cccace1e296993b699cbf2300384c60f576f", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2477, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2477, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This looks like a mixed cluster, not a duplicate set. Issue 13244 is about Tapas tokenization behavior and is unrelated to the other two. Issues 15980 and 16618 both involve model-loading/download failures, but they describe different concrete problems (private-model auth/error messaging vs. loading bert-base-uncased), so I would not merge them as duplicates.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 16618 is the broadest and most central model-loading report in the cluster, but it still does not match the other items closely enough to serve as a true duplicate anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "16618 is the best representative only in the loose sense that it is the most general and discussion-heavy model-loading issue; the cluster as a whole is too heterogeneous for a strong canonical duplicate choice.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Different problem domains: Tapas tokenization vs. generic model loading. The titles do not indicate the same underlying bug." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "Both mention model download/loading, but one is specifically a private-model login/error-message issue and the other is a failure to load bert-base-uncased; these are different failure modes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:24643", + "issue:30277", + "issue:34634", + "issue:34928", + "issue:35545", + "issue:36010", + "issue:36331", + "issue:39290", + "issue:41093", + "issue:41720", + "issue:41762", + "issue:42915", + "issue:43039", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43541", + "issue:43596", + "issue:43638", + "issue:43716", + "issue:43792", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43901", + "issue:43975", + "issue:44292", + "issue:44322", + "issue:44457", + "issue:44509", + "issue:44512", + "issue:44560", + "issue:44661", + "issue:44805", + "issue:44841", + "issue:44863", + "issue:44918", + "issue:45070", + "issue:45081", + "issue:45084", + "issue:45161", + "issue:45237", + "issue:45362", + "issue:45464", + "issue:45507", + "issue:45685", + "issue:45800" + ], + "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:44509|issue:44512", + "issue:35545|issue:36010", + "issue:43638|issue:44805", + "issue:43596|issue:44805", + "issue:44805|issue:45161", + "issue:44805|issue:44918", + "issue:44661|issue:44805", + "issue:34634|issue:35545", + "issue:30277|issue:43638", + "issue:43596|issue:43638", + "issue:43975|issue:44457", + "issue:43039|issue:45161", + "issue:44292|issue:44457", + "issue:34928|issue:44805", + "issue:24643|issue:30277", + "issue:43638|issue:45161", + "issue:34928|issue:36331", + "issue:43596|issue:45161", + "issue:43716|issue:45507", + "issue:39290|issue:45081", + "issue:43828|issue:45161", + "issue:43425|issue:44292", + "issue:44918|issue:45161", + "issue:34928|issue:41720", + "issue:43716|issue:45685", + "issue:43541|issue:43716", + "issue:43828|issue:44292", + "issue:45507|issue:45685", + "issue:43716|issue:44292", + "issue:44292|issue:44918", + "issue:43039|issue:44918", + "issue:41762|issue:43638", + "issue:43827|issue:44512", + "issue:43716|issue:44560", + "issue:43828|issue:45685", + "issue:44560|issue:45685", + "issue:43039|issue:43828", + "issue:30277|issue:36331", + "issue:34928|issue:36010", + "issue:43039|issue:44805", + "issue:43901|issue:44512", + "issue:41762|issue:43596", + "issue:39290|issue:41720", + "issue:43039|issue:44560", + "issue:43866|issue:44863", + "issue:43854|issue:43866", + "issue:39290|issue:43531", + "issue:41093|issue:43638", + "issue:43425|issue:45070", + "issue:43596|issue:45800", + "issue:45362|issue:45464", + "issue:43792|issue:45070", + "issue:43404|issue:43866", + "issue:45464|issue:45685", + "issue:43541|issue:45685", + "issue:43541|issue:43828", + "issue:36010|issue:39290", + "issue:44841|issue:45084", + "issue:41093|issue:41762" + ], + "prepared_review_unit_hash": "ba8241aa3ecd1f24f14f2f526e66fdd33af86c1d9e0d662f64cb7c213489f6ac", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 68, + "serialized_chars": 47969, + "estimated_input_tokens": 11993, + "estimated_eval_tokens": 24242 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is not a single duplicate cluster: most items are unrelated, spanning model-specific runtime errors, backend/device issues, trainer API breakages, and docs regressions. Only one soft edge looks like a true duplicate pair (the ZeRO-3/Bert index-0 loading failure).", + "confidence": 0.91, + "canonical_issue_reason": "No single canonical issue; the cluster mixes many unrelated bugs across different models, backends, and documentation changes.", + "canonical_pr_reason": null, + "best_issue_reason": "No single best issue for deduplication; the only clear duplicate is a small sub-pair, while the rest are separate problems.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Different Qwen3.5 chat/runtime failures: missing config field vs streaming API crash." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Different hardware/backend bugs: AMD GPU failure vs MPS histogram op unsupported." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Different model and code path: Mistral-3 image dtype mismatch vs GPT-OSS GPU issue." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Different bug family; autocast dtype mismatch is not the same as GPT-OSS device failure." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Different root cause: AMD GPU runtime issue vs invalid method call in GraniteMoEHybrid." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Different model/precision failure vs AMD GPU device-mapping failure." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "Different failures: DeepSpeed collective mismatch vs ONNX export error." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "Different model/subsystem: ModernBERT export vs Qwen3Moe FP8 load failure." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "Different code paths: export-time failure vs auto device mapping assert." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs bugs, but they cover different removed features and pages." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "Different bug classes: export failure vs GenerationMixin import break." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Similar index-error theme, but different model/context and no clear shared root cause." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "Same broad symptom family, but ZeRO-3 Bert load bug vs generic mask-shape mismatch." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Generic mask/index error vs GPT-OSS tensor-parallel-only failure." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Different training bug: mask-shape mismatch vs Qwen3.5 embedding unpacking." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer-mapping helper bug is unrelated to the mask-shape index error." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "Different feature areas: BarkProcessor voice preset vs ModernBERT ONNX export." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Different DeepSpeed failures: collective mismatch vs Bert ZeRO-3 index error." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "Same ZeRO-3/Bert empty-shard index-0 failure; only wording differs." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "Different issues: detokenization bug vs LoRA merge/save/load inconsistency." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger Kernel cross-entropy routing bug is unrelated to GPT-OSS TP failure." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Different model/runtime failure vs LoRA weight merge mismatch." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "FSDP activation-checkpointing mismatch is not the same as the mask-shape index error." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "Different DeepSpeed bugs: weight dimensionality error vs collective mismatch." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "Bert ZeRO-3 loading bug is unrelated to GPT-OSS tensor-parallel failure." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "Activation-checkpointing/FSDP bug is unrelated to CustomTrainer API breakage." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "Different model and failure mode: Bert ZeRO-3 load error vs GPT-OSS TP issue." + }, + { + "left": "issue:43716", + "right": "issue:45507", + "accept": false, + "reason": "Different model/runtime problem vs GraniteMoEHybrid invalid-method bug." + }, + { + "left": "issue:39290", + "right": "issue:45081", + "accept": false, + "reason": "Different models and subsystems: Gemma3 config issue vs Mistral tokenizer regex crash." + }, + { + "left": "issue:43828", + "right": "issue:45161", + "accept": false, + "reason": "Different runtime bugs: autocast dtype mismatch vs GPT-OSS TP failure." + }, + { + "left": "issue:44918", + "right": "issue:45161", + "accept": false, + "reason": "Different issue classes: Qwen3.5 embedding unpacking vs GPT-OSS TP-only failure." + }, + { + "left": "issue:41762", + "right": "issue:43638", + "accept": false, + "reason": "Both are ZeRO-3 load index errors, but Gemma3 and Bert loading paths are distinct." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Both docs regressions, but they refer to different removed v5 topics." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Different model/runtime bugs: image dtype mismatch vs video StopIteration." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Different helper/API bug vs runtime mask-shape mismatch." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Different model artifacts/issues: Ovis2 checkpoint corruption vs NemotronH checkpoint loading." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "Different models and failure modes; unit-test load issue vs corrupted checkpoint." + }, + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Related sliding-window theme, but different model families and code paths." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Different errors: mask/tensor shape mismatch vs Bert ZeRO-3 index-0 failure." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Torch-version incompatibility is unrelated to the pydantic PretrainedConfig field break." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "Bert ZeRO-3 index error is unrelated to the torch/transformers version incompatibility." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Same model family, but different errors and surfaces; not enough for one duplicate." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "Whisper load failure is unrelated to the PretrainedConfig/pydantic regression." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "Different problems: untied lm_head weights vs corrupted checkpoint." + }, + { + "left": "issue:45464", + "right": "issue:45685", + "accept": false, + "reason": "Different model/API bug vs MPS backend op support failure." + }, + { + "left": "issue:43541", + "right": "issue:45685", + "accept": false, + "reason": "Grouped_mm tracing failure is unrelated to the MPS histogram error." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Different runtime bugs: torchdynamo tracing vs autocast dtype mismatch." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "ImportError for GenerationMixin is unrelated to Gemma3 sliding-window config." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Processor failure and template compilation error are different bugs." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both are failure reports, but the shape-mismatch error and Gemma3 ZeRO-3 load bug are different roots." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is conservative and grounded in the packet. It correctly treats the cluster as mixed rather than one coherent duplicate set, and the only accepted soft edge (43596/43638) is plausibly the same ZeRO-3 Bert index-out-of-bounds bug from the titles. The other soft-edge rejections are suitably cautious given the available evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 2, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:39186", + "issue:39290", + "issue:39692", + "issue:40990", + "issue:41093", + "issue:42915", + "issue:43054", + "issue:43296", + "issue:43366", + "issue:43404", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43644", + "issue:43645", + "issue:43716", + "issue:43742", + "issue:43828", + "issue:43873", + "issue:43950", + "issue:43975", + "issue:43994", + "issue:44291", + "issue:44292", + "issue:44360", + "issue:44387", + "issue:44492", + "issue:44512", + "issue:44530", + "issue:44534", + "issue:44589", + "issue:44863", + "issue:44918", + "issue:44928", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45161", + "issue:45399", + "issue:45464", + "issue:45507", + "issue:45542", + "issue:45632", + "issue:45685", + "issue:45698", + "issue:45800" + ], + "soft_pairs": [ + "issue:43296|issue:43366", + "issue:43572|issue:45070", + "issue:44387|issue:45005", + "issue:43541|issue:45161", + "issue:43828|issue:45464", + "issue:44291|issue:45070", + "issue:45020|issue:45800", + "issue:43366|issue:45464", + "issue:43531|issue:45070", + "issue:39290|issue:43742", + "issue:43541|issue:44589", + "issue:41093|issue:43596", + "issue:43054|issue:43994", + "issue:45542|issue:45800", + "issue:43716|issue:44918", + "issue:44918|issue:45685", + "issue:45632|issue:45698", + "issue:43404|issue:45005", + "issue:40990|issue:44960", + "issue:43645|issue:45399", + "issue:44530|issue:45005", + "issue:42915|issue:45399", + "issue:44360|issue:44512", + "issue:44863|issue:45005", + "issue:43873|issue:45005", + "issue:39186|issue:44918", + "issue:44492|issue:44512", + "issue:44918|issue:45507", + "issue:43975|issue:44292", + "issue:43950|issue:44534", + "issue:43644|issue:43950", + "issue:44292|issue:45507", + "issue:39692|issue:43054", + "issue:43828|issue:44928" + ], + "prepared_review_unit_hash": "98256ba92f0f391306da3c6343c126a436c80ae212ff4fdc2d73d8527f944a1a", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37550, + "estimated_input_tokens": 9388, + "estimated_eval_tokens": 19032 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of unrelated transformer bugs across different models/features. The only clear duplicate-like group is the v5 non-persistent-buffer regression; the rest are too semantically distant to merge.", + "confidence": 0.91, + "canonical_issue_reason": "No single canonical issue is appropriate here: the items span many unrelated bugs, models, and regressions, so there is no global duplicate target.", + "canonical_pr_reason": "No pull requests are present in this cluster, so there is no canonical PR.", + "best_issue_reason": "No global best issue is suitable as a representative; at most, issue 43950 is the best local representative for the non-persistent-buffer regression subgroup.", + "best_pr_reason": "No pull requests are present in this cluster, so there is no best PR.", + "soft_edge_verdicts": [ + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "Different problems: PaddleOCR-VL load failure in vLLM vs GGUF architecture support for gpt-oss." + }, + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "Both are v5 config regressions, but the concrete failures differ: missing pad_token_idx vs pydantic field handling." + }, + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Different bug classes: int4 quantization OOM vs tied-weights behavior on translation models." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Different code paths and models: grouped_mm torch-dynamo tracing error vs GPT-OSS tensor-parallel support." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "Both are dtype-related, but one is Phi-tiny-MoE autocast mismatch and the other is Qwen3.5 streaming inference failure." + }, + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "Different regressions: unexpected _is_hf_initialized argument vs PretrainedConfig pydantic field breakage." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "General version incompatibility is too broad; remote_code breakage and torch-2.4.1 incompatibility are not the same issue." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "Different models and symptoms: GGUF gpt-oss support vs Qwen3.5 chat/completions streaming failure." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Qwen3-MoE sliding_window issue is unrelated to the PretrainedConfig pydantic regression." + }, + { + "left": "issue:39290", + "right": "issue:43742", + "accept": false, + "reason": "Different model-loading failures: Gemma3TextConfig missing sliding_window_pattern vs MobileLLM key error." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Grouped_mm tracing failure and missing Float8 storage are unrelated." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Different runtime errors and mechanisms: mask shape mismatch vs deepspeed zero3/BertModel indexing failure." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": false, + "reason": "Both involve SigLIP2 quality, but the reports describe different symptoms and no shared concrete failure is clear." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "Undefined tf backend error from tensorboard-only install is unrelated to torch/transformers version incompatibility." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Different models and failures: image preprocessor dtype mismatch vs Qwen3.5 embedding unpacking in TRL SFT trainer." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "Unpacking input embeddings and MPS histogram on Int are unrelated bugs." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": false, + "reason": "Both touch remote-code/custom-module loading, but one is a leaf-directory cache collision and the other is wrong module selection after save_pretrained; not clearly the same bug." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "One is a Mistral3 tied-weight bug affecting image-text generation; the other is a broader translation-model tied-weights issue." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "Different model/benchmark reports: GPT-OSS perplexity vs GLM5 issue." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Notebook custom-model initialization and flash-attn fallback logic are separate issues." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash is unrelated to tied weights on translation models." + }, + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "Qwen3-MoE FineGrainedFP8Config failure is unrelated to flash-attn fallback blocking." + }, + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "DSA indexer ReLU discussion and a removed docs command are unrelated." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint loading failure is unrelated to translation-model tied weights." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization/offloading behavior is unrelated to tied-weight handling." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "FSDP 2-D weight error is unrelated to Qwen3.5 embedding unpacking in TRL." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "A cache-strategy typo and a removed run-command doc issue are not the same underlying change." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Qwen3.5 trainer failure and GraniteMoEHybrid invalid method are different code paths." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "DeepSeek detokenization regression and Qwen-3-8B-NVFP4 loading error are unrelated." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "Same regression class: v5 mishandling of non-persistent buffers/register_buffer(persistent=False) leading to junk values." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": true, + "reason": "Same underlying bug: Transformers 5.x corrupts non-persistent buffers during loading/initialization." + }, + { + "left": "issue:44292", + "right": "issue:45507", + "accept": false, + "reason": "Different model-specific failures: Qwen-3-8B-NVFP4 runtime error vs GraniteMoEHybrid invalid method." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "SigLIP2 appears in both, but one is a documentation example with processor/model mismatch and quantization failure, not the same as embedding-quality degradation." + }, + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Different bugs: autocast dtype mismatch on Phi-tiny-MoE vs RLHF NaNs from Qwen3.5 position_ids/SDPA fallback." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall: the cluster is mostly heterogeneous, there are no PRs in the packet, and the only clearly duplicate subgroup is the non-persistent-buffer v5 regression (43644/43950/44534). The soft-edge verdicts are conservative and do not overclaim equivalence where the packet only shows superficial similarity." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 3, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:36331", + "issue:38175", + "issue:41553", + "issue:41628", + "issue:41762", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43257", + "issue:43404", + "issue:43493", + "issue:43643", + "issue:43742", + "issue:43782", + "issue:43792", + "issue:43824", + "issue:43827", + "issue:43828", + "issue:43856", + "issue:43866", + "issue:43872", + "issue:43881", + "issue:43883", + "issue:43940", + "issue:43950", + "issue:43975", + "issue:44220", + "issue:44291", + "issue:44368", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44509", + "issue:44589", + "issue:44661", + "issue:44960", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45237", + "issue:45313", + "issue:45356", + "issue:45399", + "issue:45520", + "issue:45701", + "issue:45800" + ], + "soft_pairs": [ + "issue:43866|issue:44403", + "issue:36331|issue:41762", + "issue:41628|issue:45800", + "issue:41553|issue:43883", + "issue:43872|issue:44589", + "issue:38175|issue:43054", + "issue:41553|issue:43881", + "issue:43828|issue:45800", + "issue:43054|issue:45701", + "issue:42947|issue:44387", + "issue:44403|issue:44451", + "issue:43643|issue:45020", + "issue:43881|issue:43883", + "issue:43054|issue:43493", + "issue:43975|issue:45520", + "issue:43975|issue:45237", + "issue:44403|issue:44488", + "issue:42947|issue:43856", + "issue:42491|issue:44368", + "issue:41762|issue:42915", + "issue:43975|issue:44451", + "issue:44661|issue:45020", + "issue:43824|issue:43872", + "issue:44991|issue:45399", + "issue:45020|issue:45356", + "issue:43782|issue:44960", + "issue:43827|issue:44509", + "issue:45081|issue:45399", + "issue:43792|issue:44220", + "issue:43404|issue:43950", + "issue:43742|issue:45399", + "issue:42915|issue:45020", + "issue:43257|issue:43866", + "issue:43742|issue:45020", + "issue:41553|issue:44991", + "issue:43940|issue:45313", + "issue:43872|issue:44291" + ], + "prepared_review_unit_hash": "44a60c011a8406f58640a6c7569be07f9741c3e805cec9c8c949eccae19643e3", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 37, + "serialized_chars": 38718, + "estimated_input_tokens": 9680, + "estimated_eval_tokens": 19616 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a mixed cluster of unrelated transformers issues. A few soft pairs look like true duplicates around remote_code config loading, docs for removed pipeline tasks, Whisper fbank extraction, bitsandbytes/init_empty_weights, and SigLIP2 regression reports; the rest are only broad subsystem matches and should stay separate.", + "confidence": 0.74, + "canonical_issue_reason": "45020 is the best umbrella among the related reports: it is open, broad, and captures the remote_code/loading regression subgroup better than the more specific issues.", + "canonical_pr_reason": null, + "best_issue_reason": "45020 is the strongest triage anchor for this cluster because it is open and broad enough to absorb several of the genuine duplicate-like reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "Different bugs: corrupted checkpoint vs generic loading noise." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "CustomTrainer loss-arg regression and Gemma3 ZeRO-3 loading failure are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:45800", + "accept": false, + "reason": "AutoImageProcessor import failure is unrelated to torch/transformers version incompatibility." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Bad Voxtral tokenizer error message vs Molmo tied-weights attribute error are different code paths." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "Both are loading/type errors, but bitsandbytes _is_hf_initialized and Float8 storage issues are distinct failures." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": true, + "reason": "Both report SigLIP2 output degradation/regression; likely the same underlying implementation issue." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Voxtral tokenizer messaging issue and glm-4v loading failure are not the same bug." + }, + { + "left": "issue:43828", + "right": "issue:45800", + "accept": false, + "reason": "autocast dtype mismatch is not the same as the torch 2.4.1 / transformers 5.8 incompatibility report." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "Specific SigLIP2 embedding regression vs generic tokenization changes; too broad to merge." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "Different memory problems: LoRA checkpointing ineffectiveness vs int4 CUDA reserved-memory growth." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic loading noise and a specific ScandiBERT load failure are not the same underlying issue." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": true, + "reason": "Both concern remote_code / trust_remote_code breakage in recent transformers versions." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "glm-4v loading failure and Molmo tied-weights attribute error are different models and failures." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": true, + "reason": "Both point to the same SigLIP2 HF-vs-original implementation discrepancy." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "Detokenization bug and flash_attn import/keyerror are unrelated." + }, + { + "left": "issue:43975", + "right": "issue:45237", + "accept": false, + "reason": "Detokenization bug and AMD GPU runtime failure are unrelated." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "Generic loading noise and a specific model load failure are not duplicates." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "Both mention memory, but LoRA checkpointing vs Qwen3 MoE training inefficiency are different bugs." + }, + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "Cross-version Qwen3_moe incompatibility is not the same as a tie_word_embeddings warning." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Both are ZeRO/deepspeed-related, but Gemma3 missing-index and Qwen3 MoE weight-conversion failures are different issues." + }, + { + "left": "issue:43975", + "right": "issue:44451", + "accept": false, + "reason": "Detokenization bug and model loading failure are not the same underlying change." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "Tokenizer-mapping add-new-model-like bug is unrelated to remote_code loading regressions." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "Qwen2.5-VL import failure and bitsandbytes storage/type error are unrelated." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer load regression and flash-attn fallback blocking are different subsystems." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "Remote_code regressions and Kimi tokenizer codec/fix_mistral_regex regression are different bugs." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Qwen3VL weight_only loading error is unrelated to GLM5." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": true, + "reason": "Both report stale docs still referencing pipeline tasks removed in v5." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Mistral regex crash and flash-attn fallback gating are unrelated." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": true, + "reason": "Both are about the same Whisper fbank feature extraction path failing." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "Tied lm_head weights in Mistral3 and non-persistent buffer corruption are different bugs." + }, + { + "left": "issue:43742", + "right": "issue:45399", + "accept": false, + "reason": "MobileLLM load failure and flash-attn fallback logic are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:45020", + "accept": false, + "reason": "Qwen3 MoE ZeRO loading/weight conversion and remote_code breakage are different issues." + }, + { + "left": "issue:43257", + "right": "issue:43866", + "accept": false, + "reason": "Qwen3 MoE weight-conversion failure and checkpoint corruption are unrelated." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "MobileLLM key error and remote_code regression are unrelated." + }, + { + "left": "issue:41553", + "right": "issue:44991", + "accept": false, + "reason": "Voxtral tokenizer error messaging and EMBEDDIA tokenizer loading regression are different bugs." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Both are ZeRO-3 load failures, but different Qwen model families and no concrete shared bug evidence." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "Both describe the same init_empty_weights / _is_hf_initialized incompatibility during model loading." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays mostly conservative. The accepted soft edges are supported by clear title-level overlap or the same documented regression subgroup (remote_code, docs removal, Whisper fbank, bitsandbytes/init_empty_weights, SigLIP2). The rejected edges are generally reasonable separations rather than overclaims." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 4, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:41093", + "issue:41553", + "issue:42617", + "issue:42915", + "issue:43257", + "issue:43317", + "issue:43366", + "issue:43531", + "issue:43550", + "issue:43596", + "issue:43638", + "issue:43645", + "issue:43646", + "issue:43824", + "issue:43828", + "issue:43854", + "issue:43872", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44155", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:44928", + "issue:45030", + "issue:45305", + "issue:45313", + "issue:45325", + "issue:45357", + "issue:45538", + "issue:45584", + "issue:45588" + ], + "soft_pairs": [ + "issue:43854|issue:44863", + "issue:42915|issue:43872", + "issue:44928|issue:45588", + "issue:41553|issue:44488", + "issue:43257|issue:45305", + "issue:36331|issue:41093", + "issue:43872|issue:45538", + "issue:44530|issue:44863", + "issue:41093|issue:44805", + "issue:44479|issue:44560", + "issue:24643|issue:30064", + "issue:43366|issue:43828", + "issue:43366|issue:43531", + "issue:43550|issue:44910", + "issue:43645|issue:43646", + "issue:43950|issue:45325", + "issue:43366|issue:44589", + "issue:44560|issue:44918", + "issue:42617|issue:43366", + "issue:42915|issue:43957", + "issue:43824|issue:43957", + "issue:36010|issue:42915", + "issue:38175|issue:43994", + "issue:43646|issue:43950", + "issue:44155|issue:45584", + "issue:44877|issue:45030", + "issue:44560|issue:44805", + "issue:43638|issue:44661", + "issue:43596|issue:44661", + "issue:43317|issue:44756", + "issue:38175|issue:39692", + "issue:45313|issue:45357" + ], + "prepared_review_unit_hash": "f0a94ceaff6bd7bbdb5d44190a659882ce543e3093b03832fd5193cbaa368cf5", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37463, + "estimated_input_tokens": 9366, + "estimated_eval_tokens": 18988 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly noisy and contains several separate bug themes; only a few pairs look like true duplicates or near-duplicates. The clearest duplicate is the Transformers 5.0 custom model initialization regression, with a couple of other plausible same-bug pairs around SigLIP2, Qwen video input, and a mask-shape IndexError family.", + "confidence": 0.58, + "canonical_issue_reason": "Issue 43646 is the broadest, most canonical wording for the custom-model initialization regression; 43645 is the notebook-specific variant of the same bug.", + "canonical_pr_reason": null, + "best_issue_reason": "43646 is the best single representative issue in this set because it describes the underlying regression generically and is the cleanest anchor for the duplicate pair.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "Different failures: GLM-4.7 test loading vs NemotronH checkpoint loading." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "Different concrete bugs: Qwen3Moe FP8 failure vs bitsandbytes constructor incompatibility." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "Both mention Qwen/attention problems, but the symptoms and code paths differ (RLHF NaNs vs flash_attention s_aux handling)." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Different loading issues: bad Voxtral tokenizer error vs a model load failure for cjvt/sleng-bert." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Both involve DeepSpeed, but one is Qwen3 MoE weight conversion and the other is gradient averaging with GAS." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "Unexpected trainer kwarg vs a tensor-mask shape mismatch; not the same bug." + }, + { + "left": "issue:43872", + "right": "issue:45538", + "accept": false, + "reason": "Unrelated: bitsandbytes init kwarg error vs CLIPTokenizer max-length behavior." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "Different subsystems and symptoms: PagedAttentionCache group-type crash vs NemotronH checkpoint loading." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": true, + "reason": "Same IndexError family with the same mask-vs-tensor shape mismatch; 44805 looks like a later report of the same off-by-one indexing bug." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": true, + "reason": "Same Qwen video-input regression theme; 44560 is a specific StopIteration manifestation of the broader 5.3.0 video bug." + }, + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "Completely different bugs: DeepSpeed weight dimensionality vs void segmentation map processing." + }, + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "GGUF gpt-oss support and Phi-tiny-MoE autocast dtype mismatch are unrelated." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "Different issues: GGUF architecture support vs Qwen3-MoE sliding_window behavior." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different failures in different models/code paths: Bamba torch.compile SDPA vs Qwen3.5 flash-attention memory access." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "Same underlying custom-model initialization regression; 43645 is just the notebook-specific description of the 43646 bug." + }, + { + "left": "issue:43950", + "right": "issue:45325", + "accept": false, + "reason": "Different regressions: non-persistent buffer corruption vs Qwen2.5-VL rope-index scaling." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "Unrelated: GGUF support request vs missing Float8 storage type." + }, + { + "left": "issue:44560", + "right": "issue:44918", + "accept": false, + "reason": "Different Qwen3.5/video vs TRL embedding unpacking bugs." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "3d_parallel.py runtime failure is unrelated to GGUF model support." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "Different failures: Qwen3Moe FP8 vs meta-device loading regression." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "Importing Qwen2_5_VL is a different issue from meta-device model loading." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "GenerationMixin import error is unrelated to Qwen3Moe FP8 loading." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": true, + "reason": "Same SigLIP2 bad-output bug reported through different APIs; both point to the same model/processor mismatch symptom." + }, + { + "left": "issue:43646", + "right": "issue:43950", + "accept": false, + "reason": "Custom model init regression and buffer corruption are different bugs." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "Different audio/Whisper failures with different code paths." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": false, + "reason": "Both are config-validation related, but they are not the same model or failure mode." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Different error classes and domains: video StopIteration vs mask shape mismatch." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "Both are loading/registration issues, but the causes are different (DeepSpeed/Bert index error vs tokenizer-mapping restriction)." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Different underlying bugs: ZeRO-3 zero-sized tensor indexing vs tokenizer-mapping lookup failure." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "Different loading problems: dequantized model offload failure vs mmap/OOM behavior on Strix Halo." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "Related SigLIP2 context, but 39692 bundles multiple documentation errors and is not clearly the same single bug." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "Different Qwen3.5 DeepSpeed load/save issues: ZeRO-3 language_model weights vs incorrect visual encoder keys on save_pretrained." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays appropriately cautious. The accepted pairs are framed as plausible same-bug matches rather than overstated certainties, and the rejected pairs are mostly conservative distinctions between different failure modes. The canonical issue choice around 43645/43646 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": 2, + "component_count": 16, + "review_unit_index": 5, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:36010", + "issue:39404", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43335", + "issue:43404", + "issue:43425", + "issue:43454", + "issue:43644", + "issue:43645", + "issue:43708", + "issue:43720", + "issue:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44075", + "issue:44164", + "issue:44291", + "issue:44387", + "issue:44402", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44661", + "issue:44683", + "issue:44756", + "issue:44910", + "issue:44960", + "issue:44977", + "issue:44991", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45406", + "issue:45588", + "issue:45663", + "issue:45704", + "issue:45759" + ], + "soft_pairs": [ + "issue:43644|issue:44534", + "issue:44960|issue:44977", + "issue:44451|issue:44488", + "issue:42491|issue:43720", + "issue:43708|issue:44075", + "issue:43854|issue:43931", + "issue:41720|issue:42915", + "issue:42491|issue:45276", + "issue:43856|issue:43873", + "issue:43257|issue:43854", + "issue:39692|issue:43994", + "issue:42915|issue:44756", + "issue:42491|issue:43931", + "issue:43856|issue:44387", + "issue:44661|issue:44683", + "issue:43644|issue:43645", + "issue:44910|issue:45588", + "issue:45362|issue:45406", + "issue:42491|issue:45292", + "issue:44488|issue:44991", + "issue:36010|issue:39404", + "issue:43404|issue:43454", + "issue:45276|issue:45335", + "issue:44164|issue:44291", + "issue:44402|issue:44488", + "issue:39692|issue:44960", + "issue:43335|issue:45663", + "issue:43425|issue:43957", + "issue:42947|issue:45704", + "issue:45663|issue:45759", + "issue:43873|issue:44387", + "issue:43854|issue:45216", + "issue:44387|issue:44756", + "issue:44291|issue:44756" + ], + "prepared_review_unit_hash": "d61e1ead56283f20fff914da5e7df9c6c60ff67a47fe81c9c67a9df580b5bf38", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 34, + "serialized_chars": 37320, + "estimated_input_tokens": 9330, + "estimated_eval_tokens": 18916 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous, with two separate duplicate-like families: one around non-persistent buffers being filled with junk, and one around `resize_token_embeddings` not propagating to output/decoder embeddings. The rest of the pairs are only loosely related and should stay separate.", + "confidence": 0.77, + "canonical_issue_reason": "Issue #45292 is the best canonical issue for the `resize_token_embeddings` family: it states the generic propagation bug in a model-agnostic way and cleanly subsumes the model-specific variants (#45276, #45335).", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #45292 is the strongest representative of the `resize_token_embeddings` subgroup because it is the broadest, most model-agnostic description. The cluster also contains a separate duplicate-like subgroup around non-persistent buffers (#43644/#44534), so #45292 is not representative of the entire cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Same bug with near-identical wording: non-persistent buffers are filled with junk in Transformers v5." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "Both involve Qwen/flash-attention, but the concrete failures and model behavior differ; not the same bug." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": false, + "reason": "Both are loading failures, but they concern different models and no shared root cause is shown." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "Different code paths and failure modes: Qwen3 MoE LoRA compatibility vs BitNet packed-weight loading." + }, + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Unrelated bugs: checkpoint resume step calculation vs optimizer arguments not being used." + }, + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Different model-loading failures for different architectures; not the same underlying issue." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Both are Qwen3-related, but one is auto device mapping/cuda assert and the other is FP8 config failure." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "Different bugs: LoRA model incompatibility vs token-embedding resize propagation." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "Both concern memory, but one is training memory usage and the other is quantization/offloading behavior." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Different subsystems and failure modes: accelerate/deepspeed weight conversion vs model loading in tests." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but one is a broken docs example and the other is a pipeline/AutoModel result-quality issue." + }, + { + "left": "issue:42915", + "right": "issue:44756", + "accept": false, + "reason": "No concrete overlap beyond being performance/loading issues; different models and mechanisms." + }, + { + "left": "issue:42491", + "right": "issue:43931", + "accept": false, + "reason": "Distinct bugs affecting different Qwen variants and different loading/training paths." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both concern memory, but one is Qwen3 MoE training memory usage and the other is int4 CUDA reserved-memory OOM." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "Unrelated: tokenizer-mapping logic vs compiled flex_attention on newer torch." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "Different regressions: junk non-persistent buffers vs notebook custom-model initialization." + }, + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Both mention flash attention, but the crash conditions and root causes are different." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Different bugs in different model stacks: Qwen chat crash vs Gemma4Processor missing `_tokenizer`." + }, + { + "left": "issue:42491", + "right": "issue:45292", + "accept": false, + "reason": "Different issues: versioned Qwen3 MoE LoRA incompatibility vs generic embedding-resize propagation." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Both tokenizer-loading failures, but for different model repos with no evidence of the same root cause." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "Import error for `GenerationMixin` is unrelated to Whisper pipeline `return_language` behavior." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both involve untied `lm_head`, but they are model-specific implementations and not clearly the same concrete bug." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": true, + "reason": "Same underlying bug: `resize_token_embeddings` does not update model-specific downstream embeddings/output weights." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "Different save/load regressions: extra_state handling vs unexpected `_is_hf_initialized` during `init_empty_weights`." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "A tokenizer vocab mismatch is not the same as the separate model-loading failure in `sleng-bert`." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "Different model families and different reported failures; no shared code-path bug is evident." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "Different model architectures and failure modes: sparse-layer creation vs FSDP2/shared_kv_states rebuilding." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Torch-version compatibility is too broad here; one is general incompatibility, the other is meta-device loading breakage." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Different training/runtime issues: gradient checkpointing with LoRA vs apex RMSNorm memory leak." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "Different composite-model problems: FSDP2 attention state handling vs `from_config` not unwrapping `text_config`." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Both are quantization/memory related, but the failure mechanisms are different." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "Different Qwen versions and different lifecycle stages: loading in tests vs checkpoint saving correctness." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "Different problems: int4 quantization memory regression vs mmap-related OOM on Strix Halo." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "No meaningful overlap beyond both affecting model loading." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The summary is mostly grounded, but `best_issue_reason` overstates the evidence by calling #45292 the representative of the \u201conly clearly duplicate issue family\u201d even though the packet also contains a clearly duplicate buffer-junk family (#43644/#44534). The resize_token_embeddings canonicalization looks reasonable, and the accepted soft-edge pairs are conservative, but that specific framing is not supported by the cluster." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 6, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:39404", + "issue:40990", + "issue:41720", + "issue:42947", + "issue:43257", + "issue:43299", + "issue:43381", + "issue:43425", + "issue:43611", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43716", + "issue:43828", + "issue:43854", + "issue:43856", + "issue:43901", + "issue:43906", + "issue:43931", + "issue:43957", + "issue:44155", + "issue:44164", + "issue:44265", + "issue:44292", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44509", + "issue:44530", + "issue:44589", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44877", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44928", + "issue:44960", + "issue:45084", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45588", + "issue:45663", + "issue:45684", + "issue:45698", + "issue:45701" + ], + "soft_pairs": [ + "issue:44292|issue:44912", + "issue:42947|issue:43381", + "issue:45684|issue:45698", + "issue:45276|issue:45292", + "issue:43381|issue:44928", + "issue:44164|issue:44756", + "issue:45588|issue:45663", + "issue:41720|issue:44155", + "issue:44479|issue:44877", + "issue:43646|issue:43708", + "issue:43854|issue:44960", + "issue:43957|issue:44756", + "issue:45292|issue:45335", + "issue:44402|issue:44451", + "issue:43645|issue:44164", + "issue:43828|issue:43957", + "issue:43257|issue:44863", + "issue:43425|issue:43828", + "issue:44589|issue:44756", + "issue:44479|issue:45663", + "issue:39404|issue:45701", + "issue:43906|issue:44530", + "issue:43611|issue:43645", + "issue:43931|issue:44863", + "issue:43425|issue:43716", + "issue:44530|issue:45084", + "issue:43901|issue:44509", + "issue:39404|issue:44912", + "issue:44863|issue:45663", + "issue:44265|issue:44898", + "issue:43299|issue:43931", + "issue:44479|issue:44805", + "issue:43856|issue:44756", + "issue:43425|issue:44918", + "issue:44589|issue:45084", + "issue:40990|issue:42947" + ], + "prepared_review_unit_hash": "33e7ca2ffc966510e98149188c689be5840cbd6880fa142255934837fd9f0553", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38389, + "estimated_input_tokens": 9598, + "estimated_eval_tokens": 19452 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a mix of unrelated Transformers/model-specific regression issues that only look similar textually. The only clear duplicate family is the resize_token_embeddings/output_embeddings bug across Gemma4, generic output_embeddings, and t5gemma variants.", + "confidence": 0.93, + "canonical_issue_reason": "issue:45292 is the broadest, most general statement of the one real duplicate family: resize_token_embeddings not updating output embeddings. The other members are model-specific variants of the same underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45292 best represents the cluster because it captures the shared concrete bug without narrowing to a single architecture, making it the best canonical issue for the accepted duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Both are quantization/loading complaints, but they involve different models and different failures (runtime error vs bf16 fallback), so not the same bug." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "PEFT LoRA gradient checkpointing ineffective is a different problem from checkpointing being disallowed in eval mode." + }, + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "Save/pretrained permission propagation and loading the wrong custom module after save are different save/load bugs." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same resize_token_embeddings regression: embeddings/output heads are not updated; 45276 is a Gemma4-specific instance of the broader 45292 bug." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Eval-mode checkpointing restriction is unrelated to the Qwen3.5 RLHF NaN/SDPA fallback issue." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "extra_state save/load handling and Strix Halo mmap OOM are unrelated subsystems." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "Flash-attention sink-less crash and Gemma4 FSDP2 shared_kv_states KeyError are different concrete code-path failures." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "cudaErrorAssert on auto device mapping and AudioFlamingo3 token leakage are unrelated bugs." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Video-input regression for Qwen VL models and strict config blocking granite_speech loading are different issues." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model initialization in notebooks is not the same as resume_from_checkpoint max_steps being miscomputed." + }, + { + "left": "issue:43854", + "right": "issue:44960", + "accept": false, + "reason": "Different model families and failures: GLM-4.7-Flash test loading vs GLM5 loading." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "meta-device loading breakage and mmap/OOM are distinct loading-path problems." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Same underlying resize_token_embeddings bug: output/decoder embeddings are not updated after resizing." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Tokenizer vocab-size mismatch and inability to load ScandiBERT are not the same concrete defect." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Notebook custom-model initialization regression is unrelated to extra_state save/from_pretrained handling." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Autocast dtype mismatch and meta-device loading failures are different code paths." + }, + { + "left": "issue:43257", + "right": "issue:44863", + "accept": false, + "reason": "Qwen3 MoE conversion/load issue and NemotronH checkpoint loading failure are separate model-loading bugs." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Torch 2.10 incompatibility is not the same as the Phi-tiny-MoE autocast dtype mismatch." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Float8 storage lookup failure and Strix Halo mmap OOM are unrelated runtime/storage problems." + }, + { + "left": "issue:44479", + "right": "issue:45663", + "accept": false, + "reason": "Qwen VL video regression and Gemma4 FSDP2 KeyError are different failures." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper return_language pipeline breakage and tokenizer changes across versions are not the same bug." + }, + { + "left": "issue:43906", + "right": "issue:44530", + "accept": false, + "reason": "A reproduction of a prior issue is not the same as the Qwen3.5 PagedAttentionCache linear_attention crash." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "Both are Transformers 5 custom-model regressions, but one is base_model_prefix loading and the other is notebook model initialization; too different to merge." + }, + { + "left": "issue:43931", + "right": "issue:44863", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch and NemotronH checkpoint loading are different model-specific loading failures." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "Torch 2.10 incompatibility is unrelated to the Mistral-3 preprocessor/model dtype mismatch." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash and 'Can't compile non template nodes' are different errors in different subsystems." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "Docs mentioning return_all_scores and docs mentioning removed pipeline tasks are separate documentation issues." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper pipeline return_language and MXFP4 quantization fallback are unrelated." + }, + { + "left": "issue:44863", + "right": "issue:45663", + "accept": false, + "reason": "NemotronH checkpoint loading and Gemma4 shared_kv_states rebuild are different bugs." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "torch.export failures with torch_compilable_check and Perceiver non-default resolution failures are unrelated." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Both involve Qwen3-VL-style models, but one is a MoE loading regression and the other is a weight-shape mismatch; not the same defect." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "Video-input regression and mask/tensor shape mismatch are different symptoms with no clear shared code path." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "Qwen3 MoE training memory usage and Strix Halo mmap OOM are different memory problems in different contexts." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch 2.10 incompatibility is unrelated to Qwen3.5 embedding unpacking in TRL SFT trainer." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage object lookup and template-node compilation errors are different failures." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "High perplexity on gpt-oss-20b and gradient checkpointing ineffective with PEFT LoRA are unrelated issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: issue:45292 is the clearest canonical issue for the resize_token_embeddings/output_embeddings family, and the two accepted soft pairs are well-supported by the issue titles/descriptions. The other soft-edge rejections are conservative and do not overclaim shared bugs." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 7, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:32090", + "issue:34567", + "issue:36032", + "issue:39404", + "issue:40444", + "issue:40990", + "issue:42915", + "issue:43317", + "issue:43404", + "issue:43421", + "issue:43425", + "issue:43452", + "issue:43482", + "issue:43541", + "issue:43582", + "issue:43596", + "issue:43611", + "issue:43632", + "issue:43643", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43883", + "issue:43994", + "issue:44079", + "issue:44164", + "issue:44206", + "issue:44220", + "issue:44451", + "issue:44466", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44671", + "issue:44683", + "issue:44811", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45507", + "issue:45584", + "issue:45685", + "issue:45759" + ], + "soft_pairs": [ + "issue:43425|issue:45685", + "issue:39404|issue:44671", + "issue:44671|issue:44991", + "issue:43994|issue:44079", + "issue:36032|issue:43452", + "issue:44164|issue:45759", + "issue:43596|issue:44683", + "issue:44991|issue:45081", + "issue:32090|issue:39404", + "issue:44164|issue:44479", + "issue:44488|issue:45081", + "issue:40990|issue:43421", + "issue:43582|issue:43632", + "issue:43883|issue:44488", + "issue:43611|issue:43644", + "issue:44206|issue:44220", + "issue:43716|issue:45084", + "issue:44488|issue:44987", + "issue:34567|issue:40444", + "issue:43643|issue:44530", + "issue:44811|issue:45584", + "issue:43644|issue:43646", + "issue:43883|issue:44991", + "issue:36032|issue:42915", + "issue:44451|issue:44987", + "issue:43541|issue:45084", + "issue:43452|issue:43482", + "issue:42915|issue:44530", + "issue:43317|issue:43482", + "issue:44849|issue:45325", + "issue:43404|issue:44466", + "issue:36032|issue:43645", + "issue:45084|issue:45507" + ], + "prepared_review_unit_hash": "98202ebf9447becc4f9e01933fe2749f2f0d39660ec3bb6fc78f8637804b8aca", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37019, + "estimated_input_tokens": 9255, + "estimated_eval_tokens": 18766 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly superficial similarity across unrelated regression reports. The only remotely close pair is 44206/44220, but the evidence is still too thin to treat it as a confirmed duplicate without shared targets, filenames, or explicit linkage.", + "confidence": 0.66, + "canonical_issue_reason": "Issue 44206 is the most concrete audio-preprocessing report in the set, with the clearest crash path and symptoms, but it should only be treated as the best representative of a weakly related pair, not as a proven duplicate anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44206 is the strongest issue candidate overall because it is specific and actionable, and it is the closest thing here to a shared underlying bug among the soft-similarity pairs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43425", + "right": "issue:45685", + "accept": false, + "reason": "Torch version compatibility and an MPS runtime TypeError are different failures." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline language return and CamemBERT masked-LM predictions are unrelated code paths." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "Both are model/tokenizer regressions, but they affect different models and different failure modes." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "Bad SigLIP2 outputs and ModelOutput key assignment are different underlying bugs." + }, + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "T5 tokenizer method conflicts do not match GGUF loading failures." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "Extra_state save/load handling is distinct from composite Qwen config unwrapping." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "ZeRO3/BertModel init and compiled flex_attention on newer torch are unrelated regressions." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer load failure and Mistral regex/backend_tokenizer errors are different issues." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer GPU broadcast NoneType and Whisper return_language are unrelated." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "Extra_state serialization and Qwen video-input regression are different failures." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "These are different model-loading bugs with different subsystems involved." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "Perplexity behavior and runtime special-token post-processing are not the same bug." + }, + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Apple Silicon allocator warmup and `_is_hf_initialized` are unrelated." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Missing tied-weights metadata and a specific model load failure are different issues." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "Both are v5 regressions, but they affect different loading/initialization mechanics." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": false, + "reason": "Both look audio-related, but this packet only shows title-level similarity and no shared target, file, or deterministic evidence of the same crash." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Image preprocessor dtype mismatch and template compilation errors are unrelated." + }, + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "These are different model-specific loading regressions." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "Config field loss with trust_remote_code is not the same as PagedAttentionCache linear_attention errors." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Both are Whisper-related, but decode flag handling and empty-transcription generation are different bugs." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Non-persistent buffer junk and custom model initialization are not the same concrete failure." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Tied-weights attribute errors and tokenizer loading regressions do not share the same root cause." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "T5 tokenizer loading and Qwen3Moe FP8 config failures are unrelated." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Different checkpoint-loading regressions with no shared code-path evidence." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Grouped-mm tracing errors and non-template-node compilation errors are unrelated." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": false, + "reason": "Both involve loading, but one is a broad API break and the other is a model-specific v5 failure." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "FP8 configuration failure and linear_attention cache issues are different Qwen/MoE bugs." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "GPU+CPU offload dequantized loading and GGUF loading failures are different problems." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Qwen3.5 hidden-state behavior and Qwen2.5-VL rope index scaling are distinct regressions." + }, + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Untied lm_head weights and device-dependent serialization of tied weights are not the same bug." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "Tokenizer method conflicts and Jupyter custom-model initialization breakage are different issues." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "Template compilation errors and GraniteMoEHybrid invalid-method calls are unrelated." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The review is mostly grounded, but it overstates the evidence for the one accepted soft pair. The packet only shows title-level soft similarity between issue:44206 and issue:44220, with no deterministic accept or shared targets/files, so calling it a true duplicate / the only real duplicate relationship is stronger than the evidence supports. A more conservative verdict would be to mark that pair as possible or uncertain duplicate rather than accepted." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 8, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:36032", + "issue:40990", + "issue:41553", + "issue:42915", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43645", + "issue:43646", + "issue:43742", + "issue:43818", + "issue:43819", + "issue:43856", + "issue:43994", + "issue:44062", + "issue:44079", + "issue:44220", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44479", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44610", + "issue:44683", + "issue:44756", + "issue:44811", + "issue:44877", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45127", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45663" + ], + "soft_pairs": [ + "issue:43262|issue:43377", + "issue:43646|issue:44534", + "issue:43994|issue:44610", + "issue:44987|issue:44991", + "issue:43322|issue:43540", + "issue:44403|issue:45081", + "issue:43377|issue:43819", + "issue:45216|issue:45357", + "issue:44451|issue:45081", + "issue:43645|issue:44534", + "issue:44265|issue:44610", + "issue:42915|issue:45084", + "issue:44220|issue:44683", + "issue:44987|issue:45081", + "issue:43404|issue:45127", + "issue:41553|issue:43582", + "issue:43632|issue:44756", + "issue:43296|issue:43531", + "issue:44062|issue:44987", + "issue:36032|issue:43742", + "issue:42915|issue:44291", + "issue:43531|issue:44877", + "issue:43611|issue:44534", + "issue:40990|issue:44811", + "issue:42915|issue:44589", + "issue:44479|issue:45325", + "issue:43317|issue:43856", + "issue:43421|issue:44568", + "issue:43818|issue:45357", + "issue:44079|issue:45663" + ], + "prepared_review_unit_hash": "0a58aa9477d8e649a17c32d8692c24ea9f6ab27b98107ebead28dee6ad2225b5", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36779, + "estimated_input_tokens": 9195, + "estimated_eval_tokens": 18646 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab-bag of distinct model- or feature-specific regressions. The only clear duplicate thread is the Qwen3.5 `save_pretrained` checkpoint bug; most other soft pairs are too broad or too different in code path to merge.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 45216 is the best representative of the only clear duplicate thread here: it states the Qwen3.5 `save_pretrained` regression at a higher level, while 45357 narrows it to the visual-encoder key mapping symptom.", + "canonical_pr_reason": null, + "best_issue_reason": "45216 is the strongest anchor for deduping this batch because it captures a concrete, reproducible serialization regression and has a later follow-up that looks like the same bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Different bugs: audio chat-template sampling-rate default vs MIMI batching/padding-mask mismatch." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "One is custom model initialization; the other is non-persistent buffer contents in v5." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Different model families and symptoms: SigLIP2 nonsense outputs vs OmDet-Turbo input-size mismatch." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Both are loading failures, but on different models and likely different code paths; not the same concrete bug." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Llava Next load-time segfault is unrelated to Qwen3OmniMoe video processing ValueError." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Generic load-time noise is not the same as a tokenizer patch crashing on Mistral regex fixing." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "Batch-vs-single padding behavior in MIMI is unrelated to DAC latent/forward mismatch." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Same Qwen3.5 `save_pretrained` serialization regression; 45357 looks like a later, more specific manifestation of the same incorrect checkpoint-key handling." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "Different models and failures: ScandiBERT loading vs Mistral tokenizer regex patch crash." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Notebook custom-model initialization regression is not the same bug as junk-filled non-persistent buffers." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "Export/checking logic failure is unrelated to processor image-size mismatch." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "Qwen3Moe FP8 loading failure and template-node compilation error are different issues." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Audio feature extraction bug and compiled flex_attention regression do not share the same code path." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Different tokenizer/model-loading regressions on different models." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Weight-tying bug in Mistral3 is unrelated to LoRA merge collapse after vocabulary extension." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Bad tokenizer error message vs Apple Silicon allocator TypeError are unrelated." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "_is_hf_initialized / init regression and Strix Halo mmap OOM are unrelated." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "PaddleOCR-VL loading failure and Qwen3-MoE sliding_window issue are distinct model bugs." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "AddedToken special-arg duplication is unrelated to model/tokenizer loading regressions on named checkpoints." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "T5Tokenizer special-token conflict and MobileLLM key error are different loading problems." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "Different failure modes: Qwen3Moe FP8 config issue vs `_is_hf_initialized` argument error." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "MoE sliding_window behavior and strict config preventing granite_speech loading are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "base_model_prefix loading breakage is not the same as non-persistent buffer corruption." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "Perplexity regression and Whisper batch_decode special-token handling are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "Qwen3Moe FP8 failure and missing Float8 storage type are separate low-level issues." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Both involve Qwen VL/video behavior, but the titles point to different symptoms and input modes; not clearly the same concrete bug." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "device_map offload failure and Qwen3 MoE training memory inefficiency are different problems." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Runtime special-token post-processor sync is not the same bug as BOS/EOS not being added for mdeberta-v3-base." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "Video-LLaVA temporal-attention/weight-sharing bug is unrelated to Qwen3.5 save_pretrained key serialization." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key assignment bug and Gemma4 FSDP2 shared_kv_states KeyError are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The only accepted soft-edge pair, 45216/45357, is reasonably supported by the titles as the same Qwen3.5 save_pretrained regression with a more specific follow-up symptom. The remaining false verdicts 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": 2, + "component_count": 16, + "review_unit_index": 9, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:33666", + "issue:35707", + "issue:36032", + "issue:40444", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43296", + "issue:43377", + "issue:43425", + "issue:43454", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43792", + "issue:43828", + "issue:43881", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44265", + "issue:44291", + "issue:44387", + "issue:44403", + "issue:44466", + "issue:44488", + "issue:44534", + "issue:44589", + "issue:44617", + "issue:44683", + "issue:44829", + "issue:44877", + "issue:44898", + "issue:44991", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45381", + "issue:45479", + "issue:45491", + "issue:45538", + "issue:45542", + "issue:45800" + ], + "soft_pairs": [ + "issue:43425|issue:44403", + "issue:42175|issue:45542", + "issue:36032|issue:41720", + "issue:43632|issue:44534", + "issue:33666|issue:40444", + "issue:44683|issue:45020", + "issue:43792|issue:44589", + "issue:44403|issue:44991", + "issue:43482|issue:45538", + "issue:44829|issue:45479", + "issue:40444|issue:45325", + "issue:43482|issue:43525", + "issue:44291|issue:44589", + "issue:43377|issue:45491", + "issue:41720|issue:45081", + "issue:44265|issue:45072", + "issue:42175|issue:43296", + "issue:43611|issue:43646", + "issue:43632|issue:45072", + "issue:44079|issue:44534", + "issue:43531|issue:44291", + "issue:43994|issue:44898", + "issue:43454|issue:44466", + "issue:43881|issue:44488", + "issue:42915|issue:43482", + "issue:43828|issue:44683", + "issue:44079|issue:45072", + "issue:43644|issue:44079", + "issue:43792|issue:43957", + "issue:40444|issue:45381", + "issue:45325|issue:45381", + "issue:43296|issue:45800", + "issue:44387|issue:44617", + "issue:35707|issue:40444", + "issue:44291|issue:44877" + ], + "prepared_review_unit_hash": "7b6cec7d3b4f4222ed8ca3541091fe59e07ce299d9cb8c0a30d984f800b25d19", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38174, + "estimated_input_tokens": 9544, + "estimated_eval_tokens": 19344 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose group of Transformers v5 regression reports, with one clear duplicate-like pair around TensorFlow/backend installation. Most other pairs are only thematically similar, not the same bug.", + "confidence": 0.78, + "canonical_issue_reason": "43632 is the best representative because it captures a concrete Transformers v5 model-initialization regression that sits near several other loading/init failures in the set.", + "canonical_pr_reason": null, + "best_issue_reason": "43632 is also the strongest global issue choice: it is concrete, clearly bug-focused, and closest to the recurring v5 loading/initialization breakages across the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Unrelated failures: one is torch version incompatibility, the other is a loading-noise/reporting issue." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": true, + "reason": "Both describe the same backend packaging bug where installing the torch extra leaves TensorFlow backend support broken/undefined." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Different models and failures: T5 tokenizer load conflict vs Qwen3 auto device-map CUDA assert." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "Both mention v5 regressions, but they are different concrete bugs: `_is_hf_initialized` handling vs non-persistent buffer serialization." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Both are Qwen2-VL/Qwen2.5-VL training issues, but the reported failures are different and not clearly the same code-path bug." + }, + { + "left": "issue:44683", + "right": "issue:45020", + "accept": false, + "reason": "Different scope: torch>=2.9 flex_attention compilation failure vs remote_code loading regressions." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Different root causes and symptoms: Whisper model runtime failure vs missing Float8 storage type." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Unrelated: logging noise on load vs tokenizer loading failure for a specific model." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "Different components: GGUF/model loading under v5 vs CLIPTokenizer max length behavior." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Different training pathologies: flash_attention_3 degenerate training vs binary classification zero-loss issue." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Both concern Qwen2.5-VL, but one is multi-image iterable-dataset finetuning and the other is rope/index scaling for still images." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "Different bugs: GGUF loading failure vs missing `pad_token_id` on Llama4Config." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "Both are loading/init-related, but one is an unexpected `_is_hf_initialized` argument and the other is a float8 storage lookup failure." + }, + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "Different model families and different symptoms: MIMI padding-mask discrepancy vs Gemma3 NaN embeddings with sliding-window attention." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "Different code paths: Qwen3 auto device mapping CUDA assert vs Mistral tokenizer regex patch crash." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "Different failures: torch.export with torch_compilable_check vs bfloat16 dtype mismatches in CI." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "Different packages and failure modes: missing TensorFlow backend in install vs PaddleOCR-VL loading failure in vLLM." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 model-loading regressions, but the concrete bugs differ: base_model_prefix loading vs custom initialization behavior." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "Different issues: initialization-flag breakage vs dtype mismatch in inference." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "Different problems: ModelOutput key assignment vs junk-filled non-persistent buffers." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Different failures: Qwen3-MoE sliding-window behavior vs init_empty_weights argument error." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Different models and symptoms: SigLIP2 nonsensical outputs vs Perceiver resize/interpolation failure." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "Related to lm_head/tied weights, but one is broken tying causing garbage generation and the other is serialization inconsistency by device." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "Different models and failures: glm-4v-9b load failure vs cjvt/sleng-bert load failure." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Different model stacks and causes: FineGrainedFP8Config failure vs GGUF loading failure." + }, + { + "left": "issue:43828", + "right": "issue:44683", + "accept": false, + "reason": "Different numerical-path bugs: autocast dtype mismatch vs compiled flex_attention failure on newer torch." + }, + { + "left": "issue:44079", + "right": "issue:45072", + "accept": false, + "reason": "Different subsystems: ModelOutput key handling vs inference dtype mismatch." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "Different regressions: non-persistent buffer values vs ModelOutput key assignment." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Different loading problems affecting different models and mechanisms." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Same model family, but one is image finetuning with multiple images and the other is video position-id handling; not the same bug." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL positional-id issues, but one is still-image temporal scaling and the other is video vision_position_ids; too different to merge as one bug." + }, + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "Different problems and versions: PaddleOCR-VL loading failure vs torch 2.4.1/transformers 5.8.0 incompatibility." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both end in OOM, but one is reserved-memory growth under int4 quantization and the other is a model-specific CUDA OOM." + }, + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Different generation/training failures: progressive generation with inputs_embeds/past_key_values vs multi-image iterable finetuning." + }, + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "Different init/config issues: unexpected `_is_hf_initialized` argument vs strict config rejecting granite_speech." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative about the soft-edge pairs. The single accepted pair (42175/45542) is plausibly the same TensorFlow-backend packaging bug, and the rest of the verdicts avoid overstating similarity. The representative issue choice is also reasonable for this cluster\u2019s v5 loading/init regressions." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 10, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43404", + "issue:43454", + "issue:43526", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43632", + "issue:43701", + "issue:43708", + "issue:43720", + "issue:43746", + "issue:43792", + "issue:43819", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44186", + "issue:44265", + "issue:44423", + "issue:44466", + "issue:44610", + "issue:44683", + "issue:44756", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45325", + "issue:45357", + "issue:45406", + "issue:45563", + "issue:45584", + "issue:45632", + "issue:45684", + "issue:45701", + "issue:45759" + ], + "soft_pairs": [ + "issue:43526|issue:43883", + "issue:43606|issue:44265", + "issue:43317|issue:43873", + "issue:44186|issue:44265", + "issue:43701|issue:45325", + "issue:43881|issue:44991", + "issue:43632|issue:45563", + "issue:43606|issue:43819", + "issue:43262|issue:43746", + "issue:44610|issue:45325", + "issue:43454|issue:43550", + "issue:43377|issue:44610", + "issue:40990|issue:45584", + "issue:39692|issue:40444", + "issue:43454|issue:45072", + "issue:43377|issue:44898", + "issue:43572|issue:45759", + "issue:44466|issue:45357", + "issue:43296|issue:44991", + "issue:43454|issue:43720", + "issue:43873|issue:45701", + "issue:43792|issue:44683", + "issue:41720|issue:44977", + "issue:45632|issue:45684", + "issue:43404|issue:45357", + "issue:44186|issue:45072", + "issue:44423|issue:45406", + "issue:34634|issue:35707", + "issue:43577|issue:43708", + "issue:35707|issue:41720", + "issue:43873|issue:44756" + ], + "prepared_review_unit_hash": "2081391a90b079d364efa34e72eb3b1b970c6f560c46639256b479c55f15d5ba", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36685, + "estimated_input_tokens": 9172, + "estimated_eval_tokens": 18600 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is a heterogeneous mix of unrelated Transformers issues. The soft edges mostly reflect broad topical similarity (model loading, offloading, multimodal serving, tokenization, training) rather than the same underlying bug or change, so none should be merged.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "Different failures in different models: BEiT Fast label reduction vs Molmo tied-weights attribute error." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "One is Bark CPU-offload device mismatch; the other is a torch.export/torch_compilable_check export failure." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "Both mention offloading/quantization, but one is dequantized model loading on gpu+cpu offload and the other is a broader quantization offload complaint." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Unrelated code paths: LayoutLMv2 tokenization/padding vs torch.export failing on torch_compilable_check." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint resume key mismatch is unrelated to Qwen2.5-VL temporal rope indexing." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "Different models and layers: GLM-4V loading failure vs tokenizer loading regression for est-roberta." + }, + { + "left": "issue:43632", + "right": "issue:45563", + "accept": false, + "reason": "Transformers v5 init-flag regression is unrelated to a paged generate warning about num_return_sequences." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "Bark CPU offload device mismatch is unrelated to DAC.from_latents versus forward mismatch." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio chat-template sample-rate default bug is unrelated to GraniteSpeech PEFT checkpoint loading." + }, + { + "left": "issue:44610", + "right": "issue:45325", + "accept": false, + "reason": "Processor image-size mismatch is unrelated to Qwen2.5-VL rope index scaling." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "lm_head tying bug in AyaVision is unrelated to Bamba torch.compile/SDPA failure." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "MIMI padding-mask batching bug is unrelated to OmDet-Turbo processor size mismatch." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "High perplexity on GPT-OSS is unrelated to Whisper empty-transcription generation after align_special_tokens." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "SigLIP2 docs example errors are unrelated to Qwen2.5-VL finetuning on iterable multi-image batches." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "AyaVision weight-tying bug and SwitchTransformers/TimmWrapper bfloat16 dtype mismatch are different bugs." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Different models and fixes: MIMI batching/padding mask vs Perceiver non-default-resolution interpolation." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "StableLm pad_token_idx regression is unrelated to Qwen composite config unwrapping in from_config." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Serialization of lm_head.weight by device is unrelated to Qwen3.5 visual encoder key saving." + }, + { + "left": "issue:43296", + "right": "issue:44991", + "accept": false, + "reason": "PaddleOCR-VL loading failure in vLLM is unrelated to tokenizer loading for est-roberta." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "Different tied-weight/loading issues: AyaVision lm_head tying vs BitNet packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Quantization/offload behavior is unrelated to a tokenization-version regression." + }, + { + "left": "issue:43792", + "right": "issue:44683", + "accept": false, + "reason": "Whisper runtime failure and compiled flex_attention failure are unrelated code paths." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Both are Qwen-family reports, but one is auto device mapping cudaErrorAssert and the other is flash-attention generation corruption." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Cache-path collisions for trust_remote_code are unrelated to read-only permissions propagated by save_pretrained/register_for_auto_class." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Mistral3 lm_head tying bug is unrelated to Qwen3.5 visual encoder key serialization." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer crash on NER/padding is unrelated to dtype mismatches in bfloat16 inference." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Both affect serve, but the concrete bugs differ: multimodal continuous batching str-to conversion vs Gemma4Processor missing _tokenizer." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "BarkProcessor voice_preset failure is unrelated to progressive generation with inputs_embeds and past_key_values." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "Blip2 dtype loading issue is unrelated to Trainer max_steps miscalculation on checkpoint resume." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Progressive generation with embeddings/past_key_values is unrelated to Qwen3 auto device mapping cudaErrorAssert." + }, + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Quantization offload behavior is unrelated to disabling mmap on Strix Halo to avoid OOM." + } + ] + }, + "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": 16, + "review_unit_index": 11, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:30064", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:43295", + "issue:43316", + "issue:43317", + "issue:43322", + "issue:43388", + "issue:43421", + "issue:43450", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43572", + "issue:43606", + "issue:43720", + "issue:43749", + "issue:43782", + "issue:43873", + "issue:43881", + "issue:44186", + "issue:44292", + "issue:44361", + "issue:44387", + "issue:44479", + "issue:44492", + "issue:44509", + "issue:44534", + "issue:44610", + "issue:44671", + "issue:44743", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44918", + "issue:44991", + "issue:45059", + "issue:45161", + "issue:45406", + "issue:45464", + "issue:45701", + "issue:45704" + ], + "soft_pairs": [ + "issue:43749|issue:44534", + "issue:43526|issue:43881", + "issue:44877|issue:44991", + "issue:42617|issue:42915", + "issue:43450|issue:45701", + "issue:43550|issue:43720", + "issue:36331|issue:42915", + "issue:44610|issue:44898", + "issue:44186|issue:44610", + "issue:36331|issue:42617", + "issue:44387|issue:45704", + "issue:44479|issue:44918", + "issue:43316|issue:44913", + "issue:40990|issue:45701", + "issue:43550|issue:43606", + "issue:43295|issue:45701", + "issue:42915|issue:43317", + "issue:41720|issue:43322", + "issue:41720|issue:45161", + "issue:44671|issue:44743", + "issue:43531|issue:43572", + "issue:44292|issue:44479", + "issue:41720|issue:42617", + "issue:36010|issue:42617", + "issue:44186|issue:44361", + "issue:43540|issue:43782", + "issue:45406|issue:45464", + "issue:41720|issue:43873", + "issue:43421|issue:45701", + "issue:30064|issue:34634", + "issue:44492|issue:44509", + "issue:43388|issue:45059", + "issue:43606|issue:43720", + "issue:42915|issue:43322" + ], + "prepared_review_unit_hash": "b74cd2a20fdcfe702b50c8b5de7b9a8c0c5099ffaff25dd2ec87c9a859754ad2", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "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": "This cluster is mostly an over-grouping of unrelated Transformers issues that only share broad themes like loading, offloading, quantization, tokenizer/processor regressions, or model-specific failures. I would not merge any soft pairs as duplicates.", + "confidence": 0.45, + "canonical_issue_reason": "issue:43873 is the broadest representative of the shared offloading/quantization loading-regression theme and is more general than the model- or processor-specific reports around it.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best single anchor for this cluster because it captures the common loading/offloading problem without being tied to one narrow model or component.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "Different bugs: FSDP CPU RAM efficient loading vs non-persistent buffers being filled with junk." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "Unrelated components: BeitImageProcessorFast label reduction vs glm-4v-9b loading failure." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Config strictness for granite_speech is not the same as tokenizer loading for EMBEDDIA/est-roberta." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "Different failures: 3d_parallel.py runtime issue vs Qwen3Moe FineGrainedFP8Config loading." + }, + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Video processor batching shape bug is separate from broad tokenization changes." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "Different code paths: torch.compile/SDPA in Bamba vs packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "Trainer API signature mismatch is unrelated to Qwen3Moe FP8 configuration/loading." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Both vision-related, but OmDet-Turbo 640px output mismatch is a different bug from Perceiver non-default-resolution handling." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Different models and symptoms: LayoutLMv2 tokenizer padding/truncation crash vs OmDet-Turbo processor size mismatch." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "CustomTrainer.compute_loss() API break is unrelated to the 3d_parallel.py execution failure." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "Int4 quantization OOM and apex FusedRMSNorm memory leak are both memory-related but not the same bug." + }, + { + "left": "issue:44479", + "right": "issue:44918", + "accept": false, + "reason": "Separate problems: video-input regression in VL models vs Qwen3.5 embedding unpacking with TRL SFT trainer." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Both involve config behavior, but Gemma3TextConfig API mismatch and GPTNeoX rotary_pct reload are different fields and failures." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "High perplexity on a model is not the same as tokenizer version-dependent tokenization changes." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Bamba compile/SDPA failure and bark-small CPU offload device mismatch are different model/runtime bugs." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "Processor.tokenizer regression in v4.57.5 is distinct from general tokenization changes across versions." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "Qwen3Moe FP8 config failure is separate from dequantized model loading with gpu+cpu offload." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Different models and symptoms: Qwen3 auto device-map cuda assert vs Llava Next segfault on load." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Qwen3 auto device mapping and GPT-OSS MoE tensor-parallel failure are different distributed execution issues." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "CamemBERT masked LM prediction regression is unrelated to Qwen3.5 recurrent-state caching." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and StableLmConfig pad_token_idx are separate configuration issues." + }, + { + "left": "issue:44292", + "right": "issue:44479", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime error is not the same as the VL video-input regression." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Qwen3 auto device mapping assert is different from the 3d_parallel.py failure." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "GenerationMixin import error and 3d_parallel.py failure are unrelated issues." + }, + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenizers and different failure modes: LayoutLMv2 crash vs MLuke tokenizer AttributeError." + }, + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Qwen3OmniMoe video-input ValueError is unrelated to Qwen3VL weight_only=True loading error." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer in serve is different from Qwen3.5 streaming inference failure." + }, + { + "left": "issue:41720", + "right": "issue:43873", + "accept": false, + "reason": "Qwen3 auto device-map cuda assert is a model-specific crash, not the same as generic offloading-with-quantization behavior." + }, + { + "left": "issue:43421", + "right": "issue:45701", + "accept": false, + "reason": "Runtime post-processor refresh for special tokens is much narrower than general tokenization changes." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Image processor void segmentation maps and BarkProcessor voice_preset are unrelated." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache strategies is not the same as stale docs mentioning removed pipeline tasks." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "Metric gathering label truncation and SAM3 PCS text/bbox behavior are different bugs." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "CPU offload device mismatch and packed-weight unpacking during loading are different root causes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 12, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:33666", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:39692", + "issue:41720", + "issue:42886", + "issue:42915", + "issue:43381", + "issue:43421", + "issue:43525", + "issue:43531", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43746", + "issue:43782", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44062", + "issue:44186", + "issue:44368", + "issue:44451", + "issue:44514", + "issue:44545", + "issue:44589", + "issue:44610", + "issue:44625", + "issue:44671", + "issue:44843", + "issue:44898", + "issue:44918", + "issue:44936", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45588", + "issue:45704", + "issue:45759" + ], + "soft_pairs": [ + "issue:43550|issue:43653", + "issue:44514|issue:44545", + "issue:43531|issue:45759", + "issue:43606|issue:43653", + "issue:43421|issue:43927", + "issue:34567|issue:35707", + "issue:44368|issue:45357", + "issue:43873|issue:45704", + "issue:41720|issue:43746", + "issue:44898|issue:45072", + "issue:44451|issue:44991", + "issue:43421|issue:43653", + "issue:45588|issue:45759", + "issue:44062|issue:44451", + "issue:33666|issue:35707", + "issue:44987|issue:45042", + "issue:42886|issue:44843", + "issue:43531|issue:44589", + "issue:43421|issue:44186", + "issue:39692|issue:44610", + "issue:43381|issue:43844", + "issue:44918|issue:44936", + "issue:45081|issue:45356", + "issue:44625|issue:45479", + "issue:43525|issue:43531", + "issue:43782|issue:45381", + "issue:43994|issue:44671", + "issue:43381|issue:45704", + "issue:34634|issue:41720", + "issue:39692|issue:43873", + "issue:42915|issue:45759" + ], + "prepared_review_unit_hash": "ca3a147c018812ccd5ffabd286071e49f03ba671e0593fe84fb93fad94fb2a9c", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36675, + "estimated_input_tokens": 9169, + "estimated_eval_tokens": 18594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most items are unrelated transformer/model bugs. I found two real duplicate pairs: the Qwen2_5_VL batched padding=False crash, and the HF_HUB_OFFLINE tokenizer load failure tied to _patch_mistral_regex. The rest are distinct code paths and should stay separate.", + "confidence": 0.79, + "canonical_issue_reason": "issue:44843 is the clearest root-cause report for the offline tokenizer regression: it names the offending _patch_mistral_regex/model_info path and the HF_HUB_OFFLINE failure mode, making it the best canonical issue among the accepted duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44843 is the most actionable representative bug in the set because it pinpoints the concrete failing code path rather than only the symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "Different bugs: torch.compile/SDPA model failure vs tokenizer special-token registration/empty decode." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Same processor and same batched apply_chat_template crash when padding=False; wording differs but underlying bug matches." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "Different Qwen config issues: sliding_window behavior vs from_config not unwrapping text_config." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to tokenizer mask-token registration." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "One is a feature/runtime-update request for tokenizer post-processing; the other is a save/load custom token ID regression." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState token counter bug is unrelated to progressive generation with inputs_embeds/past_key_values." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "A warning about tie_word_embeddings is not the same as incorrect visual-encoder keys being saved." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization/offload behavior is unrelated to the T5 apex RMSNorm memory leak." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Qwen3 auto device-map cuda assert and GraniteSpeech PEFT checkpoint loading are different failures." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Perceiver resolution/interpolation bug is unrelated to bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Different artifacts and failure modes: model loading vs tokenizer loading for different repositories." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Runtime special-token/post-processor update feature request is not the same as BigBirdTokenizer mask-token decode failure." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention sink-less crash and Qwen config unwrapping are separate code paths." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "AddedToken special-argument conflict is unrelated to loading a specific model repo." + }, + { + "left": "issue:33666", + "right": "issue:35707", + "accept": false, + "reason": "Multi-GPU training for Qwen2-VL is unrelated to progressive generation with inputs_embeds/past_key_values." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "Model load regression and PIL image-processor torchvision dependency are different problems." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": true, + "reason": "Same offline tokenizer failure: 42886 reports the symptom, 44843 identifies the exact _patch_mistral_regex/model_info cause." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Sliding_window behavior for Qwen3-MoE is unrelated to missing Float8 storage support." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Tokenizer post-processor update behavior is not the same as LayoutLMv2Tokenizer crashes on NER/padding." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "SigLIP2 doc example errors and OmDet-Turbo processor size mismatch are distinct issues." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Gradient checkpointing in eval mode is unrelated to gradient blow-up under ZeRO-3." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "TRL SFT input-embedding unpacking and trainer.evaluate() after train are different trainer bugs." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Both touch fix_mistral_regex, but one is a loading crash and the other is a Kimi-K2.5 codec/warning regression; not the same bug." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and single_label_classification zero-loss behavior are separate classification issues." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Missing pad_token_id on Llama4Config is unrelated to Qwen3-MoE sliding_window behavior." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Qwen3VL weight_only loading and Qwen2.5-VL video vision_position_ids are different multimodal bugs." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "SigLIP2 incorrect outputs and CamemBERT masked-LM regression are different model-specific failures." + }, + { + "left": "issue:43381", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing eval-mode restriction is unrelated to the T5 apex RMSNorm memory leak." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "BarkProcessor voice_preset and Qwen3 auto device-mapping cuda assert are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "SigLIP2 documentation bugs are unrelated to quantization/offloading behavior." + }, + { + "left": "issue:42915", + "right": "issue:45759", + "accept": false, + "reason": "Qwen3Moe FineGrainedFP8Config failure is not the same as Qwen 3.5/3.6 text_config unwrapping." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet, and the soft-edge calls stay conservative. The two accepted duplicate pairs are well-supported by the issue titles, and the remaining rejections avoid overclaiming sameness across clearly different bugs." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 13, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:42915", + "issue:43334", + "issue:43335", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43575", + "issue:43577", + "issue:43596", + "issue:43618", + "issue:43643", + "issue:43653", + "issue:43697", + "issue:43716", + "issue:43720", + "issue:43746", + "issue:43756", + "issue:43782", + "issue:43818", + "issue:43819", + "issue:43873", + "issue:44008", + "issue:44062", + "issue:44077", + "issue:44186", + "issue:44220", + "issue:44448", + "issue:44488", + "issue:44561", + "issue:44625", + "issue:44661", + "issue:44671", + "issue:44811", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45183", + "issue:45381", + "issue:45479", + "issue:45704", + "issue:45759" + ], + "soft_pairs": [ + "issue:43596|issue:44220", + "issue:45183|issue:45479", + "issue:43618|issue:43697", + "issue:43540|issue:44008", + "issue:44220|issue:44661", + "issue:43720|issue:43819", + "issue:44077|issue:44625", + "issue:43575|issue:45704", + "issue:30064|issue:36010", + "issue:43335|issue:43756", + "issue:43653|issue:43720", + "issue:43334|issue:45759", + "issue:43716|issue:44008", + "issue:42915|issue:43746", + "issue:43746|issue:45381", + "issue:43746|issue:43873", + "issue:43550|issue:45072", + "issue:44186|issue:44898", + "issue:43550|issue:43819", + "issue:43421|issue:44811", + "issue:44843|issue:45081", + "issue:43577|issue:43818", + "issue:44062|issue:44488", + "issue:43531|issue:43643", + "issue:43643|issue:44877", + "issue:43653|issue:43819", + "issue:44448|issue:44671", + "issue:43540|issue:45381", + "issue:43782|issue:44977", + "issue:44561|issue:45020" + ], + "prepared_review_unit_hash": "0fa4e38e93c91273f6d19c3978e8ce17cab74ab9e3b5c4583916b69e401e5277", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 37048, + "estimated_input_tokens": 9262, + "estimated_eval_tokens": 18780 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The candidate pairs are mostly superficially similar but describe different models, components, or failure modes; none look like safe duplicates or mergeable PR-equivalents. The cluster is heterogeneous, with a few repeated themes around loading/configuration regressions, but the concrete bugs differ.", + "confidence": 0.95, + "canonical_issue_reason": "issue:45020 is the broadest, most central tracker in this set: it describes a general class of recent transformers regressions affecting remote-code models, with active discussion and open status, making it the best canonical issue candidate.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the best global issue candidate because it is high-level, user-facing, and broad enough to subsume related remote-code breakages better than the more model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "Both are runtime errors, but one is a Deepspeed Zero3/BertModel indexing bug and the other is an audio feature extraction issue; different code paths and fixes." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "These are both validation/shape-related, but one is about input_ids vs inputs_embeds XOR checks and the other is about single-label classification degenerating to zero loss; not the same bug." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "CLIP attentions assignment and RTDetrV2 output drift in v5 are unrelated model behaviors with different failure surfaces and likely different fixes." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "One is a Qwen3OmniMoe video-input ValueError, the other is a Gemma 3n variable-name collision in forward(); different models and distinct defects." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Audio fbank extraction and add-new-model-like tokenizer mapping behavior are unrelated subsystems; no shared underlying bug." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "Both mention BitNet-like loading issues, but one is about packed weights during accelerate loading and the other is DAC.from_latents vs forward with missing STE; not the same change." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "PatchTSMixer post_init policy and Qwen3.5 num_labels propagation are separate configuration issues affecting different models." + }, + { + "left": "issue:43575", + "right": "issue:45704", + "accept": false, + "reason": "Qwen2 TP loading OOM and T5 apex.FusedRMSNorm memory leak are both memory-related, but they concern different models and different failure mechanisms." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Image processor void segmentation map handling and GenerationMixin import failure are unrelated bugs in different parts of the stack." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "SwitchTransformers sparse-layer creation and Smollm3 RoPE layer dropping are both model-architecture issues, but they are not the same underlying defect." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "BigBirdTokenizer special-token decode behavior and BitNet packed-weight loading are separate tokenizer vs model-loading regressions." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "Qwen3-VL missing pad_token_id and Qwen 3.5/3.6 from_config text_config unwrapping are related to composite configs, but they describe different attribute propagation problems." + }, + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Mistral-3 image preprocessor dtype mismatch and Gemma 3n audio_mel_mask collision are unrelated multimodal issues with different root causes." + }, + { + "left": "issue:42915", + "right": "issue:43746", + "accept": false, + "reason": "Qwen3Moe FineGrainedFP8Config failure and GraniteSpeech PEFT adapter loading are distinct model-loading problems." + }, + { + "left": "issue:43746", + "right": "issue:45381", + "accept": false, + "reason": "GraniteSpeech PEFT checkpoint loading and Qwen2.5-VL video vision_position_ids are unrelated; one is checkpoint restoration, the other is video position indexing." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "GraniteSpeech local checkpoint loading and offloading with quantization are different loading paths and different bugs." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Both involve Bamba/Transformer inference and dtype/compile issues, but one is torch.compile with SDPA and the other is bfloat16 mismatches in different models; not mergeable into one fix." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "LayoutLMv2Tokenizer padding/truncation crashes and Perceiver interpolation failures are unrelated preprocessing/model-usage bugs." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Bamba torch.compile+SDPA failure and DAC.from_latents mismatch are different code paths and different model families." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Runtime special-token post-processor updates and Whisper batch_decode skip_special_tokens handling are tokenizer/processor behaviors, but they are different mechanisms and fixes." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Mistral regex offline hub access and Mistral tokenizer backend_tokenizer crash both touch Mistral loading, but one is hub/network access and the other is tokenizer API compatibility; not the same bug." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "BLIP2 dtype propagation and Video-LLaVA missing temporal attention/shared weights are separate multimodal model problems." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "AddedToken special-argument duplication and failing to load a specific model checkpoint are unrelated tokenizer construction vs model loading bugs." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Sliding_window behavior in Qwen3-MoE and trust_remote_code missing fields in AutoConfig are not the same underlying issue." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "Both are config-loading regressions, but one is trust_remote_code field loss and the other is strict config blocking granite_speech; different failure modes." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and DAC latent reconstruction mismatch are unrelated tokenizer vs audio codec issues." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Pegasus v4/v5 output drift and CamemBERT masked LM prediction changes are both version regressions, but they affect different model architectures and behaviors." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Qwen3OmniMoe video ValueError and Qwen2.5-VL video position id bug are both video-related, but they are different models and different internal failures." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3VL weight_only loading error and Qwen3.5 flash-attention generation issue are separate Qwen-family bugs with different triggers." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "Both involve remote-code compatibility regressions, but one is specifically about is_torch_fx_available removal and the other is a broader recent remote_code breakage report; related theme, not the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 14, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:30064", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:42175", + "issue:43262", + "issue:43322", + "issue:43329", + "issue:43404", + "issue:43452", + "issue:43526", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43644", + "issue:43650", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43749", + "issue:43782", + "issue:43819", + "issue:43825", + "issue:43827", + "issue:43906", + "issue:43994", + "issue:44060", + "issue:44186", + "issue:44368", + "issue:44442", + "issue:44485", + "issue:44509", + "issue:44610", + "issue:44671", + "issue:44805", + "issue:44821", + "issue:44857", + "issue:44912", + "issue:44936", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45406", + "issue:45538", + "issue:45600", + "issue:45685", + "issue:45701" + ], + "soft_pairs": [ + "issue:43825|issue:43994", + "issue:43322|issue:45685", + "issue:44821|issue:45084", + "issue:30064|issue:36331", + "issue:43825|issue:43827", + "issue:44485|issue:44509", + "issue:44186|issue:44442", + "issue:43577|issue:43582", + "issue:43606|issue:45072", + "issue:34634|issue:36010", + "issue:43452|issue:44991", + "issue:44805|issue:44936", + "issue:44610|issue:45072", + "issue:42175|issue:45081", + "issue:43720|issue:45072", + "issue:45081|issue:45406", + "issue:44671|issue:45701", + "issue:43650|issue:43906", + "issue:44060|issue:44368", + "issue:43404|issue:43782", + "issue:43582|issue:44857", + "issue:43262|issue:43819", + "issue:43526|issue:43697", + "issue:43644|issue:43749", + "issue:45084|issue:45538", + "issue:43329|issue:45600", + "issue:43653|issue:44912" + ], + "prepared_review_unit_hash": "8fb7d7f9854c833947ca3feca987cb54ac82c115d19e066ca27c590703c34df4", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35997, + "estimated_input_tokens": 9000, + "estimated_eval_tokens": 18256 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: most items are unrelated bug reports that only share broad transformer/model-loading vocabulary. None of the soft pairs look like the same concrete bug/change, so I would not deduplicate any of them.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "v5 pipeline error-message cleanup vs a SigLIP2 model quality/reporting issue; different code paths and symptoms." + }, + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Llava Next loading segfault vs an MPS histogram/int dtype limitation; unrelated failures." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "AutoImageProcessor URL loading vs tokenizer template compilation; different subsystems and fixes." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Image processor void segmentation maps vs trainer.compute_loss signature mismatch; unrelated." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "Both concern v5 pipeline removals, but one is an error message and the other is docs cleanup, not the same bug." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "GLM-5 RoPE implementation discussion vs docs mentioning removed pipeline tasks; unrelated." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "Two different tokenizer crashes in different classes and load paths; too far apart to be the same bug." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "BLIP2 dtype persistence bug vs Apple Silicon allocator TypeError; unrelated." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "CPU offload device mismatch vs bfloat16 dtype mismatch; different concrete failures." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "BarkProcessor voice preset bug vs GenerationMixin import error; unrelated." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "gguf_file loading regression vs a specific tokenizer load failure in v5; different underlying causes." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "Mask shape IndexError vs trainer.evaluate() after train lifecycle bug; unrelated." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Processor output size mismatch vs dtype mismatch; not the same issue." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Packaging/backend install problem vs Mistral tokenizer regex patch crash; unrelated." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Packed-weight unpacking during accelerate loading vs bfloat16 dtype mismatch; different bugs." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Different attribute errors in different processor/tokenizer classes; not one shared fix." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "Specific CamemBERT MLM regression vs a broad tokenization-version complaint; too vague to merge." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "Placeholder issue vs isolated reproduction of another issue; no common underlying bug." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Both mention Qwen tied-weight warnings, but one is a concrete wrong tying bug and the other is a config warning in fine-tuning." + }, + { + "left": "issue:43404", + "right": "issue:43782", + "accept": false, + "reason": "Mistral3 tied-weights bug vs Qwen3VL weight_only loading error; unrelated." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError vs CUDA AMP loss crash; unrelated." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio chat-template sampling-rate default vs DAC latent-forward mismatch; different models and code paths." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug vs RTDetrV2 output drift in v5; not the same concrete defect." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Non-persistent buffer initialization issue vs FSDP RAM-efficient loading regression; related to loading broadly, but not the same bug." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "Template compilation error vs CLIPTokenizer model_max_length constant; unrelated." + }, + { + "left": "issue:43329", + "right": "issue:45600", + "accept": false, + "reason": "Multimodal token counting video-branch bug vs stale auto_mappings config references; different parts of the codebase." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "BigBirdTokenizer special-token registration bug vs MXFP4 quantization fallback; 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": 16, + "review_unit_index": 15, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:34567", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:40990", + "issue:42915", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43450", + "issue:43452", + "issue:43454", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43582", + "issue:43638", + "issue:43643", + "issue:43645", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43854", + "issue:43927", + "issue:43931", + "issue:44291", + "issue:44479", + "issue:44554", + "issue:44568", + "issue:44589", + "issue:44610", + "issue:44811", + "issue:44849", + "issue:44912", + "issue:44936", + "issue:44977", + "issue:45042", + "issue:45356", + "issue:45357", + "issue:45488", + "issue:45701", + "issue:45759" + ], + "soft_pairs": [ + "issue:43638|issue:44936", + "issue:43746|issue:44811", + "issue:38175|issue:44610", + "issue:44589|issue:45042", + "issue:43854|issue:44977", + "issue:43454|issue:43577", + "issue:36010|issue:43645", + "issue:43526|issue:44554", + "issue:44568|issue:45488", + "issue:34634|issue:36331", + "issue:43819|issue:43927", + "issue:43257|issue:43329", + "issue:39692|issue:44912", + "issue:43720|issue:44610", + "issue:43749|issue:44291", + "issue:43550|issue:44912", + "issue:43262|issue:44811", + "issue:44568|issue:45356", + "issue:43531|issue:45042", + "issue:43643|issue:45759", + "issue:36010|issue:36331", + "issue:34567|issue:40990", + "issue:43329|issue:43450", + "issue:43329|issue:44479", + "issue:44849|issue:45701", + "issue:43257|issue:43931", + "issue:43697|issue:45357", + "issue:43452|issue:43582", + "issue:42915|issue:43742" + ], + "prepared_review_unit_hash": "ad9a8a7a57f9faab700aed9bad61a1b5faf90dd128a667a4e281c5a830361de2", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36722, + "estimated_input_tokens": 9181, + "estimated_eval_tokens": 18618 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous. The only accepted duplicate edge is the Qwen video regression pair (43329 + 44479); the Qwen3 MoE loading issue (43257) is not video-related and should not be grouped with it.", + "confidence": 0.89, + "canonical_issue_reason": "issue:44479 is the best canonical issue for the only clear duplicate family: it describes the broad user-facing video-input regression across multiple Qwen VLM variants, while issue:43329 is the narrower helper-level diagnosis.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44479 is the strongest global issue anchor because it captures the broadest symptom set in the accepted family and is the most user-visible report in the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Different training failures: deepspeed/zero3 index error vs trainer.evaluate() breaking after train()." + }, + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "Different audio bugs: GraniteSpeech PEFT checkpoint loading vs Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Different models and symptoms: SigLIP2 zero probabilities vs OmDet-Turbo processor size mismatch." + }, + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Unrelated failures: float8 storage lookup error vs PIL backend image processors requiring torchvision." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "Different model families and failure modes: GLM-4.7 load issue vs Qwen3.5 flash-attention generation regression." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "Different model-specific bugs: AyaVision weight tying vs Blip2 dtype preservation." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "Both are runtime/loading-related, but one is a GenerationMixin import break and the other is a Jupyter custom-model init regression." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "Completely different subsystems: Beit image label reduction vs MPS attention correctness." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different tokenizer classes and different concrete behaviors." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "No shared code path: BarkProcessor voice_preset vs CustomTrainer.compute_loss signature mismatch." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "Different bugs: DAC latent reconstruction mismatch vs DiaConfig losing custom token IDs on save/load." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Different Qwen3-related problems: MoE weight conversion/loading with accelerate+deepspeed vs multimodal video token counting helper bug." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "Both mention quantization/docs, but they are different models and different failure modes." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "Unrelated: BitNet accelerate-loading packed weights vs OmDet-Turbo processor input-size mismatch." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "Both are loading-related, but one is FSDP CPU RAM efficient loading and the other is init_empty_weights/_is_hf_initialized handling." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "Different problems: torch.compile with SDPA vs MXFP4 quantization fallback." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Both involve audio processors, but one is chat-template sampling-rate defaults and the other is batch_decode special-token stripping." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Tokenizer regressions in different tokenizers with different concrete breakages." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Completely different areas: Qwen3-MoE sliding_window vs PIL backend torchvision dependency." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "Both are config-loading issues, but one is trust_remote_code missing fields and the other is composite Qwen config text_config unwrapping." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "Unrelated: GenerationMixin import failure vs CustomTrainer.compute_loss kwarg mismatch." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "Trainer token-count bookkeeping is unrelated to the GPT-OSS perplexity regression." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Both are video-related, but they describe different bugs: missing helper variables in multimodal token counting vs batched video processor shape mismatch." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": true, + "reason": "Same underlying Qwen video regression: 43329 identifies the broken video-path helper, and 44479 reports the broader user-visible failure across Qwen VLM variants." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Similar tokenization theme, but the reports concern different version/regression contexts and different failure descriptions." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Different Qwen3 issues: MoE loading/conversion failure vs Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "Different regression types: RTDetrV2 output divergence vs Qwen3.5 visual encoder key saving bug." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "Different loading/runtime bugs: gguf_file path handling vs Apple Silicon caching allocator warmup TypeError." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Different Qwen3-MoE-related failures: FineGrainedFP8Config vs MobileLLM key error during load." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster summary is mostly grounded, but at least one soft-edge rationale overstates shared evidence. In particular, the `issue:43257` vs `issue:43329` note says \"Both are video-related,\" but `43257` is about Qwen3 MoE weight conversion/loading with accelerate+deepspeed and is not video-related per the packet. That makes the verdict rationale non-conservative. The accepted Qwen video pair (`43329` + `44479`) looks reasonable, but the review should be rejected because not every soft-edge explanation stays strictly within the packet evidence." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 16, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:33357", + "issue:34567", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43257", + "issue:43317", + "issue:43452", + "issue:43493", + "issue:43526", + "issue:43540", + "issue:43577", + "issue:43643", + "issue:43646", + "issue:43696", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43824", + "issue:43906", + "issue:43927", + "issue:43994", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44560", + "issue:44737", + "issue:44756", + "issue:44821", + "issue:44898", + "issue:44912", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45137", + "issue:45375", + "issue:45381", + "issue:45405", + "issue:45479", + "issue:45538", + "issue:45588" + ], + "soft_pairs": [ + "issue:41720|issue:43540", + "issue:44291|issue:45042", + "issue:33357|issue:34567", + "issue:44479|issue:45381", + "issue:43317|issue:45538", + "issue:44987|issue:45538", + "issue:43540|issue:44560", + "issue:42915|issue:44821", + "issue:43927|issue:44912", + "issue:43540|issue:44479", + "issue:43257|issue:43756", + "issue:42915|issue:45538", + "issue:43577|issue:45072", + "issue:43257|issue:43696", + "issue:45137|issue:45405", + "issue:38175|issue:44898", + "issue:44737|issue:45588", + "issue:44062|issue:44991", + "issue:43742|issue:44291", + "issue:43906|issue:45381", + "issue:43646|issue:43749", + "issue:43749|issue:44756", + "issue:34567|issue:36331", + "issue:44062|issue:45081", + "issue:43643|issue:45375", + "issue:43493|issue:43994", + "issue:36010|issue:42175", + "issue:43452|issue:44488", + "issue:42175|issue:43824", + "issue:43526|issue:45479", + "issue:44291|issue:44821" + ], + "prepared_review_unit_hash": "06d530cb1f590d6d7a7f6bdfd8a7a58d6f0c7f0e17ff0b640088a22f75effb08", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36566, + "estimated_input_tokens": 9142, + "estimated_eval_tokens": 18540 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a grab bag of unrelated closed issues across loading, tokenizer, dtype, quantization, and multimodal/video regressions. The only near-duplicate-looking subgroup is the Qwen video-input regression pair; everything else is too different to merge as duplicates.", + "confidence": 0.81, + "canonical_issue_reason": "issue:44479 is the broadest and most central bug report here: it describes a v5.3.0 video-input regression across several Qwen VL variants, and likely subsumes the more specific Qwen2.5-VL video-position-id report.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44479 is the best representative issue for the cluster because it covers the clearest shared regression theme and is the only one that plausibly generalizes to another item in the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "Both mention Qwen3, but one is auto device mapping cuda asserts and the other is a video-input ValueError; different failures and code paths." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "One is an init_empty_weights / unexpected argument bug, the other is a PIL image processor torchvision dependency regression." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "MacOS bus error with CLIP is unrelated to TrainerState.num_input_tokens_seen not updating." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "Both are Qwen VL video-input regressions in v5.3.0, and the newer report looks like a specific symptom of the broader video-position-id regression." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "Dequantized model loading with device_map=auto is unrelated to CLIPTokenizer's model_max_length bug." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Loading physical-intelligence/fast on transformers>=5.1.0 is a model-loading regression, not the CLIPTokenizer max-length issue." + }, + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "Both are Qwen video-related, but one is a video-input processing ValueError and the other is a StopIteration during Qwen3-vl-embedding loading; not clearly the same concrete bug." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "FineGrainedFP8Config failure for Qwen3Moe is unrelated to AutoImageProcessor-from-URL loading." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "DiaConfig save/load token ID loss is unrelated to MXFP4 quantization falling back to bf16." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Related theme, but one is a specific video-processing ValueError in Qwen3OmniMoe while the other is a broader Qwen VL video regression; not enough evidence they are the same bug." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "MOE weight conversion with accelerate+deepspeed is unrelated to Smollm3 RoPE layer dropping." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "Qwen3Moe FineGrainedFP8Config failure is unrelated to CLIPTokenizer max-length handling." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Both involve dtype, but one is BLIP2 loading with half/bfloat16 and the other is SwitchTransformers/TimmWrapperModel inference mismatch; different concrete failures." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "MOE weight conversion under accelerate+deepspeed is unrelated to GPT-oss-20b CUDA OOM." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque underflow is unrelated to the MIN_PEFT_VERSION bump." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Siglip2 zero probabilities and Perceiver non-default-resolution failure are different model-specific correctness bugs." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "XLNet CPU arange-device bug is unrelated to flash_attention.py crashing on s_aux=None." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "AddedToken special-argument TypeError is unrelated to tokenizer loading failure for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "MobileLLM key error and init_empty_weights unexpected _is_hf_initialized argument are different loading failures." + }, + { + "left": "issue:43906", + "right": "issue:45381", + "accept": false, + "reason": "An isolated reproduction of another issue is not the same as the Qwen2.5-VL video-position-id regression." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Custom model initialization breakage in Transformers 5.0.0 is unrelated to FSDP_CPU_RAM_EFFICIENT_LOADING." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "Both are memory/loading related, but FSDP CPU RAM efficient loading and Strix Halo mmap OOM are different problems." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "TrainerState token counting not updating is unrelated to CustomTrainer.compute_loss receiving an unexpected keyword argument." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "AddedToken special kwarg duplication is unrelated to the Mistral regex patch crashing on backend_tokenizer." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "AutoConfig missing fields with trust_remote_code=True is unrelated to a missing deepstack_visual_indexes field in Qwen3_5MoeVisionConfig." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": false, + "reason": "SigLIP2 JAX-vs-HF discrepancy and nonsensical AutoModel/pipeline results are different correctness reports, not the same bug." + }, + { + "left": "issue:36010", + "right": "issue:42175", + "accept": false, + "reason": "GenerationMixin import failure is unrelated to TensorFlow being absent from the torch extra backend." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "gguf_file breaking AutoTokenizer/AutoModel loading is unrelated to the cjvt/sleng-bert load failure." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "Missing TensorFlow backend in a torch install is unrelated to Qwen2_5_VLForConditionalGeneration import failure." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels returning one label is unrelated to the zero-loss classification bug." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "Unexpected _is_hf_initialized argument during init_empty_weights is unrelated to loading AutoImageProcessor from URL." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall and conservatively phrased. The only clearly plausible duplicate/subsumption in the packet is the Qwen video-input pair (44479/45381); the rest are distinct by title and failure mode. Minor note: phrases like \"unrelated\" are a bit absolute, but not overstated here given the available evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 17, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29127", + "issue:34689", + "issue:38175", + "issue:41762", + "issue:42175", + "issue:43295", + "issue:43452", + "issue:43531", + "issue:43577", + "issue:43582", + "issue:43643", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43756", + "issue:43819", + "issue:44466", + "issue:44484", + "issue:44488", + "issue:44514", + "issue:44561", + "issue:44568", + "issue:44589", + "issue:44625", + "issue:44704", + "issue:44792", + "issue:44811", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44936", + "issue:44964", + "issue:44977", + "issue:45005", + "issue:45042", + "issue:45081", + "issue:45216", + "issue:45290", + "issue:45325", + "issue:45356", + "issue:45381", + "issue:45405", + "issue:45600" + ], + "soft_pairs": [ + "issue:44792|issue:44964", + "issue:28282|issue:34689", + "issue:45042|issue:45081", + "issue:45405|issue:45600", + "issue:43756|issue:45325", + "issue:43742|issue:44589", + "issue:28282|issue:29127", + "issue:44484|issue:44704", + "issue:43531|issue:44821", + "issue:43697|issue:45216", + "issue:44514|issue:45381", + "issue:43643|issue:44561", + "issue:44625|issue:44877", + "issue:43582|issue:44977", + "issue:43452|issue:43577", + "issue:43720|issue:44898", + "issue:43452|issue:44843", + "issue:44514|issue:45290", + "issue:44466|issue:45005", + "issue:43819|issue:44811", + "issue:42175|issue:44488", + "issue:43295|issue:45356", + "issue:41762|issue:44936", + "issue:43653|issue:44568", + "issue:24643|issue:29127", + "issue:38175|issue:43295" + ], + "prepared_review_unit_hash": "7f2d5c1e3fc5bc229cc7db2b39f5871f5203dcbc7392739182cd17c70d3bf2f1", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 35262, + "estimated_input_tokens": 8816, + "estimated_eval_tokens": 17888 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Nearly all soft pairs are only superficially related by subsystem or model family and should stay separate. The one plausible duplicate is the tied-weights serialization/save_pretrained regression pair (44466/45005).", + "confidence": 0.84, + "canonical_issue_reason": "issue:44466 is the clearest concrete report of the v5 tied-weights serialization bug, with device-dependent behavior and a specific `lm_head.weight` symptom.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44466 is the best global issue candidate because it is specific, reproducible, and appears to capture the underlying tied-weights regression more precisely than the broader follow-up report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Both are multimodal/model-loading problems, but they concern different models and failures: a Janus test case versus Phi-4 multimodal loadability." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "One is a missing-PyTorch ImportError, the other is a Llama 3.2 vision loading regression; the symptoms and code paths are unrelated." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "These are different bugs: torchvision dependency handling for PIL image processors versus a Mistral regex/tokenizer crash." + }, + { + "left": "issue:45405", + "right": "issue:45600", + "accept": false, + "reason": "A dependency-version bump and a removed-config reference in auto_mappings are separate CI/issues, not the same bug." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Both touch RoPE/position handling in different models, but one is about missing RoPE layers in Smollm3 and the other about Qwen2.5-VL temporal position scaling." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "MobileLLM key lookup failure and a Float8 storage TypeError are distinct loading/runtime failures." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "These involve completely different topics: PyTorch import detection versus clearer error messaging for LayoutLMv3 box inputs." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "Max shard size defaults and processor kwargs propagation are unrelated save/load API issues." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and loading AutoImageProcessor from URL are not the same underlying problem." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "RTDetrV2 output drift in v5 and Qwen3.5 checkpoint serialization are separate regressions." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL-related, but one is a chat-template batching crash and the other is incorrect video vision position IDs; different code paths." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Missing fields from `AutoConfig` with `trust_remote_code` and removal of `is_torch_fx_available` breaking remote-code models are different failures." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "A Qwen3.5 `num_labels` propagation bug is not the same as strict config blocking granite_speech loading." + }, + { + "left": "issue:43582", + "right": "issue:44977", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError and Qwen3.5 flash-attention generation issues are unrelated." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "GGUF loading breakage and BLIP-2 dtype propagation are different model-loading bugs." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "BitNet packed-weight loading and Perceiver non-default-resolution failures do not share the same code-path problem." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "GGUF loading failures and offline Mistral-regex hub calls are different tokenizer/model-loading regressions." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Both involve `apply_chat_template`, but the reported failures are different conditions: batching/padding versus assistant tool-call messages with no content." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": true, + "reason": "Both point to the same tied-weights/save_pretrained regression in v5; 44466 is the more specific device-dependent serialization report." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "DAC latent conversion mismatch and Whisper batch_decode special-token handling are unrelated bugs." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "A missing TensorFlow backend from a pip extra and a model-loading failure for `cjvt/sleng-bert` are not the same issue." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Processor/tokenizer regression and Kimi-K2.5 codec/fix_mistral_regex regression are different problems." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "DeepSpeed ZeRO-3 loading IndexError and trainer.evaluate() after train are separate trainer/model loading issues." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "BigBird special-token registration and MdeBERTa add_special_tokens behavior are different tokenizer bugs." + }, + { + "left": "issue:24643", + "right": "issue:29127", + "accept": false, + "reason": "A DeepSpeed training `weight` dimensionality error and a LayoutLMv3 box-info message issue are unrelated." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 and a processor.tokenizer regression are not the same bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet, and the soft-edge rejections are conservative. The 44466/45005 pair is the only one that plausibly describes the same tied-weights/save_pretrained regression, and the summary does not overstate the weaker matches. " + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 18, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:28282", + "issue:30064", + "issue:32090", + "issue:33666", + "issue:34567", + "issue:35141", + "issue:36010", + "issue:41720", + "issue:43122", + "issue:43262", + "issue:43295", + "issue:43329", + "issue:43452", + "issue:43493", + "issue:43504", + "issue:43525", + "issue:43531", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43825", + "issue:43906", + "issue:43927", + "issue:44361", + "issue:44442", + "issue:44451", + "issue:44464", + "issue:44479", + "issue:44610", + "issue:44743", + "issue:44821", + "issue:44843", + "issue:44913", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45325", + "issue:45356", + "issue:45440", + "issue:45588" + ], + "soft_pairs": [ + "issue:44361|issue:44442", + "issue:43653|issue:44361", + "issue:30064|issue:33666", + "issue:44464|issue:44743", + "issue:43577|issue:43697", + "issue:44843|issue:44991", + "issue:44479|issue:44936", + "issue:41720|issue:43906", + "issue:44977|issue:45356", + "issue:43688|issue:45440", + "issue:43531|issue:43742", + "issue:43653|issue:43927", + "issue:43577|issue:45588", + "issue:44464|issue:44977", + "issue:43550|issue:43927", + "issue:43550|issue:44361", + "issue:32090|issue:34567", + "issue:43504|issue:43550", + "issue:43504|issue:43720", + "issue:43122|issue:43906", + "issue:32090|issue:33666", + "issue:28282|issue:35141", + "issue:43504|issue:44610", + "issue:43452|issue:44451", + "issue:43688|issue:43749", + "issue:43329|issue:45325", + "issue:43742|issue:44821", + "issue:33666|issue:36010", + "issue:41720|issue:44464", + "issue:43262|issue:43295", + "issue:43756|issue:44913", + "issue:43493|issue:43825", + "issue:43525|issue:43572" + ], + "prepared_review_unit_hash": "39cf0dd8dbb710d492916152516702fb311c2ed421e9b5057d46f72555346f74", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37169, + "estimated_input_tokens": 9293, + "estimated_eval_tokens": 18842 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: it mixes unrelated bug reports across tokenizers, config loading, multimodal processors, training, generation, and model-specific regressions. The soft pairs mostly share only broad subsystem similarity, not the same concrete defect or fix path, so none should be merged as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Different failures in different tokenizers: MLukeTokenizer task handling vs AutoTokenizer loading FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "BigBirdTokenizer special-token registration bug is unrelated to MLukeTokenizer's task AttributeError." + }, + { + "left": "issue:30064", + "right": "issue:33666", + "accept": false, + "reason": "Image processor void segmentation maps and Qwen2-VL multi-GPU training are unrelated subsystems." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Both involve generation, but one is compiled forward inconsistency while the other is recurrent-state reset in modular_qwen3_5.py." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "Blip2 dtype loading and RTDetrV2 output drift are different model-specific bugs." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Offline model_info patching in Mistral regex is unrelated to est-roberta tokenizer loading regression." + }, + { + "left": "issue:44479", + "right": "issue:44936", + "accept": false, + "reason": "Video-input regression in Qwen VL models is unrelated to trainer.evaluate() failing after train()." + }, + { + "left": "issue:41720", + "right": "issue:43906", + "accept": false, + "reason": "Qwen3 auto device mapping cuda assert is not the same as an isolated reproduction of another issue." + }, + { + "left": "issue:44977", + "right": "issue:45356", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug and Kimi-K2.5 tokenizer codec regression are different code paths." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Auxiliary-loss normalization in OLMoE/GPT Oss is not the same as DeepseekV3MoE divergence from remote implementation." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and MobileLLM-125M load KeyError do not share a concrete fix path." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "BigBirdTokenizer decode output bug and DiaConfig token-ID loss on reload are separate persistence/tokenization issues." + }, + { + "left": "issue:43577", + "right": "issue:45588", + "accept": false, + "reason": "Blip2 dtype retention and flash_attention crash on s_aux=None are unrelated." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Compiled-forward chunked generation instability is different from Qwen3.5 flash-attention generation failure." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "Bamba-9B-v2 torch.compile SDPA failure is unrelated to DiaConfig custom token ID loss." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "Bamba torch.compile/SDPA issue and MLukeTokenizer task AttributeError are unrelated." + }, + { + "left": "issue:32090", + "right": "issue:34567", + "accept": false, + "reason": "Trainer broadcast NoneType error and num_input_tokens_seen not updating are different trainer bugs." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Beit legacy-field loading and Bamba SDPA compilation failure are different model/load paths." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Beit pretrained preset loading and BitNet packed-weight unpacking during accelerate loading are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:43906", + "accept": false, + "reason": "Tokenization changes across versions and an isolated reproduction of issue 38071 are not the same defect." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "Trainer _gpu_broadcast_one NoneType failure is not the same as Qwen2-VL multi-GPU training support." + }, + { + "left": "issue:28282", + "right": "issue:35141", + "accept": false, + "reason": "Missing PyTorch ImportError and embedding reinitialization after resize are unrelated." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "Beit legacy-field loading bug and OmDet-Turbo processor size mismatch are different issues." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "GGUF from_pretrained breakage and inability to load ScandiBERT are separate tokenizer/model-loading regressions." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Aux-loss normalization and FSDP CPU RAM efficient loading are different problems." + }, + { + "left": "issue:43329", + "right": "issue:45325", + "accept": false, + "reason": "Undefined helpers in multimodal token counting and still-image rope-index scaling are distinct Qwen-VL bugs." + }, + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "MobileLLM-125M load KeyError and AutoImageProcessor URL loading failure are unrelated loading paths." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Qwen2-VL multi-GPU training and GenerationMixin import failure are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "Auto device mapping cuda assert and compiled-forward generation inconsistency are different failures." + }, + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and processor.tokenizer regression are separate processor bugs." + }, + { + "left": "issue:43756", + "right": "issue:44913", + "accept": false, + "reason": "Smollm3 RoPE-layer mismatch and GPTNeoX rotary_pct reload drift are only loosely related to RoPE and not the same bug." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 HF/JAX numerical discrepancy and pipeline translation error-message regression are unrelated." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both are config padding-field regressions, but they affect different models and attributes and do not look like one shared fix." + } + ] + }, + "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": 16, + "review_unit_index": 19, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:28282", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:36331", + "issue:39692", + "issue:43257", + "issue:43278", + "issue:43295", + "issue:43299", + "issue:43334", + "issue:43493", + "issue:43504", + "issue:43540", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43716", + "issue:43720", + "issue:43756", + "issue:43824", + "issue:43844", + "issue:43927", + "issue:44291", + "issue:44361", + "issue:44423", + "issue:44466", + "issue:44479", + "issue:44488", + "issue:44554", + "issue:44589", + "issue:44625", + "issue:44734", + "issue:44843", + "issue:44849", + "issue:44964", + "issue:44977", + "issue:45083", + "issue:45200", + "issue:45229", + "issue:45357", + "issue:45405", + "issue:45520", + "issue:45701", + "issue:45704" + ], + "soft_pairs": [ + "issue:44849|issue:45357", + "issue:43720|issue:43927", + "issue:43606|issue:43927", + "issue:28282|issue:32090", + "issue:43540|issue:43716", + "issue:28282|issue:33357", + "issue:43334|issue:43572", + "issue:43577|issue:44554", + "issue:43688|issue:45704", + "issue:33666|issue:36331", + "issue:39692|issue:43493", + "issue:43720|issue:44361", + "issue:44964|issue:45405", + "issue:44479|issue:44625", + "issue:43540|issue:44977", + "issue:43493|issue:43844", + "issue:44488|issue:44843", + "issue:43756|issue:45083", + "issue:45200|issue:45229", + "issue:43278|issue:43844", + "issue:43257|issue:43299", + "issue:43646|issue:44466", + "issue:43824|issue:44589", + "issue:39692|issue:43295", + "issue:43688|issue:45701", + "issue:33357|issue:33666", + "issue:44589|issue:45520", + "issue:43504|issue:43606", + "issue:44423|issue:44734", + "issue:43572|issue:44291", + "issue:43927|issue:44361", + "issue:43504|issue:43653" + ], + "prepared_review_unit_hash": "e744884dba59f2f4762cec9a801a17408d0824e0c548f5b46c9580e3ace3d091", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37618, + "estimated_input_tokens": 9405, + "estimated_eval_tokens": 19066 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is highly heterogeneous. A few issues share a broad \u201ctransformers version/regression around loading/custom model code\u201d theme, but none of the soft pairs look like true duplicates or the same fixable code-path bug.", + "confidence": 0.27, + "canonical_issue_reason": "issue:43646 is the broadest representative of the repeated version-regression/custom-model-loading complaints; the rest are narrower model- or subsystem-specific failures.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the most representative single issue for the cluster\u2019s loose theme of Transformers 5 breaking model initialization/loading behavior.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Different Qwen3.5 regressions: hidden-states output vs incorrect save_pretrained visual keys." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "Unrelated bugs: BitNet accelerate-loading unpacking vs DiaConfig losing custom token IDs." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to config token-ID persistence." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "Different failures: missing PyTorch import vs Trainer GPU broadcast TypeError." + }, + { + "left": "issue:43540", + "right": "issue:43716", + "accept": false, + "reason": "Both are multimodal issues, but on different models and different code paths/errors." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "Missing PyTorch import and Mac bus error are unrelated." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Both mention missing pad-token fields, but they affect different configs/models and do not read as the same bug." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "BLIP2 dtype propagation and MPS attention correctness are different issues." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "MoE auxiliary-loss normalization is unrelated to a T5 RMSNorm memory leak." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Multi-GPU training support request and custom compute_loss keyword regression are not the same bug." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "Docs/example errors and SigLIP2 HF-vs-JAX implementation discrepancy are distinct." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "BitNet packed-weight loading is unrelated to MLukeTokenizer task AttributeError." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Model-loading failure and an unreleased PEFT version bump are not the same underlying problem." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Video-input regression across Qwen models is separate from num_labels propagation in Qwen3.5 config." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Video-input processing crash and flash-attention generation bug are different code paths." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "SigLIP2 implementation discrepancy is unrelated to ZeRO-3 gradient growth." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "Model loading failure and offline-mode model_info() regression are different bugs." + }, + { + "left": "issue:43756", + "right": "issue:45083", + "accept": false, + "reason": "Smollm3 RoPE-layer dropping and qwen3_omni_moe feature-length behavior are unrelated." + }, + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Text-only mm_token_type_ids defaulting and multi-GPU inference CUDA OOM are not the same issue." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Embedding dtype drift in eval and ZeRO-3 gradient anomalies are different problems." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Both involve Qwen MoE, but one is accelerate/deepspeed weight conversion and the other is dev loading breakage." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Custom model initialization regression and tied-weight serialization differences are separate failures." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "Importing a missing model class is unrelated to the Float8 storage-object lookup error." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "Docs example mistakes are not the same as the processor.tokenizer / image-to-tokenizer regression." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "Aux-loss normalization and tokenization changes are unrelated." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "Mac bus error on a pretrained CLIP model is unrelated to Qwen2-VL multi-GPU training." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Float8 storage lookup and flash_attn Python 3.13 import key error are different issues." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Legacy field model-load failure and Bark CPU offload device mismatch are unrelated." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Continuous-batching multimodal crash and KV-cache continuation indexing bug are different server paths." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig and unexpected _is_hf_initialized in init_empty_weights are unrelated." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "Token-ID persistence and tokenizer AttributeError on tasks are different bugs." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "Legacy-field model loading and BigBirdTokenizer special-token registration are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 20, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:15354", + "issue:29942", + "issue:30333", + "issue:38175", + "issue:39401", + "issue:43232", + "issue:43278", + "issue:43295", + "issue:43334", + "issue:43493", + "issue:43525", + "issue:43540", + "issue:43575", + "issue:43606", + "issue:43688", + "issue:43701", + "issue:43844", + "issue:43873", + "issue:43994", + "issue:44112", + "issue:44315", + "issue:44464", + "issue:44466", + "issue:44493", + "issue:44521", + "issue:44554", + "issue:44568", + "issue:44610", + "issue:44625", + "issue:44792", + "issue:44849", + "issue:44898", + "issue:44964", + "issue:45072", + "issue:45084", + "issue:45127", + "issue:45198", + "issue:45200", + "issue:45290", + "issue:45325", + "issue:45356", + "issue:45405", + "issue:45412", + "issue:45446", + "issue:45520", + "issue:45584", + "issue:45701", + "issue:45759" + ], + "soft_pairs": [ + "issue:43575|issue:45405", + "issue:44625|issue:45759", + "issue:44792|issue:45405", + "issue:43606|issue:44112", + "issue:44521|issue:45290", + "issue:43994|issue:44493", + "issue:44464|issue:45584", + "issue:29942|issue:30333", + "issue:43844|issue:44964", + "issue:43688|issue:43873", + "issue:15354|issue:38175", + "issue:45084|issue:45520", + "issue:44568|issue:45701", + "issue:44554|issue:45446", + "issue:43540|issue:45200", + "issue:45072|issue:45198", + "issue:43232|issue:44849", + "issue:39401|issue:44568", + "issue:43278|issue:43701", + "issue:44315|issue:44849", + "issue:43295|issue:44610", + "issue:44554|issue:45356", + "issue:45200|issue:45325", + "issue:43334|issue:43525", + "issue:44466|issue:45127", + "issue:43493|issue:43701", + "issue:44898|issue:45412" + ], + "prepared_review_unit_hash": "cecfff152a77325070ccffe5a2e3723f46000170f8ce191b6577d484d6eb78fd", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35906, + "estimated_input_tokens": 8977, + "estimated_eval_tokens": 18210 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly heterogeneous model- and subsystem-specific bugs, not a tight duplicate cluster. The soft-similarity pairs are generally false positives; no PRs are present.", + "confidence": 0.93, + "canonical_issue_reason": "No single issue clearly subsumes the others; if forced to pick an umbrella-style representative, issue 45701 is the broadest, but it is still too vague to be a true canonical duplicate target.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45701 is the most general and reusable representative among the set, with the broadest scope around tokenizer/version behavior, though it does not actually match the other items closely.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "Different problems: TP OOM on model load vs unreleased PEFT version requirement." + }, + { + "left": "issue:44625", + "right": "issue:45759", + "accept": false, + "reason": "Both involve Qwen config plumbing, but one is num_labels propagation and the other is unwrapping text_config from composite configs; not the same bug." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "Janus image-generation test failure is unrelated to a PEFT version bump." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "Different models and failures: CPU offload device mismatch vs a stale CI test override." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Both hit apply_chat_template, but one is zero assistant masks for multimodal inputs and the other is a crash on tool-call assistant messages." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "Different symptoms and code paths: bad SigLIP2 outputs vs unexpected position_id keys." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Chunked generation with compiled forward is unrelated to Whisper empty-transcription generation after align_special_tokens." + }, + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention CI failures and MLflow job-state reporting are unrelated." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "DeepSpeed ZeRO-3 gradient inflation is unrelated to loading Phi-4 multimodal models." + }, + { + "left": "issue:43688", + "right": "issue:43873", + "accept": false, + "reason": "Aux-loss normalization in MoE models is unrelated to quantized offloading behavior." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "TorchScript GeneratorExp export support and SigLIP2 zero-probability outputs are different issues." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Non-template-node compile error and flash_attn import KeyError are unrelated." + }, + { + "left": "issue:44568", + "right": "issue:45701", + "accept": false, + "reason": "Specific BOS/EOS token regression for one tokenizer is not the same as a broad tokenization-version-change report." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "MPS attention correctness and a PyTorch version check for AuxRequest import are different bugs." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Video-input processing failure in Qwen3OmniMoe is unrelated to Gemma 4 token-type defaults." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "BFloat16 dtype mismatches in inference are unrelated to Wav2Vec2 save/tokenization failures." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation kwargs after sync_gpus and Qwen3.5 hidden-state output bugs are unrelated." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Wrong offset_mapping for Qwen3 tokenizer is not the same as missing BOS/EOS insertion for mDeBERTa." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "BF16-to-FP32 dtype change during eval is unrelated to resume_from_checkpoint key mismatch." + }, + { + "left": "issue:44315", + "right": "issue:44849", + "accept": false, + "reason": "Liger Kernel not applied with model_init is unrelated to Qwen3.5 output_hidden_states behavior." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Processor/tokenizer access regression is different from OmDet-Turbo\u2019s image-size mismatch." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "MPS attention correctness and Kimi-K2.5 tokenizer regression are unrelated." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Gemma 4 mm_token_type_ids defaults and Qwen2.5-VL rope-index scaling are different multimodal bugs." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": false, + "reason": "Both are pad_token_id AttributeErrors, but they affect different model/config classes and likely need separate fixes." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "Inconsistent serialization of lm_head.weight is unrelated to LoRA merge collapse with extended vocabulary." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 JAX-vs-HF discrepancy is unrelated to checkpoint resume key mismatch." + }, + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Perceiver resolution/interpolation failure is unrelated to RT-DETR memory not being released." + } + ] + }, + "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": 16, + "review_unit_index": 21, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:37428", + "issue:38175", + "issue:42915", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43425", + "issue:43450", + "issue:43493", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43653", + "issue:43819", + "issue:43824", + "issue:43825", + "issue:43827", + "issue:43844", + "issue:43856", + "issue:43873", + "issue:43901", + "issue:43957", + "issue:43975", + "issue:43994", + "issue:44112", + "issue:44206", + "issue:44360", + "issue:44442", + "issue:44464", + "issue:44514", + "issue:44521", + "issue:44561", + "issue:44610", + "issue:44792", + "issue:44857", + "issue:44936", + "issue:44987", + "issue:45083", + "issue:45127", + "issue:45198", + "issue:45245", + "issue:45290", + "issue:45381", + "issue:45412", + "issue:45563", + "issue:45584", + "issue:45593" + ], + "soft_pairs": [ + "issue:37428|issue:43957", + "issue:43572|issue:44987", + "issue:43873|issue:44464", + "issue:43295|issue:43450", + "issue:43572|issue:43957", + "issue:43824|issue:43975", + "issue:43844|issue:44936", + "issue:43295|issue:43873", + "issue:44514|issue:44521", + "issue:44442|issue:45198", + "issue:38175|issue:45412", + "issue:43278|issue:43856", + "issue:43295|issue:43994", + "issue:38175|issue:43493", + "issue:37428|issue:43824", + "issue:42915|issue:43824", + "issue:43827|issue:43901", + "issue:43653|issue:44112", + "issue:43450|issue:43493", + "issue:43450|issue:43575", + "issue:44360|issue:44561", + "issue:44792|issue:45563", + "issue:43819|issue:45083", + "issue:45127|issue:45245", + "issue:43262|issue:43450", + "issue:44206|issue:45584", + "issue:43819|issue:44112", + "issue:44610|issue:44857", + "issue:37428|issue:43425", + "issue:44857|issue:45593", + "issue:43278|issue:43825", + "issue:43606|issue:45083", + "issue:45290|issue:45381" + ], + "prepared_review_unit_hash": "6552d82c24b2c529d057f7b4575a0bf446e000f4b50bf0a80c7ded158a6f2535", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "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": "The set is mostly heterogeneous and the soft edges are too broad to treat as true duplicates. There is a small chat-template/multimodal subgroup, but even there the paired issues describe different failure modes, so I would not merge any of the suggested edges.", + "confidence": 0.27, + "canonical_issue_reason": "No issue cleanly represents the whole set. If forced to pick one representative, issue 44521 is the broadest open, user-facing report in the only clearly related subgroup (multimodal `apply_chat_template` behavior), but it is not a duplicate of the others.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44521 is the best single representative among these because it has the broadest framing and strongest overlap with the chat-template/multimodal subset, though the overall cluster is too mixed for a strong canonical choice.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "Different failures: an ImportError for `_flash_supports_window_size` versus meta-device model loading regressions. Same library, not the same bug." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "Different model/config breakages with no clear shared code-path; one is `StableLmConfig.pad_token_idx`, the other is a load failure for a specific model on v5.1+." + }, + { + "left": "issue:43873", + "right": "issue:44464", + "accept": false, + "reason": "Quantization/offload behavior and compiled-forward chunked generation are unrelated symptoms and code paths." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Processor/tokenizer regression is not the same bug as batched video shape handling." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Different regressions: missing config field versus meta-device loading failures; no evidence of one underlying issue." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "An import/export failure for Qwen2.5-VL is not the same as DeepSeek detokenization corruption." + }, + { + "left": "issue:43844", + "right": "issue:44936", + "accept": false, + "reason": "Training-gradient instability with ZeRO-3 is a different problem from `trainer.evaluate()` failing after training." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor API regression and offloading/quantization behavior are separate subsystems and bugs." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Both involve `apply_chat_template`, but one is a batched-input crash with `padding=False` and the other is all-zero assistant masks; related area, different concrete bug." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Different tokenizer/model families and different failure modes (`AutoTokenizer` load vs save/tokenization failure)." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 zero probabilities and RT-DETR memory not released are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Embedding dtype mismatch during eval is not the same as Qwen3 MoE memory usage." + }, + { + "left": "issue:43295", + "right": "issue:43994", + "accept": false, + "reason": "Custom processor/tokenizer regression and SigLIP2 output quality issues are different bugs." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": false, + "reason": "Both mention SigLIP2, but one is zero probabilities in a model run and the other is a discrepancy versus JAX implementation; not clearly the same concrete defect." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "Import errors for different symbols and models; no shared underlying bug." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "FP8 config failure for Qwen3 MoE versus missing Qwen2.5-VL export are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs/API transition issues around v5, but they concern different pipeline behaviors and should not be merged." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "Tokenizer special-token registration and GraniteSpeech device override tests are different issues." + }, + { + "left": "issue:43450", + "right": "issue:43493", + "accept": false, + "reason": "Video batched-shape bug and SigLIP2 implementation discrepancy do not share the same failure mode." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Batched video shape handling is unrelated to Qwen2-57B tensor-parallel OOM." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "DSA indexer ReLU discussion and removal of `is_torch_fx_available` breakage are not the same bug." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "Janus image-generation test failure is unrelated to paged `generate()` warning behavior." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "DAC latent reconstruction mismatch and Qwen3 Omni MoE feature-length helper behavior are different code paths." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA merge/collapsed model behavior is unrelated to category-count limits." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sample-rate defaulting and video batched shape handling are different processor bugs." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Feature-extractor argument regression and Whisper empty-transcription failure are unrelated." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "DAC forward/latents mismatch and GraniteSpeech CI device override are separate issues." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "OmDet-Turbo preprocessing size mismatch and LwDetrImageLoss AMP crash are different bugs." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "ImportError for a flash-attention helper and Torch 2.10 incompatibility are not the same defect." + }, + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "AMP/CUDA loss crash versus missing auxiliary losses when denoising is off are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Embedding dtype change and pipeline translation error-message regression are different problems." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "CPU-offload device mismatch in Bark and a Qwen3 Omni MoE helper behavior issue are unrelated." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Assistant-message tool-call crash in `apply_chat_template` and Qwen2.5-VL video `vision_position_ids` mismatch are different bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 22, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:29127", + "issue:29942", + "issue:32090", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:40444", + "issue:43335", + "issue:43388", + "issue:43404", + "issue:43504", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43653", + "issue:43701", + "issue:43720", + "issue:43819", + "issue:43931", + "issue:44112", + "issue:44292", + "issue:44351", + "issue:44448", + "issue:44462", + "issue:44493", + "issue:44561", + "issue:44568", + "issue:44743", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45083", + "issue:45200", + "issue:45229", + "issue:45310", + "issue:45459", + "issue:45478", + "issue:45491", + "issue:45520", + "issue:45588", + "issue:45593", + "issue:45712" + ], + "soft_pairs": [ + "issue:44448|issue:44743", + "issue:45200|issue:45588", + "issue:43335|issue:43550", + "issue:44462|issue:44561", + "issue:44561|issue:44779", + "issue:37428|issue:44351", + "issue:43611|issue:43931", + "issue:43388|issue:43526", + "issue:39692|issue:43526", + "issue:43819|issue:44493", + "issue:43335|issue:43653", + "issue:29942|issue:32090", + "issue:44855|issue:45083", + "issue:45310|issue:45478", + "issue:43404|issue:45200", + "issue:43550|issue:44112", + "issue:39692|issue:44493", + "issue:43335|issue:45083", + "issue:43335|issue:43504", + "issue:44448|issue:44779", + "issue:44292|issue:45520", + "issue:38175|issue:44493", + "issue:43335|issue:45072", + "issue:44977|issue:45229", + "issue:43335|issue:45593", + "issue:43720|issue:44112", + "issue:44448|issue:44977", + "issue:45459|issue:45712", + "issue:44568|issue:45491", + "issue:29127|issue:32090", + "issue:43606|issue:44857", + "issue:40444|issue:43526", + "issue:34689|issue:35141", + "issue:44857|issue:45072", + "issue:43701|issue:45083" + ], + "prepared_review_unit_hash": "9aa6aac0bb1fa2d3b540fb4296da362c82c0f006b36ee77beb66468d23891552", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 35, + "serialized_chars": 37901, + "estimated_input_tokens": 9476, + "estimated_eval_tokens": 19208 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated issue set; the only clear duplicate is the Qwen3.5 Moe from_pretrained regression, which appears as an exact title match in issues 45310 and 45478.", + "confidence": 0.94, + "canonical_issue_reason": "Issue 45310 is the earlier, better-supported report of the exact Qwen3.5 Moe from_pretrained regression and has an inbound reference, making it the best canonical issue.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45310 is the strongest representative issue here because it is the original of the only exact duplicate pair and is phrased as a concrete, reproducible bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Different models and failure modes: Pegasus output regression vs Qwen3.5 recurrent-state reset." + }, + { + "left": "issue:45200", + "right": "issue:45588", + "accept": false, + "reason": "Different code paths: Gemma4 token-type handling vs flash_attention integration crash on s_aux=None." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "Different subsystems and bugs: SwitchTransformers config layer count vs Bamba torch.compile/SDPA failure." + }, + { + "left": "issue:44462", + "right": "issue:44561", + "accept": false, + "reason": "Unrelated regressions: tokenizer.json loading vs removal of is_torch_fx_available affecting trust_remote_code." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "Both are v5 regressions, but one is a remote-code API removal issue and the other is a Deepseek tokenizer correctness bug." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and likely different fixes." + }, + { + "left": "issue:43611", + "right": "issue:43931", + "accept": false, + "reason": "Different loading failures for different models: base_model_prefix handling vs Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Both mention labels, but one is metric gathering and the other is Beit image-processor label reduction." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "Documentation/example failures are not the same bug as BeitImageProcessorFast label reduction." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "Different symptoms and components: DAC latent reconstruction vs unexpected position-id key handling." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "Different models and bugs: SwitchTransformers sparse-layer creation vs BigBirdTokenizer special-token registration." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention test failures and Trainer _gpu_broadcast_one NoneType errors are unrelated." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Python parsing/import issue in DebertaV2 vs qwen3_omni_moe feature-length helper behavior." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Exact same title and same Qwen3.5 Moe from_pretrained error report; clearly the same underlying bug." + }, + { + "left": "issue:43404", + "right": "issue:45200", + "accept": false, + "reason": "Different generation/training bugs: lm_head weight tying vs mm_token_type_ids defaulting." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "Different failures in different models: torch.compile SDPA issue vs stale CI device-override test." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 docs/example problems are unrelated to position-id key regressions." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "Different model families and code paths; no shared concrete bug." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "SwitchTransformers config bug and Beit legacy-field loading bug are unrelated." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Both are transformer v5 regressions, but they affect different models and causes." + }, + { + "left": "issue:44292", + "right": "issue:45520", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime error is not the same as a Python 3.13 flash_attn import KeyError." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "Different issues: zero probabilities in SigLIP2 vs unexpected position-id keys." + }, + { + "left": "issue:43335", + "right": "issue:45072", + "accept": false, + "reason": "SwitchTransformers config bug is unrelated to bfloat16 dtype mismatches in inference." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Different models and failures: flash-attention generation bug vs CUDA OOM on multi-GPU inference." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "Sparse-layer config bug and missing auxiliary losses are unrelated." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "Packed-weight loading failure is unrelated to a stale CI device-override test." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Different models and symptoms; one is output drift, the other is flash-attention generation failure." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "Tokenizer error masking and leftover dummy classes are separate maintenance issues." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Tokenizer special-token addition bug and Gemma3 NaN embeddings are unrelated." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "Error-message clarity in layoutlmv3 is not the same as a Trainer broadcast TypeError." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "Device-mismatch CPU offload bug and float16 AMP crash are different failures." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "Multi-image iterable-dataset finetuning bug is unrelated to Beit label reduction." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Model-loading regression for Llama 3.2 vision and post_init reinitialization are distinct bugs." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "Different model/runtime issues: LwDetr AMP crash vs SwitchTransformers dtype mismatch." + }, + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Checkpoint resume key mismatch is unrelated to qwen3_omni_moe feature-length behavior." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: 45310 and 45478 are the only exact title match, and 45310 is earlier with an inbound reference. The soft-edge verdicts are conservative and do not overstate evidence; the lone true duplicate call is well supported." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 23, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:25251", + "issue:29942", + "issue:30064", + "issue:32090", + "issue:35141", + "issue:36010", + "issue:37428", + "issue:42915", + "issue:42994", + "issue:43065", + "issue:43066", + "issue:43278", + "issue:43381", + "issue:43412", + "issue:43421", + "issue:43550", + "issue:43653", + "issue:43873", + "issue:43927", + "issue:43976", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44361", + "issue:44448", + "issue:44451", + "issue:44462", + "issue:44466", + "issue:44479", + "issue:44493", + "issue:44743", + "issue:44857", + "issue:44871", + "issue:44977", + "issue:45083", + "issue:45198", + "issue:45245", + "issue:45372", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45520", + "issue:45563", + "issue:45588", + "issue:45698" + ], + "soft_pairs": [ + "issue:43927|issue:45698", + "issue:45372|issue:45588", + "issue:43550|issue:45083", + "issue:44977|issue:45563", + "issue:43412|issue:45588", + "issue:43066|issue:44462", + "issue:43873|issue:44493", + "issue:45245|issue:45563", + "issue:29942|issue:35141", + "issue:45372|issue:45459", + "issue:36010|issue:37428", + "issue:44448|issue:44466", + "issue:43278|issue:43381", + "issue:25251|issue:32090", + "issue:44857|issue:45588", + "issue:43065|issue:45412", + "issue:43653|issue:44743", + "issue:44186|issue:45198", + "issue:44451|issue:45520", + "issue:44361|issue:45198", + "issue:43976|issue:45440", + "issue:44265|issue:45198", + "issue:43994|issue:45412", + "issue:44871|issue:44977", + "issue:43421|issue:45491", + "issue:42915|issue:42994", + "issue:29942|issue:30064", + "issue:44206|issue:44479", + "issue:43065|issue:43994" + ], + "prepared_review_unit_hash": "09e4521981c1d1f90387584150bc47f175cfd3533569445e7b1aa38193ab27c2", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35812, + "estimated_input_tokens": 8953, + "estimated_eval_tokens": 18162 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is not a true duplicate set; the items span many unrelated bugs and regressions, so all soft edges should be rejected. There are no PRs in the input, and no single issue cleanly represents the whole set.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue fits: the cluster is heterogeneous across tokenization, generation, quantization, vision, import, and trainer/runtime bugs.", + "canonical_pr_reason": null, + "best_issue_reason": "No global best issue is suitable because none of the issues subsumes the others or represents one underlying defect.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "Both involve save/load behavior around custom tokens or modules, but one is about token IDs being lost and the other is about loading the wrong custom module; different failure modes." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "ImportError in Gemma 4 processor loading from a missing dependency is unrelated to a flash-attention crash on sink-less models." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "torch.compile SDPA failure in Bamba-9B-v2 and a helper output-length bug in qwen3_omni_moe are distinct code paths." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Qwen3.5 flash-attention generation regression and a stale num_return_sequences warning in paged generate() are not the same bug." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Executorch export failure due to missing training guard/erfinv support is unrelated to the s_aux=None flash_attention.py crash." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Wrong tokenizer decoder type and ignoring tokenizer.json are both tokenizer issues, but they affect different loading/serialization logic." + }, + { + "left": "issue:43873", + "right": "issue:44493", + "accept": false, + "reason": "Quantization offloading behavior and an unexpected position-id key error are different symptoms with no clear shared code-path." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Category-count runtime error and a stale generation warning are unrelated." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention 2 test failures and embedding reinitialization in post_init are unrelated regressions." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "Missing ReasoningEffort import and protobuf decode-error masking affect different tokenizer/processor paths." + }, + { + "left": "issue:36010", + "right": "issue:37428", + "accept": false, + "reason": "Both are import errors, but they are for different symbols, modules, and likely fixes." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "Pegasus output differences across versions and inconsistent lm_head.weight serialization are separate model-behavior issues." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Embedding dtype drift in evaluate() and gradient-checkpointing incompatibility in eval mode are distinct." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "top_k nesting output change in pipeline and Trainer NoneType broadcast error are unrelated." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "AMP/CUDA crash in LwDetrImageLoss is unrelated to the sink-less flash_attention None crash." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder and RT-DETR memory leak on deletion are different problems." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and qwen3_5 recurrent-state reset are unrelated." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "LayoutLMv2Tokenizer NER/padding crash and Wav2Vec2 save_pretrained/tokenization failure are different model/tokenizer bugs." + }, + { + "left": "issue:44451", + "right": "issue:45520", + "accept": false, + "reason": "ScandiBERT loading failure and flash_attn KeyError on Python 3.13 are unrelated." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "MLukeTokenizer task-time AttributeError and Wav2Vec2 save_pretrained/tokenization failure do not share the same underlying defect." + }, + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Python version compatibility problem and DeepseekV3MoE divergence are unrelated." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export failure with torch_compilable_check and Wav2Vec2 save_pretrained/tokenization failure are different code paths." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "Siglip2 nonsense outputs and RT-DETR memory retention are not the same bug." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "Gemma-3 eos_token_id configuration mismatch and Qwen3.5 flash-attention generation regression are separate issues." + }, + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "TokenizerBackend post-processor update behavior and Gemma3 NaN embeddings on GPU are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "Qwen3Moe FP8 failure during inference/training and quantized model saving failure are different quantization bugs." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "Flash Attention 2 test failures and void segmentation map processing are unrelated." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Unsupported center arg in LasrFeatureExtractor and Qwen video input regression are different failures." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "Sam3PixelDecoder dummy Conv2d and Siglip2 AutoModel/pipeline 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": 2, + "component_count": 16, + "review_unit_index": 24, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:29127", + "issue:29942", + "issue:32090", + "issue:33357", + "issue:34567", + "issue:36010", + "issue:39692", + "issue:41720", + "issue:42915", + "issue:43262", + "issue:43278", + "issue:43329", + "issue:43334", + "issue:43381", + "issue:43452", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43653", + "issue:43701", + "issue:43723", + "issue:43756", + "issue:43824", + "issue:43873", + "issue:43931", + "issue:44186", + "issue:44206", + "issue:44279", + "issue:44442", + "issue:44464", + "issue:44554", + "issue:44610", + "issue:44734", + "issue:44743", + "issue:44811", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45103", + "issue:45229", + "issue:45245", + "issue:45372", + "issue:45406", + "issue:45412", + "issue:45440", + "issue:45491", + "issue:45588", + "issue:45593" + ], + "soft_pairs": [ + "issue:44734|issue:45406", + "issue:33357|issue:39692", + "issue:29127|issue:34567", + "issue:44610|issue:45412", + "issue:45103|issue:45372", + "issue:43873|issue:45229", + "issue:43756|issue:45440", + "issue:43723|issue:44554", + "issue:45372|issue:45406", + "issue:43278|issue:43575", + "issue:43381|issue:43701", + "issue:43278|issue:43329", + "issue:43653|issue:44442", + "issue:45440|issue:45593", + "issue:41720|issue:45229", + "issue:29942|issue:36010", + "issue:32090|issue:33357", + "issue:43334|issue:43531", + "issue:43526|issue:44186", + "issue:44977|issue:45245", + "issue:43824|issue:44279", + "issue:43526|issue:44857", + "issue:42915|issue:43931", + "issue:43526|issue:44610", + "issue:44743|issue:44811", + "issue:44442|issue:45072", + "issue:43262|issue:44206", + "issue:43452|issue:45588", + "issue:43575|issue:43701", + "issue:44464|issue:45491" + ], + "prepared_review_unit_hash": "06a6e03128fae22379544ad63841cff0e045ee1290318171009ea6a79bb25d93", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36804, + "estimated_input_tokens": 9201, + "estimated_eval_tokens": 18658 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the soft pairs are mostly surface-similar but refer to different models, components, and failure modes. I found no true duplicate/merge-worthy PR-style overlap, and no single issue cleanly represents the whole set.", + "confidence": 0.96, + "canonical_issue_reason": "No clear canonical duplicate source: the items span unrelated bugs across training, tokenizers, multimodal processors, generation, and model loading.", + "canonical_pr_reason": null, + "best_issue_reason": "If a single anchor is needed, issue:45229 is the strongest standalone report because it has the most discussion and describes a high-impact multi-GPU inference OOM regression.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "Different Gemma4-related failures: KV-cache continuation tensor indexing vs missing `_tokenizer` on processor load/serve." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "Both mention CLIP/SigLIP-style usage, but one is a MacOS bus error in a community CLIP model and the other is a docs example mismatch/quantization failure." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error messaging vs TrainerState token counting update bug; unrelated subsystems." + }, + { + "left": "issue:44610", + "right": "issue:45412", + "accept": false, + "reason": "Processor input-size mismatch for OmDet-Turbo vs RT-DETR memory not being released; different bugs." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "auto_docstring annotation crash vs Gemma4 loading import failure; no shared code path." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Both are memory/loading-adjacent, but one is quantization/offloading behavior and the other is a model-specific multi-GPU OOM." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Different model families and different architectural issues: Smollm3 RoPE layer handling vs DeepSeekV3 MoE divergence." + }, + { + "left": "issue:43723", + "right": "issue:44554", + "accept": false, + "reason": "Tokenizer loading regression in v5 vs MPS attention correctness with mismatched value/query head dims." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Both touch Gemma4 loading, but the root causes differ: missing `ReasoningEffort` import vs missing `_tokenizer` attribute in serve." + }, + { + "left": "issue:43278", + "right": "issue:43575", + "accept": false, + "reason": "Embedding dtype drift during eval vs Qwen2-57B tensor-parallel OOM; not the same bug." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Eval-mode gradient checkpointing restriction vs resume-from-checkpoint key mismatch; unrelated trainer behaviors." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Different issues entirely: dtype change in evaluation vs multimodal token counting video-branch undefined names." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "Tokenizer special-token registration bug vs AutoTokenizer failing on a FastSpeech2Conformer tokenizer; different failure modes." + }, + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "DeepSeekV3MoE implementation divergence vs D-FINE auxiliary-loss behavior when denoising is off." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Different model-specific GPU problems: Qwen3 auto device-map assert vs Gemma4 multi-GPU OOM." + }, + { + "left": "issue:29942", + "right": "issue:36010", + "accept": false, + "reason": "Flash Attention 2 test failures and a GenerationMixin import error are unrelated." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "Trainer NoneType broadcast error vs MacOS CLIP bus error; no shared bug." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Qwen3-VL missing `pad_token_id` vs Qwen3-MoE sliding-window issue; different components and symptoms." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug vs LayoutLMv2Tokenizer crashes on NER/padding; not the same defect." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Qwen3.5 flash-attention generation problems vs category-count overflow; unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Specific missing `Qwen2_5_VLForConditionalGeneration` import vs generic dependency issue; too broad and not the same root cause." + }, + { + "left": "issue:43526", + "right": "issue:44857", + "accept": false, + "reason": "Beit label reduction bug vs LwDetrImageLoss AMP/CUDA crash; different model paths." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3Moe FP8 config failure vs Qwen3-VL weight shape mismatch; distinct model-loading errors." + }, + { + "left": "issue:43526", + "right": "issue:44610", + "accept": false, + "reason": "Different processor bugs: Beit label reduction vs OmDet-Turbo image size mismatch." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Qwen3.5 recurrent-state reset with cache vs Whisper batch_decode ignoring skip_special_tokens; unrelated." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "FastSpeech2ConformerTokenizer loading bug vs dtype mismatch in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Audio chat-template sample-rate default bug vs LasrFeatureExtractor unsupported argument crash; different paths." + }, + { + "left": "issue:43452", + "right": "issue:45588", + "accept": false, + "reason": "gguf_file loading regression vs flash_attention `s_aux=None` crash; unrelated." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "Qwen2 tensor-parallel OOM vs checkpoint resume key mismatch; different problems." + }, + { + "left": "issue:44464", + "right": "issue:45491", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency vs Gemma3 NaN embeddings from sliding-window/all-padding behavior." + } + ] + }, + "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": 16, + "review_unit_index": 25, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:29127", + "issue:33290", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:35141", + "issue:38175", + "issue:39692", + "issue:42907", + "issue:43295", + "issue:43316", + "issue:43381", + "issue:43441", + "issue:43493", + "issue:43525", + "issue:43550", + "issue:43575", + "issue:43611", + "issue:43646", + "issue:43761", + "issue:43827", + "issue:43873", + "issue:43931", + "issue:43994", + "issue:44077", + "issue:44079", + "issue:44188", + "issue:44206", + "issue:44387", + "issue:44442", + "issue:44479", + "issue:44568", + "issue:44743", + "issue:44898", + "issue:44964", + "issue:44977", + "issue:45072", + "issue:45216", + "issue:45229", + "issue:45290", + "issue:45440", + "issue:45479", + "issue:45491", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45698", + "issue:45759" + ], + "soft_pairs": [ + "issue:43441|issue:45588", + "issue:29127|issue:35141", + "issue:43646|issue:44077", + "issue:43575|issue:44387", + "issue:45072|issue:45229", + "issue:44977|issue:45584", + "issue:33357|issue:38175", + "issue:45588|issue:45600", + "issue:44479|issue:45290", + "issue:43761|issue:44079", + "issue:44964|issue:45440", + "issue:43316|issue:43525", + "issue:43931|issue:45216", + "issue:33290|issue:34567", + "issue:42907|issue:45698", + "issue:34689|issue:43611", + "issue:44442|issue:44898", + "issue:43761|issue:44188", + "issue:39692|issue:45491", + "issue:43381|issue:43873", + "issue:43493|issue:43827", + "issue:43550|issue:44743", + "issue:43295|issue:45584", + "issue:43994|issue:44206", + "issue:44568|issue:45479", + "issue:43316|issue:45759" + ], + "prepared_review_unit_hash": "7cd5289733dfe4a5679224f87d2069fdd78ac97bf6ef64b2e098b69806c20de5", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35673, + "estimated_input_tokens": 8919, + "estimated_eval_tokens": 18094 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: most pairs share only broad themes like v5 regressions, OOMs, or model-loading failures, but they do not appear to be the same underlying bug/change. I found no safe duplicate merges among the soft edges.", + "confidence": 0.44, + "canonical_issue_reason": "issue:43646 is the broadest, most representative regression report in the set ('Transformers 5.0.0 breaks custom model initialization'), so it best serves as a cluster anchor despite the overall heterogeneity.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the best single issue to represent the cluster because it is a generic v5 regression report and not tied to one narrow model or subsystem.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Both involve FlashAttention, but one is a Ministral v5 RC generation failure and the other is an AttributeError in flash_attention.py for sink-less models; different root causes." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and token-embedding reinitialization on resize are unrelated bugs." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "One is a broad custom model initialization regression; the other is a PatchTSTMixer-specific post_init policy issue." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both mention OOM, but one is tensor-parallel loading of Qwen2-57B and the other is increased reserved memory under int4 quantization; different failure modes." + }, + { + "left": "issue:45072", + "right": "issue:45229", + "accept": false, + "reason": "dtype mismatch CI failures and Gemma4 multi-GPU CUDA OOM are not the same bug." + }, + { + "left": "issue:44977", + "right": "issue:45584", + "accept": false, + "reason": "Qwen3.5 flash-attention generation issues and Whisper empty-transcription failure after align_special_tokens are unrelated." + }, + { + "left": "issue:33357", + "right": "issue:38175", + "accept": false, + "reason": "A MacOS bus error on CLIP and zero probabilities on SigLIP2 are distinct model/runtime problems." + }, + { + "left": "issue:45588", + "right": "issue:45600", + "accept": false, + "reason": "FlashAttention crash on s_aux=None and auto_mappings references to removed configs affect different code paths." + }, + { + "left": "issue:44479", + "right": "issue:45290", + "accept": false, + "reason": "Video-input regression for Qwen multimodal models and chat-template crashes on tool-call messages are separate regressions." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "Missing hidden_states in CLIPVisionModel.forward and ModelOutput key assignment when previous value is None are different defects." + }, + { + "left": "issue:44964", + "right": "issue:45440", + "accept": false, + "reason": "Phi-4 multimodal loading failure and DeepseekV3MoE divergence from remote implementation are unrelated." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "A config API discrepancy and a missing pad_token_id attribute are not the same underlying issue." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Weight-shape mismatch on load and incorrect save_pretrained checkpoint output are different lifecycle stages and bugs." + }, + { + "left": "issue:33290", + "right": "issue:34567", + "accept": false, + "reason": "Adafactor OOM in DeepSpeed and TrainerState token-count tracking are unrelated." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "Saving dequantized models and loading the wrong custom module after save_pretrained concern different serialization paths." + }, + { + "left": "issue:34689", + "right": "issue:43611", + "accept": false, + "reason": "Llama 3.2 Vision loading breakage and base_model_prefix loading regression are separate model-loading failures." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer loading failure and Perceiver non-default-resolution failure are unrelated." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "A forward-output bug in CLIPVisionModel and attention-kernel divergence under torch.compile are different problems." + }, + { + "left": "issue:39692", + "right": "issue:45491", + "accept": false, + "reason": "SigLIP2 docs/example issues and Gemma3 mixed-length NaN embeddings are not the same bug." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Gradient checkpointing in eval mode and quantization offloading behavior are unrelated." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and docs referencing removed pipeline APIs are distinct." + }, + { + "left": "issue:43550", + "right": "issue:44743", + "accept": false, + "reason": "torch.compile SDPA failure in Bamba and recurrent-state reset in Qwen3.5 are different code-path bugs." + }, + { + "left": "issue:43295", + "right": "issue:45584", + "accept": false, + "reason": "Processor/tokenizer access regression and Whisper empty-transcription failure are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:44206", + "accept": false, + "reason": "Bad outputs from SigLIP2 AutoModel/pipeline and a LaserFeatureExtractor argument crash are different issues." + }, + { + "left": "issue:44568", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer special-token addition regression and degenerate zero-loss classification behavior do not share the same root cause." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Gemma3TextConfig API mismatch and Qwen multimodal from_config text_config unwrapping are separate config-handling bugs." + } + ] + }, + "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": 16, + "review_unit_index": 26, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:25251", + "issue:28282", + "issue:29127", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:36010", + "issue:41628", + "issue:41950", + "issue:42175", + "issue:43232", + "issue:43295", + "issue:43316", + "issue:43381", + "issue:43388", + "issue:43404", + "issue:43441", + "issue:43475", + "issue:43540", + "issue:43550", + "issue:43606", + "issue:43643", + "issue:43701", + "issue:43720", + "issue:43742", + "issue:43825", + "issue:43856", + "issue:43873", + "issue:44077", + "issue:44162", + "issue:44190", + "issue:44442", + "issue:44479", + "issue:44496", + "issue:44514", + "issue:44792", + "issue:44855", + "issue:44871", + "issue:44877", + "issue:44913", + "issue:45127", + "issue:45137", + "issue:45200", + "issue:45372", + "issue:45584", + "issue:45656" + ], + "soft_pairs": [ + "issue:32090|issue:33290", + "issue:43720|issue:44442", + "issue:43643|issue:44913", + "issue:41628|issue:45372", + "issue:43873|issue:44871", + "issue:44190|issue:44496", + "issue:34689|issue:36010", + "issue:43381|issue:43856", + "issue:44855|issue:45137", + "issue:44077|issue:44479", + "issue:33290|issue:45656", + "issue:29127|issue:33357", + "issue:43540|issue:44514", + "issue:43381|issue:43742", + "issue:43316|issue:44877", + "issue:33290|issue:34689", + "issue:42175|issue:43316", + "issue:25251|issue:28282", + "issue:43404|issue:43441", + "issue:44162|issue:44792", + "issue:43232|issue:43388", + "issue:43295|issue:43825", + "issue:41950|issue:43475", + "issue:44871|issue:45584", + "issue:43550|issue:43701", + "issue:43606|issue:43701", + "issue:45127|issue:45200", + "issue:43295|issue:44871", + "issue:33290|issue:35141" + ], + "prepared_review_unit_hash": "b730d94e30325e4ed69e90dd8ed9b27c75c735c90bcb3b06ad8eedcb82dd9de4", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36563, + "estimated_input_tokens": 9141, + "estimated_eval_tokens": 18538 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: the paired issues only overlap at a broad subsystem level (pipelines, model loading, deepspeed, multimodal/video), but each describes a different concrete bug or regression. No soft edge should be merged.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical duplicate stands out; these issues span unrelated code paths in loading, config, generation, training, and pipeline behavior.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41950 is the most central/open pipeline-related report and is the closest thing to an anchor, but it still does not duplicate the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Different failures: Trainer NoneType broadcast vs deepspeed/adafactor OOM." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "Unrelated loading problems: BitNet packed-weight loading vs tokenizer class import/load failure." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "Both are config-related, but one is missing fields with trust_remote_code and the other is rotary_pct not persisting on reload." + }, + { + "left": "issue:41628", + "right": "issue:45372", + "accept": false, + "reason": "Different import errors for different symbols and different model families." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization/offloading behavior and Gemma eos_token_id mismatch are distinct bugs." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Dataset loading in a training script is unrelated to model config recognition in AutoModel." + }, + { + "left": "issue:34689", + "right": "issue:36010", + "accept": false, + "reason": "One is a model-loading regression; the other is a missing GenerationMixin import." + }, + { + "left": "issue:43381", + "right": "issue:43856", + "accept": false, + "reason": "Gradient checkpointing in eval mode and MoE memory usage are different issues." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 IndentationError in torch.jit parsing is unrelated to DeepSpeed ZeRO3 deque failure." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "Post-init handling for patchtsmixer is unrelated to the Qwen video-input regression." + }, + { + "left": "issue:33290", + "right": "issue:45656", + "accept": false, + "reason": "Both involve deepspeed/optimizer behavior, but OOM from adafactor is not the same as double optimizer step calls." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and MacOS CLIP bus error are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Both are multimodal/video processor issues, but they fail in different APIs and code paths." + }, + { + "left": "issue:43381", + "right": "issue:43742", + "accept": false, + "reason": "Gradient checkpointing eval-mode restriction is unrelated to MobileLLM loading key errors." + }, + { + "left": "issue:43316", + "right": "issue:44877", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy and granite_speech strict config loading are different config-schema problems." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "Deepspeed OOM and Llama 3.2 vision model loading are not the same bug." + }, + { + "left": "issue:42175", + "right": "issue:43316", + "accept": false, + "reason": "TensorFlow backend packaging and Gemma3TextConfig API mismatch are unrelated." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline top_k nesting behavior has nothing to do with PyTorch import availability." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Mistral3 weight tying and Ministral flash-attention failure are separate model-specific regressions." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 breakage and janus image-generation test failure are different model/test problems." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation kwargs update after sync_gpus and gather_for_metrics label dropping are different training/eval bugs." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Custom processor.tokenizer regression is unrelated to the pipeline translation-task error message." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Video-classification pipeline choosing image processors is not the same as Sam3VisionEncoderOutput missing fpn_position_embeddings." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Gemma eos_token_id inconsistency and Whisper empty-transcription generation failure are different issues." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "torch.compile/SDPA failure and resume_from_checkpoint key mismatch are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU-offload device mismatch and resume_from_checkpoint key mismatch are unrelated." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "LoRA merge collapse with extended vocabulary and mm_token_type_ids defaulting are different model-behavior bugs." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Processor/tokenizer API regression and Gemma eos_token_id mismatch are unrelated." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "Deepspeed/adafactor OOM is not the same as embedding reinitialization after resize." + } + ] + }, + "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": 16, + "review_unit_index": 27, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:33290", + "issue:36331", + "issue:41720", + "issue:42617", + "issue:43065", + "issue:43335", + "issue:43452", + "issue:43575", + "issue:43644", + "issue:43646", + "issue:43673", + "issue:43742", + "issue:43761", + "issue:43827", + "issue:43873", + "issue:43906", + "issue:43931", + "issue:43994", + "issue:44060", + "issue:44077", + "issue:44190", + "issue:44242", + "issue:44464", + "issue:44479", + "issue:44496", + "issue:44521", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44734", + "issue:44743", + "issue:44792", + "issue:44855", + "issue:44871", + "issue:44877", + "issue:44913", + "issue:44936", + "issue:44964", + "issue:44977", + "issue:45127", + "issue:45245", + "issue:45290", + "issue:45357", + "issue:45381", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45698" + ], + "soft_pairs": [ + "issue:43761|issue:44479", + "issue:33290|issue:36331", + "issue:43575|issue:43742", + "issue:43065|issue:44617", + "issue:43452|issue:44496", + "issue:42617|issue:43575", + "issue:43646|issue:44913", + "issue:43673|issue:44743", + "issue:44936|issue:45698", + "issue:43761|issue:44242", + "issue:43575|issue:43873", + "issue:44077|issue:44534", + "issue:44479|issue:44734", + "issue:44464|issue:44871", + "issue:44617|issue:45584", + "issue:43931|issue:44060", + "issue:43065|issue:44610", + "issue:43906|issue:45290", + "issue:43827|issue:43994", + "issue:44792|issue:44977", + "issue:44534|issue:44913", + "issue:41720|issue:45245", + "issue:44190|issue:44964", + "issue:44792|issue:45600", + "issue:44855|issue:45588", + "issue:44521|issue:45381", + "issue:43335|issue:44855", + "issue:45127|issue:45357", + "issue:44077|issue:44877", + "issue:43644|issue:44913" + ], + "prepared_review_unit_hash": "b6bceb6b257993b2adcf6bc2f3710f4c9dda6e4740b429e1cba036ff94ec0d28", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36638, + "estimated_input_tokens": 9160, + "estimated_eval_tokens": 18576 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the issues cover unrelated bugs across training, loading, generation, multimodal inputs, configs, and docs. I do not see any true duplicate pairs among the soft-edge candidates, so all soft edges are rejected. No PRs are present in the cluster.", + "confidence": 0.91, + "canonical_issue_reason": "No single issue cleanly represents the cluster because the items are not one underlying bug. If forced to pick the broadest/most central metadata-wise, issue 43873 has the most discussion and inbound references, but it still does not match the rest of the items.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43873 is the most central by activity and inbound links, so it is the best available anchor only from a triage/metadata standpoint. It is not a good semantic canonical for the cluster because the surrounding issues are unrelated.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "Both are model-generation regressions in broad terms, but they target different model families and different failure modes." + }, + { + "left": "issue:33290", + "right": "issue:36331", + "accept": false, + "reason": "OOM in DeepSpeed/Adafactor is unrelated to the custom compute_loss signature TypeError." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "One is tensor-parallel OOM on a large Qwen model; the other is a config/key error when loading MobileLLM." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "Same Sam3 family, but one is a dummy Conv2d concern and the other is a CUDA OOM in Sam3Video." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "Both are loading failures, but one is about gguf_file handling and the other is about an unrecognized model/config mismatch." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "These are different runtime failures with different entrypoints and symptoms; not the same bug." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "Buffer initialization junk and GPTNeoX rotary_pct reload behavior are distinct config/runtime issues." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Both mention cache behavior, but they affect different code paths and model implementations." + }, + { + "left": "issue:44936", + "right": "issue:45698", + "accept": false, + "reason": "Trainer lifecycle failure is unrelated to loading the wrong custom module after save_pretrained." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "A CLIPVision hidden-states regression is not the same as missing load-balancing loss in MoE routing." + }, + { + "left": "issue:43575", + "right": "issue:43873", + "accept": false, + "reason": "OOM on a specific model load is not the same as quantization/offloading not working as expected." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "A patchtsmixer post_init constraint issue is unrelated to the non-persistent buffer junk regression." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Both are generation-related regressions, but they involve different subsystems and different concrete bugs." + }, + { + "left": "issue:44464", + "right": "issue:44871", + "accept": false, + "reason": "Chunked generation inconsistency and eos_token_id config mismatch are unrelated problems." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "Sam3Video OOM and Whisper empty-transcription generation failure are unrelated." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Shape mismatch on loading Qwen3-VL is not the same as an incorrect tied-weights warning in Qwen3-Next." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "Both touch vision/model components, but one concerns a dummy Conv2d and the other an input-size mismatch in OmDet-Turbo." + }, + { + "left": "issue:43906", + "right": "issue:45290", + "accept": false, + "reason": "An isolated reproduction of an older issue is not the same as a chat template crash on tool-call assistant messages." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs still referencing pipeline() is not the same bug as SigLIP2 producing nonsensical outputs with pipeline." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "A failing Janus image-generation test and a Qwen3.5 flash-attention generation bug are different issues." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "The non-persistent buffer regression and GPTNeoXConfig reload behavior do not share the same root cause." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "CUDA assert on A800 with auto device mapping is unrelated to the category-count runtime limit." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Local dataset loading in a training script is unrelated to loading a multimodal Phi-4 model." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "A Janus model test failure is unrelated to stale auto_mappings references to removed configs." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Python 3.13 import/parsing failure in DebertaV2 is unrelated to flash_attention crashing on s_aux=None." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "Both are multimodal/template issues, but one is assistant-mask generation and the other is video vision_position_ids." + }, + { + "left": "issue:43335", + "right": "issue:44855", + "accept": false, + "reason": "A SwitchTransformers sparse-layer bug is unrelated to the DebertaV2 import/parsing failure." + }, + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "LoRA merge collapse with extended vocabulary is unrelated to incorrect visual encoder keys saved by Qwen3.5." + }, + { + "left": "issue:44077", + "right": "issue:44877", + "accept": false, + "reason": "PatchTSMixer post_init policy and strict config loading for granite_speech are different problems." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer junk and rotary_pct reverting on reload are different regressions with different scopes." + } + ] + }, + "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": 16, + "review_unit_index": 28, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:36246", + "issue:38175", + "issue:38617", + "issue:41628", + "issue:41950", + "issue:43065", + "issue:43116", + "issue:43295", + "issue:43335", + "issue:43404", + "issue:43479", + "issue:43540", + "issue:43550", + "issue:43643", + "issue:43650", + "issue:43761", + "issue:43976", + "issue:44016", + "issue:44112", + "issue:44186", + "issue:44190", + "issue:44220", + "issue:44246", + "issue:44464", + "issue:44466", + "issue:44496", + "issue:44561", + "issue:44683", + "issue:44734", + "issue:44743", + "issue:44855", + "issue:44913", + "issue:44933", + "issue:44938", + "issue:44964", + "issue:45003", + "issue:45216", + "issue:45245", + "issue:45276", + "issue:45362", + "issue:45375", + "issue:45406", + "issue:45446", + "issue:45542", + "issue:45563", + "issue:45584", + "issue:45600", + "issue:45632" + ], + "soft_pairs": [ + "issue:44938|issue:45542", + "issue:43976|issue:44855", + "issue:43550|issue:44190", + "issue:43761|issue:44743", + "issue:43540|issue:44734", + "issue:45276|issue:45584", + "issue:44112|issue:45600", + "issue:43650|issue:44220", + "issue:43295|issue:45276", + "issue:43335|issue:44190", + "issue:43116|issue:44190", + "issue:45245|issue:45584", + "issue:43404|issue:45216", + "issue:43479|issue:45375", + "issue:36246|issue:41628", + "issue:44186|issue:44734", + "issue:44933|issue:45362", + "issue:44464|issue:45563", + "issue:38175|issue:43065", + "issue:44466|issue:44913", + "issue:44561|issue:45632", + "issue:44683|issue:45446", + "issue:44734|issue:45632", + "issue:44246|issue:45003", + "issue:43643|issue:44496", + "issue:36246|issue:38617", + "issue:45406|issue:45632", + "issue:45563|issue:45584", + "issue:44964|issue:45600", + "issue:41950|issue:44016" + ], + "prepared_review_unit_hash": "a27762a12c0f13f2ea06dc6fe543b38379a92261f006f7156e94a55556253557", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36803, + "estimated_input_tokens": 9201, + "estimated_eval_tokens": 18658 + }, + "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 (imports, config loading, generation, CI, and model-specific bugs). I don\u2019t see any soft edge pair that is the same underlying bug/change, so none should be merged.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue really fits: the cluster is heterogeneous and the soft pairs span different models, code paths, and failure modes. If forced to pick one representative issue, 44561 is the broadest regression-style report, but it does not subsume the rest.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44561 is the most broadly scoped, systemic regression report in the set (trust_remote_code models breaking due to a removed helper) and is the closest thing to a cluster anchor, though the cluster is not actually cohesive.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Different failures: Python import/load problem vs tensorboard-without-TensorFlow backend error." + }, + { + "left": "issue:43976", + "right": "issue:44855", + "accept": false, + "reason": "Both involve Python versions, but one is a package compatibility/load issue and the other is a Python 3.13 parsing/IndentationError in a model import path." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "Unrelated subsystems: torch.compile/SDPA on Bamba vs local dataset loading in an image classification example." + }, + { + "left": "issue:43761", + "right": "issue:44743", + "accept": false, + "reason": "Both are state-related regressions, but they affect different models and different mechanisms (hidden states vs recurrent cache state)." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Video input validation in Qwen3OmniMoe is unrelated to KV-cache tensor indexing in transformers serve." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Gemma4 embedding resize bug and Whisper empty-transcription generation failure are distinct code paths." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "Both touch CI, but one is a stale GraniteSpeech test and the other is a broken auto_mappings reference to removed configs." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "No overlap beyond being issue reports; one is placeholder text, the other is an audio feature extraction bug." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Custom processor/tokenizer regression is unrelated to Gemma4 token embedding resize behavior." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug and image-classification dataset loading are different bugs." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "Example script multi-label output issue is not the same as local dataset loading in a separate training script." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "Different failures: category-count runtime limit vs Whisper empty-transcription handling." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Both mention tied weights/serialization, but they involve different models and different failure modes." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Different multimodal config bugs: default init when None vs a missing deepstack_visual_indexes field being dropped." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they are for different symbols and different import surfaces." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "Tokenizer padding/truncation crash is unrelated to serve KV-cache continuation indexing." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "Missing image_utils import is unrelated to Qwen3.5 chat crashes." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Chunked generation inconsistency under compilation and a stale warning about num_return_sequences are not the same bug." + }, + { + "left": "issue:38175", + "right": "issue:43065", + "accept": false, + "reason": "Unexpected zero probabilities in a SigLIP model has nothing to do with Sam3PixelDecoder's dummy Conv2d." + }, + { + "left": "issue:44466", + "right": "issue:44913", + "accept": false, + "reason": "One is inconsistent lm_head.weight serialization by device; the other is rotary_pct silently reverting on reload." + }, + { + "left": "issue:44561", + "right": "issue:45632", + "accept": false, + "reason": "Both involve trust_remote_code, but one is a removed helper breaking remote models and the other is a cache-path collision for local models." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Both touch flex_attention, but one is a torch>=2.9 compiled-fail issue and the other is an incorrect version check for AuxRequest import." + }, + { + "left": "issue:44734", + "right": "issue:45632", + "accept": false, + "reason": "Serve KV-cache tensor indexing and trust_remote_code cache-path collisions are unrelated problems." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Import slowdown and unsafe sys.modules access may be adjacent, but they are not the same concrete bug." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "Missing fields from trust_remote_code configs and unrecognized model_type in config.json are different config-loading failures." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Both are missing-import reports, but they refer to different symbols and different fixes." + }, + { + "left": "issue:45406", + "right": "issue:45632", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer is unrelated to trust_remote_code cache-path collisions." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "A stale warning and a generation failure after align_special_tokens are different issues." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "Phi-4 multimodal model loading failure and stale auto_mappings references to removed configs do not look like the same root cause." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "Pipeline misrouting to image processors is unrelated to a notebook syntax error." + } + ] + }, + "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": 16, + "review_unit_index": 29, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:41950", + "issue:42175", + "issue:42907", + "issue:43295", + "issue:43299", + "issue:43381", + "issue:43388", + "issue:43441", + "issue:43479", + "issue:43525", + "issue:43526", + "issue:43531", + "issue:43577", + "issue:43673", + "issue:43688", + "issue:43824", + "issue:43827", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44075", + "issue:44112", + "issue:44220", + "issue:44242", + "issue:44279", + "issue:44336", + "issue:44351", + "issue:44368", + "issue:44462", + "issue:44496", + "issue:44933", + "issue:44964", + "issue:45020", + "issue:45042", + "issue:45081", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45335", + "issue:45357", + "issue:45405", + "issue:45584", + "issue:45593", + "issue:45600", + "issue:45759" + ], + "soft_pairs": [ + "issue:43388|issue:43688", + "issue:43295|issue:45335", + "issue:44112|issue:44964", + "issue:42907|issue:43441", + "issue:43854|issue:44964", + "issue:44496|issue:45042", + "issue:43673|issue:44336", + "issue:43531|issue:45600", + "issue:43299|issue:45216", + "issue:44016|issue:44220", + "issue:43526|issue:43901", + "issue:42175|issue:44279", + "issue:45335|issue:45584", + "issue:43381|issue:45593", + "issue:44075|issue:44368", + "issue:43299|issue:45357", + "issue:44496|issue:45759", + "issue:43901|issue:44496", + "issue:44242|issue:45200", + "issue:43577|issue:45216", + "issue:41950|issue:45020", + "issue:44016|issue:44062", + "issue:43976|issue:45405", + "issue:43388|issue:43827", + "issue:44351|issue:45278", + "issue:43479|issue:43525", + "issue:44462|issue:45081", + "issue:43824|issue:44933" + ], + "prepared_review_unit_hash": "484124b200f6079599e090c5de4e43f90388f07c04c6e06e4b84e2d92710ce74", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35575, + "estimated_input_tokens": 8894, + "estimated_eval_tokens": 18044 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly separate regressions across different models, loaders, and utilities; they share a broad Transformers-v5 regression theme but not the same underlying bug. I would not merge any of the suggested soft pairs.", + "confidence": 0.9, + "canonical_issue_reason": "issue:45020 is the broadest open umbrella for recent-version regressions breaking remote_code model loading, so it is the best representative issue for this mixed cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 best captures the cluster\u2019s common theme without being overly narrow to one model family or code path.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "Different bugs: one drops label elements in distributed metric gathering; the other normalizes auxiliary loss for MoE models." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "Both involve regressions, but one is a processor/tokenizer API break and the other is resize_token_embeddings not updating decoder embeddings." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "CI stale test failure is unrelated to loading Phi-4 multimodal with latest transformers." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Different model families and failures: saving dequantized Ministral/Devstral vs FlashAttention load failure for Ministral-3." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Same general family only; one is a unit-test loading issue for GLM-4.7-Flash, the other is a Phi-4 multimodal load regression." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "Different problems: unrecognized config/model type vs PIL image processors incorrectly requiring torchvision." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Generation cache missing in chunked prefill is unrelated to ANSI codes emitted by loading_report." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to stale auto_mappings references breaking CI." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Both are Qwen-related regressions, but one is loading MoE models while the other is incorrect save_pretrained output for Qwen3.5." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error and _torch_extract_fbank_features() are unrelated bugs." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "BEiT image processor label reduction bug is unrelated to TextClassificationPipeline docs mentioning return_all_scores." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "Packaging/backend dependency issue for pip install '.[torch]' is not the same as a generic dependency issue with transformers." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "resize_token_embeddings affecting decoder.embed_tokens is unrelated to Whisper empty-transcription failure after align_special_tokens." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary losses are different training/evaluation bugs." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "Optimizer SGD args ignored is unrelated to a tie_word_embeddings warning in Qwen3.5 LoRA fine-tuning." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Both are Qwen3.5-related, but one concerns model loading and the other incorrect visual encoder keys in save_pretrained." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Unrecognized model config and AutoModelForCausalLM.from_config not unwrapping text_config are distinct config-resolution bugs." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "Docs mentioning return_all_scores is unrelated to AutoModel config recognition failure." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "Load-balancing loss omission and Gemma 4 mm_token_type_ids defaulting are different model-training issues." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "Blip2 dtype handling and Qwen3.5 save_pretrained regression are unrelated." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "Video-classification image-processor lookup is a specific pipeline bug, not the broader remote_code breakage." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error and AddedToken keyword conflict are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Python version compatibility and an unreleased MIN_PEFT_VERSION bump are different release issues." + }, + { + "left": "issue:43388", + "right": "issue:43827", + "accept": false, + "reason": "gather_for_metrics label handling is unrelated to docs still referencing removed pipeline() APIs." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "HybridCache import failure is not the same as broad import errors after upgrading to 5.5.0." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Phi4MultimodalConfig default init bug and Llama4Config missing pad_token_id are different config-class issues." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json is unrelated to _patch_mistral_regex crashing on a Tokenizer object." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Missing Qwen2_5_VL export and nonexistent image_utils import are separate import/export problems." + } + ] + }, + "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": 16, + "review_unit_index": 30, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:32090", + "issue:35141", + "issue:38175", + "issue:39692", + "issue:41950", + "issue:42994", + "issue:43065", + "issue:43122", + "issue:43295", + "issue:43299", + "issue:43317", + "issue:43441", + "issue:43450", + "issue:43650", + "issue:43653", + "issue:43723", + "issue:43784", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:43994", + "issue:44162", + "issue:44261", + "issue:44263", + "issue:44279", + "issue:44297", + "issue:44361", + "issue:44462", + "issue:44466", + "issue:44521", + "issue:44617", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:44898", + "issue:44987", + "issue:45276", + "issue:45290", + "issue:45325", + "issue:45397", + "issue:45405", + "issue:45542", + "issue:45593", + "issue:45600", + "issue:45712" + ], + "soft_pairs": [ + "issue:32090|issue:35141", + "issue:44683|issue:45542", + "issue:44279|issue:45405", + "issue:39692|issue:43901", + "issue:43295|issue:44779", + "issue:43299|issue:43854", + "issue:44297|issue:44462", + "issue:38175|issue:44617", + "issue:41950|issue:43650", + "issue:44462|issue:44779", + "issue:44521|issue:45325", + "issue:43976|issue:44279", + "issue:43976|issue:44987", + "issue:45600|issue:45712", + "issue:43450|issue:44279", + "issue:44466|issue:45276", + "issue:43065|issue:44898", + "issue:42994|issue:43317", + "issue:44261|issue:44263", + "issue:43650|issue:45290", + "issue:43441|issue:44683", + "issue:44623|issue:45397", + "issue:43723|issue:44162", + "issue:44623|issue:44749", + "issue:44671|issue:44779", + "issue:43653|issue:45593", + "issue:43784|issue:45542", + "issue:44361|issue:45593", + "issue:43901|issue:43994", + "issue:43122|issue:43650", + "issue:44987|issue:45542" + ], + "prepared_review_unit_hash": "ce2b37c2a696b79aaedbcd899788eb4a669d830b771034c2cccc8cfbfb0ec4ae", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36284, + "estimated_input_tokens": 9071, + "estimated_eval_tokens": 18398 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is a loose grab-bag of unrelated issues across tokenization, processor/model loading, quantization, attention kernels, docs, and CI. None of the suggested soft pairs look like the same concrete bug or change.", + "confidence": 0.91, + "canonical_issue_reason": "No true canonical duplicate stands out; the issues are heterogeneous. If one must be chosen as the representative report, issue 44521 is the clearest high-signal regression with an explicit multimodal API failure and active discussion.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44521 is the strongest representative issue in the set because it has a concrete repro, clear impact, and ongoing discussion; however, it is not a duplicate hub for the other items.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer broadcast TypeError and token-embedding reinitialization are different bugs in different code paths." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "FlashAttention/torch compatibility and tensorboard-without-TF backend errors are unrelated." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "Both are dependency-related, but one is a vague dependency complaint while the other is a specific unreleased PEFT version bump." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "Both mention docs, but one is a SigLIP2 example with model/processor issues and quantization failure; the other is a TextClassificationPipeline doc mismatch." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer regressions affecting different APIs/models and failure modes." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Different model-loading failures for different model families; not the same underlying bug." + }, + { + "left": "issue:44297", + "right": "issue:44462", + "accept": false, + "reason": "Tokenizer save metadata mismatch and AutoTokenizer ignoring repository tokenizer.json are distinct serialization/load problems." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "Unexpected probabilities vs CUDA OOM are unrelated inference quality and memory issues." + }, + { + "left": "issue:41950", + "right": "issue:43650", + "accept": false, + "reason": "Video-classification processor selection and an empty title issue have no substantive overlap." + }, + { + "left": "issue:44462", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer bugs: repository tokenizer.json loading versus Deepseek tokenization changes in v5." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "Both are multimodal/VL-related, but one is assistant mask generation and the other is temporal position-id scaling; different code paths." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "Python-version incompatibility and a generic dependency issue are not the same bug." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "Different failures: one is Python support, the other is loading a specific model on newer transformers." + }, + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Both are CI/repo hygiene, but they involve different files and different failure mechanisms." + }, + { + "left": "issue:43450", + "right": "issue:44279", + "accept": false, + "reason": "Video processor batch shape bug and a dependency issue are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Both touch tied/output embeddings, but one is device-dependent serialization and the other is Gemma4 resize_token_embeddings propagation." + }, + { + "left": "issue:43065", + "right": "issue:44898", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder and Perceiver non-default-resolution failure are unrelated model bugs." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Quantized model saving failure and dequantized model loading with offload are different code paths." + }, + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "Different GLM-related bugs: missing rms_norm_eps precision issue versus torch.split return handling." + }, + { + "left": "issue:43650", + "right": "issue:45290", + "accept": false, + "reason": "An empty-title issue and a chat-template crash on tool-call assistant messages are unrelated." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "Both involve attention kernels, but they are different kernels, models, and failure modes." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "Processor.save_pretrained missing files and Gemma-4 zero3 from_pretrained are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "Tokenizer loading regression and ESM2 model breakage are different problems." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "Processor save files and tokenizer/filtering slowdown are unrelated." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "CamemBERT masked-LM prediction regression and Deepseek tokenization regression are different bugs." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "BigBirdTokenizer special-token decoding bug and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "Sentence-transformers import NameError and tensorboard/TensorFlow backend error are different dependency/import issues." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "MLukeTokenizer task AttributeError and D-FINE auxiliary-loss omission are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "Docs mention of return_all_scores versus actual SigLIP2 incorrect outputs are different issue types and code paths." + }, + { + "left": "issue:43122", + "right": "issue:43650", + "accept": false, + "reason": "Tokenizer behavior change across versions and an empty-title issue are unrelated." + }, + { + "left": "issue:44987", + "right": "issue:45542", + "accept": false, + "reason": "Model-loading regression and tensorboard/TensorFlow backend error 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": 16, + "review_unit_index": 31, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:36246", + "issue:41628", + "issue:41950", + "issue:42491", + "issue:42757", + "issue:43262", + "issue:43299", + "issue:43352", + "issue:43450", + "issue:43475", + "issue:43525", + "issue:43673", + "issue:43761", + "issue:43824", + "issue:43881", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44162", + "issue:44188", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44291", + "issue:44295", + "issue:44297", + "issue:44336", + "issue:44351", + "issue:44373", + "issue:44462", + "issue:44485", + "issue:44496", + "issue:44521", + "issue:44561", + "issue:44589", + "issue:44617", + "issue:44857", + "issue:44871", + "issue:44964", + "issue:45200", + "issue:45250", + "issue:45356", + "issue:45362", + "issue:45397", + "issue:45593", + "issue:45759" + ], + "soft_pairs": [ + "issue:44016|issue:44291", + "issue:44297|issue:45397", + "issue:43262|issue:44521", + "issue:44188|issue:44336", + "issue:43673|issue:43761", + "issue:43475|issue:44016", + "issue:43525|issue:45759", + "issue:43525|issue:43881", + "issue:43450|issue:44617", + "issue:44589|issue:45362", + "issue:43901|issue:44246", + "issue:42491|issue:43299", + "issue:44336|issue:44373", + "issue:41628|issue:43976", + "issue:16998|issue:22355", + "issue:44242|issue:45593", + "issue:43352|issue:45250", + "issue:44295|issue:45397", + "issue:44246|issue:44496", + "issue:44462|issue:45356", + "issue:41950|issue:44220", + "issue:44336|issue:44485", + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:43673|issue:44561", + "issue:43824|issue:43976", + "issue:44246|issue:44871", + "issue:43976|issue:44964", + "issue:44242|issue:44857", + "issue:44162|issue:45200" + ], + "prepared_review_unit_hash": "cf84359d19b1747fc5f8bdba4df3bcfb6279ecff605a1195c32ddb261481a6d4", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36485, + "estimated_input_tokens": 9122, + "estimated_eval_tokens": 18500 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly unrelated Transformers issues that only share broad subsystem keywords. None of the soft pairs look like true duplicates of the same bug/change, so I would not merge any of them.", + "confidence": 0.73, + "canonical_issue_reason": "Issue 43824 has the strongest external signal and community impact in this set (multiple inbound references), making it the best anchor issue among these heterogeneous reports, though it is not a duplicate umbrella for the rest.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43824 is the most visible and broadly referenced report here, so it is the best single issue for triage and follow-up.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Completely different bugs: notebook syntax error vs. init_empty_weights argument handling." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Different failure modes and model families; tokenizer_class metadata mismatch is unrelated to gemma-4 ZeRO3 loading." + }, + { + "left": "issue:43262", + "right": "issue:44521", + "accept": false, + "reason": "Both involve multimodal chat/template code, but one is audio sampling-rate defaults and the other is assistant mask generation; not the same bug." + }, + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "Attention-kernel divergence under torch.compile is unrelated to ANSI escaping in loading reports." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Generation cache handling during chunked_prefill is unrelated to CLIPVisionModel hidden_states regression." + }, + { + "left": "issue:43475", + "right": "issue:44016", + "accept": false, + "reason": "SAM3 video encoder attribute error vs. notebook syntax error; no shared code path." + }, + { + "left": "issue:43525", + "right": "issue:45759", + "accept": false, + "reason": "Both are config-related, but one is a missing pad_token_id on Llama4Config and the other is composite Qwen text_config unwrapping; different defects." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Different model-specific loading issues with different missing attributes and symptoms." + }, + { + "left": "issue:43450", + "right": "issue:44617", + "accept": false, + "reason": "A batched-shape bug in video processors is not the same as a CUDA OOM in Sam3Video." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Float8 storage lookup failure is a serialization/runtime issue, not the same as a Qwen3.5 chat crash." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "Docs mismatch for return_all_scores is unrelated to import latency." + }, + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both mention Qwen/MoE compatibility, but they affect different model types and loading paths; not clearly the same bug." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "ANSI codes in loading_report and a wrong docstring for position_ids are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "Missing AutoImageProcessor export is unrelated to a Python version compatibility regression." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "A model_max_length question and a missing transformers.onnx module are unrelated." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "Both are loss-related, but they are different models and different gating conditions; not the same concrete bug." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "Generic Flash Attention 2.0 discussion is not the same as a specific model explicitly lacking FA2 support." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "Position_ids buffer access error is unrelated to gemma-4 ZeRO3 loading." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Import performance issue is unrelated to model-type resolution failure." + }, + { + "left": "issue:44462", + "right": "issue:45356", + "accept": false, + "reason": "Tokenizer ignores tokenizer.json vs. codec handling regression and warning text; different tokenizer bugs." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification processor selection and audio feature extraction are different pipelines with different failures." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "Loading-report ANSI output is unrelated to GLM-5 RoPE implementation." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are ImportErrors, but they involve different missing symbols in different packages." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Different missing imports/exported symbols; no evidence of the same underlying packaging bug." + }, + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Cache handling in chunked_prefill and removal of is_torch_fx_available are separate regressions." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Missing Qwen2_5_VLForConditionalGeneration export is unrelated to a Python 3.9/3.10 support regression." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Import slowness is unrelated to Gemma-3 eos_token_id configuration inconsistency." + }, + { + "left": "issue:43976", + "right": "issue:44964", + "accept": false, + "reason": "Python-version compatibility and Phi-4-multimodal loading failure are different issues." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Different loss implementations and different runtime conditions; not the same bug." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "ESM2 breakage and Gemma 4 mm_token_type_ids defaults are unrelated model-specific issues." + } + ] + }, + "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": 16, + "review_unit_index": 32, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:36246", + "issue:41950", + "issue:42886", + "issue:43065", + "issue:43232", + "issue:43257", + "issue:43262", + "issue:43408", + "issue:43441", + "issue:43475", + "issue:43526", + "issue:43540", + "issue:43596", + "issue:43653", + "issue:43901", + "issue:44162", + "issue:44246", + "issue:44295", + "issue:44336", + "issue:44368", + "issue:44451", + "issue:44466", + "issue:44496", + "issue:44749", + "issue:44779", + "issue:44877", + "issue:44977", + "issue:44987", + "issue:44995", + "issue:45003", + "issue:45081", + "issue:45137", + "issue:45200", + "issue:45229", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45375", + "issue:45397", + "issue:45459", + "issue:45561", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45656", + "issue:45701", + "issue:45759" + ], + "soft_pairs": [ + "issue:45081|issue:45362", + "issue:44779|issue:45292", + "issue:44295|issue:44749", + "issue:43441|issue:43540", + "issue:45292|issue:45584", + "issue:44877|issue:45375", + "issue:44779|issue:45701", + "issue:45292|issue:45701", + "issue:44987|issue:45459", + "issue:43901|issue:45003", + "issue:45561|issue:45563", + "issue:43232|issue:44336", + "issue:41950|issue:44451", + "issue:36246|issue:43596", + "issue:43065|issue:43408", + "issue:43526|issue:45003", + "issue:45375|issue:45759", + "issue:43257|issue:45137", + "issue:44496|issue:45003", + "issue:44995|issue:45200", + "issue:43262|issue:44246", + "issue:44466|issue:45335", + "issue:43475|issue:43540", + "issue:42886|issue:45229", + "issue:44995|issue:45588", + "issue:45397|issue:45656", + "issue:43653|issue:44162", + "issue:44368|issue:45563", + "issue:44368|issue:44977" + ], + "prepared_review_unit_hash": "28ed0275ff8949ae0639fce186478a544963a2a9fe8079bdfd6806ba92dc43e3", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35988, + "estimated_input_tokens": 8997, + "estimated_eval_tokens": 18250 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: it mixes unrelated tokenizer, config, generation, DeepSpeed, SAM, audio, docs, and performance issues. No soft pair looks like the same underlying bug/change, and there are no PRs here.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue stands out because the items do not share one underlying defect; choosing any single issue would be arbitrary.", + "canonical_pr_reason": null, + "best_issue_reason": "No single issue is a good global representative for duplicate triage; the cluster is too broad and unrelated.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45081", + "right": "issue:45362", + "accept": false, + "reason": "Different models and failure modes: Mistral tokenizer regex patch crash vs Qwen3.5 chat crash." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "One is tokenization regression in DeepSeek; the other is resize_token_embeddings not updating output embeddings." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "Unrelated bugs: position_ids buffer access error vs a post-upgrade performance slowdown." + }, + { + "left": "issue:43441", + "right": "issue:43540", + "accept": false, + "reason": "Different subsystems and symptoms: FlashAttention on Ministral-3 vs video input validation in Qwen3OmniMoe." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "Embedding resize bug and Whisper empty-transcription failure are separate code paths." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Both mention config strictness, but they target different model configs and different missing-field behaviors." + }, + { + "left": "issue:44779", + "right": "issue:45701", + "accept": false, + "reason": "Both concern tokenization changes, but one is a specific DeepSeek regression and the other is a broad version-dependent tokenization complaint." + }, + { + "left": "issue:45292", + "right": "issue:45701", + "accept": false, + "reason": "Embedding resize behavior is unrelated to a broad tokenization-version regression." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "Loading failure for a specific model is unrelated to protobuf exception masking in tokenizer decode errors." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Docs mismatch for TextClassificationPipeline is unrelated to unsafe sys.modules access in modeling_utils." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "pytest-xdist file race and stale generate() warning are distinct issues." + }, + { + "left": "issue:43232", + "right": "issue:44336", + "accept": false, + "reason": "Generation kwargs handling after sync_gpus is unrelated to ANSI formatting in loading_report." + }, + { + "left": "issue:41950", + "right": "issue:44451", + "accept": false, + "reason": "Video-classification pipeline processor lookup and ScandiBERT model loading are unrelated." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "ImportError for a missing processor class and a DeepSpeed/BertModel index error are different bugs." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "Both involve SAM3, but one is a dummy layer implementation issue and the other is a model-type mismatch warning." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "BEiT label reduction bug is unrelated to unsafe sys.modules access." + }, + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Both touch composite Qwen config handling, but one drops a field under @strict while the other fails to unwrap text_config in from_config." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Different DeepSpeed problems: missing MOE weight conversion vs ZeRO3 deque underflow." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model/config loading is unrelated to sys.modules access in model utilities." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "A stale cache crash on second forward is not the same as missing mm_token_type_ids defaults." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio chat template sampling-rate defaulting and slow imports are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Serialization of tied weights depending on device is different from resize_token_embeddings not updating decoder embeddings." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "Sam3 vision encoder missing an attribute is unrelated to Qwen3OmniMoe video-input validation." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Offline cache loading for tokenizers and Gemma4 multi-GPU OOM are unrelated." + }, + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "Different failure points: stale indexer cache on second forward vs flash_attention.py crashing on sink-less models." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Gemma4 ZeRO3 from_pretrained and double optimizer step under DeepSpeed are separate problems." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "BigBirdTokenizer special-token registration bug is unrelated to ESM2 being broken." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "A tie_word_embeddings warning in Qwen3.5 fine-tuning is unrelated to a stale num_return_sequences warning." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "Both are Qwen3.5-related, but one is a config warning and the other is a flash-attention generation failure." + } + ] + }, + "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": 16, + "review_unit_index": 33, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:41628", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:42994", + "issue:43232", + "issue:43452", + "issue:43475", + "issue:43482", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43704", + "issue:43827", + "issue:43931", + "issue:44038", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44297", + "issue:44361", + "issue:44368", + "issue:44623", + "issue:44661", + "issue:44734", + "issue:44749", + "issue:44869", + "issue:44933", + "issue:45003", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45200", + "issue:45230", + "issue:45310", + "issue:45356", + "issue:45375", + "issue:45381", + "issue:45397", + "issue:45459", + "issue:45478", + "issue:45561", + "issue:45600", + "issue:45656" + ], + "soft_pairs": [ + "issue:44060|issue:44077", + "issue:45137|issue:45600", + "issue:43673|issue:44734", + "issue:45003|issue:45561", + "issue:43931|issue:44368", + "issue:44933|issue:45459", + "issue:44869|issue:45356", + "issue:44297|issue:44623", + "issue:44361|issue:44749", + "issue:43646|issue:44162", + "issue:43452|issue:45003", + "issue:44623|issue:45656", + "issue:43673|issue:43827", + "issue:43232|issue:43673", + "issue:28282|issue:30990", + "issue:42994|issue:43482", + "issue:44749|issue:45656", + "issue:44749|issue:45310", + "issue:43475|issue:43643", + "issue:45230|issue:45397", + "issue:44038|issue:44623", + "issue:42371|issue:45092", + "issue:43475|issue:45381", + "issue:45081|issue:45103", + "issue:41628|issue:44933", + "issue:42673|issue:43704", + "issue:42994|issue:43525", + "issue:43475|issue:44661", + "issue:45375|issue:45478", + "issue:45310|issue:45375", + "issue:42222|issue:45200" + ], + "prepared_review_unit_hash": "b6bcce7cba1c5a25f0fee41eeedba27ad2f49b634e73084333853a14a123a9df", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36943, + "estimated_input_tokens": 9236, + "estimated_eval_tokens": 18728 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most items are distinct model- or subsystem-specific regressions. Only two soft edges look like true duplicates: the repeated Qwen3.5-Moe from_pretrained error and the repeated Qwen3ForCausalLM VRAM leak. The rest share only broad themes (generation, tokenizer, save/load, v5 regressions) and should stay separate.", + "confidence": 0.93, + "canonical_issue_reason": "issue:43646 is the best umbrella issue: it describes a broad Transformers v5 custom-model initialization regression, which is the closest cluster-wide anchor among the issues here.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the most general and reusable issue to represent the cluster; the other candidates are narrower, model-specific fallout or unrelated bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "Both are model-specific warnings/errors, but they are different bugs in different code paths (tie-word-embeddings warning vs optional post_init handling)." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque handling and auto_mappings cleanup are unrelated bugs despite both affecting training/tooling." + }, + { + "left": "issue:43673", + "right": "issue:44734", + "accept": false, + "reason": "Both concern generation/KV-cache code, but one is a missing cache field in chunked_prefill and the other is a tensor-indexing crash in /v1/responses continuation; not the same defect." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "Unsafe sys.modules access and the xdist captured_info.txt race are different failures in different subsystems." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch and a Qwen3.5 LoRA tie-word-embeddings warning are not the same underlying bug." + }, + { + "left": "issue:44933", + "right": "issue:45459", + "accept": false, + "reason": "A missing import from image_utils is unrelated to tokenizer error masking when protobuf is absent." + }, + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Whisper timestamp decode crash and Kimi-K2.5 tokenizer codec/regression handling are different concrete tokenizer bugs." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "Both are save_pretrained-related, but one is tokenizer metadata mismatch and the other is processor file omission; too different to merge." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "MLukeTokenizer task AttributeError and the reported performance slowdown after upgrade are unrelated." + }, + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "44162 looks like a specific model breakage, but there is not enough evidence that it is the same concrete failure as the broad custom-model initialization regression." + }, + { + "left": "issue:43452", + "right": "issue:45003", + "accept": false, + "reason": "GGUF from_pretrained breakage and sys.modules access safety are separate code-path problems." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "Processor save_pretrained file output and a double optimizer-step bug under deepspeed are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:43827", + "accept": false, + "reason": "Generation cache handling and docs still referencing removed pipeline() APIs are not the same bug." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "These are both generation-related, but one is a sync_gpus kwargs update issue and the other is a chunked_prefill cache regression." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "ImportError about missing PyTorch and Sentence Transformers hanging on load are different user-facing failures." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Quantized saving failure and Qwen2.5-GGUF loading failure in v5 are different directions and different bugs." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "A performance regression after upgrading versions is not the same as deepspeed causing optimizer.step() to run twice." + }, + { + "left": "issue:44749", + "right": "issue:45310", + "accept": false, + "reason": "Slow filtering after upgrade and Qwen3.5-Moe from_pretrained error are unrelated regressions." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "SAM3 vision output attribute loss and trust_remote_code returning missing config fields are different problems." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "A placeholder bug report and Gemma-4 ZeRO3 from_pretrained failure are not the same underlying issue." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe v5 issues and missing processor save artifacts are different bugs." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 settings deprecation and InternVL2 meta-init incompatibility are unrelated." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "SAM3 encoder output missing fields and Qwen2.5-VL video position IDs are different model-specific defects." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "Tokenizer regex patching and auto_docstring future-annotations handling are unrelated crash paths." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Both are import-related, but one is AutoImageProcessor import availability and the other is a missing image_utils import; not the same concrete bug." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Same title and same reported behavior: Qwen3ForCausalLM leaks VRAM in multiple dataloader threads." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized model saving failure and missing pad_token_id on Llama4Config are different issues." + }, + { + "left": "issue:43475", + "right": "issue:44661", + "accept": false, + "reason": "A SAM3 output attribute bug and add-new-model-like failing on TOKENIZER_MAPPING_NAMES are unrelated." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Both are Qwen3.5-Moe-related, but one is a specific config field being dropped and the other is a from_pretrained failure; not enough to treat as the same bug." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "The from_pretrained error and the missing deepstack_visual_indexes config field are related by model family only, not by the same concrete failure." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "VitPose breakage and Gemma-4 token-type defaults are unrelated model-specific bugs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster choice is grounded in the packet, and the soft-edge decisions are conservative overall. The only accepted duplicate pair is well-supported by identical titles/behavior, and the false rejections avoid over-claiming sameness from broad thematic overlap. Minor note: the summary mentions the repeated Qwen3.5-Moe from_pretrained error as a true duplicate, but that exact pair is not one of the listed soft edges; still, this does not materially undermine the evaluation." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 34, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:33290", + "issue:35141", + "issue:36010", + "issue:41628", + "issue:42175", + "issue:42831", + "issue:42898", + "issue:42994", + "issue:43232", + "issue:43278", + "issue:43377", + "issue:43479", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43901", + "issue:44038", + "issue:44060", + "issue:44079", + "issue:44164", + "issue:44188", + "issue:44222", + "issue:44242", + "issue:44295", + "issue:44351", + "issue:44373", + "issue:44448", + "issue:44496", + "issue:44568", + "issue:44623", + "issue:44734", + "issue:44779", + "issue:44792", + "issue:44964", + "issue:45092", + "issue:45103", + "issue:45125", + "issue:45137", + "issue:45310", + "issue:45356", + "issue:45405", + "issue:45561", + "issue:45600", + "issue:45656", + "issue:45712" + ], + "soft_pairs": [ + "issue:42831|issue:44222", + "issue:41628|issue:44351", + "issue:44222|issue:45310", + "issue:42994|issue:44164", + "issue:43377|issue:44242", + "issue:36010|issue:41628", + "issue:42175|issue:45712", + "issue:44038|issue:45310", + "issue:44964|issue:45137", + "issue:44496|issue:45561", + "issue:45561|issue:45600", + "issue:42898|issue:44568", + "issue:30990|issue:35141", + "issue:43901|issue:45561", + "issue:43232|issue:44792", + "issue:43232|issue:44188", + "issue:44188|issue:44373", + "issue:43278|issue:45137", + "issue:44623|issue:45310", + "issue:43784|issue:43824", + "issue:30333|issue:33290", + "issue:44734|issue:45103", + "issue:43723|issue:44568", + "issue:42898|issue:45356", + "issue:42898|issue:44779", + "issue:44060|issue:45125", + "issue:45405|issue:45712", + "issue:43874|issue:45092", + "issue:42898|issue:44448", + "issue:43479|issue:44079", + "issue:44188|issue:44964", + "issue:44295|issue:45656" + ], + "prepared_review_unit_hash": "40571a51d021bb4a9facc6fe18684887c1c4e12bcc574d86b76966c1cfaf0cfa", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37269, + "estimated_input_tokens": 9318, + "estimated_eval_tokens": 18892 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a heterogeneous collection of unrelated Transformers issues (tokenizer regressions, multimodal import/load failures, quantization, DeepSpeed, docs, and packaging). None of the soft-edge pairs look like the same underlying bug/change, so they should all remain separate.", + "confidence": 0.79, + "canonical_issue_reason": "No single true canonical fits this mixed cluster; if one issue must be chosen, #43723 is the broadest user-facing regression here (AutoTokenizer.from_pretrained loading in v5), but it does not actually subsume the rest.", + "canonical_pr_reason": null, + "best_issue_reason": "#43723 is the best representative only because it is the most general, well-scoped tokenizer-loading regression among the set; the other issues are too diverse to serve as duplicates of it.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "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 failure for MoE; different bug classes." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they concern different symbols (`AutoImageProcessor` vs `HybridCache`) and different code paths." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "Both involve MoE/FP8 and loading/saving, but one is a save_pretrained bug and the other is a from_pretrained error; not the same defect." + }, + { + "left": "issue:42994", + "right": "issue:44164", + "accept": false, + "reason": "Both are save/load problems, but one is quantized-model saving and the other is handling `extra_state`; too different to merge." + }, + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "One is a missing padding-mask bug in MIMI encoder outputs; the other is load-balancing loss logic when `output_router_logits=False`." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Different missing imports (`GenerationMixin` vs `AutoImageProcessor`) and different failure surfaces." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Both touch packaging/dependency behavior, but one is about missing TensorFlow in a torch install and the other is dummy classes leaking into the namespace." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Same model family, but one is a Transformers 5.0 Qwen3-VL-Moe bug and the other is a Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Completely different areas: multimodal model loading versus DeepSpeed ZeRO3 deque underflow." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Unrecognized model/config loading versus a pytest-xdist file race; unrelated bugs." + }, + { + "left": "issue:45561", + "right": "issue:45600", + "accept": false, + "reason": "Both are CI/test-related, but one is a parallel test race and the other is stale auto_mappings references." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Both are tokenizer regressions in v5, but they affect different behaviors (`clean_up_tokenization_spaces` vs `add_special_tokens`) and different models." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "Sentence-transformers loading hang is unrelated to token embedding resizing/reinitialization." + }, + { + "left": "issue:43901", + "right": "issue:45561", + "accept": false, + "reason": "Docs mismatch for `return_all_scores` versus a test utility race; not the same underlying issue." + }, + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Generation kwarg update after `sync_gpus` is unrelated to janus image-generation test failure." + }, + { + "left": "issue:43232", + "right": "issue:44188", + "accept": false, + "reason": "Both are generation/attention code, but one is a `sync_gpus` generation bug and the other is torch.compile branching in attention kernels." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "Attention-kernel divergence is unrelated to a wrong docstring for `position_ids`." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "dtype change during evaluate versus a ZeRO3 deque pop error; unrelated." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Processor save_pretrained missing files and Qwen3.5 MoE from_pretrained failure are different save/load issues." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but one is a third-party sentence-transformers compatibility issue and the other is a missing model class export." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "MLFlow reporting on training failure is unrelated to DeepSpeed OOM with Adafactor." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Serving KV-cache indexing crash and auto_docstring annotation handling are unrelated subsystems." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer loading in v5 and missing BOS/EOS insertion are distinct tokenizer bugs." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer regressions, but they concern different models and different codec/regex behavior." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "General tokenizer behavior change versus DeepSeek-specific incorrect tokenization; not the same bug." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Both are Qwen3.5 MoE-related, but one is a tied-weights warning and the other is missing tensor-parallel plan support." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "A version pin to unreleased PEFT is unrelated to leftover dummy classes in repo checks." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Both involve multimodal model loading, but one is a missing image-patch method and the other is remote-code/meta-init incompatibility." + }, + { + "left": "issue:42898", + "right": "issue:44448", + "accept": false, + "reason": "Tokenizer cleanup behavior in v5 versus Pegasus output differences between v4 and v5; related theme, different concrete bug." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Config default initialization and ModelOutput key assignment are different code paths and failure modes." + }, + { + "left": "issue:44188", + "right": "issue:44964", + "accept": false, + "reason": "Attention-kernel divergence under compile is unrelated to Phi-4 multimodal model loading." + }, + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "Position_ids buffer access and DeepSpeed double optimizer step are different runtime bugs with no shared 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": 2, + "component_count": 16, + "review_unit_index": 35, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:22355", + "issue:28282", + "issue:30333", + "issue:30990", + "issue:34689", + "issue:35141", + "issue:36331", + "issue:38175", + "issue:41628", + "issue:42175", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42898", + "issue:42994", + "issue:43295", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:43873", + "issue:43976", + "issue:44206", + "issue:44222", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44514", + "issue:44623", + "issue:44869", + "issue:44871", + "issue:44913", + "issue:44933", + "issue:44945", + "issue:44977", + "issue:45092", + "issue:45103", + "issue:45310", + "issue:45357", + "issue:45406", + "issue:45459", + "issue:45656", + "issue:45712" + ], + "soft_pairs": [ + "issue:42898|issue:43295", + "issue:44933|issue:45712", + "issue:42175|issue:44933", + "issue:41628|issue:45459", + "issue:44222|issue:44297", + "issue:44297|issue:45656", + "issue:43976|issue:44933", + "issue:43479|issue:44869", + "issue:44206|issue:45103", + "issue:42831|issue:44623", + "issue:44222|issue:44623", + "issue:30333|issue:36331", + "issue:41628|issue:42757", + "issue:30333|issue:34689", + "issue:43723|issue:44977", + "issue:44297|issue:45310", + "issue:28282|issue:30333", + "issue:30333|issue:35141", + "issue:44263|issue:44360", + "issue:44222|issue:44484", + "issue:43873|issue:44945", + "issue:44483|issue:45406", + "issue:44263|issue:45092", + "issue:44373|issue:44485", + "issue:43479|issue:43644", + "issue:44514|issue:45103", + "issue:44295|issue:44623", + "issue:43723|issue:44871", + "issue:45092|issue:45357", + "issue:44261|issue:45092", + "issue:43479|issue:44913", + "issue:30990|issue:38175", + "issue:22355|issue:28282", + "issue:42886|issue:42994" + ], + "prepared_review_unit_hash": "a5aba265cf0e7fbb56eb585d9afb44d3040b518553c02cf74815689adb2f9763", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37655, + "estimated_input_tokens": 9414, + "estimated_eval_tokens": 19084 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: it mixes unrelated import errors, save/load regressions, training/runtime bugs, and a few API/serialization issues. None of the soft pairs look like the same underlying defect, so I would not merge any of them as duplicates.", + "confidence": 0.69, + "canonical_issue_reason": "If one issue must anchor the cluster, issue 44933 is the cleanest narrow import/export failure and the closest representative of the small API-surface breakage subset. That said, the overall set is too diverse for a strong duplicate canonical.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44933 is the best single representative only because it is concrete, scoped, and closest to several other import/export-themed reports; it is not a true umbrella for the whole cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42898", + "right": "issue:43295", + "accept": false, + "reason": "Different regressions: tokenization cleanup behavior vs processor/tokenizer access in a multimodal custom model path." + }, + { + "left": "issue:44933", + "right": "issue:45712", + "accept": false, + "reason": "Both touch import/export surfaces, but one is a missing import in image_utils and the other is dummy class leakage in a different file; not the same bug." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Packaging/backend dependency omission is unrelated to a missing import from image_utils." + }, + { + "left": "issue:41628", + "right": "issue:45459", + "accept": false, + "reason": "AutoImageProcessor import failure and protobuf-related tokenizer error handling are different code paths." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "Both involve save_pretrained, but one is FP8/MoE save failure and the other is tokenizer metadata mismatch." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer save metadata bug is unrelated to a deepspeed optimizer step-count issue." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python-version compatibility failure does not match a missing internal import." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "A multimodal config default-init bug is unrelated to Whisper timestamp decoding crashing on a trailing replacement character." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "Feature-extractor argument regression and auto_docstring annotation parsing crash are different failures." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FP8 accuracy problem is not the same as processor.save_pretrained missing files." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "Different save_pretrained bugs: FP8/MoE serialization versus missing processor assets." + }, + { + "left": "issue:30333", + "right": "issue:36331", + "accept": false, + "reason": "MLflow reporting on failed training jobs is unrelated to a CustomTrainer compute_loss signature break." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Different import failures from different dependencies; one is transformers API exposure, the other is huggingface_hub symbol availability." + }, + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "Training-job tracking and Llama 3.2 vision model loading are unrelated regressions." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer loading in v5 and Qwen3.5 flash-attention generation failures are distinct model/runtime issues." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer save metadata mismatch is not the same as Qwen3.5 MoE from_pretrained failing on newer transformers." + }, + { + "left": "issue:28282", + "right": "issue:30333", + "accept": false, + "reason": "Missing PyTorch dependency for AutoModel is unrelated to MLflow job reporting on failure." + }, + { + "left": "issue:30333", + "right": "issue:35141", + "accept": false, + "reason": "Training-job reporting and output-embedding reinitialization during token resize are different bugs." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "Both mention the DSA indexer, but one is about torch.split return handling and the other about a missing ReLU; not the same defect." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "FP8 MoE save failure and max_shard_size default question are not the same issue." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Quantization offloading behavior and pipeline-parallelism output correctness are separate runtime problems." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "OpenAI chat endpoint request rejection and Gemma4Processor missing _tokenizer in serve are different failures." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "DSA indexer return-value issue is unrelated to remote-code checkpoints failing meta initialization in v5." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "Docstring correctness and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Config initialization for Phi4Multimodal is not the same as junk-filled non-persistent buffers in Transformers 5.0.0." + }, + { + "left": "issue:44514", + "right": "issue:45103", + "accept": false, + "reason": "Batched apply_chat_template crash and auto_docstring annotation crash affect different subsystems." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "Reading position_ids after buffer registration is unrelated to processor save_pretrained missing files." + }, + { + "left": "issue:43723", + "right": "issue:44871", + "accept": false, + "reason": "Tokenizer loading in v5 and Gemma eos_token_id inconsistency are different model/config issues." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "Old InternVL2 meta-init incompatibility and Qwen3.5 visual encoder key saving are separate checkpoint/serialization bugs." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "Missing rms_norm_eps precision issue in MLA q_a_layernorm is unrelated to remote-code meta initialization compatibility." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Phi4Multimodal config initialization and GPTNeoXConfig rotary_pct reloading are different config serialization bugs." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "SentenceTransformers loading stall and zero probabilities in siglip2-base are different model-runtime symptoms." + }, + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Missing transformers.onnx module and AutoModel requiring PyTorch are separate missing-dependency/import issues." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Offline tokenizer cache loading and quantized model saving failure are different behaviors." + } + ] + }, + "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": 16, + "review_unit_index": 36, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:29127", + "issue:30990", + "issue:33357", + "issue:41628", + "issue:42371", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43232", + "issue:43408", + "issue:43475", + "issue:43576", + "issue:43723", + "issue:43756", + "issue:43784", + "issue:43824", + "issue:43937", + "issue:44038", + "issue:44060", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44442", + "issue:44485", + "issue:44514", + "issue:44521", + "issue:44625", + "issue:44704", + "issue:44743", + "issue:44779", + "issue:44829", + "issue:44945", + "issue:45200", + "issue:45245", + "issue:45278", + "issue:45290", + "issue:45310", + "issue:45593", + "issue:45600", + "issue:45753" + ], + "soft_pairs": [ + "issue:41628|issue:43824", + "issue:42831|issue:45310", + "issue:44263|issue:44373", + "issue:44945|issue:45753", + "issue:42831|issue:44038", + "issue:42757|issue:45278", + "issue:43576|issue:43784", + "issue:42886|issue:43066", + "issue:44117|issue:44704", + "issue:43232|issue:45593", + "issue:42371|issue:44263", + "issue:30990|issue:33357", + "issue:43756|issue:44060", + "issue:44704|issue:45290", + "issue:42907|issue:44945", + "issue:43122|issue:44514", + "issue:44295|issue:44297", + "issue:43232|issue:45245", + "issue:43723|issue:44779", + "issue:43723|issue:44442", + "issue:43408|issue:43475", + "issue:43784|issue:45600", + "issue:44261|issue:44373", + "issue:43122|issue:44521", + "issue:42907|issue:43937", + "issue:41628|issue:45278", + "issue:44360|issue:44485", + "issue:43232|issue:44743", + "issue:44625|issue:44829", + "issue:43116|issue:45200", + "issue:43122|issue:44779", + "issue:29127|issue:30990", + "issue:44038|issue:44297" + ], + "prepared_review_unit_hash": "d78fba7e157338a690a75b506207bee0924de36e3c1adf4df45be14dc97b3cd5", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37250, + "estimated_input_tokens": 9313, + "estimated_eval_tokens": 18882 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab-bag of unrelated Transformers v5 regression reports. The soft-similarity pairs are mostly thematic matches only; I would not merge any of them as true duplicates.", + "confidence": 0.74, + "canonical_issue_reason": "issue:45278 is the broadest umbrella report in the set (many import errors after upgrading to v5.5.0), so it is the best single issue to represent the cluster\u2019s main regression theme.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the most representative issue because it spans the cluster\u2019s dominant failure mode\u2014package import regressions after a v5 upgrade\u2014and is broader than the narrower, model-specific or API-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but one is root-level AutoImageProcessor export and the other is Qwen2.5-VL class export; different missing symbols and likely different fixes." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "FineGrainedFP8 accuracy drift and Qwen3.5-MoE from_pretrained failure are unrelated bug classes." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "One is a torch.split return-value issue in a DSA indexer; the other is a docstring problem for position_ids." + }, + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Pipeline-parallelism output corruption and a Qwen3_5 image-specific infinite loop are different runtime failures." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "These are different model-family bugs: FP8 accuracy regression vs Qwen3-VL-Moe behavior in v5." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "The second is a broad import-regression report, but the first is a specific missing is_offline_mode symbol; not enough evidence of the same underlying bug." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "Broken env command and a NameError while importing sentence-transformers are unrelated." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Offline cache loading failure and wrong tokenizer decoder type are tokenizer bugs, but not the same failure or code path." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None and AutoProcessor kwargs forwarding are different loader issues." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation kwargs state handling and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 settings guidance and a torch.split/indexer bug do not match." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Sentence-Transformers loading hang and a MacOS bus error with a CLIP model are different problems." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Smollm3 RoPE-layer dropping and Qwen3-Next tied-weights warnings are unrelated model-implementation issues." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "AutoProcessor cached_file kwargs forwarding and apply_chat_template crashing on tool-call messages are separate API bugs." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Saving dequantized Ministrals/Devstrals and pipeline-parallelism output errors are not the same code path." + }, + { + "left": "issue:43122", + "right": "issue:44514", + "accept": false, + "reason": "Different tokenization across versions and a batched multimodal apply_chat_template crash are distinct regressions." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "Reading position_ids after buffer registration and tokenizer_class mismatch on save are unrelated serialization issues." + }, + { + "left": "issue:43232", + "right": "issue:45245", + "accept": false, + "reason": "Generation sync_gpus handling and a category-cardinality runtime error are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44779", + "accept": false, + "reason": "Both involve tokenizer loading in v5, but one is AutoTokenizer load failure and the other is incorrect Deepseek tokenization output; not the same bug." + }, + { + "left": "issue:43723", + "right": "issue:44442", + "accept": false, + "reason": "AutoTokenizer load failure and FastSpeech2ConformerTokenizer load failure are related only at a very high level." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "SAM3 model-type warning and missing fpn_position_embeddings attribute are different SAM 3 issues." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "Sentence-transformers import NameError and stale auto_mappings references to removed Sam3LiteText configs are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "A precision issue in MLA q_a_layernorm and a docstring issue for position_ids are not duplicates." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Both involve apply_chat_template, but one is a version-to-version tokenization change and the other is all-zero assistant masks for multimodal inputs." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Saving dequantized models and invalid GenerationConfig for GLM-5 are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "The latter is a broad import-regression report, but the former is a specific missing AutoImageProcessor export; not enough to call them the same issue." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "A missing ReLU in the DSA indexer and GLM-5 RoPE implementation issues are different." + }, + { + "left": "issue:43232", + "right": "issue:44743", + "accept": false, + "reason": "Generation state update logic and Qwen3.5 recurrent-state reset under cache are separate bugs." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "num_labels propagation in Qwen3.5 and flash_attention_3 training degeneration are unrelated." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Multi-label example-script behavior and Gemma 4 mm_token_type_ids defaults are different concerns." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Tokenization drift across versions and Deepseek tokenizer correctness regressions are not the same underlying bug." + }, + { + "left": "issue:29127", + "right": "issue:30990", + "accept": false, + "reason": "A LayoutLMv3 error message clarity issue is unrelated to Sentence Transformers loading hang." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe bug report and tokenizer save_pretrained class mismatch are different." + } + ] + }, + "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": 16, + "review_unit_index": 37, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:33453", + "issue:36296", + "issue:38617", + "issue:39401", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42757", + "issue:42831", + "issue:42907", + "issue:43064", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43441", + "issue:43452", + "issue:43493", + "issue:43673", + "issue:43784", + "issue:43824", + "issue:43825", + "issue:43874", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44075", + "issue:44222", + "issue:44295", + "issue:44351", + "issue:44485", + "issue:44488", + "issue:44521", + "issue:44704", + "issue:44779", + "issue:44936", + "issue:44938", + "issue:44964", + "issue:44977", + "issue:45083", + "issue:45216", + "issue:45278", + "issue:45446", + "issue:45478", + "issue:45563", + "issue:45753" + ], + "soft_pairs": [ + "issue:42907|issue:44964", + "issue:42907|issue:43452", + "issue:44038|issue:44295", + "issue:39401|issue:44488", + "issue:36296|issue:43116", + "issue:43824|issue:44351", + "issue:44038|issue:44222", + "issue:43066|issue:44779", + "issue:39401|issue:44521", + "issue:43122|issue:44977", + "issue:45216|issue:45478", + "issue:42757|issue:44351", + "issue:43992|issue:44704", + "issue:42907|issue:44222", + "issue:43784|issue:45446", + "issue:43937|issue:44485", + "issue:42222|issue:43441", + "issue:43976|issue:45446", + "issue:42371|issue:43874", + "issue:43493|issue:45753", + "issue:44075|issue:44936", + "issue:42831|issue:45753", + "issue:43116|issue:43825", + "issue:44038|issue:45478", + "issue:44222|issue:45478", + "issue:43116|issue:43493", + "issue:43064|issue:43673", + "issue:45083|issue:45446", + "issue:43937|issue:45563", + "issue:42175|issue:44938", + "issue:33453|issue:39401", + "issue:39401|issue:44779", + "issue:38617|issue:45278", + "issue:22355|issue:29127" + ], + "prepared_review_unit_hash": "6b631e4eb0e48ffa2e7e10566df912707e6da446a7c1f2ee9fc0ab5e5282b49e", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37298, + "estimated_input_tokens": 9325, + "estimated_eval_tokens": 18906 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the soft pairs span unrelated import errors, tokenizer regressions, multimodal/model-specific bugs, training issues, and warning/message issues. I reject all soft edges because none clearly share the same underlying bug or code change.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue stands out because the items do not form one duplicate family; they cover many distinct bugs across different APIs, models, and subsystems.", + "canonical_pr_reason": null, + "best_issue_reason": "No best issue for the cluster overall: it is not coherent enough to justify choosing one issue as the representative duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Different failures: saving/dequantizing a model versus loading a multimodal model." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Unrelated code paths: dequantized save failure vs gguf/AutoTokenizer loading." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Different problems: Qwen3-VL-Moe behavior versus position_ids buffer access." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Tokenizer offset-mapping bug is unrelated to loading a specific model checkpoint." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor-parallel training bug and empty multi-label classification results are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and different underlying fixes." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Same broad model family, but one is a model loading bug and the other is an FP8 save_pretrained issue." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer regressions affecting different models and behaviors." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Offset mapping regression is unrelated to assistant mask generation for multimodal inputs." + }, + { + "left": "issue:43122", + "right": "issue:44977", + "accept": false, + "reason": "Different symptoms: tokenization change across versions versus flash-attention generation failure." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5, but one is about save_pretrained checkpoint contents and the other about from_pretrained loading." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Different missing imports from different packages; no shared code-path bug is evident." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Model weight loading bug and AutoProcessor kwargs forwarding are separate issues." + }, + { + "left": "issue:42907", + "right": "issue:44222", + "accept": false, + "reason": "Both mention saving, but one is dequantized model saving and the other is FP8 moe save_pretrained." + }, + { + "left": "issue:43784", + "right": "issue:45446", + "accept": false, + "reason": "Sentence-transformers import failure and a PyTorch version check bug are unrelated." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "Invalid GenerationConfig and RoPE implementation discussion are not the same bug." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "Different models and failures: vitpose breakage versus Ministral flash-attention support." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python support/version constraint issue is separate from the flex_attention import check." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 settings guidance and a missing image-patch-count method are unrelated." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and Qwen3_5 image looping are different bugs." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "Optimizer-argument handling is different from train/evaluate state corruption." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "FP8 accuracy issue and an infinite-loop image case are unrelated." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Example-script classification behavior and pipeline translation error messaging are separate issues." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Both are Qwen3.5/Moe-related, but one is a Qwen3-VL-Moe bug and the other is a from_pretrained load failure." + }, + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "Both are Moe-related save/load issues, but they affect different APIs and likely different fixes." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Classification example output bug and SigLIP2 model discrepancy are unrelated." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "FSDP2 optimizer-state bug and missing GenerationMixin cache are separate training/generation issues." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "Helper function behavior in qwen3_omni_moe is unrelated to the flex_attention import check." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "Invalid generation config and a stale warning about num_return_sequences are different problems." + }, + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Backend dependency packaging issue versus Python 3.14 load failure; no shared bug." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Tokenizer loading regression and Qwen3 offset mapping issue are not the same defect." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenization bugs, but they affect different models and different behaviors." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "Specific missing import and broad many-import-errors report are too coarse to treat as one bug." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "No shared underlying bug: missing transformers.onnx module vs an error-message clarity issue." + } + ] + }, + "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": 16, + "review_unit_index": 38, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:43012", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43381", + "issue:43408", + "issue:43421", + "issue:43450", + "issue:43577", + "issue:43742", + "issue:43827", + "issue:43874", + "issue:43937", + "issue:43986", + "issue:44075", + "issue:44117", + "issue:44220", + "issue:44261", + "issue:44291", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44493", + "issue:44496", + "issue:44829", + "issue:44945", + "issue:44964", + "issue:44977", + "issue:45071", + "issue:45072", + "issue:45081", + "issue:45216", + "issue:45278", + "issue:45310", + "issue:45405", + "issue:45446", + "issue:45459", + "issue:45464", + "issue:45542", + "issue:45588", + "issue:45656" + ], + "soft_pairs": [ + "issue:44117|issue:45588", + "issue:43827|issue:45446", + "issue:43381|issue:44945", + "issue:43937|issue:44977", + "issue:45405|issue:45446", + "issue:43450|issue:43986", + "issue:39401|issue:43421", + "issue:44117|issue:45081", + "issue:42673|issue:45072", + "issue:38617|issue:41628", + "issue:43408|issue:43742", + "issue:44829|issue:45216", + "issue:44291|issue:45071", + "issue:42222|issue:45542", + "issue:44410|issue:45310", + "issue:44484|issue:45310", + "issue:42175|issue:42222", + "issue:43937|issue:44496", + "issue:44483|issue:45464", + "issue:44075|issue:45656", + "issue:43066|issue:43122", + "issue:44964|issue:45278", + "issue:43012|issue:43577", + "issue:42371|issue:44261", + "issue:38617|issue:44220", + "issue:43097|issue:44493", + "issue:43874|issue:44261", + "issue:42222|issue:45459" + ], + "prepared_review_unit_hash": "9831ece046567207fefb48eacc2ee6ad9ad40f42d1939da2a9ae306515f671e3", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36085, + "estimated_input_tokens": 9022, + "estimated_eval_tokens": 18300 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated transformer bugs/regressions across tokenizers, model loading, precision, and optimizer behavior. The soft pairs are all false-positive duplicates; none look like the same concrete bug or mergeable change.", + "confidence": 0.9, + "canonical_issue_reason": "No single canonical issue is a safe choice here: the set is heterogeneous, and the titles point to distinct failure modes rather than one underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "If one issue must be kept as the broadest representative, issue:45278 is the most umbrella-like report ('many import errors after update'), but it still does not actually unify the other issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "Different failures: tokenizer mapping fallback vs flash-attention crash on missing s_aux. No shared code-path bug is evident." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs mentioning pipeline() and a flex_attention import-version check are unrelated." + }, + { + "left": "issue:43381", + "right": "issue:44945", + "accept": false, + "reason": "Gradient checkpointing in eval mode and pipeline-parallelism output errors are different runtime problems." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "GLM-5 GenerationConfig validation is not the same as Qwen3.5 flash-attention generation failure." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "A released-version pin issue and a PyTorch import/version check bug are separate dependency problems." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Batched video shape handling and AutoProcessor crash without torchvision are distinct video-processing issues." + }, + { + "left": "issue:39401", + "right": "issue:43421", + "accept": false, + "reason": "Wrong offset mapping for Qwen3 tokenizer is not the same as runtime post-processor updates for special tokens." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer mapping None handling and Mistral regex patch crashes are different tokenizer code paths." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "VRAM leak in Qwen3ForCausalLM and bfloat16 dtype mismatches are unrelated model/runtime bugs." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but for different symbols and likely different causes; not the same bug." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "SAM3 model-type warning and MobileLLM loading KeyError are unrelated model-loading issues." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "FlashAttention-3 training degeneration and Qwen3.5 save_pretrained regression do not share a concrete fix path." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "init_empty_weights/_is_hf_initialized and PretrainedConfig type-checking breakage are separate v5 regressions." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "VitPose model breakage and tensorboard-without-TF backend errors are different backend/model issues." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Missing qkv/gate projections in qwen3next is not the same as Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "save_pretrained max_shard_size behavior is unrelated to Qwen3.5 MoE loading." + }, + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "TF not installed in a torch-only extra and VitPose breakage are separate dependency/model problems." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "Invalid GenerationConfig for GLM-5 and unrecognized model/config.json loading are different errors." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Both involve chat/completions, but one is an API request-format regression and the other is a streaming inference failure on a specific model." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "Unused SGD arguments and double optimizer.step() under deepspeed are different optimizer issues." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Wrong tokenizer decoder type and changed tokenization output across versions are related thematically but not shown to be the same concrete bug." + }, + { + "left": "issue:44964", + "right": "issue:45278", + "accept": false, + "reason": "A specific model-loading failure and a broad import-errors report are not the same underlying issue." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "PyTorch warning on compiling bfloat16 models and BLIP2 dtype retention are distinct precision/dtype bugs." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 settings guidance and MLA rms_norm_eps precision mismatch are unrelated." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "ImportError for layer_type_validation and _torch_extract_fbank_features() are unrelated failures." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "Removed tie_embeddings_and_encoder_decoder and unexpected position-id keys are different v5 regressions." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Missing get_number_of_image_patches and missing rms_norm_eps are separate multimodal/model-config issues." + }, + { + "left": "issue:42222", + "right": "issue:45459", + "accept": false, + "reason": "VitPose model breakage and protobuf-hidden tokenizer errors are not the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 39, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:38617", + "issue:41628", + "issue:42371", + "issue:42673", + "issue:42757", + "issue:42831", + "issue:43012", + "issue:43064", + "issue:43097", + "issue:43335", + "issue:43408", + "issue:43441", + "issue:43452", + "issue:43502", + "issue:43577", + "issue:43873", + "issue:43874", + "issue:43906", + "issue:43931", + "issue:43976", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44062", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44315", + "issue:44351", + "issue:44368", + "issue:44410", + "issue:44464", + "issue:44484", + "issue:44485", + "issue:44704", + "issue:44829", + "issue:44945", + "issue:44964", + "issue:44977", + "issue:44987", + "issue:45071", + "issue:45103", + "issue:45245", + "issue:45341", + "issue:45375", + "issue:45478", + "issue:45542", + "issue:45588" + ], + "soft_pairs": [ + "issue:44484|issue:45478", + "issue:44263|issue:44485", + "issue:44410|issue:45478", + "issue:43441|issue:44829", + "issue:43012|issue:43408", + "issue:41628|issue:45071", + "issue:45071|issue:45542", + "issue:42673|issue:44977", + "issue:42371|issue:42831", + "issue:42673|issue:43906", + "issue:44117|issue:44987", + "issue:43097|issue:43452", + "issue:43986|issue:44484", + "issue:43335|issue:45071", + "issue:43931|issue:44038", + "issue:43976|issue:44945", + "issue:43408|issue:44368", + "issue:43874|issue:44263", + "issue:44829|issue:45245", + "issue:38617|issue:42757", + "issue:43064|issue:44315", + "issue:44464|issue:45071", + "issue:44062|issue:44117", + "issue:43097|issue:45588", + "issue:44410|issue:45375", + "issue:38617|issue:44351", + "issue:44704|issue:45103", + "issue:42673|issue:43577", + "issue:44038|issue:44410", + "issue:44261|issue:44485", + "issue:44964|issue:45341", + "issue:43986|issue:44117", + "issue:43931|issue:44410", + "issue:43502|issue:43992", + "issue:43097|issue:43873" + ], + "prepared_review_unit_hash": "3b488282b8f625eb1b6fead3af826e79dfd51b01d54279301e730a3c7992ce81", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38166, + "estimated_input_tokens": 9542, + "estimated_eval_tokens": 19340 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a loose grab bag of unrelated Transformers issues across imports, model loading, training, quantization, and config bugs. None of the soft pairs look like the same underlying defect, so I would not collapse any of them.", + "confidence": 0.95, + "canonical_issue_reason": "No clear canonical issue: the items do not form one duplicate family and instead cover many unrelated regressions and feature bugs.", + "canonical_pr_reason": "No pull requests are present in the cluster.", + "best_issue_reason": "No single issue is a good global representative; the cluster is too heterogeneous to safely use one issue as the canonical duplicate target.", + "best_pr_reason": "No pull requests are present, so there is no best PR candidate.", + "soft_edge_verdicts": [ + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "Different bugs: save_pretrained shard sizing vs Qwen3.5 MoE loading failure." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "Unrelated failures: torch.split return handling vs GLM-5 RoPE implementation." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "Different model/config problems: missing qwen3next projections vs Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "Both touch flash attention, but one is a model load failure and the other is degenerate training; not the same bug." + }, + { + "left": "issue:43012", + "right": "issue:43408", + "accept": false, + "reason": "A PyTorch precision warning and a model-type warning are different issues." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "ImportError for AutoImageProcessor is unrelated to PretrainedConfig type checking." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Type-checking regression and undefined TensorFlow backend error are unrelated." + }, + { + "left": "issue:42673", + "right": "issue:44977", + "accept": false, + "reason": "VRAM leak in dataloader threads is not the same as Qwen3.5 flash-attention generation failure." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 API settings guidance and FineGrainedFP8 accuracy loss are different problems." + }, + { + "left": "issue:42673", + "right": "issue:43906", + "accept": false, + "reason": "The VRAM leak report does not match the isolated reproduction of issue 38071." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer mapping None handling and loading physical-intelligence/fast are different failure modes." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "Removed tie_embeddings API and gguf_file loading breakage are unrelated." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Missing torchvision crash in AutoProcessor is unrelated to save_pretrained shard size." + }, + { + "left": "issue:43335", + "right": "issue:45071", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug is unrelated to type-checking regressions." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "Different Qwen-related load issues with different symptoms and likely different code paths." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python version compatibility and pipeline-parallelism output corruption are unrelated." + }, + { + "left": "issue:43408", + "right": "issue:44368", + "accept": false, + "reason": "A SAM model-type mismatch warning is unrelated to a Qwen fine-tuning warning about tie_word_embeddings." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Different GLM bugs in different components: missing image-patch method vs torch.split return values." + }, + { + "left": "issue:44829", + "right": "issue:45245", + "accept": false, + "reason": "FlashAttention-3 training degeneration and category-cardinality limits are unrelated." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols and different underlying causes." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "FSDP2/PEFT optimizer-state corruption is not the same as Liger Kernel not being applied with model_init." + }, + { + "left": "issue:44464", + "right": "issue:45071", + "accept": false, + "reason": "Compiled-forward generation inconsistency is unrelated to PretrainedConfig type checking." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "AddedToken keyword duplication and tokenizer mapping None handling are separate tokenizer bugs." + }, + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "Removed tie_embeddings API and flash_attention s_aux=None crash are unrelated." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Missing qwen3next projections and missing Qwen3_5MoeVisionConfig field are different model/config issues." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Different missing imports; same general theme, but not the same concrete bug." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "Passing kwargs to cached_file and auto_docstring future-annotations crash are unrelated." + }, + { + "left": "issue:42673", + "right": "issue:43577", + "accept": false, + "reason": "VRAM leak in multi-thread loading and BLIP2 dtype retention are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Both mention Qwen models, but one is a version/load compatibility issue and the other is a missing-projection architecture bug." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "Missing rms_norm_eps precision issue and GLM-5 RoPE implementation are different bugs." + }, + { + "left": "issue:44964", + "right": "issue:45341", + "accept": false, + "reason": "Phi-4 multimodal load failure and a testing_utils bug are unrelated." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "Video-model AutoProcessor crash without torchvision is unrelated to tokenizer mapping None handling." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Different Qwen model families and different load-path failures." + }, + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "local_files_only network leak and missing embed_tokens.weight are unrelated." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "Removed tie_embeddings API and quantization offloading 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": 2, + "component_count": 16, + "review_unit_index": 40, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:36683", + "issue:38617", + "issue:42617", + "issue:42757", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43299", + "issue:43352", + "issue:43381", + "issue:43475", + "issue:43502", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43618", + "issue:43761", + "issue:43824", + "issue:43825", + "issue:43867", + "issue:43901", + "issue:43906", + "issue:43937", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44230", + "issue:44368", + "issue:44380", + "issue:44483", + "issue:44484", + "issue:44514", + "issue:44655", + "issue:44683", + "issue:45216", + "issue:45290", + "issue:45341", + "issue:45372", + "issue:45381", + "issue:45561", + "issue:45588", + "issue:45753" + ], + "soft_pairs": [ + "issue:43066|issue:43906", + "issue:43986|issue:45290", + "issue:43986|issue:45372", + "issue:45341|issue:45561", + "issue:43986|issue:45341", + "issue:43097|issue:43901", + "issue:43825|issue:44655", + "issue:42617|issue:45341", + "issue:45341|issue:45588", + "issue:43992|issue:44484", + "issue:42886|issue:43381", + "issue:43208|issue:43550", + "issue:43012|issue:44368", + "issue:43012|issue:43550", + "issue:42907|issue:45753", + "issue:43531|issue:45341", + "issue:42907|issue:44655", + "issue:43208|issue:43526", + "issue:44483|issue:44514", + "issue:42757|issue:43824", + "issue:43012|issue:44683", + "issue:43618|issue:43761", + "issue:44483|issue:45290", + "issue:43064|issue:44380", + "issue:44038|issue:44230", + "issue:38617|issue:43502", + "issue:44483|issue:45381", + "issue:43299|issue:43867", + "issue:43901|issue:44655", + "issue:36683|issue:43475", + "issue:43867|issue:45216", + "issue:42994|issue:44655", + "issue:43352|issue:43937", + "issue:43066|issue:45381" + ], + "prepared_review_unit_hash": "756a42d4a40647e424a91863b083395f8df588ec0ab78bceb33b82b2092bdb68", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 34, + "serialized_chars": 37049, + "estimated_input_tokens": 9263, + "estimated_eval_tokens": 18782 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous v5-era bug reports and import/runtime regressions; only the CLIPVisionModel output regressions look like the same underlying change. No PRs are present.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 43097 is the broadest, clearest representative of the cluster\u2019s dominant theme: an explicit Transformers v5 breaking-change regression with user-visible API fallout.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43097 is the best overall issue representative because it captures a concrete, high-impact v5 compatibility break rather than a narrow model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Both involve v5 behavior, but one is a tokenizer decoder-type bug and the other is an unrelated isolated reproduction; no same code-path evidence." + }, + { + "left": "issue:43986", + "right": "issue:45290", + "accept": false, + "reason": "Different failures: missing torchvision for AutoProcessor vs chat template crashing on tool-call assistant messages." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "Both are import/dependency-related, but they affect different components and different missing symbols; not the same bug." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "Both mention testing utilities, but one is a vague bug report and the other is a specific xdist file race; insufficient evidence of identity." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "A processor-loading dependency error is unrelated to a generic testing_utils bug." + }, + { + "left": "issue:43097", + "right": "issue:43901", + "accept": false, + "reason": "One is an API removal regression; the other is documentation lag about return_all_scores." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "Pipeline task support messaging and saving Pipeline objects are different problems." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "3d_parallel.py execution failure is unrelated to a testing_utils issue." + }, + { + "left": "issue:45341", + "right": "issue:45588", + "accept": false, + "reason": "A testing_utils bug and a flash-attention sink-less-model crash are unrelated." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Missing embed_tokens weights on load and max_shard_size default behavior in save_pretrained are different issues." + }, + { + "left": "issue:42886", + "right": "issue:43381", + "accept": false, + "reason": "Offline cache loading and eval-mode gradient checkpointing are unrelated." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "Different model families and different failure modes; not the same underlying bug." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "Both are warnings, but one is a PyTorch precision warning and the other is a model config warning about tie_word_embeddings." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "A compile/precision warning is unrelated to a Bamba SDPA compile failure." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Quantized model saving and Qwen image generation looping are unrelated." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "A Qwen3-MoE sliding_window bug is unrelated to testing_utils." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Dequantized model saving and saving Pipeline objects are different save_pretrained problems." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "xLSTM training bugs and a BeitImageProcessorFast label reduction bug are unrelated." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "OpenAI-style chat completion request rejection is not the same as a Qwen2_5_VLProcessor batching crash." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different packages and different missing names; no shared underlying change is evident." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "A bfloat16 compile warning and a flex_attention failure on torch>=2.9 are different code paths." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": true, + "reason": "Both report a CLIPVisionModel v5 regression where optional outputs are not populated correctly; attentions/hidden_states are missing from the same forward output path." + }, + { + "left": "issue:44483", + "right": "issue:45290", + "accept": false, + "reason": "Different APIs and symptoms: chat completion request validation vs chat template handling of tool-call assistant messages." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "FSDP2/PEFT optimizer-state corruption and GPT2 attention-scaling under SDPA/FlashAttention are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Both mention Qwen3 VL/MoE, but one is a Transformers loading bug and the other is fp8 inference support; not the same bug." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "A missing configuration_utils import and API calls despite local_files_only=True are unrelated." + }, + { + "left": "issue:44483", + "right": "issue:45381", + "accept": false, + "reason": "Chat-completions request handling and Qwen2.5-VL video vision_position_ids are different bugs." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Qwen3VL MoE loading breakage and a sorted state_dict load error are not clearly the same underlying issue." + }, + { + "left": "issue:43901", + "right": "issue:44655", + "accept": false, + "reason": "Documentation mismatch for return_all_scores is unrelated to Pipeline serialization." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Gemma3Config missing vocab_size and SAM3 vision output missing fpn_position_embeddings are different model attribute regressions." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "Both involve loading/saving state_dicts, but one is a sorted-load error and the other is incorrect saved checkpoints; too little to treat as the same bug." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Quantized model saving failure and saving Pipeline objects are different save_pretrained paths." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "FA2 support not implemented for NemotronHForCausalLM is unrelated to an invalid GenerationConfig in GLM-5." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative about the soft edges. The only accepted pair, issue:43618 and issue:43761, is plausibly the same CLIPVisionModel v5 output regression, and the other pairwise rejections do not overclaim identity. No PRs are present, so the mergeability constraint is not an issue." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 41, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:28282", + "issue:33453", + "issue:36010", + "issue:36296", + "issue:36683", + "issue:41669", + "issue:42757", + "issue:42890", + "issue:43064", + "issue:43097", + "issue:43299", + "issue:43352", + "issue:43441", + "issue:43525", + "issue:43575", + "issue:43618", + "issue:43756", + "issue:43867", + "issue:43881", + "issue:43906", + "issue:43976", + "issue:43992", + "issue:44077", + "issue:44117", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44263", + "issue:44355", + "issue:44360", + "issue:44380", + "issue:44484", + "issue:44485", + "issue:44492", + "issue:44655", + "issue:44855", + "issue:44857", + "issue:44938", + "issue:45216", + "issue:45310", + "issue:45357", + "issue:45375", + "issue:45440", + "issue:45478", + "issue:45479", + "issue:45561", + "issue:45588", + "issue:45753" + ], + "soft_pairs": [ + "issue:44263|issue:45440", + "issue:36010|issue:42757", + "issue:44220|issue:44655", + "issue:43756|issue:44485", + "issue:36683|issue:43525", + "issue:43618|issue:45216", + "issue:43906|issue:45753", + "issue:36683|issue:44242", + "issue:36683|issue:43881", + "issue:43299|issue:45753", + "issue:43352|issue:43976", + "issue:43441|issue:44380", + "issue:43352|issue:44380", + "issue:43992|issue:45310", + "issue:44380|issue:45479", + "issue:41669|issue:44246", + "issue:43064|issue:44492", + "issue:43867|issue:45375", + "issue:43618|issue:44380", + "issue:43352|issue:45478", + "issue:44355|issue:44855", + "issue:36683|issue:44117", + "issue:42890|issue:45561", + "issue:43097|issue:43352", + "issue:43867|issue:45357", + "issue:43976|issue:44938", + "issue:43992|issue:45478", + "issue:44360|issue:45440", + "issue:43064|issue:44655", + "issue:36296|issue:43575", + "issue:44077|issue:44380", + "issue:43618|issue:44857", + "issue:43618|issue:45588", + "issue:28282|issue:33453", + "issue:43867|issue:44484" + ], + "prepared_review_unit_hash": "3db5c8667f7f0ded7b1bb17242445473c19aad809106aedab920236f327562e0", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 37790, + "estimated_input_tokens": 9448, + "estimated_eval_tokens": 19152 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most proposed edges are false positives across unrelated model/config/version bugs. The only convincing duplicate-like pair is the import-time slowdown report versus the generic slow-import report; the rest differ in concrete model, code path, or failure mode.", + "confidence": 0.63, + "canonical_issue_reason": "issue:41669 is the most specific and actionable issue in the only convincing duplicate pair: it names the concrete import-time regression cause ('import *' in models) and the measurable slowdown, while issue:44246 is the broader symptom report.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41669 is the best global issue representative because it is specific, root-cause oriented, and can subsume the more generic slow-import report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44263", + "right": "issue:45440", + "accept": false, + "reason": "Different bugs in different model families; one is a torch.split / indexer issue, the other is a DeepSeekV3MoE implementation divergence." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols and different dependency breakages; not the same underlying bug." + }, + { + "left": "issue:44220", + "right": "issue:44655", + "accept": false, + "reason": "Unrelated failures: audio feature extraction versus saving Pipeline objects." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Both mention RoPE, but they are model-specific implementation bugs with different symptoms and targets." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Both are missing config attributes, but for different models and different fields; too generic to treat as one bug." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "Different problems: CLIPOutput attentions assignment versus Qwen3.5 checkpoint save regression." + }, + { + "left": "issue:43906", + "right": "issue:45753", + "accept": false, + "reason": "The first is a reproduction of another issue, while the second is a specific Qwen3_5 image infinite-loop report; not enough evidence they are the same bug." + }, + { + "left": "issue:36683", + "right": "issue:44242", + "accept": false, + "reason": "Different model behavior: missing Gemma3Config field versus MoE load-balancing loss not being added." + }, + { + "left": "issue:36683", + "right": "issue:43881", + "accept": false, + "reason": "Different loading failures affecting different models and different config/runtime paths." + }, + { + "left": "issue:43299", + "right": "issue:45753", + "accept": false, + "reason": "Both involve Qwen-related multimodal/model loading behavior, but one is a v5.0.0 loading regression and the other is an image-specific infinite loop." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "Unrelated issues: Flash Attention support for a model versus Python version compatibility for the package." + }, + { + "left": "issue:43441", + "right": "issue:44380", + "accept": false, + "reason": "Both touch attention backends, but they are different concrete failures in different models." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Different model/code-path problems; one is model support for FA2, the other is GPT2 attention scaling being ignored." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "Different models and different failure modes: missing embed_tokens.weight versus Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "No shared underlying bug; one is attention scaling with backends, the other is degenerate zero loss in sequence classification." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": true, + "reason": "Both report the same underlying import-time performance regression in transformers; one is the concrete cause/root-cause report and the other is the generic symptom report." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Completely different topics: distributed optimizer state corruption versus a cache strategy typo." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "Different loading/serialization bugs affecting different Qwen-related configs and fields." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "Different issues despite both involving model internals; no shared concrete code-path failure." + }, + { + "left": "issue:43352", + "right": "issue:45478", + "accept": false, + "reason": "Same broad area as another FlashAttention issue, but different model and no clear shared concrete bug." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "One is about compiled Python files, the other about a Python 3.13 torch.jit parsing/indentation problem." + }, + { + "left": "issue:36683", + "right": "issue:44117", + "accept": false, + "reason": "Different loader assumptions in different parts of the stack; not the same bug." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "Both are test-related, but one is missing set_seed and the other is an xdist race on a captured-info file." + }, + { + "left": "issue:43097", + "right": "issue:43352", + "accept": false, + "reason": "Unrelated: an API removal/regression versus a Flash Attention support limitation." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "Different save/load regressions involving different Qwen3.5 components and different key handling." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both are Python-version issues, but they affect different versions and appear to be separate compatibility problems." + }, + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Different model families and different from_pretrained failures." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "Different code paths: DSA indexer math versus DeepSeekV3 implementation divergence." + }, + { + "left": "issue:43064", + "right": "issue:44655", + "accept": false, + "reason": "Unrelated distributed-training state bug versus pipeline serialization issue." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Both mention tensor parallelism, but one is a training bug and the other is an OOM on model loading." + }, + { + "left": "issue:44077", + "right": "issue:44380", + "accept": false, + "reason": "Different bugs in different model implementations; no evidence they share the same failure mechanism." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "Different failures: missing attentions assignment versus AMP/CUDA crash in a loss function." + }, + { + "left": "issue:43618", + "right": "issue:45588", + "accept": false, + "reason": "Both touch attention-related code, but one is a CLIPOutput regression and the other is a FlashAttention crash on sink-less models." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "ImportError for missing PyTorch is unrelated to tokenizer loading regression." + }, + { + "left": "issue:43867", + "right": "issue:44484", + "accept": false, + "reason": "Model loading error versus a save_pretrained shard-size question; not the same bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall: the only accepted duplicate-like pair is well supported by the issue titles and descriptions, and the other soft-edge rejections stay conservative rather than overclaiming equivalence. The summary does not overstate the evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 42, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:42548", + "issue:42757", + "issue:42831", + "issue:42890", + "issue:43010", + "issue:43299", + "issue:43502", + "issue:43576", + "issue:43618", + "issue:43723", + "issue:43824", + "issue:43827", + "issue:43874", + "issue:43901", + "issue:44062", + "issue:44075", + "issue:44112", + "issue:44222", + "issue:44230", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44368", + "issue:44387", + "issue:44485", + "issue:44556", + "issue:44655", + "issue:44749", + "issue:44938", + "issue:44961", + "issue:44964", + "issue:44977", + "issue:45071", + "issue:45125", + "issue:45278", + "issue:45341", + "issue:45375", + "issue:45468", + "issue:45478", + "issue:45600", + "issue:45656" + ], + "soft_pairs": [ + "issue:43618|issue:44977", + "issue:43901|issue:44938", + "issue:44655|issue:44938", + "issue:42890|issue:44964", + "issue:42890|issue:44112", + "issue:42890|issue:45600", + "issue:43502|issue:44062", + "issue:42890|issue:45341", + "issue:43010|issue:44075", + "issue:42548|issue:44351", + "issue:44222|issue:45125", + "issue:43502|issue:44485", + "issue:44355|issue:44749", + "issue:30990|issue:31795", + "issue:45125|issue:45478", + "issue:44387|issue:44556", + "issue:45125|issue:45375", + "issue:43010|issue:45278", + "issue:43723|issue:44556", + "issue:30333|issue:33453", + "issue:43010|issue:44368", + "issue:43502|issue:44222", + "issue:43502|issue:44961", + "issue:44315|issue:44360", + "issue:42831|issue:43874", + "issue:30333|issue:36296", + "issue:44355|issue:44655", + "issue:44222|issue:44230", + "issue:44355|issue:45071", + "issue:43827|issue:44961", + "issue:33453|issue:34689", + "issue:43010|issue:45656", + "issue:43299|issue:44556", + "issue:42757|issue:43576", + "issue:33453|issue:36296", + "issue:43576|issue:45071", + "issue:43576|issue:43824", + "issue:44961|issue:45468" + ], + "prepared_review_unit_hash": "08b28b7e20f8d7590a7225874567e8ee850f8545e6c8ea155d83a739a076fa97", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 38, + "serialized_chars": 38498, + "estimated_input_tokens": 9625, + "estimated_eval_tokens": 19506 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a false-positive grouping: the issues span unrelated areas like loading/import regressions, documentation cleanup, FP8, tensor parallelism, testing, and optimizers. I do not see a single underlying bug/change that would justify merging any pair, and there is no PR evidence here.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43618", + "right": "issue:44977", + "accept": false, + "reason": "Different model/output bug classes: CLIPOutput attentions assignment vs Qwen3.5 flash-attention generation behavior." + }, + { + "left": "issue:43901", + "right": "issue:44938", + "accept": false, + "reason": "One is a docs mismatch about return_all_scores; the other is a Python 3.14 import/load failure." + }, + { + "left": "issue:44655", + "right": "issue:44938", + "accept": false, + "reason": "Saving Pipeline objects and Python 3.14 import/load failure are unrelated failure modes." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "Different test/model areas: SamHQ integration test seeding flakiness vs Phi-4 multimodal model loading." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "A missing set_seed in one model test is not the same as a stale GraniteSpeech device-override CI failure." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "Unrelated: SamHQ test randomness vs auto_mappings references to removed Sam3LiteText configs." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "Local-files-only network leakage and AddedToken argument duplication are different code paths." + }, + { + "left": "issue:42890", + "right": "issue:45341", + "accept": false, + "reason": "Both mention testing, but the specific SamHQ seeding issue is not clearly the same as an unspecified testing_utils bug." + }, + { + "left": "issue:43010", + "right": "issue:44075", + "accept": false, + "reason": "Cache/layer no_grad decoration is unrelated to SGD optimizer arguments not being used." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Different missing exports: PreTrainedModel vs HybridCache." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 save_pretrained for MoE models is not the same bug as missing _tp_plan for tensor parallelism." + }, + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "Network access despite local_files_only and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Compiled Python file errors and tokenizer slowdown after upgrade are different problems." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Loading hang and documentation confusion around model.forward arguments do not share a concrete root cause." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5 MoE, but one is tensor-parallel metadata and the other is a from_pretrained load error; too broad to merge." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Int4 quantization memory/OOM regression and checkpoint reload failure after v5 upgrades are different regressions." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Missing _tp_plan and missing deepstack_visual_indexes are separate config issues." + }, + { + "left": "issue:43502", + "right": "issue:45278", + "accept": false, + "reason": "Local file handling and broad import failures after 4.57->5.5 are not the same bug." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Tokenizer loading in v5 and generic checkpoint reload after upgrade are only loosely related." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLflow reporting after training failure and tokenizer loading regression are unrelated." + }, + { + "left": "issue:43010", + "right": "issue:44368", + "accept": false, + "reason": "no_grad decoration for cache updates is unrelated to a tie_word_embeddings warning during LoRA fine-tuning." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "Network requests with local_files_only and FP8 save_pretrained for MoE are different bugs." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "The 'racoon' issue has no visible relation to local_files_only network behavior." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger Kernel application in model_init is unrelated to a DSA indexer/ReLU discussion." + }, + { + "left": "issue:42831", + "right": "issue:43874", + "accept": false, + "reason": "FineGrainedFP8 accuracy and GLM46V image patch counting are different code paths." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLflow reporting after failed training and tensor-parallel training bugs are not the same underlying issue." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Compiled Python file errors and Pipeline save_pretrained failures are different concerns." + }, + { + "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 concrete bug." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled file execution errors and PretrainedConfig type-checking breaks are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Pipeline-docs cleanup and the 'racoon' issue are unrelated." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Tokenizer loading regression and Llama 3.2 vision model loading breakage are different failures." + }, + { + "left": "issue:43010", + "right": "issue:45656", + "accept": false, + "reason": "Cache update no_grad decoration is unrelated to deepspeed causing double optimizer steps." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Qwen3VL MoE loading regression and checkpoint reload regression after v5.2/v5.3 are not clearly the same root cause." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Missing is_offline_mode import and a broken transformers env command are different packaging/CLI issues." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "CLI breakage and PretrainedConfig type-checking are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "Broken env command and missing Qwen2_5_VLForConditionalGeneration export are different import/CLI failures." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "No meaningful relation between the 'racoon' issue and Gemma-4 audio positional encoding." + } + ] + }, + "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": 16, + "review_unit_index": 43, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:31795", + "issue:36296", + "issue:38617", + "issue:41628", + "issue:41669", + "issue:42548", + "issue:42890", + "issue:43010", + "issue:43232", + "issue:43352", + "issue:43408", + "issue:43526", + "issue:43576", + "issue:43673", + "issue:43827", + "issue:43901", + "issue:43976", + "issue:44038", + "issue:44077", + "issue:44246", + "issue:44261", + "issue:44273", + "issue:44295", + "issue:44297", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44373", + "issue:44393", + "issue:44492", + "issue:44556", + "issue:44623", + "issue:44704", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:44938", + "issue:44961", + "issue:45003", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45341", + "issue:45478", + "issue:45542", + "issue:45753" + ], + "soft_pairs": [ + "issue:31795|issue:38617", + "issue:43010|issue:43976", + "issue:41628|issue:42548", + "issue:38617|issue:43576", + "issue:44908|issue:44961", + "issue:43576|issue:44938", + "issue:43010|issue:43232", + "issue:36296|issue:43901", + "issue:41628|issue:44355", + "issue:36296|issue:43526", + "issue:44261|issue:44360", + "issue:44273|issue:45003", + "issue:44623|issue:45230", + "issue:44273|issue:45542", + "issue:44393|issue:45753", + "issue:43576|issue:44556", + "issue:44393|issue:45071", + "issue:41669|issue:43673", + "issue:44038|issue:45125", + "issue:44556|issue:45092", + "issue:44373|issue:44961", + "issue:44704|issue:44908", + "issue:42890|issue:44908", + "issue:44246|issue:44273", + "issue:43408|issue:44315", + "issue:43976|issue:44393", + "issue:43352|issue:44393", + "issue:44297|issue:45230", + "issue:43576|issue:44908", + "issue:44908|issue:45071", + "issue:44295|issue:45230", + "issue:43673|issue:44492", + "issue:41669|issue:44351", + "issue:41669|issue:45341", + "issue:41669|issue:44492", + "issue:44315|issue:44829", + "issue:44393|issue:45478", + "issue:44077|issue:44861", + "issue:43827|issue:44908" + ], + "prepared_review_unit_hash": "07f242530ebf0041cef551fa15be47c0e2a66a21d43f6367c1a58f3c4b632f08", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 39, + "serialized_chars": 38754, + "estimated_input_tokens": 9689, + "estimated_eval_tokens": 19634 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose collection of unrelated transformer issues with a few superficial overlaps around imports, caching, and v5 regressions. None of the soft pairs look like the same concrete bug/change, so I reject all soft edges. The broadest representative issue is the import/performance regression around `import *` usage.", + "confidence": 0.84, + "canonical_issue_reason": "issue:41669 is the broadest and most central representative of the cluster\u2019s recurring import/loading regressions, with concrete impact and multiple nearby import-related complaints.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41669 best captures the cluster\u2019s most common underlying theme (import-time/loading regressions) and is more representative than the narrower model-specific or doc-only issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Docs wording for forward args vs an ImportError for a missing symbol; different problem types and code paths." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Cache mutation/no_grad concern vs Python version compatibility failure; unrelated." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Different missing imports (`AutoImageProcessor` vs `PreTrainedModel`) with separate causes." + }, + { + "left": "issue:38617", + "right": "issue:43576", + "accept": false, + "reason": "Missing import symbol vs broken `transformers env` CLI; not the same bug." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs bug vs unrelated placeholder issue title; no shared underlying change." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "CLI env command regression vs checkpoint reload incompatibility; different functionality." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "Both are generation/cache-related, but one is about `@torch.no_grad` on updates and the other about `_update_model_kwargs_for_generation` after GPU sync; not the same fix." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor parallel training bug vs documentation mismatch for pipeline args; unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44355", + "accept": false, + "reason": "Import error for `AutoImageProcessor` vs errors running compiled Python files; different failure modes." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor parallel training bug vs `reduce_labels` returning one label; no concrete overlap." + }, + { + "left": "issue:44261", + "right": "issue:44360", + "accept": false, + "reason": "MLA layernorm epsilon precision issue vs DSA indexer/ReLU comment; different subsystems and symptoms." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": false, + "reason": "Lazy loading bug vs unsafe `sys.modules` access; related area, but not the same underlying defect." + }, + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "Processor save/load missing files vs a generic bug report title; insufficient evidence of sameness." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading regression vs undefined TF backend when tensorboard is installed; unrelated." + }, + { + "left": "issue:44393", + "right": "issue:45753", + "accept": false, + "reason": "Qwen3-VL bbox hallucination/error vs infinite loop on a specific image; different concrete bugs." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Broken env command vs checkpoint reload compatibility after upgrade; different code paths." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "Vision model bbox issue vs `PretrainedConfig` type-checking regression; unrelated." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import-star performance problem vs missing GenerationMixin cache in chunked prefill; adjacent theme, but not the same bug." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Qwen3-VL-Moe bug vs missing `_tp_plan` for tensor parallelism; different model features and fixes." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Checkpoint reload breakage across versions vs remote-code/meta-init incompatibility for InternVL2; not the same issue." + }, + { + "left": "issue:44373", + "right": "issue:44961", + "accept": false, + "reason": "Wrong docstring for `position_ids` vs unrelated placeholder title; no evidence of a shared bug." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "`cached_file` kwargs forwarding bug vs scheduler `timescale` forwarding bug; both forwarding-related but different APIs and fixes." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "Test flakiness from missing seeds vs scheduler kwargs bug; unrelated." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Import slowness vs lazy loading failure; same broad area but not the same concrete defect." + }, + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "Model-type warning on instantiation vs Liger Kernel not applied with `model_init`; different behaviors." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Python version incompatibility vs Qwen3-VL visual hallucination/error; unrelated." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Flash Attention 2 support error vs Qwen3-VL bbox hallucination; different model/runtime issues." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "Tokenizer class mismatch in save_pretrained vs generic bug report title; no match." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "CLI env regression vs scheduler kwargs bug; unrelated." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwargs forwarding vs config type-checking regression; different code paths." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "`position_ids` buffer access error vs generic bug report title; no concrete overlap." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Missing generation cache vs typo in cache strategies; different fixes." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-star performance regression vs missing `HybridCache` import; related import area, but not the same bug." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Import-time regression vs a little bug in `testing_utils.py`; no shared underlying issue." + }, + { + "left": "issue:41669", + "right": "issue:44492", + "accept": false, + "reason": "Import-star slowdown vs cache-strategy typo; different problems." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel application with `model_init` vs degenerate training under flash_attention_3; different model-training bugs." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3-VL image output bug vs Qwen3.5 Moe `from_pretrained` error; different models and failure modes." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "`post_init` allowance issue vs tied-weight-key AttributeError; unrelated." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs still referencing removed pipeline APIs vs scheduler kwargs handling; different scopes." + } + ] + }, + "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": 16, + "review_unit_index": 44, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-385", + "nodes": [ + "issue:30333", + "issue:31515", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:37428", + "issue:38617", + "issue:41669", + "issue:42371", + "issue:42913", + "issue:43066", + "issue:43298", + "issue:43502", + "issue:43519", + "issue:43525", + "issue:43572", + "issue:43576", + "issue:43606", + "issue:43704", + "issue:43827", + "issue:44230", + "issue:44261", + "issue:44263", + "issue:44273", + "issue:44315", + "issue:44360", + "issue:44485", + "issue:44492", + "issue:44623", + "issue:44655", + "issue:44861", + "issue:44908", + "issue:44998", + "issue:45092", + "issue:45230", + "issue:45290", + "issue:45310", + "issue:45341", + "issue:45468" + ], + "soft_pairs": [ + "issue:31795|issue:33453", + "issue:41669|issue:44908", + "issue:42371|issue:44492", + "issue:44655|issue:44861", + "issue:44360|issue:45468", + "issue:43576|issue:44273", + "issue:45230|issue:45310", + "issue:43827|issue:44492", + "issue:44273|issue:44908", + "issue:43519|issue:44485", + "issue:41669|issue:43704", + "issue:43519|issue:44908", + "issue:31795|issue:34689", + "issue:31795|issue:37428", + "issue:44315|issue:45092", + "issue:43704|issue:44485", + "issue:44485|issue:45468", + "issue:42371|issue:43704", + "issue:43502|issue:43519", + "issue:43519|issue:44623", + "issue:43519|issue:43704", + "issue:44263|issue:45468", + "issue:43606|issue:45468", + "issue:44261|issue:45468", + "issue:43704|issue:45341", + "issue:42913|issue:43066", + "issue:44908|issue:44998", + "issue:44861|issue:45290", + "issue:43525|issue:44861", + "issue:41669|issue:44998", + "issue:43704|issue:44998", + "issue:43066|issue:44998", + "issue:43572|issue:44861", + "issue:31515|issue:38617", + "issue:44360|issue:44998", + "issue:44230|issue:44623", + "issue:30333|issue:31515", + "issue:31515|issue:36296", + "issue:31515|issue:31795", + "issue:44230|issue:45310", + "issue:31515|issue:33453", + "issue:43298|issue:44230" + ], + "prepared_review_unit_hash": "50881226b04dcaf01a66949e534287454847ffa2ba9f6c2bd46607047cf6d53f", + "component_budget": { + "node_count": 385, + "item_count": 385, + "soft_pair_count": 1445, + "serialized_chars": 744694, + "estimated_input_tokens": 186174, + "estimated_eval_tokens": 372604 + }, + "budget": { + "node_count": 40, + "item_count": 40, + "soft_pair_count": 42, + "serialized_chars": 34967, + "estimated_input_tokens": 8742, + "estimated_eval_tokens": 17740 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a loose collection of unrelated Transformers bugs and docs issues. The soft pairs share broad themes like loading, tokenizers, or v5 regressions, but not the same concrete failure mode, so none should be merged as duplicates.", + "confidence": 0.74, + "canonical_issue_reason": "issue:45092 is the most representative and actionable loading-regression report here: it names a specific v5 meta-initialization incompatibility, a clear affected class of checkpoints, and a plausible shared code path with nearby loading bugs.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45092 is the best single issue to anchor the cluster because it is specific, technically grounded, and broad enough to represent the main loading-compatibility regression theme without being overly narrow to one model.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "One is a documentation confusion report; the other is a tokenizer-loading regression. Same broad area, different problem." + }, + { + "left": "issue:41669", + "right": "issue:44908", + "accept": false, + "reason": "Import-star performance cleanup and scheduler kwargs handling are unrelated code paths." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 behavior settings and a cache-strategies typo are not the same bug." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Saving Pipeline objects and tied-weight-key handling fail in different parts of serialization." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "A missing ReLU in one DSA indexer and Gemma4 audio positional encoding are different model-implementation issues." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Broken env command and lazy-loading problems are unrelated features." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "Both mention general bugs, but only one is a specific from_pretrained failure; the other is too vague to match." + }, + { + "left": "issue:43827", + "right": "issue:44492", + "accept": false, + "reason": "Outdated pipeline docs and a cache-strategy typo are different scopes and symptoms." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading behavior and inverse_sqrt scheduler kwargs are unrelated." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Timestamp calculation in a processor and GLM-5 RoPE math are not the same defect." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import-star usage/performance and VRAM leakage in multi-threaded dataloaders are different problems." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Processor timestamp math and scheduler parameter passing do not share a concrete bug." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Documentation confusion about forward args does not duplicate a model-loading regression." + }, + { + "left": "issue:31795", + "right": "issue:37428", + "accept": false, + "reason": "A doc issue about model.forward is unrelated to a flash-attention import error." + }, + { + "left": "issue:44315", + "right": "issue:45092", + "accept": false, + "reason": "Liger-kernel application via model_init and meta-initialization checkpoint loading are different mechanisms." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "VRAM leakage and RoPE implementation bugs are distinct code-path failures." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "GLM-5 RoPE math and Gemma4 audio positional encoding are separate architecture-specific issues." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 configuration warning and VRAM leakage are unrelated." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "Unexpected network requests with local_files_only and a timestamp bug in a processor are not duplicates." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Processor timestamp calculations and missing processor save files are different save/load bugs." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Timestamp miscalculation and dataloader-thread VRAM leaks are unrelated." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "torch.split return handling in one indexer does not match Gemma4 audio encoding behavior." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch and Gemma4 audio positional encoding are different model-runtime issues." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "A missing rms_norm_eps precision issue is not the same as an audio positional encoding bug." + }, + { + "left": "issue:43704", + "right": "issue:45341", + "accept": false, + "reason": "VRAM leakage in dataloader threads and a testing_utils bug do not match." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Both concern tokenizer behavior across v4/v5, but one is a broad behavior change report and the other is a specific decoder-type error." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwargs handling and an unrelated unemployment issue are clearly not duplicates." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Tied-weight key attribute handling and chat-template tool-call crashes are different areas of the codebase." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_id and tied-weight-key crashes are separate config/serialization defects." + }, + { + "left": "issue:41669", + "right": "issue:44998", + "accept": false, + "reason": "Import-star cleanup and an unrelated unemployment issue have no overlap." + }, + { + "left": "issue:43704", + "right": "issue:44998", + "accept": false, + "reason": "VRAM leaks in training and an unrelated unemployment issue are not duplicates." + }, + { + "left": "issue:43066", + "right": "issue:44998", + "accept": false, + "reason": "Tokenizer decoder type in v5 and an unrelated unemployment issue are not the same bug." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_idx in a config and tied-weight-key crashes affect different subsystems." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "A slow checkpoint load and an import error for layer_type_validation are unrelated." + }, + { + "left": "issue:44360", + "right": "issue:44998", + "accept": false, + "reason": "DSA indexer implementation issues are unrelated to the unemployment placeholder issue." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 support for a model family and missing save_pretrained files are different topics." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLflow reporting on failed training jobs and slow checkpoint loading are unrelated." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "Slow checkpoint loading and tensor-parallel training bugs are different failures." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Checkpoint-loading performance and forward-API documentation confusion are not the same issue." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 support work and a Qwen3.5 MoE from_pretrained error are distinct tasks." + }, + { + "left": "issue:31515", + "right": "issue:33453", + "accept": false, + "reason": "Slow checkpoint loading is not the same as a tokenizer-loading regression." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Continuous batching audio-model support and FP8 inference support are different feature gaps." + } + ] + }, + "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": 16, + "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": "b5fea8e9a0d6664c58b60317903a64d8a05aeae154b894602d0b8a793b2a723b", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1590, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1590, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are not duplicates: one requests Flash Attention 2 support for more architectures, the other requests GGUF support for more architectures. They target different feature areas and underlying code paths.", + "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 enhancement requests (Flash Attention 2 vs GGUF support) with no shared underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 4, + "component_count": 16, + "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-27088-4", + "nodes": [ + "issue:27088", + "issue:43834", + "issue:44365", + "issue:44370" + ], + "soft_pairs": [ + "issue:27088|issue:44370", + "issue:27088|issue:44365", + "issue:27088|issue:43834", + "issue:43834|issue:44370", + "issue:43834|issue:44365", + "issue:44365|issue:44370" + ], + "prepared_review_unit_hash": "8476dcdfa0ed61b3f136c2774154fe309c9b9308b91d0bf003a7e86c95c71be4", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4453, + "estimated_input_tokens": 1114, + "estimated_eval_tokens": 2484 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4453, + "estimated_input_tokens": 1114, + "estimated_eval_tokens": 2484 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are all docs-translation issues, but they are not the same underlying request: one is a concrete Turkish translation issue, while the others are generic template-style placeholders for arbitrary languages. The similarity is mostly boilerplate, so I would not merge them as duplicates.", + "confidence": 0.92, + "canonical_issue_reason": "Issue #27088 is the only concrete, language-specific request ([i18n-TR] Turkish). The others are generic placeholders and do not describe the same exact translation target.", + "canonical_pr_reason": null, + "best_issue_reason": "#27088 is the most actionable representative of the cluster because it names a specific language and a specific docs translation task, unlike the templated generic entries.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Same broad docs-translation theme, but one is Turkish-specific and the other is a generic language placeholder; not the same underlying issue." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "The titles/body structure match only as a template; the actual target language differs, so this is not a duplicate." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "A concrete Turkish translation request is not the same as a generic placeholder issue for an unspecified language." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": false, + "reason": "Both are template-like translation requests, but they are placeholders rather than the same concrete translation task." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": false, + "reason": "High textual similarity comes from shared boilerplate, not from the same underlying language-specific request." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": false, + "reason": "These are nearly identical templates, but without a shared concrete target language they should not be treated as duplicates of 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": 5, + "component_count": 16, + "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-36817-14", + "nodes": [ + "issue:36817", + "issue:37015", + "issue:41115", + "issue:41211", + "issue:42503", + "issue:42738", + "issue:43671", + "issue:43837", + "issue:43909", + "issue:44121", + "issue:44205", + "issue:45027", + "issue:45177", + "issue:45322" + ], + "soft_pairs": [ + "issue:36817|issue:41211", + "issue:36817|issue:43671", + "issue:37015|issue:43837", + "issue:37015|issue:44121", + "issue:37015|issue:43909", + "issue:37015|issue:42503", + "issue:37015|issue:45027", + "issue:41115|issue:43837", + "issue:41115|issue:44121", + "issue:41115|issue:42503", + "issue:41115|issue:45027", + "issue:41211|issue:43671", + "issue:42503|issue:43837", + "issue:42503|issue:43671", + "issue:42503|issue:42738", + "issue:42738|issue:45027", + "issue:43671|issue:43837", + "issue:43671|issue:45027", + "issue:43671|issue:45177", + "issue:43671|issue:44121", + "issue:43837|issue:45322", + "issue:43837|issue:43909", + "issue:43909|issue:44205", + "issue:43671|issue:43909", + "issue:44121|issue:45322", + "issue:43837|issue:44121", + "issue:43837|issue:45027", + "issue:43837|issue:45177" + ], + "prepared_review_unit_hash": "7d57b4f4777b51a6759b26c325bce7dc0b49113a630a4287cefbea30fa11818c", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 28, + "serialized_chars": 17650, + "estimated_input_tokens": 4413, + "estimated_eval_tokens": 9082 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 28, + "serialized_chars": 17650, + "estimated_input_tokens": 4413, + "estimated_eval_tokens": 9082 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are mostly one-off model addition requests across different model families and modalities. They are loosely similar at the \u201cadd a model/config support\u201d level, but not duplicates of the same underlying bug or change.", + "confidence": 0.94, + "canonical_issue_reason": "issue:37015 ('Add NeoBERT') is a clear, generic model-add request and sits near the center of the similarity set, but it is still distinct from the other model-specific requests.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:37015 is the best representative issue for this cluster because it is open, straightforward, and broadly matches the repeated pattern of requesting new model support without being tied to a narrow edge case.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "Different model requests ('EuroBert' vs 'DEIMv2'); same broad category, but not the same underlying change." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "'EuroBert' config support is unrelated to Qwen3-TTS support; different model and modality." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "'NeoBERT' vs 'Qwen3-ASR' are different model families and different tasks." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "A NeoBERT addition is not the same change as adding OpenAI weight-sparse transformer support." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "'NeoBERT' and 'LFM2.5 Audio 1.5B' are different model requests, with no shared code-path problem." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "'NeoBERT' and 'ModernVBERT' are different model additions; similarity is only at the generic config-support level." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "'NeoBERT' is unrelated to 'Voxtral-4B-TTS-2603'; different model and audio/TTS path." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 architecture support and Qwen3-ASR support are separate model integrations." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 architecture support is not the same as weight-sparse transformer support." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "'MiniCPM3' and 'ModernVBERT' are distinct model requests." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "MiniCPM3 architecture work is unrelated to Voxtral TTS support." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "'DEIMv2' and 'Qwen3-TTS' are different model additions and likely different code paths." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT model support is separate from Qwen3-ASR support." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT and Qwen3-TTS are different models/tasks; no same underlying bug." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "'ModernVBERT models' and 'BERT-like models with RoPE' overlap conceptually, but they are not the same concrete change." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "A generic RoPE request for BERT-like models is not the same as adding Voxtral TTS support." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "Qwen3-TTS and Qwen3-ASR share a family name, but they are different modalities and separate support requests." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Both are audio-related, but TTS support for Qwen3 and Voxtral TTS are distinct model integrations." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-TTS and DeepSeek-OCR2 are unrelated model requests." + }, + { + "left": "issue:43671", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-TTS support is not the same as adding a weight-sparse transformer architecture." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "Qwen3-ASR support and EUPE encoder support are different model changes." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "ASR support and LFM2.5 Audio support are related only by being audio models, not the same change." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "LFM2.5 Audio support and SAM3-LiteText are different model families." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS and LFM2.5 Audio 1.5B are separate audio model requests, not duplicates." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Weight-sparse transformer support and EUPE encoder support target different architectures." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-ASR support is unrelated to adding OpenAI weight-sparse transformer support." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral-4B-TTS-2603 are different audio model requests." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-ASR and DeepSeek-OCR2 are unrelated model additions." + } + ] + }, + "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": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-37934-6", + "nodes": [ + "issue:37934", + "issue:42936", + "issue:43240", + "issue:43668", + "issue:43813", + "issue:44031" + ], + "soft_pairs": [ + "issue:37934|issue:43668", + "issue:42936|issue:43240", + "issue:42936|issue:44031", + "issue:42936|issue:43813", + "issue:42936|issue:43668", + "issue:43240|issue:43668", + "issue:43668|issue:44031", + "issue:43668|issue:43813" + ], + "prepared_review_unit_hash": "73bf6ad4adb85a211f3e1ccc2beab8fe300da25fd0bb9687de6ece9ee1b13d87", + "component_budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 8, + "serialized_chars": 6293, + "estimated_input_tokens": 1574, + "estimated_eval_tokens": 3404 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 8, + "serialized_chars": 6293, + "estimated_input_tokens": 1574, + "estimated_eval_tokens": 3404 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster is only loosely similar on wording, but the issues describe unrelated bugs across different models/configs/utilities. I would not merge any of the soft pairs as duplicates.", + "confidence": 0.18, + "canonical_issue_reason": "issue:44031 is the most concrete and broadly actionable bug report in the set, with the most discussion and a clear regression-style description.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44031 is the best representative issue because it is specific, reproducible in intent, and not just a typo or type-hint cleanup.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "One is about Llama4TextL2Norm possibly using RMS norm, the other is a ModernBERTConfig type-hint problem; different code paths and bug types." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Mask2Former ignore_value not being used is unrelated to kwargs not being forwarded to a loss function; only broad training-loss context overlaps." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "Mask2Former loss handling and tokenizer regex warnings are unrelated subsystems with different failure modes." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "A missing ignore_value usage in Mask2Former is not the same as a typo in a quantization attribute name." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "Mask2Former loss behavior and a ModernBERT config type annotation issue are unrelated." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Forwarding kwargs to loss calculation is a runtime behavior bug; incorrect norm_eps typing is a config typing issue." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "ModernBERTConfig norm_eps typing and tokenizer regex warnings are unrelated concerns." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "These are separate small issues in different areas: config typing versus a string typo in quantization code." + } + ] + }, + "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": 16, + "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": "6713cdc4c094c1720e2741cce0d098f986096d7a6ae74c24a8a80db4dfc521ab", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are not duplicates: one is about block decoding behavior, while the other is a feature/support request for DeepSeek v3.2. They do not describe the same underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "If one had to be kept as the more actionable standalone item, #43037 is clearer and more specific; however, these issues should not be clustered as duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "The titles and implied scopes are unrelated: block decoding is an implementation/behavior topic, while DeepSeek v3.2 support is a model compatibility feature request. They do not appear to be the same bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 8, + "component_count": 16, + "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": "6c1ea85627d6344b0ff1098379f46bf8680483470be4944f3bb9938bd2993306", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1583, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1583, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related only at a high level: both mention custom trainers and Trainer behavior, but they target different problems. One is about `remove_unused_column` configurability, the other about FLOPS estimation for custom trainers.", + "confidence": 0.94, + "canonical_issue_reason": "Issue #43598 is the better anchor because it is the more general Trainer-customizability concern and is phrased as a behavior change rather than a narrow metric-specific fix.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #43598 is the best standalone issue to keep for this cluster because it captures the broader custom-trainer customization theme more cleanly than the FLOPS-specific report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "Do not treat as duplicates: they share the same Trainer/custom-trainer context, but one concerns `remove_unused_column` behavior and the other concerns FLOPS counting. Different underlying bugs, different code paths, and not a single merged fix." + } + ] + }, + "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": 16, + "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": "ff52469a3c880ce0a151dbbe97de0e742168deae07d23d198aa63ed0591c0ab9", + "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 two issues are about different model families and different AutoModel entry points, so they are not duplicates.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue: one request is for Qwen3-Omni registration in AutoModel/AutoModelForConditionalGeneration, the other is for AutoModelForSequenceClassification support for Qwen3.5.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43684 is the broader, more central model-registration request, but it should not be treated as a duplicate of 44405.", + "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 target different models and auto classes, so they should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 10, + "component_count": 16, + "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": "be4346cea43a9ab2f8f119c822518ec9aff176f0c9f700001d6e20a054c2a4b6", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4266, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4266, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly low-signal noise: three near-identical closed issues titled \"huggingface \" and one unrelated issue titled \"Fett\". The three huggingface issues look similar or likely related, but the packet only provides soft similarity evidence, so I would not confirm them as duplicates; \"Fett\" is unrelated.", + "confidence": 0.78, + "canonical_issue_reason": "43845 is the earliest of the repeated \"huggingface \" issues and is a reasonable representative for the similar low-signal reports, though the evidence is not strong enough to confirm exact duplication.", + "canonical_pr_reason": null, + "best_issue_reason": "43845 is the best issue representative because it is the earliest of the three nearly identical \"huggingface \" reports and has no distinguishing evidence against it.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "Different titles and no supporting body content or shared targets; only weak similarity is present." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "Different titles and no supporting body content or shared targets; not enough evidence for the same underlying issue." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "\"Fett\" is unrelated to the repeated \"huggingface \" issues and lacks evidence of the same underlying problem." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": false, + "reason": "The titles are the same, but the packet only shows soft similarity with no deterministic acceptance or concrete shared evidence, so this is not enough to confirm a duplicate." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": false, + "reason": "They look similar at a very high level, but there is no concrete evidence that they describe the same issue, so the edge should not be accepted." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": false, + "reason": "Only soft similarity is present; that is insufficient to treat these as the same issue." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster summary is mostly grounded, but the soft-edge handling is too strong. In the packet, every pair has only `soft_similarity`, and `deterministic_accept` is false for all of them. That makes the statements that 43845/43846/43859 are duplicates too confident. It is fine to say the three `huggingface ` issues are similar or likely related, but not enough evidence is shown to accept those edges as same-issue duplicates. The `Fett` issue being unrelated is reasonable." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 11, + "component_count": 16, + "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": "b7defc7b9b413d1537f8cfcc020d2889c8eb923f17c28a08bd665def2d3050eb", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are mostly unrelated low-information spam issues; there is no real duplicate cluster. The only reasonable grouping signal is superficial title similarity, not shared bug content.", + "confidence": 0.93, + "canonical_issue_reason": "Issue 44371 is the earliest generic \"\" report and is as representative as any other for the spam-only subset.", + "canonical_pr_reason": null, + "best_issue_reason": "44371 is the least-specific, earliest item and best fits as a placeholder representative for this noisy cluster, though it is not a true duplicate canonical.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44371", + "right": "issue:44572", + "accept": false, + "reason": "Both are spam-labeled, but there is no shared underlying problem or concrete issue description to indicate the same bug/change." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Same as above: identical-looking spam titles do not establish a duplicate underlying issue." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "One is generic spam and the other is a documentation enhancement suggestion; they are clearly different in intent and content." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "Generic spam vs a named quote-like title; no evidence they describe the same problem." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "Documentation suggestion and spam are not the same underlying issue." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "Documentation suggestion and spam are unrelated despite superficial low-information similarity." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Both are titled \"\", but that is only a spam pattern; no concrete shared bug/change is visible." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder vs a specific named title; no shared issue substance." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder and a specific named title are not enough to treat these 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": 12, + "component_count": 16, + "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": "5630e0b17106fee10a55a0f7e1dc2949c15354aef95e66e0c0911b79030c4a30", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1645, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1645, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are effectively duplicates: they have the same title, near-identical body size, and no distinguishing metadata beyond one having an inbound reference. Issue #45145 is the better canonical candidate.", + "confidence": 0.98, + "canonical_issue_reason": "#45145 is the stronger canonical issue because it has an inbound reference, suggesting other discussion already points to it.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 best represents the cluster since it appears to be the referenced/central copy of the duplicated report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": true, + "reason": "Same title and highly similar content; no evidence they describe different underlying problems." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both issues have the same title, very similar body lengths, and no file/target links. The inbound reference on #45145 is a reasonable basis for treating it as the canonical copy. The duplicate verdict is appropriately conservative for a soft-similarity pair." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 16, + "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": 16, + "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": "37bf3d74887375b3ef196de2e72d8fd7f6bdcc56ec0af2a0374a5c84b5501db6", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1669, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1669, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are both 4-bit inference issues, but they describe different failure modes and likely different code paths: one is a Gemma4/audio-module uint8 `torch.finfo()` TypeError, the other is a BitsAndBytesConfig regression where `llm_int8_skip_modules` removes the default `lm_head` exclusion and triggers an AssertionError. I would not deduplicate them.", + "confidence": 0.96, + "canonical_issue_reason": "None \u2014 the two issues do not appear to be the same underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45674 is the better representative because it describes a clearer, more general BitsAndBytesConfig regression affecting 4-bit inference.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45672", + "right": "issue:45674", + "accept": false, + "reason": "Different bugs: 45672 is a Gemma4/audio-module `torch.finfo()` TypeError on uint8 weights, while 45674 is a BitsAndBytesConfig `llm_int8_skip_modules`/`lm_head` exclusion regression causing an AssertionError in 4-bit inference. Shared quantized inference context is too broad to merge." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 15, + "component_count": 16, + "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": "45797cdc7e1dca785be5e88e9c0551fd59918545ff104d9c0ed22294787222ed", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1476, + "estimated_input_tokens": 369, + "estimated_eval_tokens": 994 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1476, + "estimated_input_tokens": 369, + "estimated_eval_tokens": 994 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two closed issues do not look like duplicates of a software problem; they appear to be unrelated name/date placeholders with no substantive issue content.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a strong canonical representative, but if forced, issue 45762 is slightly earlier and equally sparse; however, there is no meaningful bug/report content to canonicalize.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45762", + "right": "issue:45763", + "accept": false, + "reason": "Different personal-name titles with different dates and no body or shared problem description; 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": 16, + "component_count": 16, + "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": "0d42dd80081cf6f91ed964b5146cded90f5a5a33a687c3d0203f68f72876e131", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two closed issues with unrelated, low-information personal-name/date titles and no shared content, targets, or filenames. They do not look like duplicates or the same underlying report.", + "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": "The titles reference different people and different dates, with no body text, targets, filenames, or other evidence of a shared bug or request. This does not appear to be the same issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 1, + "component_count": 26, + "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:45795" + ], + "soft_pairs": [ + "pull_request:39895|pull_request:43424", + "pull_request:41855|pull_request:43823", + "pull_request:41855|pull_request:43424", + "pull_request:42668|pull_request:43823", + "pull_request:43251|pull_request:43823", + "pull_request:43251|pull_request:44827", + "pull_request:43424|pull_request:44827", + "pull_request:43424|pull_request:43823", + "pull_request:43823|pull_request:45655", + "pull_request:43823|pull_request:43995", + "pull_request:43995|pull_request:45213", + "pull_request:43995|pull_request:44827", + "pull_request:44827|pull_request:45170", + "pull_request:44827|pull_request:45795", + "pull_request:44827|pull_request:45655", + "pull_request:44827|pull_request:45213", + "pull_request:43823|pull_request:45213", + "pull_request:43823|pull_request:45795", + "pull_request:43251|pull_request:43254" + ], + "prepared_review_unit_hash": "3460822e8fd9721d900972f61ee57dfb39a73e56ed8b9e8a72ad255bfb3f8101", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 19, + "serialized_chars": 35200, + "estimated_input_tokens": 8800, + "estimated_eval_tokens": 17856 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 19, + "serialized_chars": 35200, + "estimated_input_tokens": 8800, + "estimated_eval_tokens": 17856 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a collection of unrelated pull requests. The only likely duplicate/change overlap is the pair of PRs fixing `fixed_cross_entropy` (`43251` and `43254`), which address the same underlying loss-function kwargs gap; the rest appear to be separate model/docs/test work.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43251 is the best representative for the cross-entropy kwargs fix: it directly targets issue 43240 and provides the more explicit implementation. PR 43254 is a narrower alternate patch for the same bug.", + "best_issue_reason": null, + "best_pr_reason": "PR 43251 is the strongest global representative because it has an explicit issue link and the clearest statement of the intended behavior change in `fixed_cross_entropy`.", + "soft_edge_verdicts": [ + { + "left": "pull_request:39895", + "right": "pull_request:43424", + "accept": false, + "reason": "Unrelated PRs: one adds VideoPrism, the other adds an Executorch dynamic-shape test." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Different areas: Mistral tokenizer method additions vs adding a new MobileLLM model." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Tokenizer API patch vs Executorch test; no shared code-path bug." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor robustness changes are unrelated to adding MobileLLM." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "Loss-function kwargs fix vs new MobileLLM model addition." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "Cross-entropy helper fix and Mistral4 test/model work are different changes." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "Executorch dynamic-shape testing is unrelated to Mistral4 fixes." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "Different artifacts and code paths: export test vs new model implementation." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "Adding MobileLLM and fixing config resolution order are unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "New model addition vs Falcon output-interface refactor; not the same bug/change." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon refactor and model-creation skill docs are unrelated." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Different model families and different changes; no same underlying issue." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 test fixes and CLIP-like `layrnorm` rename mapping are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45795", + "accept": false, + "reason": "Mistral4 test/model changes do not match docs for audio/video processors." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "Mistral4 fixes vs config-resolution order fix are separate issues." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Model fixes vs a skill/documentation PR are not the same code-path change." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "New model implementation and model-creation skill docs are not duplicates." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45795", + "accept": false, + "reason": "MobileLLM model addition and docs for audio/video processors are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 test/model changes vs model-creation skill docs are unrelated." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Both PRs fix the same underlying `fixed_cross_entropy` kwargs support gap in `src/transformers/loss/loss_utils.py`, and they could plausibly be merged into one patch." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted pair, PR 43251 and PR 43254, clearly targets the same `fixed_cross_entropy` kwargs issue in the same file and could plausibly be merged into one bugfix PR. The other pairwise rejections are supported by the diffs and do not overstate relatedness." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 2, + "component_count": 26, + "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": "e9931d629496dfe3663839c7b619327c155122eac0533f0fa8eeec1dd59cbdfe", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7226, + "estimated_input_tokens": 1807, + "estimated_eval_tokens": 3870 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7226, + "estimated_input_tokens": 1807, + "estimated_eval_tokens": 3870 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two related VibeVoice PRs, but they target different scopes: the first implements VibeVoice, while the second adds VibeVoice Realtime plus an acoustic tokenizer. They share substantial model scaffolding, but not the same concrete change set, so they should not be merged as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #42781 is the broader and newer VibeVoice effort, adding the realtime path and acoustic tokenizer on top of the base model work, so it is the better representative of this cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR #42781 is the strongest cluster representative because it covers the expanded VibeVoice feature set and supersets much of the base integration work.", + "soft_edge_verdicts": [ + { + "left": "Implement VibeVoice", + "right": "Add VibeVoice Realtime", + "accept": false, + "reason": "Shared filenames show the same VibeVoice scaffolding, but the PRs implement different scopes: base VibeVoice integration versus a realtime variant with an acoustic tokenizer. That is related work, 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": 3, + "component_count": 26, + "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": "15c624f9e2f9edc0a7e510d30a0ec401bb75d413813b7a28239299498a83bce4", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two pull requests are unrelated: one adds TimesFM covariate forecasting support, while the other adds DINOv3 image classification support. They do not appear to be duplicate changes or the same underlying bug fix.", + "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:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Reject: these target different model families, different user-facing features, and different code paths; they are not plausibly mergeable into one PR as the same change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 4, + "component_count": 26, + "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": 26, + "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": "352464f829c5c7682c8fa58b9f4172bea57022050fe283ab00c256a47dbd8a6e", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10428, + "estimated_input_tokens": 2607, + "estimated_eval_tokens": 5470 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10428, + "estimated_input_tokens": 2607, + "estimated_eval_tokens": 5470 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are three unrelated model-addition PRs with similar scaffolding, not duplicates of the same change. The shared files are mostly registry/docs boilerplate; the actual model code, processors, and tests target different models.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: the cluster is not a duplicate set. Each PR adds a different model (DeepSeek OCR, Isaac, Kimi2-6) with distinct code paths and tests, so none should represent the others.", + "best_issue_reason": null, + "best_pr_reason": "No best PR for deduplication. If forced to pick a representative, it would still be arbitrary because the PRs are semantically unrelated beyond shared template files.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41797", + "right": "pull_request:45186", + "accept": false, + "reason": "Different models and implementations: DeepSeek OCR vs Isaac. Shared auto-registry/docs files are boilerplate and do not indicate the same underlying change." + }, + { + "left": "pull_request:45186", + "right": "pull_request:45630", + "accept": false, + "reason": "Different model additions: Isaac vs Kimi2-6. They share only generic framework wiring, not the same concrete bug or feature." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 6, + "component_count": 26, + "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": "7451ae42f4d3255edfe904ba4c5c295f0af070ac527a0ba41c359339ef94e59d", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5934, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5934, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs both update GitHub Actions workflows, but they target different action families and different upgrade reasons, so they look related rather than duplicates.", + "confidence": 0.78, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 42975 is the better canonical representative: it has the clearer, more specific goal (Node 24 compatibility) and updates core workflow actions across many files.", + "best_issue_reason": null, + "best_pr_reason": "PR 42975 is the best overall representative of the cluster because its scope is broader and more clearly framed than PR 42976, which is mainly a docker/build-push-action version bump.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "Both are workflow dependency upgrades, but they change different concrete actions and do not appear to fix the same underlying bug or be safely mergeable into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 7, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-42978-4", + "nodes": [ + "pull_request:42978", + "pull_request:43448", + "pull_request:43451", + "pull_request:45490" + ], + "soft_pairs": [ + "pull_request:42978|pull_request:43448", + "pull_request:42978|pull_request:43451", + "pull_request:42978|pull_request:45490", + "pull_request:43448|pull_request:43451", + "pull_request:43448|pull_request:45490", + "pull_request:43451|pull_request:45490" + ], + "prepared_review_unit_hash": "ee02c42ce084d67ffc982db7f8b8ef6d7313d26326dc6c750bfaa94308aaa48a", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 14078, + "estimated_input_tokens": 3520, + "estimated_eval_tokens": 7296 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 14078, + "estimated_input_tokens": 3520, + "estimated_eval_tokens": 7296 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are only superficially similar: they are separate model-addition changes for ViT NEPA, Molmo, Molmo2, and CTSM. They share standard Transformers scaffolding (docs, auto-mapping, tests), but each targets a different model and code path, so this is not a duplicate cluster.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "If a representative is needed, #43451 is the strongest anchor because it is the most extensive and heavily reviewed PR, but it is not a duplicate of the others.", + "best_issue_reason": null, + "best_pr_reason": "#43451 is the best cluster representative by scope and review activity, but only as an exemplar of a model-addition PR; it should not be treated as the same change as the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42978", + "right": "pull_request:43448", + "accept": false, + "reason": "Different model additions (ViT NEPA vs Molmo). Shared auto-mapping/docs/test scaffolding is not enough to treat them as the same fix or merge into one PR." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "ViT NEPA and Molmo2 are distinct models with different configs, processors, and implementation files; they do not address the same underlying change." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "Different model families and code paths (ViT NEPA vs CTSM); only generic repository integration files overlap." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are related by naming, but these are separate model introductions with different model/processing stacks and cannot be merged as one duplicate PR." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo and CTSM are unrelated model additions; overlapping docs/auto-registry changes are incidental." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM are distinct features touching different model implementations; they are not the same bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 8, + "component_count": 26, + "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": "63cf15d43b6a40f0f894f6c482c8f3ebb559a5b542a8e18246eeb833479af5af", + "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": "These three pull requests do not look like duplicates of one another: one is a Whisper tokenizer/test fix, one is a clearly broken test-only change, and one is a separate FP8 dequantization fix. The only similarity is superficial model/test overlap.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:44815 is the best representative of the set because it is the most coherent, substantive bug fix with a clear code-path change and accompanying tests; the other two are either unrelated or test-only noise.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:44815 is the strongest standalone PR here: it targets a concrete dequantization/loading issue, spans the relevant implementation files, and looks like a real fix rather than a test tweak.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Whisper tokenizer decode fix vs unrelated CLIP/ViT test edits; no shared bug or code path." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper batch/timestamp decoding bug is unrelated to FP8 dequantization/loading logic." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "Test-only changes in vision models are not the same underlying problem as the FP8 dequantization 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": 9, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43488-2", + "nodes": [ + "pull_request:43488", + "pull_request:43532" + ], + "soft_pairs": [ + "pull_request:43488|pull_request:43532" + ], + "prepared_review_unit_hash": "cee35efe8962a9e421767ef7de6d7e7683237372175210fec9d8d04674e9db79", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4562, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4562, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two PRs are both intentionally non-mergeable test/demonstration changes, but they target different areas: one tweaks packaging/version metadata, while the other rewires CI/workflow and testing utilities. The shared Vit test file is incidental and not evidence of the same underlying fix.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "Not the same concrete change: 43488 is about setup/version/dependency metadata, while 43532 changes GitHub workflows plus testing utilities. The overlap in a Vit test file is incidental, and these 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": 10, + "component_count": 26, + "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": "a464f95e8fd489e4706cb6d5b349e58d7fadb8470381c9de593504b69d86d3dd", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6083, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6083, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related but distinct PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They touch overlapping docs and registration files, but the model support, tests, and filenames show different concrete features, so they should not be deduplicated or merged as one change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR for a duplicate set: these are separate pipeline additions with different targets and code paths, not the same underlying change.", + "best_issue_reason": null, + "best_pr_reason": "PR #43613 is the broader representative of the cluster because it adds more model integrations and tests, but it is still not a duplicate of #43612.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Same broad theme and shared docs/registration updates, but they implement different pipelines with different task/model/test files and different concrete functionality." + } + ] + }, + "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": 26, + "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": "38e9015a7dbccab547382d5732f4cb6cbd9bc2f8520457364fdccbfda8199583", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4292, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4292, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are both Trainer.py refactors, but they address different behaviors: one changes gradient-accumulation loss scaling, the other factors out signature-column filtering for custom trainers. They are not the same underlying fix, so the soft duplicate link 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": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Different code paths and goals: 43651 changes loss normalization during training_step, while 43663 changes data-column filtering based on model signatures. Same file, but not the same bug or mergeable change set." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 12, + "component_count": 26, + "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-43656-3", + "nodes": [ + "pull_request:43656", + "pull_request:43836", + "pull_request:43842" + ], + "soft_pairs": [ + "pull_request:43656|pull_request:43842", + "pull_request:43836|pull_request:43842" + ], + "prepared_review_unit_hash": "73ed8cd93ea341612b011e0e4f2bdae8c54695002b624c56aadb19432c1ba459", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 5992, + "estimated_input_tokens": 1498, + "estimated_eval_tokens": 3252 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 5992, + "estimated_input_tokens": 1498, + "estimated_eval_tokens": 3252 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three PRs target the same serve CLI TypeAdapter/annotation NameError family in src/transformers/cli/serve.py. PR 43842 is the most focused fix; the others are broader or earlier variants of the same underlying problem.", + "confidence": 0.84, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43842 is the most directly targeted and minimal fix for the TypeAdapter NameError when pydantic is absent, with a clear runtime-safety change and less unrelated churn than the others.", + "best_issue_reason": null, + "best_pr_reason": "PR 43842 best represents the cluster because it addresses the concrete failure mode explicitly and cleanly in the same code path, without the broader annotation refactor noise present in the other PRs.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": true, + "reason": "Both are fixing the same serve CLI TypeAdapter NameError in the same file; 43656 is a broader annotations-related variant, but it plausibly overlaps with and could be merged with the targeted pydantic-safety fix in 43842." + }, + { + "left": "pull_request:43836", + "right": "pull_request:43842", + "accept": true, + "reason": "Both PRs aim at the same concrete TypeAdapter annotation/NameError issue in cli/serve.py. 43836 is the broader wrapper-style patch, while 43842 is a smaller, more explicit version of the same fix and could plausibly be combined." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: all three PRs are in src/transformers/cli/serve.py, all are tied to the same TypeAdapter/annotation NameError family, and 43842 is indeed the most narrowly targeted fix. The soft-edge acceptances are conservative enough because each pair plausibly merges into one bugfix PR without clear contradiction or unrelated scope." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 13, + "component_count": 26, + "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": 26, + "review_unit_index": 1, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43996", + "pull_request:43997", + "pull_request:44007", + "pull_request:44010", + "pull_request:44013", + "pull_request:44017", + "pull_request:44018", + "pull_request:44019", + "pull_request:44024", + "pull_request:44044", + "pull_request:44056", + "pull_request:44059", + "pull_request:44066", + "pull_request:44068", + "pull_request:44071", + "pull_request:44072", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44085", + "pull_request:44086", + "pull_request:44098", + "pull_request:44101", + "pull_request:44116", + "pull_request:44129", + "pull_request:44154", + "pull_request:44161", + "pull_request:44722" + ], + "soft_pairs": [ + "pull_request:43996|pull_request:44085", + "pull_request:43996|pull_request:44044", + "pull_request:44066|pull_request:44085", + "pull_request:44007|pull_request:44072", + "pull_request:44072|pull_request:44722", + "pull_request:44066|pull_request:44072", + "pull_request:44013|pull_request:44044", + "pull_request:44066|pull_request:44086", + "pull_request:44066|pull_request:44071", + "pull_request:44018|pull_request:44068", + "pull_request:44066|pull_request:44068", + "pull_request:44018|pull_request:44066", + "pull_request:44019|pull_request:44722", + "pull_request:44086|pull_request:44722", + "pull_request:44019|pull_request:44085", + "pull_request:44018|pull_request:44019", + "pull_request:44019|pull_request:44071", + "pull_request:44068|pull_request:44116", + "pull_request:44068|pull_request:44722", + "pull_request:44071|pull_request:44722", + "pull_request:44018|pull_request:44116", + "pull_request:44018|pull_request:44086", + "pull_request:44018|pull_request:44085", + "pull_request:44068|pull_request:44086", + "pull_request:44071|pull_request:44085", + "pull_request:44059|pull_request:44086", + "pull_request:44085|pull_request:44086", + "pull_request:44059|pull_request:44722", + "pull_request:44098|pull_request:44101", + "pull_request:44018|pull_request:44071", + "pull_request:44017|pull_request:44722", + "pull_request:44017|pull_request:44161", + "pull_request:44017|pull_request:44073", + "pull_request:44068|pull_request:44071", + "pull_request:44071|pull_request:44086", + "pull_request:43997|pull_request:44073", + "pull_request:44017|pull_request:44071", + "pull_request:44073|pull_request:44101", + "pull_request:44059|pull_request:44071", + "pull_request:44059|pull_request:44068", + "pull_request:44056|pull_request:44101", + "pull_request:44018|pull_request:44059", + "pull_request:44073|pull_request:44161", + "pull_request:44101|pull_request:44161", + "pull_request:43997|pull_request:44017", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:43997|pull_request:44018", + "pull_request:44056|pull_request:44161", + "pull_request:43997|pull_request:44161", + "pull_request:44074|pull_request:44161", + "pull_request:43997|pull_request:44071", + "pull_request:44010|pull_request:44071", + "pull_request:44010|pull_request:44073", + "pull_request:44018|pull_request:44129", + "pull_request:43997|pull_request:44074", + "pull_request:43997|pull_request:44129", + "pull_request:44010|pull_request:44018", + "pull_request:44071|pull_request:44129", + "pull_request:43996|pull_request:44161", + "pull_request:43996|pull_request:44017", + "pull_request:44010|pull_request:44044", + "pull_request:44044|pull_request:44074", + "pull_request:44066|pull_request:44722", + "pull_request:44085|pull_request:44722", + "pull_request:44085|pull_request:44129", + "pull_request:43996|pull_request:44007", + "pull_request:43996|pull_request:44129", + "pull_request:44044|pull_request:44129", + "pull_request:44044|pull_request:44154", + "pull_request:44024|pull_request:44076" + ], + "prepared_review_unit_hash": "f71f2370d0c8b506074c8d5b2f21a825a99d3bd8706feb80868693c287d3b25a", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84497, + "estimated_input_tokens": 21125, + "estimated_eval_tokens": 42506 + }, + "budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 72, + "serialized_chars": 76932, + "estimated_input_tokens": 19233, + "estimated_eval_tokens": 38722 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of independent, model-specific output-tracing refactors. The only clear duplicate groups are the GPT-Neo pair (44018/44068) and the GPT-J/CodeGen family (44066/44085/44722).", + "confidence": 0.82, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44722 is the strongest representative of the GPT-J/CodeGen tracing refactor: it covers the same concrete code path as the earlier GPT-J PRs, includes both related files, and is the most complete version tied to the umbrella issue #43979.", + "best_issue_reason": null, + "best_pr_reason": "44722 is the best single PR to represent the duplicate GPT-J/CodeGen change set because it is the most comprehensive and directly subsumes the earlier GPT-J-focused attempts.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Different models and code paths (CVT/FNet vs GPT-J); same umbrella topic is too broad to merge." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "Different model families and different refactors; only share a broad standardized-output theme." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": true, + "reason": "Both edit GPT-J output tracing in the same file and target the same underlying change; one is just a broader version with CodeGen included." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "RegNet/ResNet-style vision backbones are separate concrete code paths from EfficientNet; not the same bug/change." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet output tracing and GPT-J/CodeGen output tracing are unrelated model-specific changes." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen and EfficientNet are different implementations; no shared concrete fix." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "MobileNetV2 and DeBERTa are unrelated model stacks; same refactor pattern is not enough." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen output tracing vs MGP-STR output tracing are different model-specific changes." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen and MPT touch different code paths and should not be merged as duplicates." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44068", + "accept": true, + "reason": "Same GPT-Neo file and same output-tracing refactor; these are duplicate fixes for the same code path." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J/CodeGen versus GPT-Neo are different model implementations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "Different models (GPT-Neo vs GPT-J/CodeGen); only the refactor style is similar." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet and GPT-J/CodeGen are unrelated concrete changes." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR and GPT-J/CodeGen are different code paths; not mergeable as one PR." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44085", + "accept": false, + "reason": "ResNet output tracing and GPT-J output tracing are separate model fixes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet are unrelated model implementations." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet and MPT are distinct model-specific refactors." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert/XLM are unrelated code paths." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo is not the same underlying change as GPT-J/CodeGen, despite similar output-tracing mechanics." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT and GPT-J/CodeGen touch different model internals." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert/XLM are different model families and fixes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are different concrete changes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo and GPT-J are separate code paths." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are unrelated model-specific refactors." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44086", + "accept": false, + "reason": "MPT and MGP-STR are different models; not the same change." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT2 output tracing and MGP-STR output tracing are separate fixes." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J and MGP-STR are different model code paths." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT2 refactor is unrelated to GPT-J/CodeGen." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT and Flaubert/XLM are different model implementations." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44722", + "accept": false, + "reason": "Segformer and GPT-J/CodeGen are unrelated model-specific refactors." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "Segformer and LongT5 are different code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44073", + "accept": false, + "reason": "Segformer and VisualBert are separate model fixes." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are different model families." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44073", + "accept": false, + "reason": "RegNet and VisualBert are unrelated concrete changes." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44017", + "accept": false, + "reason": "RegNet and Segformer are separate model-specific refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are different model code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet and GPT-Neo are unrelated fixes." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44161", + "accept": false, + "reason": "VisualBert and LongT5 are different model implementations." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet and Flaubert/XLM are unrelated model-specific refactors." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44059", + "accept": false, + "reason": "MPNet and GPT2 are different code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are different concrete changes." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 are unrelated model fixes." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44161", + "accept": false, + "reason": "FNet/CVT and LongT5 are different model families." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44017", + "accept": false, + "reason": "FNet/CVT and Segformer are separate concrete code paths." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBert and DeBERTa are unrelated model-specific changes." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa and TextNet are different model implementations." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBert and VisualBert are unrelated refactors." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44059", + "accept": false, + "reason": "GPT-Neo and GPT2 are different model code paths." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44076", + "accept": false, + "reason": "TextNet and ImageGPT are separate model-specific changes." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet and TextNet are unrelated concrete fixes." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44101", + "accept": false, + "reason": "MPNet and Flaubert/XLM are different models." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44071", + "accept": false, + "reason": "TextNet and MPT are unrelated model-specific refactors." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44018", + "accept": false, + "reason": "SqueezeBert and GPT-Neo are different code paths." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet and MPT are unrelated model fixes." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBert and DeBERTa are different model families." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa and TextNet are separate concrete changes." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: it sticks to the model-specific output-tracing refactors and only treats the GPT-Neo and GPT-J/CodeGen sequences as duplicate families. The accepted soft edges are conservative because they involve the same model/file and look like broader-vs-subset versions of the same bugfix. The rejected pairs are also appropriately limited to different model families or unrelated code paths." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 26, + "review_unit_index": 2, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43997", + "pull_request:44010", + "pull_request:44056", + "pull_request:44073", + "pull_request:44074", + "pull_request:44129" + ], + "soft_pairs": [ + "pull_request:44073|pull_request:44074", + "pull_request:44056|pull_request:44073", + "pull_request:44056|pull_request:44074", + "pull_request:43997|pull_request:44010", + "pull_request:44010|pull_request:44129" + ], + "prepared_review_unit_hash": "3671a9ca317cef380c6c3a4f0136839b04a078538852c315c32c18929129d0be", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84497, + "estimated_input_tokens": 21125, + "estimated_eval_tokens": 42506 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 5, + "serialized_chars": 13065, + "estimated_input_tokens": 3267, + "estimated_eval_tokens": 6790 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These PRs all follow the same broad theme\u2014migrating model forwards to standardized output tracing\u2014but they are model-specific refactors in different code paths, not duplicate fixes for one underlying bug. No soft edge should be merged as the same change.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "44129 is the best representative PR because it has an explicit issue target and a concrete, self-contained SpeechT5 refactor with clear scope.", + "best_issue_reason": null, + "best_pr_reason": "44129 is the strongest standalone PR: explicit issue linkage, clear problem statement, and the most complete/grounded implementation among the candidates.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "Same refactor pattern, but different models and different implementation details/tests; not the same underlying bug or mergeable into one PR." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "Both are output-tracing migrations, but MPNet and VisualBert are separate model code paths with distinct changes." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "Different model families and separate refactors; similarity is only at the decorator/migration pattern level." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet vs SqueezeBert are unrelated model implementations; these are parallel cleanup refactors, not one shared fix." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "Different models and code paths; they share a migration style but 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": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43998-13", + "nodes": [ + "pull_request:43998", + "pull_request:43999", + "pull_request:44000", + "pull_request:44001", + "pull_request:44002", + "pull_request:44003", + "pull_request:44004", + "pull_request:44025", + "pull_request:44026", + "pull_request:44027", + "pull_request:44028", + "pull_request:44029", + "pull_request:44030" + ], + "soft_pairs": [ + "pull_request:43998|pull_request:44002", + "pull_request:43998|pull_request:44000", + "pull_request:43998|pull_request:44001", + "pull_request:43998|pull_request:44003", + "pull_request:43998|pull_request:44029", + "pull_request:43999|pull_request:44001", + "pull_request:43999|pull_request:44002", + "pull_request:43999|pull_request:44025", + "pull_request:43999|pull_request:44026", + "pull_request:43999|pull_request:44027", + "pull_request:44000|pull_request:44001", + "pull_request:44000|pull_request:44002", + "pull_request:44000|pull_request:44003", + "pull_request:44000|pull_request:44025", + "pull_request:44001|pull_request:44002", + "pull_request:44001|pull_request:44003", + "pull_request:44001|pull_request:44025", + "pull_request:44002|pull_request:44029", + "pull_request:44002|pull_request:44003", + "pull_request:44003|pull_request:44028", + "pull_request:44003|pull_request:44004", + "pull_request:43998|pull_request:44004", + "pull_request:44001|pull_request:44004", + "pull_request:44002|pull_request:44004", + "pull_request:44000|pull_request:44004", + "pull_request:44025|pull_request:44026", + "pull_request:44025|pull_request:44027", + "pull_request:44025|pull_request:44028", + "pull_request:44025|pull_request:44029", + "pull_request:44025|pull_request:44030", + "pull_request:44026|pull_request:44029", + "pull_request:44026|pull_request:44027", + "pull_request:44026|pull_request:44028", + "pull_request:44026|pull_request:44030", + "pull_request:44027|pull_request:44028", + "pull_request:44027|pull_request:44029", + "pull_request:44027|pull_request:44030", + "pull_request:44028|pull_request:44029", + "pull_request:44028|pull_request:44030", + "pull_request:44029|pull_request:44030" + ], + "prepared_review_unit_hash": "360f58a728175ce26e42ad2615ca3efc6f1db7e33516338a9518351f5d115e45", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40169, + "estimated_input_tokens": 10043, + "estimated_eval_tokens": 20342 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40169, + "estimated_input_tokens": 10043, + "estimated_eval_tokens": 20342 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All items are distinct model-specific refactors around output tracing/can_return_tuple. They share a common pattern, but they do not target the same concrete bug or change and should not be deduplicated.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44003 is the strongest representative: it covers two closely related Mamba-family implementations and includes the core tracing/capture plumbing, making it the most substantial exemplar of this refactor cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR 44003 is the best cluster representative because it is broader than the single-model PRs and shows the main output-tracing pattern without being just a narrow variant.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Same broad refactor theme, but timm_backbone and upernet are different code paths with different output logic." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "Different model families and different forward signatures; only the refactor pattern is shared." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Both are output-tracing refactors, but timm_backbone and univnet are unrelated implementations." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "Mismatched model families; 43998 is a single backbone wrapper, while 44003 changes Mamba/FalconMamba internals." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "timm_backbone and rwkv use different tracing mechanisms and code paths; not the same change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "Both are refactors, but MobileNetV1 and UnivNet are separate models with unrelated forward contracts." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "MobileNetV1 and UperNet are different model targets; shared refactor wording is not enough." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "Different model families and different files; one is MobileNetV1, the other depth-anything variants." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "MobileNetV1 output tracing is unrelated to VisionEncoderDecoder wrapper changes." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "MobileNetV1 and SpeechEncoderDecoder are separate encoder-decoder/model wrapper code paths." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "Vision-Text Dual Encoder and UnivNet are unrelated models; only the refactor motif overlaps." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Different architectures and forward semantics; not a single bug fix." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "Vision-text dual encoder changes do not overlap with Mamba/FalconMamba tracing changes." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "Different model families and different output-capture implementations." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "UnivNet and UperNet are unrelated model code paths; same pattern, different change." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Mamba/FalconMamba and UnivNet are different implementations with different tracing plumbing." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "UnivNet is not the same bug/change as the depth-anything refactor." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "UnivNet output tracing is unrelated to CodeGen cache/output handling." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44029", + "accept": false, + "reason": "UperNet segmentation wrapper and RWKV recurrent core are different code paths." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "UperNet and Mamba/FalconMamba are separate model families; not mergeable as one PR." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44028", + "accept": false, + "reason": "Mamba-family tracing changes do not match SuperPoint's feature extraction/output refactor." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "Different model families and different refactor mechanics; CodeGen cache changes are not the same change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "Timm backbone output tracing and CodeGen cache/output typing are unrelated." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "UnivNet and CodeGen touch unrelated forward logic and output shapes." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "UperNet segmentation outputs are not the same code path as CodeGen's attention/cache logic." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "Vision-text dual encoder and CodeGen are unrelated; shared refactor theme is too broad." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "Both are refactors, but they target different wrapper/model families: depth-anything vs vision/speech encoder-decoder." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "Depth Anything and Speech Encoder-Decoder are unrelated changes, not one bug fix." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "Depth-estimation models and SuperPoint do not share the same concrete output-tracing change." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "Depth Anything output tracing is separate from RWKV's recurrent state/output handling." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "Depth-estimation refactor and DPR encoder pooling refactor are distinct changes." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "Vision/Speech encoder-decoder wrappers and RWKV internals are different code paths." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "VisionEncoderDecoder and SpeechEncoderDecoder are parallel wrappers but still separate PR targets and changes." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "Encoder-decoder wrappers are not the same change as SuperPoint feature tracing." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "VisionEncoderDecoder wrapper changes and DPR encoder changes are unrelated." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "SpeechEncoderDecoder and SuperPoint are unrelated model implementations." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "Speech encoder-decoder wrapper logic is not the same as RWKV state/output refactoring." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "SpeechEncoderDecoder and DPR touch different architectures and output contracts." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "SuperPoint and RWKV are different code paths with different output semantics." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "SuperPoint feature tracing is unrelated to DPR encoder pooling/output handling." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "RWKV recurrent tracing and DPR encoder output refactors are distinct model changes." + } + ] + }, + "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": 26, + "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": "5b0970c67d84b73184eb185c6296518accef95aa2b257ddf82ef81e8028fb20b", + "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": "Three open PRs, but they are separate model-addition implementations rather than duplicates. They share some auto/docs plumbing, yet each targets a different model family and code path.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45075 is the best representative of the cluster because it is a fuller, non-draft implementation with substantial tests and conversion support, but it is not a duplicate of the others.", + "best_issue_reason": null, + "best_pr_reason": "PR 45075 is the strongest standalone PR in this set: it looks the most complete and mature, with broader implementation coverage and high review activity.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "PenguinVL vs Voxtral TTS are different model families with different functionality and code paths; shared auto/docs files are just framework plumbing." + }, + { + "left": "pull_request:44662", + "right": "pull_request:45075", + "accept": false, + "reason": "PenguinVL and DeepSeek-OCR-2 are unrelated model additions; they touch overlapping registration files but implement different concrete models." + }, + { + "left": "pull_request:45075", + "right": "pull_request:45401", + "accept": false, + "reason": "DeepSeek-OCR-2 and Voxtral TTS solve different tasks and modify different model code, so 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": 17, + "component_count": 26, + "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": "dbe44d063893dc357fbc04a83dfc9f85ceabd52fe063fe3581db146b42bca775", + "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 PRs are related in the broad area of model/kernel fusion, but they do not look like the same concrete change. One introduces a generic Module Fusion API and tests; the other wires in n-to-1 kernel fusion through KernelConfig and integration points. I would not merge them as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45363 is the more complete, user-facing feature: it adds KernelConfig-based fusion plumbing across model loading and integrations, with a concrete example. PR #44979 is a lower-level API addition for module fusion and appears to be a separate foundational piece rather than the same change.", + "best_issue_reason": null, + "best_pr_reason": "PR #45363 is the best representative of the cluster because it is the broader end-to-end implementation and most clearly defines the actual fused-kernel workflow. PR #44979 is adjacent but not the same change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "Both concern fusion, but #44979 adds a generic Module Fusion API while #45363 implements KernelConfig-driven n-to-1 kernel fusion and integration wiring. Different abstractions, different code paths, and not plausibly one merged PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 18, + "component_count": 26, + "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": "8e1683248f8f91e0e7e8a9846ef70332eb5cc454db8a2e536855ddd992998558", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4413, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4413, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs fix the same Mistral tokenizer AttributeError by switching from backend_tokenizer.pre_tokenizer to tokenizer.pre_tokenizer in the same file. PR 45317 is the more complete implementation because it includes an added test, while 45086 is the smaller precursor.", + "confidence": 0.97, + "canonical_issue_reason": "Issue 45081 is the shared bug report both PRs explicitly target, so it is the natural canonical issue for this cluster.", + "canonical_pr_reason": "PR 45317 is the better canonical PR because it addresses the same bug fix and also adds test coverage, making it the more complete representative change.", + "best_issue_reason": "If selecting one issue node, 45081 best represents the underlying bug since both PRs are direct fixes for that exact reported problem.", + "best_pr_reason": "45317 is the best PR overall because it fixes the same concrete AttributeError path as 45086 but is more complete due to the accompanying test file.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same underlying bug fix in the same code path: both replace backend_tokenizer.pre_tokenizer with tokenizer.pre_tokenizer in _patch_mistral_regex for issue 45081. 45317 is essentially the extended version with tests, so they are merge-equivalent." + } + ] + }, + "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": 26, + "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": "277c3d8c80453e28a761c90e13983981c88b623862e2769bd0c3a0896550d9c5", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6066, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6066, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two NemotronH pull requests overlap on model files, but they are not the same change: one is a narrow conversion/mapping refactor, the other is a broader model split into NemotronH/NemotronHDense/NemotronHSparse with docs and tests.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45569 is the better canonical representative because it is the broader, more complete architectural change and includes the user-facing model split, docs, and tests. PR #45115 is a narrower internal refactor focused on conversion mappings and inheritance wiring.", + "best_issue_reason": null, + "best_pr_reason": "PR #45569 is the strongest overall PR for the cluster since it covers the more substantial end-state of the NemotronH work. PR #45115 appears to be a partial, implementation-focused step rather than the main deliverable.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "Do not merge as duplicates: #45115 targets conversion-key renames and inheritance adjustments, while #45569 introduces a broader NemotronH reorganization into separate dense/sparse models plus docs and tests. They share subsystem files but not the same concrete bug or mergeable code-path fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 20, + "component_count": 26, + "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": "34184b235bab7babd767dfbd1c42543bfc2f90d2122b9c13d01b72661b5c429c", + "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": "Two related but distinct PRs: one adds the heterogeneous config plumbing, the other adds heterogeneous model/runtime support built on top of it. They share helper files, but they are not the same underlying change and should not be merged as duplicates.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45332 is the broader end-to-end feature PR: it includes runtime/model changes, cache/masking updates, and tests, so it is the most representative single artifact for the heterogeneous-model work.", + "best_issue_reason": null, + "best_pr_reason": "PR #45332 is the best standalone representative because it exercises the full heterogeneous-model code path rather than only the configuration layer; #45333 is a narrower prerequisite-style config PR.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "Related feature work, but not the same concrete change: #45333 adds per-layer configuration support, while #45332 adds model/cache/runtime support that consumes that config. They touch different code paths 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": 21, + "component_count": 26, + "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": "f348b1de3926d44f0e704422f8a49f8bd39fd503c4e44d95b0819bbbbb65ce2b", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5514, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5514, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs are typing-related, but they are not the same change: one is a broad repo-wide type-checking cleanup, the other is a focused typing/Protocol update for modeling-related code. Shared files alone are not enough to merge them as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45425 is the more focused, self-contained change: it introduces typing aliases/Protocols and updates specific transformer/modeling code paths, whereas #45415 is a broad type-checking sweep across many files plus config changes.", + "best_issue_reason": null, + "best_pr_reason": "PR #45425 is the better representative for the cluster because it has a clearer, concrete implementation scope and stronger review activity; #45415 is more of an umbrella typing cleanup than a single bug/change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Reject: both are typing-themed and overlap on a few files, but they target different concrete changes. #45415 is a broad mypy/type-checking pass across many modules; #45425 adds specific typing protocols/aliases for modeling-utils integration. They could be merged independently and are not the same underlying fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 22, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45487-2", + "nodes": [ + "pull_request:45487", + "pull_request:45615" + ], + "soft_pairs": [ + "pull_request:45487|pull_request:45615" + ], + "prepared_review_unit_hash": "31e2ed1a51e054fee5ec2c579256755731d2d38862818f9a3750a9dca92daebc", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5159, + "estimated_input_tokens": 1290, + "estimated_eval_tokens": 2836 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5159, + "estimated_input_tokens": 1290, + "estimated_eval_tokens": 2836 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two PRs are unrelated: one fixes a model-parallel/device placement bug in AltCLIP/ChineseCLIP and similar text models, while the other only updates Qianfan OCR XPU test expectations. They do not appear to be the same underlying change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45487 is the more canonical PR because it addresses a concrete runtime bug in model code across affected models, rather than only adjusting test expectations.", + "best_issue_reason": null, + "best_pr_reason": "PR #45487 is the best representative of the cluster since it is the substantive code fix and has broader impact; PR #45615 is a narrow test-only update for a different model.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45487", + "right": "pull_request:45615", + "accept": false, + "reason": "Different models and different kinds of changes: #45487 fixes device placement in model forward code, while #45615 only adds XPU test outputs for Qianfan OCR. They do not look like the same underlying bug or mergeable change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 23, + "component_count": 26, + "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": "aef7968e10581de1f1f998011ea9ddd976c4762a3f1cbd67ddfdd464dc4b398d", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs concern expert parallelism, but they fix different failure modes: one is about DeepSpeed ZeRO-3 loading/serialization via accelerate, the other about FSDP2 rank-0 broadcast overwriting experts. They are related but not duplicates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45662 is the better representative of the cluster because it targets a concrete EP correctness bug in distributed loading/parameter materialization and includes the broader fix path across loading and MoE integration.", + "best_issue_reason": null, + "best_pr_reason": "PR #45662 is the strongest standalone change: it addresses a specific, reproducible corruption path for experts under FSDP2 and has the more substantive code-path fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Same general EP area, but different bugs and code paths: ZeRO-3/accelerate config handling vs FSDP2 rank-0 broadcast overwriting experts. They would not plausibly be merged as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 24, + "component_count": 26, + "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": "b8daa5471c31ab4185860429ad22fea64323e30161e3640608cdb3a1ddb82f79", + "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/captured_info logging collision around issue #45561, with 45645 effectively subsuming 45639 and adding CI-side log collection changes.", + "confidence": 0.87, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45645 is the better canonical representative because it contains the core xdist-safe artifact fix from 45639 and extends it to the CI workflow/notification path that consumes those logs.", + "best_issue_reason": null, + "best_pr_reason": "PR 45645 is the strongest overall fit for the cluster: it resolves the underlying captured_info collision and also updates the log aggregation path so the fix is complete end-to-end.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45639", + "right": "pull_request:45645", + "accept": true, + "reason": "Same underlying bug: xdist-safe captured_info handling for patched testing logs. 45645 includes the exact testing_utils change from 45639 plus related CI consumption updates, so the two are mergeable as one change set." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 25, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45683-2", + "nodes": [ + "pull_request:45683", + "pull_request:45734" + ], + "soft_pairs": [ + "pull_request:45683|pull_request:45734" + ], + "prepared_review_unit_hash": "eba4ca0f26b2d3be0c00e316270a9341d31af2208b44dbab31b1d3a8e81e7538", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4469, + "estimated_input_tokens": 1118, + "estimated_eval_tokens": 2492 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4469, + "estimated_input_tokens": 1118, + "estimated_eval_tokens": 2492 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs touch the same quantization helper, but they fix different bugs: one special-cases audio submodules, the other changes how default skip lists are combined with user-supplied skips. They should not be clustered as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "45734 is the better representative PR because it fixes the core skip-module composition logic in `get_modules_to_not_convert`, which is broader and more central than the narrow audio-module special case in 45683.", + "best_issue_reason": null, + "best_pr_reason": "45734 is the strongest standalone PR in the set: it addresses a concrete quantization failure mode in the main code path and its fix is directly about the helper\u2019s behavior rather than a model-specific workaround.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45683", + "right": "pull_request:45734", + "accept": false, + "reason": "Both touch quantization skip logic, but 45683 adds audio-tower/embed_audio exclusions while 45734 changes additive handling of `skip_modules`; these are different underlying bugs and would not naturally 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": 26, + "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 + } + ] +} diff --git a/snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/manifest.json b/snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..18ce4af5d0c8ed73852717b0ae974ae2c596716e --- /dev/null +++ b/snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260506t060033z", + "artifacts": { + "hybrid": "snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-05-06T06:08:43Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260506T060033Z", + "variant": "hybrid" +} diff --git a/snapshots/20260506T060033Z/manifest.json b/snapshots/20260506T060033Z/manifest.json index cbdc3c739980151ad23064fb2f4920ae997ba247..be13b9bb57d4e03d532acace38443f5c49c13663 100644 --- a/snapshots/20260506T060033Z/manifest.json +++ b/snapshots/20260506T060033Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 217 }, "extracted_at": "2026-05-06T06:00:33Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260506t060033z", + "runs": { + "hybrid-model-20260506t060033z": { + "analysis_id": "hybrid-model-20260506t060033z", + "artifacts": { + "hybrid": "snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260506T060033Z/analysis-runs/hybrid-model-20260506t060033z/manifest.json", + "model": null, + "published_at": "2026-05-06T06:08:43Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260506T060033Z", "watermark": {