diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index 895abac8c5f2cbaa8cbc23335771ba5ded2c5197..5150ad2b1d6288832298807e05130020f605cd48 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": "20260513T000101Z", - "generated_at": "2026-05-13T00:09:42Z", + "snapshot_id": "20260513T060025Z", + "generated_at": "2026-05-13T06:15:24Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ @@ -1148,6 +1148,204 @@ } ] }, + { + "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-45812-4", "summary": "Cluster of 2 issues and 2 PRs centered on issue #45812.", @@ -1473,76 +1671,6 @@ } ] }, - { - "cluster_id": "cluster-43824-3", - "summary": "Cluster of 1 issues and 2 PRs centered on issue #43824.", - "status": "open", - "confidence": 0.75, - "canonical_issue_number": 43824, - "canonical_pr_number": 43836, - "issue_numbers": [ - 43824 - ], - "pr_numbers": [ - 43836, - 43842 - ], - "evidence_types": [ - "closing_reference", - "shared_issue_target" - ], - "pr_comparisons": [ - { - "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-45081-3", "summary": "Cluster of 1 issues and 2 PRs centered on issue #45081.", @@ -1754,11 +1882,11 @@ { "left_pr_number": 45861, "right_pr_number": 45903, - "code_similarity": 0.617, - "size_similarity": 0.516, + "code_similarity": 0.506, + "size_similarity": 0.32, "file_overlap": 1.0, - "area_overlap": 0.377, - "patch_similarity": 0.546, + "area_overlap": 0.243, + "patch_similarity": 0.379, "shared_filenames": [ "src/transformers/models/qwen3_5/modeling_qwen3_5.py", "src/transformers/models/qwen3_5_moe/modeling_qwen3_5_moe.py", @@ -1779,12 +1907,16 @@ ], "right_ranges": [ [ - 146, - 154 + 106, + 114 ], [ - 161, - 183 + 145, + 158 + ], + [ + 165, + 208 ] ] }, @@ -1798,12 +1930,16 @@ ], "right_ranges": [ [ - 142, - 150 + 102, + 110 + ], + [ + 141, + 154 ], [ - 157, - 179 + 161, + 204 ] ] }, @@ -1817,12 +1953,16 @@ ], "right_ranges": [ [ - 1315, - 1323 + 1277, + 1285 ], [ - 1330, - 1352 + 1314, + 1327 + ], + [ + 1334, + 1377 ] ] }, @@ -1836,12 +1976,16 @@ ], "right_ranges": [ [ - 354, - 362 + 316, + 324 + ], + [ + 353, + 366 ], [ - 369, - 391 + 373, + 416 ] ] }, @@ -1859,12 +2003,12 @@ 38 ], [ - 270, - 292 + 269, + 315 ], [ - 311, - 319 + 331, + 344 ] ] }, @@ -1878,12 +2022,16 @@ ], "right_ranges": [ [ - 865, - 873 + 827, + 835 ], [ - 880, - 902 + 864, + 877 + ], + [ + 884, + 927 ] ] }, @@ -1918,11 +2066,11 @@ ], [ 45, - 79 + 78 ], [ - 523, - 582 + 522, + 607 ] ] } @@ -2244,6 +2392,16 @@ "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, @@ -2253,15 +2411,6 @@ "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-43824-3", - "canonical_pr_number": 43836, - "duplicate_pr_numbers": [ - 43842 - ], - "target_issue_number": 43824, - "reason": "PRs in cluster-43824-3 are treated as duplicates because they converge on issue #43824 with closing_reference, shared_issue_target evidence." - }, { "cluster_id": "cluster-43979-11", "canonical_pr_number": 44007, diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index 45f9a0f2cfbe01984efb8959e5d97b9c16360fe2..98a260e3021a6081f67aebc3b6977b6c038e2e24 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": "20260513T000101Z", - "generated_at": "2026-05-13T00:09:42Z", + "snapshot_id": "20260513T060025Z", + "generated_at": "2026-05-13T06:15:24Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -32,7 +32,6 @@ "issue:42915", "issue:43039", "issue:43425", - "issue:43531", "issue:43541", "issue:43596", "issue:43638", @@ -54,7 +53,6 @@ "issue:44841", "issue:44863", "issue:44918", - "issue:45070", "issue:45081", "issue:45161", "issue:45237", @@ -63,8 +61,10 @@ "issue:45685", "issue:45812", "issue:45823", + "issue:45859", "issue:45874", "issue:45901", + "issue:45902", "issue:45920" ], "soft_pairs": [ @@ -92,6 +92,7 @@ "issue:44292|issue:44457", "issue:34928|issue:44805", "issue:24643|issue:30277", + "issue:45859|issue:45902", "issue:43638|issue:45161", "issue:34928|issue:36331", "issue:43596|issue:45161", @@ -128,152 +129,150 @@ "issue:43866|issue:44863", "issue:43854|issue:43866", "issue:41093|issue:43638", - "issue:43792|issue:45901", - "issue:39290|issue:43531", - "issue:43425|issue:45070" + "issue:43792|issue:45901" ], - "prepared_review_unit_hash": "d8929537911f6fb7669770810bebaef505f55d787bec02412d6ed1a18de07460", + "prepared_review_unit_hash": "a2f3104273f2a6b93ba3d9fc44eeaf994fecaf3c740760c999a2eb16e1520e75", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 63, - "serialized_chars": 46570, - "estimated_input_tokens": 11643, - "estimated_eval_tokens": 23542 + "soft_pair_count": 62, + "serialized_chars": 46367, + "estimated_input_tokens": 11592, + "estimated_eval_tokens": 23440 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of unrelated bug reports. The only clear duplicate-like families are the DeepSpeed ZeRO-3 BERT index-out-of-bounds reports and the tokenizer v5 regression affecting Granite/other models. No PRs are present.", + "summary": "This cluster contains a few real duplicate families amid lots of unrelated regressions: the broad v5 tokenizer wrong-ID reports and the ZeRO-3 zero-size index-out-of-bounds reports. The mask/tensor shape mismatch issue 41093 should not be grouped with the ZeRO-3 load failures; it is a different failure mode.", "confidence": 0.71, - "canonical_issue_reason": "issue:45920 is the broadest and most representative issue here: it frames the regression as a multi-model AutoTokenizer token-ID bug in v5, not just a single-model symptom.", + "canonical_issue_reason": "Issue 45920 is the best canonical issue because it is the broadest report in the clearest duplicate family: a v5 AutoTokenizer regression affecting multiple model families, with 45812 as a narrower Granite-specific instance.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45920 is the best global anchor because it is open, explicit about the regression, and covers multiple affected model families, making it the most useful duplicate target.", + "best_issue_reason": "45920 is the best global issue candidate because it is the most general, user-visible regression in the cluster and spans the widest set of affected models.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44322", "right": "issue:45464", "accept": false, - "reason": "Different failure modes: config attribute missing vs streaming chat/completions inference error." + "reason": "Different Qwen3.5 failures and code paths: missing config attribute vs streaming chat/completions inference error." }, { "left": "issue:45237", "right": "issue:45685", "accept": false, - "reason": "Different hardware/backend problems: AMD GPU failure vs MPS histogram op not implemented for int tensors." + "reason": "Different backends and bugs: AMD GPU failure vs MPS histogram op not implemented for int tensors." }, { "left": "issue:43716", "right": "issue:45237", "accept": false, - "reason": "Different models and symptoms: image preprocessor dtype mismatch vs GPT-OSS failing on AMD GPUs." + "reason": "Different model families and symptoms: image-preprocessor dtype mismatch vs GPT-OSS on AMD GPUs." }, { "left": "issue:43828", "right": "issue:45237", "accept": false, - "reason": "Autocast dtype mismatch on Phi MoE is a different code path than AMD GPU compatibility." + "reason": "Both are dtype/device issues, but on different models and different failure points; not the same bug." }, { "left": "issue:45237", "right": "issue:45507", "accept": false, - "reason": "These describe unrelated issues: AMD GPU support vs an invalid method call in GraniteMoEHybrid." + "reason": "Different model-specific issues: AMD GPU execution failure vs GraniteMoEHybrid calling an invalid method." }, { "left": "issue:44292", "right": "issue:45237", "accept": false, - "reason": "NVFP4 runtime error on Qwen-3-8B is not the same as AMD GPU compatibility failure." + "reason": "Different hardware/model problems with no shared concrete code path." }, { "left": "issue:30277", "right": "issue:35545", "accept": false, - "reason": "DeepSpeed collective mismatch on Jamba-v01 and ModernBERT ONNX export are unrelated problems." + "reason": "DeepSpeed collective mismatch is unrelated to ModernBERT ONNX export." }, { "left": "issue:35545", "right": "issue:42915", "accept": false, - "reason": "ONNX export failure and FineGrainedFP8 model failure are different bugs." + "reason": "ONNX export failure and Qwen3Moe FP8 failure are different subsystems and symptoms." }, { "left": "issue:35545", "right": "issue:41720", "accept": false, - "reason": "ModernBERT ONNX export and Qwen3 auto device-map assert are unrelated." + "reason": "ModernBERT export and Qwen3 auto-device-map cuda assert are unrelated issues." }, { "left": "issue:44509", "right": "issue:44512", "accept": false, - "reason": "Both are docs regressions, but they target different stale references and are not clearly the same underlying change." + "reason": "Both are docs regressions in v5, but they target different removed features and likely different pages/edits." }, { "left": "issue:35545", "right": "issue:36010", "accept": false, - "reason": "ONNX export failure is unrelated to the GenerationMixin import error." + "reason": "Different problems: export-time failure vs import-time GenerationMixin breakage." }, { "left": "issue:44805", "right": "issue:45823", "accept": false, - "reason": "Mask/index shape mismatch is not the same as Gemma4 PLE device placement mismatch." + "reason": "Mask/tensor shape IndexError is unrelated to Gemma4 device mismatch during forward." }, { "left": "issue:43638", "right": "issue:44805", "accept": false, - "reason": "Both are index/shape errors, but they involve different model/setup paths and are not clearly the same bug." + "reason": "Both are IndexErrors, but one is ZeRO-3 zero-size indexing and the other is a mask shape mismatch; not the same bug." }, { "left": "issue:43596", "right": "issue:44805", "accept": false, - "reason": "ZeRO-3 BERT empty-dimension indexing and mask-shape mismatch are different failures." + "reason": "Different failure modes: ZeRO-3 zero-size indexing vs mask shape mismatch." }, { "left": "issue:44805", "right": "issue:45161", "accept": false, - "reason": "Shape mismatch during indexing is unrelated to GPT-OSS tensor-parallel behavior." + "reason": "Shape-mismatch IndexError is unrelated to GPT-OSS tensor-parallel behavior." }, { "left": "issue:44805", "right": "issue:44918", "accept": false, - "reason": "These are different errors in different training paths: mask mismatch vs embedding unpacking failure." + "reason": "Different code paths: mask indexing error vs TRL SFT/input-embedding unpacking failure." }, { "left": "issue:34634", "right": "issue:35545", "accept": false, - "reason": "BarkProcessor voice_preset handling and ModernBERT ONNX export are unrelated." + "reason": "BarkProcessor voice preset handling is unrelated to ONNX export." }, { "left": "issue:30277", "right": "issue:43638", "accept": false, - "reason": "Both mention DeepSpeed, but the collective mismatch and Bert zero-size index error are different code paths." + "reason": "Both involve DeepSpeed, but one is a collective mismatch and the other is a ZeRO-3 zero-size indexing bug." }, { "left": "issue:43596", "right": "issue:43638", "accept": true, - "reason": "Both report the same ZeRO-3/BERT index-0-out-of-bounds failure, just phrased with slightly different model descriptions." + "reason": "Same ZeRO-3 load-time index 0 is out of bounds for dimension 0 with size 0 failure; model names differ, but the bug appears identical." }, { "left": "issue:43039", @@ -285,202 +284,202 @@ "left": "issue:43975", "right": "issue:44457", "accept": false, - "reason": "Detokenization regression and LoRA merge/reload output mismatch are not the same bug." + "reason": "Incorrect detokenization and LoRA merge/save/reload inconsistency are separate bugs." }, { "left": "issue:44292", "right": "issue:44457", "accept": false, - "reason": "Qwen-3-8B NVFP4 runtime error is unrelated to LoRA save/load inconsistency." + "reason": "Qwen NVFP4 runtime error is unrelated to LoRA merge/save behavior." }, { "left": "issue:34928", "right": "issue:44805", "accept": false, - "reason": "Both are shape-related, but activation-checkpointing/FSDP recomputation and mask indexing are different failures." + "reason": "FSDP activation-checkpointing size mismatch is a different bug from the mask shape IndexError." }, { "left": "issue:24643", "right": "issue:30277", "accept": false, - "reason": "DeepSpeed weight-dimension error and Jamba collective mismatch are different training failures." + "reason": "Different DeepSpeed failures: weight must be 2-D vs collective mismatch on ranks." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": false, + "reason": "Same component family, but different concrete bugs: CPU offload incompatibility vs uninitialized-memory read after to_empty(cuda)." }, { "left": "issue:43638", "right": "issue:45161", "accept": false, - "reason": "BERT ZeRO-3 empty-index failure and GPT-OSS TP issue are unrelated." + "reason": "BERT ZeRO-3 zero-size indexing is unrelated to GPT-OSS tensor-parallel support." }, { "left": "issue:34928", "right": "issue:36331", "accept": false, - "reason": "FSDP/activation-checkpointing recomputation mismatch is unrelated to the CustomTrainer loss-signature change." + "reason": "FSDP recomputed tensor size mismatch and compute_loss() signature breakage are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "ZeRO-3 BERT loading failure is unrelated to GPT-OSS tensor-parallel behavior." }, { "left": "issue:43716", "right": "issue:45507", "accept": false, - "reason": "Image preprocessor dtype mismatch and invalid-method-call in GraniteMoEHybrid are different problems." + "reason": "Different model-specific failures with no shared code path." }, { "left": "issue:39290", "right": "issue:45081", "accept": false, - "reason": "Missing sliding_window_pattern in Gemma3TextConfig is unrelated to the Mistral tokenizer regex patch crash." + "reason": "Gemma3 config missing attribute in vLLM and Mistral tokenizer regex crash are unrelated." }, { - "left": "issue:45823", - "right": "issue:45874", + "left": "issue:43828", + "right": "issue:44292", "accept": false, - "reason": "Device mismatch in Gemma4 forward and expensive reverse embedding lookup are different symptoms and fixes." + "reason": "Different model/runtime combinations and different dtype-related failures." }, { - "left": "issue:43828", - "right": "issue:45161", + "left": "issue:43039", + "right": "issue:44918", "accept": false, - "reason": "Autocast dtype mismatch and GPT-OSS TP failure are unrelated." + "reason": "Cross-entropy dispatch under Liger Kernel and TRL SFT embedding unpacking are different bugs." }, { - "left": "issue:43425", - "right": "issue:44292", - "accept": false, - "reason": "Torch version incompatibility is not the same as the Qwen-3-8B NVFP4 runtime error." + "left": "issue:41762", + "right": "issue:43638", + "accept": true, + "reason": "Same ZeRO-3 load-time index 0 is out of bounds for dimension 0 with size 0 error; only the model wrapper differs." }, { - "left": "issue:34928", - "right": "issue:41720", + "left": "issue:43827", + "right": "issue:44512", "accept": false, - "reason": "FSDP activation-checkpointing recomputation mismatch is unrelated to Qwen3 auto device-mapping asserts." + "reason": "Both are v5 docs cleanups, but they remove references to different features and are not clearly the same change." }, { - "left": "issue:44918", - "right": "issue:45161", + "left": "issue:43716", + "right": "issue:44560", "accept": false, - "reason": "TRL SFT embedding unpacking failure and GPT-OSS TP issue are different bugs." + "reason": "Image-preprocessor dtype mismatch and Qwen3-vl video StopIteration are different issues." }, { - "left": "issue:43716", + "left": "issue:43828", "right": "issue:45685", "accept": false, - "reason": "Different backend/model issues: Mistral dtype mismatch vs MPS histogram op failure." + "reason": "Autocast dtype mismatch is unrelated to the MPS histogram kernel error." }, { - "left": "issue:44292", - "right": "issue:44918", + "left": "issue:44918", + "right": "issue:45685", "accept": false, - "reason": "These are distinct model/training failures with no shared concrete code path." + "reason": "TRL SFT embedding unpacking and MPS int-histogram failure are unrelated." }, { "left": "issue:43039", "right": "issue:43828", "accept": false, - "reason": "Both involve tensor dtype behavior, but they are different model paths and not the same defect." + "reason": "Different models and different failure points, despite both involving loss/dtype behavior." }, { "left": "issue:30277", "right": "issue:36331", "accept": false, - "reason": "DeepSpeed collective mismatch and compute_loss signature breakage are unrelated." + "reason": "DeepSpeed collective mismatch and trainer compute_loss signature mismatch are unrelated." }, { "left": "issue:34928", "right": "issue:36010", "accept": false, - "reason": "FSDP checkpointing mismatch is unrelated to the GenerationMixin import error." + "reason": "FSDP checkpointing mismatch and GenerationMixin import failure are different subsystems." }, { "left": "issue:43039", "right": "issue:44805", "accept": false, - "reason": "Liger Kernel cross-entropy dispatch and mask-shape indexing errors are different bugs." + "reason": "Liger Kernel cross-entropy issue is unrelated to the mask shape mismatch IndexError." }, { "left": "issue:43901", "right": "issue:44512", "accept": false, - "reason": "These are both docs-related, but they concern different removed/changed APIs and are not clearly the same issue." + "reason": "Docs mentioning return_all_scores and docs mentioning the removed run command are separate documentation issues." }, { "left": "issue:41762", "right": "issue:43596", - "accept": false, - "reason": "Both are DeepSpeed ZeRO-3 indexing problems, but they affect different model families and symptoms." + "accept": true, + "reason": "Same ZeRO-3 zero-size index-out-of-bounds load failure; only the model wrapper differs." }, { "left": "issue:39290", "right": "issue:41720", "accept": false, - "reason": "Sliding-window config issues and device-map cudaErrorAssert are different failure modes." + "reason": "Gemma3 config missing attribute in vLLM and Qwen3 auto-device-map cuda assert are different bugs." }, { "left": "issue:45237", "right": "issue:45901", "accept": false, - "reason": "GPT-OSS AMD GPU failure and table-question-answering crash are unrelated." + "reason": "GPT-OSS AMD GPU failure is unrelated to the table-question-answering crash." }, { "left": "issue:44841", "right": "issue:45823", "accept": false, - "reason": "Processor failure for Voxtral and Gemma4 device mismatch are unrelated." + "reason": "Voxtral processor failure and Gemma4 device mismatch are unrelated." }, { "left": "issue:43039", "right": "issue:44560", "accept": false, - "reason": "Liger Kernel cross-entropy issue and Qwen3-VL video StopIteration are different code paths." + "reason": "Liger Kernel cross-entropy issue and Qwen3-vl video StopIteration are unrelated." }, { "left": "issue:45812", "right": "issue:45920", "accept": true, - "reason": "Both report the same v5 AutoTokenizer wrong-token-ID regression; 45920 is the broader multi-model version." + "reason": "Same tokenizer regression family: AutoTokenizer returns wrong token IDs across multiple model families; 45920 is the broader report and 45812 is a narrower Granite-specific instance." }, { "left": "issue:43866", "right": "issue:44863", "accept": false, - "reason": "Checkpoint corruption and NemotronH checkpoint loading failure are distinct model-loading problems." + "reason": "Checkpoint corruption and checkpoint loading implementation failure are not the same bug." }, { "left": "issue:43854", "right": "issue:43866", "accept": false, - "reason": "Both are model-loading issues, but one is unit-test loading of GLM-4.7-Flash and the other is Ovis2 checkpoint corruption." + "reason": "Unit-test model loading failure and corrupted checkpoint are different root causes." }, { "left": "issue:41093", "right": "issue:43638", "accept": false, - "reason": "Mask/index shape mismatch and ZeRO-3 BERT empty-index errors are different bugs." + "reason": "The mask/tensor shape mismatch in 41093 is a different failure mode from the ZeRO-3 zero-size index-out-of-bounds load error in 43638." }, { "left": "issue:43792", "right": "issue:45901", "accept": false, - "reason": "Whisper runtime failure and table QA crash are unrelated." - }, - { - "left": "issue:39290", - "right": "issue:43531", - "accept": false, - "reason": "These both mention sliding window, but they involve different model/config regressions and no shared concrete fix is evident." - }, - { - "left": "issue:43425", - "right": "issue:45070", - "accept": false, - "reason": "Torch 2.10 incompatibility and PretrainedConfig pydantic-field breakage are unrelated." + "reason": "Whisper model load/run failure is unrelated to the table-question-answering crash." } ] }, "evaluator_result": { - "accept": true, - "feedback": "Grounded and conservative overall. The summary sticks to what the packet shows, the two accepted duplicate-like pairs are well supported by the issue titles, and the other soft-edge rejections are appropriately cautious rather than overstated. No PR pairs are present." + "accept": false, + "feedback": "Rejecting because one soft-edge verdict is too aggressive: issue:41093 is not grounded as a duplicate of the DeepSpeed ZeRO-3 zero-size indexing reports. Its title is a mask/tensor shape mismatch, while issue:43638 is a load-time `index 0 is out of bounds for dimension 0 with size 0` failure. Those look like different bugs, so the accepted duplicate link is not conservative enough. The tokenizer cluster acceptance (45812 vs 45920) and the ZeRO-3 pairings are plausible, but the overreach on 41093 means the overall output should not be accepted." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -508,6 +507,7 @@ "issue:43296", "issue:43366", "issue:43404", + "issue:43425", "issue:43531", "issue:43541", "issue:43572", @@ -521,9 +521,7 @@ "issue:43994", "issue:44291", "issue:44292", - "issue:44360", "issue:44387", - "issue:44512", "issue:44530", "issue:44589", "issue:44841", @@ -541,15 +539,16 @@ "issue:45632", "issue:45685", "issue:45698", + "issue:45759", "issue:45800", "issue:45823", - "issue:45859", "issue:45901", "issue:45902" ], "soft_pairs": [ + "issue:39290|issue:43531", + "issue:43425|issue:45070", "issue:43596|issue:45800", - "issue:45859|issue:45902", "issue:43975|issue:45901", "issue:45362|issue:45464", "issue:43792|issue:45070", @@ -572,9 +571,10 @@ "issue:43366|issue:45464", "issue:45020|issue:45800", "issue:43531|issue:45070", - "issue:41093|issue:43596", "issue:43541|issue:44589", + "issue:41093|issue:43596", "issue:43054|issue:43994", + "issue:45759|issue:45902", "issue:45542|issue:45800", "issue:40990|issue:44960", "issue:43716|issue:44918", @@ -582,263 +582,268 @@ "issue:45632|issue:45698", "issue:43404|issue:45005", "issue:43645|issue:45399", - "issue:44530|issue:45005", - "issue:44360|issue:44512" + "issue:44530|issue:45005" ], - "prepared_review_unit_hash": "17f32b47dc8c24854a58553b70a61217450fb31543784aee6b638028bb154d20", + "prepared_review_unit_hash": "fba4059aae310db279bfa674131424140ddba385b0d45a5017683d34a96867bd", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 36, - "serialized_chars": 38151, - "estimated_input_tokens": 9538, - "estimated_eval_tokens": 19332 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 37, + "serialized_chars": 37920, + "estimated_input_tokens": 9480, + "estimated_eval_tokens": 19216 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous set of unrelated regressions. The only plausible duplicate families are custom-module loading/cache collisions and Qwen3.5-MoE rotary-embedding device/materialization bugs; the rest are too far apart to merge.", - "confidence": 0.44, - "canonical_issue_reason": "Issue 45698 is the best representative of the strongest duplicate family: it is open, active, and specifically about wrong custom-module loading after save_pretrained, which aligns with the cache/path collision report in 45632.", + "summary": "The cluster is mostly heterogeneous transformer bug reports with no strong duplicate core. The only plausible thematic pair, 45632/45698, is still too weak to merge because the evidence is only soft similarity and the two reports are not shown to share the same concrete failure path.", + "confidence": 0.28, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 45698 is the most suitable cluster representative because it is current, concrete, and has the clearest duplicate-like counterpart; it also has more discussion activity than the other plausible family members.", + "best_issue_reason": "Issue 45020 is the broadest open umbrella for recent `remote_code` regressions, but this cluster remains mixed and does not support a stronger canonical duplicate anchor.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43596", - "right": "issue:45800", + "left": "issue:39290", + "right": "issue:43531", "accept": false, - "reason": "Both are runtime failures, but they have different root causes: Zero-3/BertModel loading vs torch/transformers version incompatibility." + "reason": "Both mention `sliding_window`, but they involve different model families and error paths." }, { - "left": "issue:45859", - "right": "issue:45902", - "accept": true, - "reason": "Same Qwen3_5MoeTextRotaryEmbedding device/materialization path; both point to broken handling of inv_freq/forward under offload or to_empty() materialization." + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Torch-version incompatibility and a `PretrainedConfig`/pydantic breakage are different failures." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "Both are runtime/version problems, but the concrete bugs are unrelated." }, { "left": "issue:43975", "right": "issue:45901", "accept": false, - "reason": "Different features and symptoms: detokenization regression vs table-question-answering crash." + "reason": "Detokenization regression and table-question-answering crash do not share a code path." }, { "left": "issue:45362", "right": "issue:45464", "accept": false, - "reason": "Same model family, but the failures are not the same concrete bug: transformers chat crash vs streaming chat/completions failure." + "reason": "Both touch Qwen3.5 chat, but one is a chat crash and the other is a streaming API failure." }, { "left": "issue:43792", "right": "issue:45070", "accept": false, - "reason": "Whisper load failure and a PretrainedConfig/pydantic breakage are unrelated." + "reason": "Whisper load failure is unrelated to the pydantic/config regression." }, { "left": "issue:44292", "right": "issue:45901", "accept": false, - "reason": "Qwen-3 NVFP4 runtime error and table-question-answering crash are unrelated code paths." + "reason": "Qwen NVFP4 runtime error and table QA crash are different model/task failures." }, { "left": "issue:43404", "right": "issue:43866", "accept": false, - "reason": "Mistral3 tied-weight bug and Ovis2 checkpoint corruption are different problems." + "reason": "`lm_head` tying and a checkpoint corruption report are not the same bug." }, { "left": "issue:39290", "right": "issue:45823", "accept": false, - "reason": "Both mention Gemma models, but one is a missing config attribute in vLLM and the other is a device mismatch in forward; not the same bug." + "reason": "Gemma3 missing-attribute and Gemma4 device-mismatch failures are different issues." }, { "left": "issue:45464", "right": "issue:45685", "accept": false, - "reason": "Qwen3.5 streaming inference failure and an MPS MoE histogram error are unrelated." + "reason": "Chat streaming failure and an MPS histogram/int error are unrelated." }, { "left": "issue:43541", "right": "issue:45685", "accept": false, - "reason": "Grouped_mm tracing failure and an MPS Int histogram error are different failures." + "reason": "Both are MoE-related, but the traced `grouped_mm` error and the MPS int histogram error are different." }, { "left": "issue:43541", "right": "issue:43828", "accept": false, - "reason": "Both involve dtype-related failures, but in different models and code paths; not one duplicate bug." + "reason": "Both are dtype-sensitive runtime errors, but they affect different models and execution paths." }, { "left": "issue:36010", "right": "issue:39290", "accept": false, - "reason": "ImportError for GenerationMixin and a Gemma3 config attribute error are unrelated." + "reason": "`GenerationMixin` import failure is unrelated to the Gemma3 `sliding_window` attribute issue." }, { "left": "issue:44841", "right": "issue:45084", "accept": false, - "reason": "Processor failure for Voxtral and a template-node compile error are unrelated." + "reason": "Processor failure and non-template-node compilation are different subsystems." }, { "left": "issue:41093", "right": "issue:41762", "accept": false, - "reason": "Both are IndexError-style reports, but one is a mask/tensor shape mismatch and the other is a Zero-3 load-time empty-index failure." + "reason": "Both are index errors, but the mask-shape mismatch and zero-sized tensor index-out-of-bounds are distinct bugs." }, { "left": "issue:45800", "right": "issue:45823", "accept": false, - "reason": "Version incompatibility and a Gemma4 device_map forward mismatch are not the same underlying bug." + "reason": "Torch/transformers incompatibility and a device placement mismatch are not the same defect." }, { "left": "issue:43296", "right": "issue:43366", "accept": false, - "reason": "PaddleOCR-VL loading in vLLM and GGUF gpt-oss support are unrelated." + "reason": "PaddleOCR-VL loading and GGUF gpt-oss support are unrelated." }, { "left": "issue:43572", "right": "issue:45070", "accept": false, - "reason": "Config-field absence and a pydantic PretrainedConfig regression are related only loosely by v5 changes, not the same bug." + "reason": "Missing `pad_token_idx` in `StableLmConfig` is separate from the pydantic `PretrainedConfig` regression." }, { "left": "issue:44387", "right": "issue:45005", "accept": false, - "reason": "CUDA reserved-memory/OOM under int4 quantization and tied-weights translation issues are distinct." + "reason": "Int4 memory/OOM behavior and tied-weights translation issues are different bugs." }, { "left": "issue:43541", "right": "issue:45161", "accept": false, - "reason": "Different MoE-related failures: tracing/grouped_mm vs tensor-parallel-only behavior on GPT-OSS." + "reason": "Grouped-`mm` tracing failure and GPT-OSS tensor-parallel-only failure are different MoE issues." }, { "left": "issue:43828", "right": "issue:45464", "accept": false, - "reason": "Different models and surfaces: autocast dtype mismatch vs streaming chat/completions failure." + "reason": "Autocast dtype mismatch and chat streaming failure are different code paths." }, { "left": "issue:44291", "right": "issue:45070", "accept": false, - "reason": "Unexpected _is_hf_initialized argument and PretrainedConfig pydantic field breakage are different init/config regressions." + "reason": "`init_empty_weights` TypeError is not the same as the `PretrainedConfig` pydantic break." }, { "left": "issue:43366", "right": "issue:45464", "accept": false, - "reason": "GGUF gpt-oss support and Qwen3.5 streaming inference failure are unrelated." + "reason": "GGUF model support and chat/completions streaming are unrelated." }, { "left": "issue:45020", "right": "issue:45800", "accept": false, - "reason": "Remote_code breakage and torch 2.4.1/transformers 5.8.0 incompatibility are too broad and not the same failure." + "reason": "`remote_code` regressions and torch/transformers incompatibility are different problems." }, { "left": "issue:43531", "right": "issue:45070", "accept": false, - "reason": "Sliding_window handling in Qwen3-MoE and PretrainedConfig pydantic breakage are different issues." + "reason": "Qwen3-MoE `sliding_window` behavior does not match the pydantic config regression." }, { - "left": "issue:41093", - "right": "issue:43596", + "left": "issue:43541", + "right": "issue:44589", "accept": false, - "reason": "Both are index errors, but one is a mask/tensor mismatch and the other is Zero-3/BertModel loading." + "reason": "Grouped-`mm` tracing failure and missing Float8 storage are different runtime bugs." }, { - "left": "issue:43541", - "right": "issue:44589", + "left": "issue:41093", + "right": "issue:43596", "accept": false, - "reason": "Grouped_mm tracing failure and Float8 storage lookup failure are unrelated." + "reason": "Both are zero-3-related index errors, but with different models, shapes, and failure mechanisms." }, { "left": "issue:43054", "right": "issue:43994", "accept": false, - "reason": "Both concern siglip2 results, but one reports weak embeddings while the other reports nonsensical AutoModel/pipeline output; not enough to \u092a\u0941\u0937\u094d\u091f\u093f same concrete bug." + "reason": "Both concern Siglip2 quality, but the symptoms differ and the data do not establish the same root cause." + }, + { + "left": "issue:45759", + "right": "issue:45902", + "accept": false, + "reason": "Composite config unwrapping and uninitialized rotary-embedding memory are different Qwen issues." }, { "left": "issue:45542", "right": "issue:45800", "accept": false, - "reason": "TensorBoard-without-TensorFlow backend error and a torch/transformers incompatibility are unrelated." + "reason": "TensorBoard backend error is unrelated to the torch compatibility issue." }, { "left": "issue:40990", "right": "issue:44960", "accept": false, - "reason": "GPT-OSS perplexity regression and a generic GLM5 issue are not the same bug." + "reason": "High perplexity on gpt-oss and a GLM5 report are not the same bug." }, { "left": "issue:43716", "right": "issue:44918", "accept": false, - "reason": "Gemma3 image-text dtype mismatch and Qwen3.5 embedding unpacking failure are different model-specific regressions." + "reason": "Gemma dtype mismatch and Qwen3.5 embedding unpacking are different model behaviors." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "FSDP weight dimensionality error and Qwen3.5 embedding unpacking failure are unrelated." + "reason": "FSDP weight-dimension error and Qwen3.5 embedding unpacking are unrelated." }, { "left": "issue:45632", "right": "issue:45698", - "accept": true, - "reason": "Both describe the same custom-module loading/caching collision class: local model paths or save_pretrained/from_pretrained cause the wrong remote code module to be resolved." + "accept": false, + "reason": "The reports are thematically similar, but the packet only supports soft similarity; the shared custom-code loading theme is not enough to confirm the same concrete bug." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Tied lm_head in Mistral3 and translation-model tied-weights issues are different model families and failure modes." + "reason": "Tied-weights issues are related at a high level, but these are not the same concrete defect." }, { "left": "issue:43645", "right": "issue:45399", "accept": false, - "reason": "Jupyter custom-model initialization breakage and flash-attn fallback gating are unrelated." + "reason": "Jupyter custom-model initialization breakage and flash-attn fallback gating are different issues." }, { "left": "issue:44530", "right": "issue:45005", "accept": false, - "reason": "PagedAttentionCache linear_attention failure and translation-model tied-weight issues are unrelated." - }, - { - "left": "issue:44360", - "right": "issue:44512", - "accept": false, - "reason": "DSA indexer discussion and removed run-command docs are unrelated." + "reason": "`linear_attention` crash and tied-weights translation issues are unrelated." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The assessment is grounded in the packet and stays conservative overall. The two accepted soft pairs are the strongest same-bug candidates, and the rejected pairs are mostly defensible separations based on different model families, surfaces, or failure modes. The canonical choice of issue 45698 is also supported by the clearer duplicate linkage to 45632 and its higher discussion activity." + "accept": false, + "feedback": "Reject: the response overstates evidence on the only accepted soft edge. In the packet, issue:45632 vs issue:45698 is marked deterministic_accept=false and supported only by soft_similarity, so treating it as a confirmed same-bug pair is too strong. The canonical_issue_reason also goes beyond the packet by claiming an explicit cache/path-collision root cause and a duplicate-anchor relationship that are not established by the provided data. The rest of the soft-edge rejections are conservative, but this one accepted edge is not." }, "evaluator_used": true, - "retried": false, - "accepted_nontrivial_soft_edge": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -861,7 +866,6 @@ "issue:41553", "issue:41628", "issue:41762", - "issue:42491", "issue:42915", "issue:42947", "issue:43054", @@ -879,7 +883,7 @@ "issue:43950", "issue:43975", "issue:44292", - "issue:44368", + "issue:44360", "issue:44387", "issue:44403", "issue:44451", @@ -899,11 +903,14 @@ "issue:45399", "issue:45507", "issue:45520", + "issue:45663", "issue:45701", "issue:45800", - "issue:45859" + "issue:45859", + "issue:45902" ], "soft_pairs": [ + "issue:44360|issue:44512", "issue:42915|issue:45399", "issue:43873|issue:45005", "issue:44863|issue:45005", @@ -918,11 +925,13 @@ "issue:36331|issue:41762", "issue:43866|issue:44403", "issue:41628|issue:45800", + "issue:43950|issue:45902", "issue:43054|issue:45701", "issue:43872|issue:44589", "issue:41553|issue:43883", "issue:38175|issue:43054", "issue:44910|issue:45859", + "issue:45663|issue:45902", "issue:41553|issue:43881", "issue:42947|issue:44387", "issue:43054|issue:43493", @@ -934,175 +943,192 @@ "issue:42947|issue:43856", "issue:43782|issue:44960", "issue:43881|issue:43883", - "issue:41762|issue:42915", - "issue:42491|issue:44368" + "issue:41762|issue:42915" ], - "prepared_review_unit_hash": "607ac9e33741735aac3865a23516a528e233b29a2d8f1e4432816968fd875bdd", + "prepared_review_unit_hash": "fccb4f311e0d441938ca68053e799f9afd09637c3f918aba80561cf77144e63a", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 32, - "serialized_chars": 36367, - "estimated_input_tokens": 9092, - "estimated_eval_tokens": 18440 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37475, + "estimated_input_tokens": 9369, + "estimated_eval_tokens": 18994 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose collection of Transformers issues around model loading/regressions, but only one soft pair is a true duplicate: the non-persistent buffer corruption regression. The rest share broad themes or model families without the same concrete bug.", - "confidence": 0.44, - "canonical_issue_reason": "issue:43950 is the clearest representative of the duplicate-worthy regression: non-persistent buffers being corrupted on load in Transformers 5.x, which is also stated almost verbatim in issue:44534.", + "summary": "The cluster is still mostly heterogeneous. Issue 43950 remains the best canonical issue for the non-persistent-buffer corruption regression, and 44534 is a clear duplicate. Issue 45902 is related in area but not conservative enough to accept as a duplicate from the packet alone, so it should stay separate/uncertain.", + "confidence": 0.79, + "canonical_issue_reason": "43950 is the clearest umbrella report for the Transformers v5 non-persistent-buffer corruption regression (`register_buffer(persistent=False)` / junk on load).", "canonical_pr_reason": null, - "best_issue_reason": "issue:43950 best captures the only clearly duplicated underlying bug in this cluster and is the most specific, central description of the regression.", + "best_issue_reason": "43950 best represents the shared bug family because it states the regression broadly and precisely, while 44534 is a direct duplicate and 45902 is only a possible related manifestation.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "Unrelated topics: DSA indexer discussion versus docs/CLI removal." + }, { "left": "issue:42915", "right": "issue:45399", "accept": false, - "reason": "Both involve model loading/fallback edge cases, but one is a FineGrainedFP8Config failure and the other is flash-attn fallback gating; different bugs." + "reason": "Different failure modes: Qwen3Moe FP8 config versus flash-attn fallback gating." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantization/offloading behavior and tied-weights warnings are unrelated code paths." + "reason": "Quantization/offloading behavior is not the same bug as tied-weights handling in translation models." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "NemotronH checkpoint loading failure is not the same bug as tied-weights handling on translation models." + "reason": "NemotronH checkpoint loading is a different defect from tied-weights issues in v5." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "A cache-strategy typo and a docs reference to a removed command are different documentation issues." + "reason": "A typo in cache strategies is unrelated to the removed CLI/docs reference." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "TRL embedding unpacking failure and GraniteMoEHybrid invalid method call are unrelated." + "reason": "Different code paths and symptoms: TRL SFT embedding unpacking versus GraniteMoEHybrid method dispatch." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "Detokenization regression and Qwen-3 NVFP4 runtime failure are different problems." + "reason": "Different runtime failures: detokenization regression versus NVFP4 run failure." }, { "left": "issue:43950", "right": "issue:44534", "accept": true, - "reason": "Same underlying regression: Transformers 5.x corrupts non-persistent buffers during load; titles describe the same bug." + "reason": "Both report the same v5 regression where non-persistent buffers are corrupted/fill with junk on load." }, { "left": "issue:43644", "right": "issue:43950", "accept": false, - "reason": "trust_remote_code missing fields and non-persistent buffer corruption are different regressions." + "reason": "Remote-code config field loss is a different bug from non-persistent-buffer corruption." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "Both mention SigLIP2, but one is a broken docs example and the other is degraded text embeddings; not the same bug." + "reason": "Both mention SigLIP2, but one is a docs/example problem and the other is an embedding-quality complaint." }, { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to Qwen3.5 gradient explosion from 3D position_ids." + "reason": "Autocast dtype mismatch is not the same bug as Qwen3.5 position-id/SDPA instability." }, { "left": "issue:36331", "right": "issue:41762", "accept": false, - "reason": "CustomTrainer compute_loss signature mismatch and Gemma3 ZeRO-3 loading IndexError are unrelated." + "reason": "Trainer API kwarg mismatch is unrelated to Gemma3 ZeRO-3 loading failures." }, { "left": "issue:43866", "right": "issue:44403", "accept": false, - "reason": "Corrupted checkpoint and noisy load-time logging are not the same issue." + "reason": "Checkpoint corruption is not the same as generic loading noise." }, { "left": "issue:41628", "right": "issue:45800", "accept": false, - "reason": "AutoImageProcessor import failure is not the same as torch/Transformers version incompatibility." + "reason": "Import/export symbol issues and torch-version incompatibility are different problems." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": false, + "reason": "Related area, but 45902 describes uninitialized `inv_freq` after `meta -> to_empty(cuda)` materialization and is not clearly the same regression as the non-persistent-buffer junk-on-load bug." }, { "left": "issue:43054", "right": "issue:45701", "accept": false, - "reason": "SigLIP2 embedding quality and tokenization changes are different behaviors." + "reason": "Tokenization/model-behavior complaints, but the symptoms and likely root causes differ." }, { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "bitsandbytes _is_hf_initialized incompatibility and Float8 storage lookup failure are different quantization/runtime errors." + "reason": "Different quantization backend errors: Int8Params kwarg mismatch versus Float8 storage lookup failure." }, { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "Bad AutoTokenizer error for Voxtral and Molmo missing all_tied_weights_keys are unrelated." + "reason": "Bad AutoTokenizer error reporting and missing tied-weights metadata are unrelated." }, { "left": "issue:38175", "right": "issue:43054", "accept": false, - "reason": "Both are SigLIP2-related, but zero probabilities and worse text embeddings are not clearly the same bug." + "reason": "Both are SigLIP2-related, but zero probabilities and lower embedding quality are not clearly the same bug." }, { "left": "issue:44910", "right": "issue:45859", "accept": false, - "reason": "Flash Attention illegal memory access from 3D position_ids and CPU offload incompatibility in Qwen3_5MoeTextRotaryEmbedding are different failures." + "reason": "Flash-attn illegal memory access from 3D position_ids is not the same as CPU-offload incompatibility in rotary embedding." + }, + { + "left": "issue:45663", + "right": "issue:45902", + "accept": false, + "reason": "Gemma4 FSDP2 shared_kv_states KeyError is unrelated to buffer materialization corruption." }, { "left": "issue:41553", "right": "issue:43881", "accept": false, - "reason": "Voxtral tokenizer loading message and glm-4v-9b loading failure are unrelated." + "reason": "Bad AutoTokenizer error reporting and GLM4V loading failure are different issues." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "Gradient checkpointing ineffectiveness with LoRA and int4 reserved-memory OOM are different training/memory issues." + "reason": "Gradient checkpointing ineffectiveness and int4 memory bloat/OOM are distinct bugs." }, { "left": "issue:43054", "right": "issue:43493", "accept": false, - "reason": "Both concern SigLIP2, but one reports worse embeddings and the other an HF vs JAX discrepancy; not enough evidence they are the same concrete bug." + "reason": "SigLIP2 embedding quality complaint and HF-vs-JAX discrepancy are related in theme but not clearly the same bug." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "Generic loading noise and inability to load a specific model are different issues." + "reason": "Generic loading noise and a specific model loading failure are not the same underlying defect." }, { "left": "issue:43643", "right": "issue:45020", "accept": false, - "reason": "AutoConfig missing fields with trust_remote_code and broad remote_code breakage are related only at a high level, not the same bug." + "reason": "Missing fields from trust_remote_code config loading is a specific bug; 45020 is a broader umbrella report." }, { "left": "issue:43975", @@ -1114,46 +1140,46 @@ "left": "issue:43975", "right": "issue:45237", "accept": false, - "reason": "DeepSeek tokenization/detokenization and GPT-OSS AMD GPU support are unrelated." + "reason": "Detokenization regression and AMD GPU support/loading are different failures." }, { "left": "issue:44403", "right": "issue:44488", "accept": false, - "reason": "Load-time noise and a specific model load failure are not the same bug." + "reason": "Different model loading reports with no clear shared code-path bug." }, { "left": "issue:42947", "right": "issue:43856", "accept": false, - "reason": "LoRA checkpointing behavior and Qwen3 MoE memory inefficiency are different training issues." + "reason": "Both mention memory, but one is LoRA checkpointing behavior and the other is Qwen3 MoE training memory usage." }, { "left": "issue:43782", "right": "issue:44960", "accept": false, - "reason": "Qwen3VL weight_only load error and GLM5 are unrelated model-loading failures." + "reason": "Qwen3VL weight_only loading error is unrelated to the GLM5 report." }, { "left": "issue:43881", "right": "issue:43883", "accept": false, - "reason": "glm-4v-9b loading failure and Molmo tied-weights attribute error are unrelated." + "reason": "Loading failure for glm-4v-9b and missing all_tied_weights_keys are different defects." }, { "left": "issue:41762", "right": "issue:42915", "accept": false, - "reason": "Gemma3 ZeRO-3 index error and Qwen3Moe FineGrainedFP8Config failure are different model-loading problems." + "reason": "Gemma3 ZeRO-3 loading error is unrelated to Qwen3Moe FP8 configuration failure." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The cluster summary is grounded in the packet and stays conservative. The only accepted soft edge, issue:43950 vs issue:44534, is well supported by the nearly identical titles and shared non-persistent-buffer corruption regression. The other pairwise rejections read cautious rather than overstated, and no PR mergeability concerns apply here." + "accept": false, + "feedback": "The cluster summary is mostly grounded, but one soft-edge verdict is not conservative enough: treating issue 45902 as a duplicate of 43950 goes beyond what the packet supports from titles alone. 45902 describes uninitialized `inv_freq` memory after `meta -> to_empty(cuda)` materialization, which may be related, but it is not clearly the same regression as the non-persistent-buffer junk-on-load bug in 43950/44534. The 44534 link is well supported; the 45902 link should be downgraded to uncertain rather than accepted." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -1176,6 +1202,7 @@ "issue:36331", "issue:41093", "issue:41553", + "issue:42491", "issue:42915", "issue:42947", "issue:43257", @@ -1194,6 +1221,7 @@ "issue:43950", "issue:44220", "issue:44291", + "issue:44368", "issue:44479", "issue:44488", "issue:44509", @@ -1208,18 +1236,18 @@ "issue:45081", "issue:45305", "issue:45313", + "issue:45325", "issue:45356", "issue:45399", "issue:45538", "issue:45588", - "issue:45663", "issue:45685", - "issue:45759", "issue:45823", "issue:45902", "issue:45923" ], "soft_pairs": [ + "issue:42491|issue:44368", "issue:45685|issue:45823", "issue:43824|issue:43872", "issue:45020|issue:45823", @@ -1228,7 +1256,6 @@ "issue:43828|issue:45823", "issue:43827|issue:44509", "issue:45081|issue:45399", - "issue:43950|issue:45902", "issue:45020|issue:45356", "issue:43404|issue:43950", "issue:43742|issue:45399", @@ -1243,7 +1270,6 @@ "issue:41553|issue:44991", "issue:42947|issue:45923", "issue:43257|issue:45305", - "issue:45759|issue:45902", "issue:36331|issue:41093", "issue:44928|issue:45588", "issue:41093|issue:44805", @@ -1251,251 +1277,245 @@ "issue:41553|issue:44488", "issue:44530|issue:44863", "issue:44479|issue:44560", + "issue:45325|issue:45902", "issue:24643|issue:30064", - "issue:43366|issue:43828", - "issue:45663|issue:45823" + "issue:43366|issue:43828" ], - "prepared_review_unit_hash": "cf6877b70d5d921d85f3ef02d73f8d14adb71573ccfb5c6f68d341c6cd4f2f30", + "prepared_review_unit_hash": "c236b628e099e50cf069dfc2fb32a015813810e1678d11bd06adc896faa53e66", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 34, - "serialized_chars": 37714, - "estimated_input_tokens": 9429, - "estimated_eval_tokens": 19114 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 38062, + "estimated_input_tokens": 9516, + "estimated_eval_tokens": 19288 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly distinct reports. Clear duplicates only for the docs regression, Whisper fbank extraction, the `_is_hf_initialized`/`init_empty_weights` incompatibility, the Qwen video-input regression, and the repeated mask-shape IndexError. The buffer/materialization pair `43950`/`45902` is too weak to merge and should stay separate.", - "confidence": 0.67, - "canonical_issue_reason": "If an umbrella issue is needed, `issue:45020` is the broadest open triage bucket for recent model-loading regressions, but this cluster is too mixed to treat as one duplicate set.", + "summary": "Mostly a heterogeneous set of recent Transformers bug reports with superficial title similarity; only a few pairs look like true duplicates (docs cleanup, mask-shape IndexError, and the `_is_hf_initialized` loading regression).", + "confidence": 0.69, + "canonical_issue_reason": "issue:45020 is the broadest umbrella-style report in the set and the closest thing to a cluster anchor, covering recent-version model-loading breakage for `remote_code` models.", "canonical_pr_reason": null, - "best_issue_reason": "`issue:45020` is the best global triage anchor because it is open and broad, though it should not be used to absorb the weaker buffer/materialization report pair.", + "best_issue_reason": "issue:45020 is the most generally representative issue here: broad, recent, and centered on model-loading regressions rather than a narrow model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "Both involve Qwen/LoRA wording, but one is a version-compatibility failure and the other is a warning about `tie_word_embeddings`; not the same bug." + }, { "left": "issue:45685", "right": "issue:45823", "accept": false, - "reason": "Different models and failure modes; only a high-level runtime/device similarity." + "reason": "Different failure modes on different paths: MPS `histogram_mps` int issue vs Gemma4 device placement mismatch." }, { "left": "issue:43824", "right": "issue:43872", "accept": false, - "reason": "Qwen2.5-VL import error vs bitsandbytes constructor regression; unrelated roots." + "reason": "Import error for a missing Qwen2.5-VL class vs bitsandbytes `Int8Params` constructor incompatibility; unrelated code paths." }, { "left": "issue:45020", "right": "issue:45823", "accept": false, - "reason": "Broad remote-code breakage is not the same concrete device-mismatch bug." + "reason": "`remote_code` loading regressions are too broad to equate with a specific Gemma4 forward device mismatch." }, { "left": "issue:44661", "right": "issue:45020", "accept": false, - "reason": "Registry helper failure is a different code path from remote-code loading regressions." + "reason": "Tokenizer-mapping failure in `add-new-model-like` is a distinct tooling bug, not the same as remote-code loading breakage." }, { "left": "issue:44991", "right": "issue:45399", "accept": false, - "reason": "Tokenizer loading failure and flash-attn fallback gating are different subsystems." + "reason": "Tokenizer loading failure for a specific model vs flash-attn fallback gating; no shared concrete failure path." }, { "left": "issue:43828", "right": "issue:45823", "accept": false, - "reason": "Autocast dtype mismatch is not the same bug as the device-mapping failure." + "reason": "Autocast dtype mismatch in Phi MoE and Gemma4 device mismatch are different runtime errors." }, { "left": "issue:43827", "right": "issue:44509", "accept": true, - "reason": "Same docs regression: removed v5 pipeline tasks are still referenced." + "reason": "Both report stale docs still referencing removed pipeline tasks after v5; same documentation cleanup issue." }, { "left": "issue:45081", "right": "issue:45399", "accept": false, - "reason": "Mistral tokenizer regex crash is unrelated to flash-attn fallback checks." - }, - { - "left": "issue:43950", - "right": "issue:45902", - "accept": false, - "reason": "Related materialization/buffer-safety themes, but different concrete failures and no evidence of one shared fix." + "reason": "Tokenizer regex crash vs flash-attn backend fallback logic; different subsystems and symptoms." }, { "left": "issue:45020", "right": "issue:45356", "accept": false, - "reason": "Remote-code breakage and tokenizer codec regression are different bugs." + "reason": "Generic `remote_code` regression vs a specific Kimi-K2.5 tokenizer/codec regression; not the same bug." }, { "left": "issue:43404", "right": "issue:43950", "accept": false, - "reason": "Tied lm_head bug is unrelated to non-persistent buffer corruption." + "reason": "One is a tied-weights bug in Mistral3, the other a `from_pretrained()` non-persistent-buffer regression; unrelated behavior." }, { "left": "issue:43742", "right": "issue:45399", "accept": false, - "reason": "Model key error and flash-attn fallback gating do not share a concrete code path." + "reason": "MobileLLM loading failure and flash-attn fallback blocking are not the same code-path." }, { "left": "issue:43257", "right": "issue:43866", "accept": false, - "reason": "Different model families with different loading failures." + "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed vs a corrupted Ovis2 checkpoint; different underlying problems." }, { "left": "issue:42915", "right": "issue:45020", "accept": false, - "reason": "FP8/Qwen3-MoE failure is not the same issue as remote-code model breakage." + "reason": "A specific LoRA incompatibility on hf4.x vs a broad `remote_code` loading regression; not the same concrete bug." }, { "left": "issue:43638", "right": "issue:45823", "accept": false, - "reason": "Zero3 index error with BERT is unrelated to Gemma4 device placement." + "reason": "DeepSpeed ZeRO-3 index error on a Bert model vs Gemma4 PLE device mismatch; unrelated." }, { "left": "issue:43792", "right": "issue:44220", - "accept": true, - "reason": "Same Whisper fbank extraction failure; the model cannot run because that path is broken." + "accept": false, + "reason": "Whisper not running is too vague to merge with a specific `_torch_extract_fbank_features()` report without stronger evidence." }, { "left": "issue:43872", "right": "issue:44291", "accept": true, - "reason": "Same `_is_hf_initialized` / `init_empty_weights` incompatibility." + "reason": "Both center on the unexpected `_is_hf_initialized` kwarg during model init/loading, indicating the same regression." }, { "left": "issue:43940", "right": "issue:45313", - "accept": true, - "reason": "Both report the same DeepSpeed ZeRO-3 weight-loading failure pattern on Qwen3-family models." + "accept": false, + "reason": "Both are ZeRO-3 weight-loading failures, but they target different model families and are not clearly the same concrete defect." }, { "left": "issue:42915", "right": "issue:43872", "accept": false, - "reason": "Qwen3-MoE FP8 failure is unrelated to the bitsandbytes regression." + "reason": "LoRA compatibility on version upgrades vs bitsandbytes `_is_hf_initialized` constructor breakage; different failures." }, { "left": "issue:43854", "right": "issue:44863", "accept": false, - "reason": "Different architectures and different model-loading problems." + "reason": "GLM-4.7 test loading issue and NemotronH checkpoint loading failure are separate model-specific bugs." }, { "left": "issue:41553", "right": "issue:44991", "accept": false, - "reason": "Bad Voxtral tokenizer error and est-roberta tokenizer failure are distinct issues." + "reason": "Tokenizer loading error for Voxtral vs tokenizer loading failure for EMBEDDIA/est-roberta; no shared concrete mechanism shown." }, { "left": "issue:42947", "right": "issue:45923", "accept": false, - "reason": "Gradient checkpointing not working with LoRA is not the same as a missing capability flag." + "reason": "Gradient checkpointing not effective with LoRA is different from a missing `supports_gradient_checkpointing` flag on a remote-code model." }, { "left": "issue:43257", "right": "issue:45305", "accept": false, - "reason": "DeepSpeed weight conversion bug and GAS gradient averaging bug are different failures." - }, - { - "left": "issue:45759", - "right": "issue:45902", - "accept": false, - "reason": "Composite-config unwrapping and buffer materialization are different code paths." + "reason": "Weight conversion failure under accelerate+deepspeed vs gradient averaging under GAS with DeepSpeed; unrelated training bugs." }, { "left": "issue:36331", "right": "issue:41093", "accept": false, - "reason": "Unexpected kwarg and mask-shape mismatch are different trainer/runtime errors." + "reason": "Unexpected `compute_loss` kwarg vs mask/tensor shape mismatch; different error classes and code paths." }, { "left": "issue:44928", "right": "issue:45588", "accept": false, - "reason": "RLHF NaN explosion and flash-attn integration AttributeError are unrelated." + "reason": "RLHF NaN explosion from Qwen3.5 position ids is not the same as flash_attention `s_aux=None` crashing." }, { "left": "issue:41093", "right": "issue:44805", "accept": true, - "reason": "Same indexed-tensor/mask shape mismatch IndexError, just with different dimensions." + "reason": "Same IndexError pattern about mask shape not matching indexed tensor shape; 44805 is a later generic reproduction of the same bug." }, { "left": "issue:43872", "right": "issue:45538", "accept": false, - "reason": "bitsandbytes regression is unrelated to CLIPTokenizer max-length handling." + "reason": "`_is_hf_initialized` loading regression and CLIPTokenizer max length constant issue are unrelated." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "Voxtral tokenizer error and sleng-bert loading failure are separate tokenizer/model-loading problems." + "reason": "Different tokenizer/model-loading failures for different models; not the same root cause." }, { "left": "issue:44530", "right": "issue:44863", "accept": false, - "reason": "Qwen3.5 cache crash and NemotronH checkpoint loading failure are unrelated." + "reason": "PagedAttentionCache linear_attention crash and NemotronH checkpoint loading failure are distinct model/runtime issues." }, { "left": "issue:44479", "right": "issue:44560", - "accept": true, - "reason": "Same Qwen3-VL video-input regression affecting the video path." + "accept": false, + "reason": "Both involve video/VL models, but one is a broad v5.3 regression and the other a specific `StopIteration` in a different embedding flow." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "RoPE/position-id scaling bug for Qwen2.5-VL vs uninitialized `inv_freq` after meta-to-empty materialization in Qwen3.5 MoE; different defects." }, { "left": "issue:24643", "right": "issue:30064", "accept": false, - "reason": "DeepSpeed training error and void-segmentation-map preprocessing are unrelated." + "reason": "DeepSpeed training `weight must be 2-D` and void segmentation map processing are unrelated issues." }, { "left": "issue:43366", "right": "issue:43828", "accept": false, - "reason": "GGUF support request and Phi MoE autocast dtype mismatch are unrelated." - }, - { - "left": "issue:45663", - "right": "issue:45823", - "accept": false, - "reason": "Gemma4 FSDP2 shared_kv_states bug is different from the device mismatch report." + "reason": "GGUF architecture support for gpt-oss is unrelated to Phi MoE autocast dtype mismatch." } ] }, "evaluator_result": { - "accept": false, - "feedback": "Most of the rejections are conservative and grounded in the packet, but at least one accepted soft pair overstates the evidence: issue:43950 vs issue:45902. From the titles alone, these look related to buffer/materialization behavior, but not clearly the same bug or a single mergeable PR without more evidence. The summary also calls them a clear duplicate too strongly. Because the review should stay conservative, I\u2019d reject this clustering as written." + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The selected anchor issue 45020 is a reasonable umbrella for broad remote-code/model-loading regressions, and the accepted soft pairs are supported by the issue titles and described failure modes. No soft-edge verdict appears overstated." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -1533,8 +1553,6 @@ "issue:43708", "issue:43720", "issue:43824", - "issue:43854", - "issue:43931", "issue:43950", "issue:43957", "issue:43994", @@ -1558,11 +1576,13 @@ "issue:45313", "issue:45357", "issue:45584", + "issue:45663", "issue:45685", "issue:45823", "issue:45834" ], "soft_pairs": [ + "issue:45663|issue:45823", "issue:44155|issue:45584", "issue:43366|issue:43531", "issue:43950|issue:45834", @@ -1592,229 +1612,222 @@ "issue:44451|issue:44488", "issue:41720|issue:42915", "issue:42491|issue:45276", - "issue:43854|issue:43931", "issue:43708|issue:44075" ], - "prepared_review_unit_hash": "9de674e634ca89c3d50bb5e417ed15336fe8a34050497880e56e965928d522a2", + "prepared_review_unit_hash": "9abd19eeba6a0164ec1d1bc13a5dc6ec8056bd1f54171f646b4713a33ead26d6", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 36838, - "estimated_input_tokens": 9210, - "estimated_eval_tokens": 18676 + "serialized_chars": 36322, + "estimated_input_tokens": 9081, + "estimated_eval_tokens": 18418 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a grab-bag of Transformers regressions and model-loading bugs rather than one duplicate cluster. A couple of pairs are near-identical, especially the Transformers 5 custom model init regressions and the non-persistent buffer corruption reports.", - "confidence": 0.87, - "canonical_issue_reason": "Issue 43646 is the broadest and most representative of the repeated Transformers 5 custom-model initialization regressions; it\u2019s the best umbrella-style issue in this set, though many other items are model-specific or unrelated.", + "summary": "Mostly unrelated issue set with a few clear duplicate pairs around the same concrete regression; the rest are broad theme matches but not the same bug.", + "confidence": 0.86, + "canonical_issue_reason": "issue:43644 is the clearest representative of the core v5 regression and has an exact duplicate (44534).", "canonical_pr_reason": null, - "best_issue_reason": "43646 has the cleanest generic title, a likely core reproduction path, and matches the main recurring theme in the cluster better than the more niche model-specific reports.", + "best_issue_reason": "It is a broad, concrete regression report with the cleanest reproduction and the strongest evidence of duplication.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Same model family, but different failures: FSDP2 KeyError in attention vs device_map='auto' forward-device mismatch." + }, { "left": "issue:44155", "right": "issue:45584", "accept": false, - "reason": "Different models and failure modes: AudioFlamingo3 track leakage vs Whisper empty-transcription generation." + "reason": "Different subsystems and symptoms: AudioFlamingo3 track leakage vs Whisper empty-transcription generation failure." }, { "left": "issue:43366", "right": "issue:43531", "accept": false, - "reason": "Different feature areas: GGUF support for gpt-oss vs Qwen3-MoE sliding_window behavior." + "reason": "Both involve Qwen-related models, but one is GGUF/gpt-oss support and the other is a sliding_window bug in Qwen3-MoE." }, { "left": "issue:43950", "right": "issue:45834", "accept": false, - "reason": "Unrelated bugs: non-persistent buffer corruption vs Kosmos2.5 long OCR index error." + "reason": "Different bugs: non-persistent buffer corruption vs an OCR long-input index error." }, { "left": "issue:43550", "right": "issue:44910", "accept": false, - "reason": "Different code paths and models: Bamba torch.compile/SDPA vs Qwen3.5 flash-attention with 3D position_ids." + "reason": "Different code paths: torch.compile with SDPA on Bamba vs Flash Attention misreading 3D position_ids on Qwen3.5." }, { "left": "issue:44560", "right": "issue:45685", "accept": false, - "reason": "Unrelated failures: Qwen3-VL video StopIteration vs MPS histogram op not implemented for MoE int tensors." + "reason": "Unrelated failures: Qwen3-VL video StopIteration vs an MPS histogram kernel not implemented for Int." }, { "left": "issue:43824", "right": "issue:43957", "accept": false, - "reason": "Both are loading issues, but one is a missing export and the other is meta-device loading breakage; not the same bug." + "reason": "ImportError for a missing class is not the same as meta-device loading breaking some models." }, { "left": "issue:43645", "right": "issue:43646", "accept": true, - "reason": "Very close duplicate regression: both report Transformers 5 breaking custom model initialization, with 43645 just adding the notebook context." + "reason": "These describe the same custom-model initialization regression; one is notebook-specific, the other is the general bug." }, { "left": "issue:43366", "right": "issue:44589", "accept": false, - "reason": "Different problems: GGUF architecture support vs float8 storage type lookup error." + "reason": "GGUF/gpt-oss support request and a Float8 storage lookup error are different issues." }, { "left": "issue:42915", "right": "issue:43957", "accept": false, - "reason": "Different subsystems and symptoms: Qwen3-MoE FP8 config failure vs meta-device loading issues." + "reason": "Qwen3-MoE FineGrainedFP8Config failure is unrelated to meta-device loading regressions." }, { "left": "issue:42617", "right": "issue:43366", "accept": false, - "reason": "3d_parallel.py runtime failure is unrelated to GGUF model architecture support." + "reason": "3d_parallel.py execution failure and GGUF gpt-oss support are different problems." }, { "left": "issue:44560", "right": "issue:44918", "accept": false, - "reason": "Both involve Qwen3-VL/Qwen3.5 multimodal issues, but the symptoms and paths differ: video StopIteration vs embedding unpacking in TRL SFT." + "reason": "Both touch Qwen3.5, but one is a video-generation StopIteration and the other is TRL SFT embedding unpacking." }, { "left": "issue:36010", "right": "issue:42915", "accept": false, - "reason": "Different APIs and failure modes: GenerationMixin import error vs Qwen3-MoE FP8 model loading." + "reason": "A missing GenerationMixin import is unrelated to a Qwen3-MoE FP8 runtime failure." }, { "left": "issue:44661", "right": "issue:45823", "accept": false, - "reason": "Tokenizer mapping registration bug vs Gemma4 PLE device mismatch during forward are unrelated." + "reason": "Tokenizer-mapping-name edge case and Gemma4 PLE device mismatch are different code-path problems." }, { "left": "issue:38175", "right": "issue:43994", - "accept": false, - "reason": "Same model family, but the reported symptoms are too different to confirm the same underlying bug from titles alone." + "accept": true, + "reason": "Same SigLIP2 model regression: unexpected zero probabilities / nonsensical outputs from AutoModel and pipeline." }, { "left": "issue:43632", "right": "issue:45834", "accept": false, - "reason": "_is_hf_initialized flag regression is unrelated to Kosmos2.5 long-input index errors." + "reason": "The _is_hf_initialized flag regression is unrelated to the long OCR index error." }, { "left": "issue:44560", "right": "issue:44805", "accept": false, - "reason": "Different error classes and models: video StopIteration vs tensor mask/index shape mismatch." + "reason": "Different Qwen3-VL issues: video StopIteration vs a generic mask/tensor shape mismatch." }, { "left": "issue:44877", "right": "issue:45030", "accept": false, - "reason": "Both mention config loading, but one is strict config gating and the other is a tiny-random GLM4V validation failure; not clearly the same bug." + "reason": "Both are config-loading failures, but they affect different model configs and likely different validation checks." }, { "left": "issue:43638", "right": "issue:44661", "accept": false, - "reason": "Different regression types: DeepSpeed ZeRO-3 Bert initialization vs tokenizer mapping model registration." + "reason": "DeepSpeed ZeRO-3 Bert init failure is not the same as add-new-model-like failing inside TOKENIZER_MAPPING_NAMES." }, { "left": "issue:43596", "right": "issue:44661", "accept": false, - "reason": "Both touch BERT/custom model loading, but the zero3 empty-tensor IndexError and tokenizer-mapping failure are different code paths." - }, - { - "left": "issue:44661", - "right": "issue:44805", - "accept": false, - "reason": "Tokenizer mapping registration failure is unrelated to the generic mask shape IndexError." + "reason": "Same as above: a ZeRO-3 BertModel index error versus a tokenizer-mapping bug." }, { "left": "issue:43317", "right": "issue:44756", "accept": false, - "reason": "Both are device/memory related, but one is dequantized model offload loading and the other is mmap OOM on Strix Halo." + "reason": "Different storage/offload problems: dequantized model loading on gpu+cpu offload vs mmap OOM on Strix Halo." }, { "left": "issue:43644", "right": "issue:44534", "accept": true, - "reason": "These are the same regression: Transformers 5 filling non-persistent buffers with junk, just worded slightly differently." + "reason": "Exact same regression about non-persistent buffers being filled with junk in Transformers v5." }, { "left": "issue:38175", "right": "issue:39692", "accept": false, - "reason": "Same SigLIP2 area, but one is bad outputs and the other is documentation example errors and quantization mismatch." + "reason": "Both are SigLIP2-related, but 39692 is a docs/example mismatch plus quantization failure, not the same concrete bug as zero probabilities." }, { "left": "issue:44960", "right": "issue:44977", "accept": false, - "reason": "No shared bug: a generic GLM5 issue vs Qwen3.5 flash-attention generation failure." + "reason": "GLM5 and Qwen3.5 Flash Attention generation failures are different model-specific issues." }, { "left": "issue:42491", "right": "issue:43720", "accept": false, - "reason": "Different model-loading regressions: Qwen3_MoE LoRA compatibility across hf versions vs BitNet packed-weight unpacking under accelerate." + "reason": "LoRA/Qwen3-MoE incompatibility across versions and BitNet packed-weight loading are unrelated." }, { "left": "issue:45313", "right": "issue:45357", "accept": false, - "reason": "Different Qwen3.5 regressions: ZeRO-3 loading of language_model vs incorrect save_pretrained visual-encoder keys." + "reason": "One is ZeRO-3 weight loading for language_model; the other is save_pretrained writing incorrect visual encoder keys." }, { "left": "issue:44451", "right": "issue:44488", "accept": false, - "reason": "Both are BERT model load failures, but the titles point to different models and no clear shared root cause." + "reason": "Both are model-loading failures, but they reference different checkpoints and there is no clear evidence of the same root cause." }, { "left": "issue:41720", "right": "issue:42915", "accept": false, - "reason": "Different hardware/runtime failures: auto device mapping cudaErrorAssert vs Qwen3-MoE FineGrainedFP8Config." + "reason": "Qwen3 auto device mapping cudaErrorAssert and Qwen3-MoE FP8 failures are different runtime bugs." }, { "left": "issue:42491", "right": "issue:45276", "accept": false, - "reason": "LoRA hf4.x/hf5.x compatibility is unrelated to gemma4 resize_token_embeddings behavior." - }, - { - "left": "issue:43854", - "right": "issue:43931", - "accept": false, - "reason": "Both are load errors, but for different models and failure modes: unit-test load regression vs weight-shape mismatch." + "reason": "Qwen3-MoE LoRA/version incompatibility is unrelated to Gemma4 resize_token_embeddings not updating tied outputs." }, { "left": "issue:43708", "right": "issue:44075", "accept": false, - "reason": "Trainer resume/max_steps logic is unrelated to optimizer SGD args being ignored." + "reason": "Trainer checkpoint-resume max_steps math and optimizer SGD args not being used are separate trainer/optimizer issues." } ] }, "evaluator_result": { "accept": true, - "feedback": "Overall grounded and conservative. The cluster reads like a mixed bag of Transformers regressions, with only the 43645/43646 and 43644/44534 pairs looking like close duplicates from the titles alone. One small caution: phrases like \u201clikely core reproduction path\u201d lean a bit beyond the packet, so keep the justification anchored to the title-level evidence." + "feedback": "The summary is grounded in the packet and stays conservative: the clear duplicate/near-duplicate pairs are identified, and the non-matches are rejected with appropriately cautious reasoning based on the issue titles. I did not see any overstatement that would require rejection." }, "evaluator_used": true, "retried": false, @@ -1884,6 +1897,7 @@ "issue:45859" ], "soft_pairs": [ + "issue:43854|issue:43931", "issue:43257|issue:43854", "issue:45488|issue:45812", "issue:42915|issue:44756", @@ -1896,16 +1910,16 @@ "issue:43856|issue:44387", "issue:44910|issue:45588", "issue:42491|issue:45292", - "issue:45276|issue:45335", "issue:36010|issue:39404", + "issue:45276|issue:45335", "issue:43404|issue:43454", "issue:44164|issue:44291", "issue:41720|issue:44155", "issue:44402|issue:44488", "issue:39692|issue:44960", "issue:43425|issue:43957", - "issue:43873|issue:44387", "issue:42947|issue:45704", + "issue:43873|issue:44387", "issue:43335|issue:45663", "issue:45663|issue:45759", "issue:45663|issue:45859", @@ -1915,245 +1929,241 @@ "issue:43854|issue:45216", "issue:44292|issue:44912", "issue:44743|issue:44960", - "issue:43957|issue:44756", - "issue:45276|issue:45292" + "issue:43957|issue:44756" ], - "prepared_review_unit_hash": "4af094b2b2f289610ddfff7af07dca56f0c3daec33c837d94513999f6443c0ac", + "prepared_review_unit_hash": "7919e3b5a55e53e37a2139c8a0a16e560ddce8c086b4ea7f97f6b88543ee379a", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 33, - "serialized_chars": 37135, - "estimated_input_tokens": 9284, - "estimated_eval_tokens": 18824 + "serialized_chars": 37140, + "estimated_input_tokens": 9285, + "estimated_eval_tokens": 18826 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a collection of unrelated Transformers bugs. The only clear duplicate family is the `resize_token_embeddings` regression across related model classes; most other pairs share only a broad subsystem, not the same concrete bug.", - "confidence": 0.87, - "canonical_issue_reason": "issue:45276 is the best anchor for the only clear duplicate family here: it describes the concrete `resize_token_embeddings` propagation bug in a specific composite model, and it subsumes the more generic/variant reports.", + "summary": "This cluster is mostly a grab bag of distinct Transformers issues (tokenization, model loading, quantization, device mapping, and model-specific architecture bugs). The soft pairs share broad themes, but none look like the same concrete underlying bug or change.", + "confidence": 0.83, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45276 is the most actionable and representative report in the cluster: it states the bug clearly, names the affected model path, and matches the closest duplicate reports.", + "best_issue_reason": "issue:44991 is the cleanest broad regression report in the set: it describes a general tokenizer-loading failure on transformers>=5.0.0, with a clear, stable title and scope.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Both are model-loading failures, but they affect different models and different failure modes (unit-test load issue vs weight-shape mismatch)." + }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Different models and failure modes: Qwen3 MoE load/conversion vs GLM-4.7-Flash loading in tests." + "reason": "Different code paths: Qwen3 MoE checkpoint conversion under accelerate/deepspeed vs GLM-4.7-Flash loading in tests." }, { "left": "issue:45488", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but they describe different concrete failures and likely different code paths." + "reason": "Both are tokenizer regressions, but one is a LlamaTokenizer pre-tokenizer override and the other is wrong token IDs for Granite models; not the same bug." }, { "left": "issue:42915", "right": "issue:44756", "accept": false, - "reason": "Unrelated: Qwen3 MoE weight conversion vs an mmap/OOM issue on Strix Halo." + "reason": "Unrelated problems: Qwen3Moe FP8 failure vs a Strix Halo mmap OOM workaround." }, { "left": "issue:42491", "right": "issue:43931", "accept": false, - "reason": "Different bugs: LoRA checkpoint incompatibility across versions vs a Qwen3-VL shape mismatch on load." + "reason": "LoRA compatibility on hf5.x is a save/load format issue, not the same as Qwen3-VL shape-mismatch loading." }, { "left": "issue:39692", "right": "issue:43994", "accept": false, - "reason": "Same general model family (SigLIP2), but one is a docs example problem and the other is a model/pipeline inference regression." + "reason": "Same model family, but one is a broken doc example and the other is a runtime inference regression; different fixes." }, { "left": "issue:43856", "right": "issue:43873", "accept": false, - "reason": "Memory-related, but not the same bug: Qwen3 MoE training efficiency vs quantized offloading behavior." + "reason": "Both mention memory, but one is Qwen3 MoE training memory use and the other is quantization/offloading behavior." }, { "left": "issue:44488", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading failures, but there is no evidence they share the same root cause or affected code path." + "reason": "Both are tokenizer-loading failures, but they involve different models and likely different tokenizer files/call paths." }, { "left": "issue:43644", "right": "issue:43645", "accept": false, - "reason": "Both mention Transformers 5.0, but one is about non-persistent buffers and the other about notebook custom-model initialization." + "reason": "One is junk-filled non-persistent buffers, the other is a Jupyter custom-model initialization issue; not the same bug." }, { "left": "issue:45362", "right": "issue:45406", "accept": false, - "reason": "Different components: chat crashes for Qwen3.5-35B vs a Gemma4Processor attribute error in serve." + "reason": "Different components and models: transformers chat crash on Qwen3.5 vs serve crashing on Gemma4Processor missing _tokenizer." }, { "left": "issue:43856", "right": "issue:44387", "accept": false, - "reason": "Both are memory/quantization themed, but one is training memory usage and the other is int4 reserved-memory OOM." + "reason": "Both are memory-related, but one is training-memory inefficiency and the other is reserved-memory growth under int4 quantization." }, { "left": "issue:44910", "right": "issue:45588", "accept": false, - "reason": "Both involve Flash Attention, but the reported crashes are different: packed-sequence misinterpretation vs `s_aux=None` attribute error." + "reason": "Both hit flash_attention, but one is a Qwen3.5 position_ids crash and the other is an s_aux=None AttributeError on sink-less models." }, { "left": "issue:42491", "right": "issue:45292", "accept": false, - "reason": "Unrelated: LoRA checkpoint compatibility vs a generic `resize_token_embeddings` output-embedding regression." - }, - { - "left": "issue:45276", - "right": "issue:45335", - "accept": true, - "reason": "Same underlying bug: `resize_token_embeddings` is not propagating updates into model-specific embedding/output modules in related architectures." + "reason": "LoRA checkpoint incompatibility is unrelated to resize_token_embeddings behavior." }, { "left": "issue:36010", "right": "issue:39404", "accept": false, - "reason": "Different errors: an import failure for `GenerationMixin` vs Whisper pipeline `return_language` behavior." + "reason": "Importing GenerationMixin and Whisper pipeline return_language are unrelated failures." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": false, + "reason": "Both involve resize_token_embeddings, but they hit different architectures and different submodules; too model-specific to treat as one bug." }, { "left": "issue:43404", "right": "issue:43454", "accept": false, - "reason": "Both are weight-tying problems, but they affect different model classes and are not clearly the same concrete bug." + "reason": "Both are lm_head tying bugs, but they are on different model classes (Mistral3 vs AyaVision) and likely separate fixes." }, { "left": "issue:44164", "right": "issue:44291", "accept": false, - "reason": "Different lifecycle issues: saving/loading `extra_state` vs `init_empty_weights` passing an unexpected argument." + "reason": "Different serialization/loading paths: extra_state handling vs init_empty_weights/_is_hf_initialized argument mismatch." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "Different code paths and symptoms: device-mapped Qwen3 CUDA assert vs batched AudioFlamingo3 leakage across tracks." + "reason": "Qwen3 auto device-map cuda assert and AudioFlamingo3 batched inference leakage are unrelated bugs." }, { "left": "issue:44402", "right": "issue:44488", "accept": false, - "reason": "Different tokenizer/model-loading problems; no evidence they share the same fix." + "reason": "Both concern specific model/tokenizer loading, but one is vocab-size token placement and the other is a separate tokenizer load failure." }, { "left": "issue:39692", "right": "issue:44960", "accept": false, - "reason": "Different models and failures; GLM5 is unrelated to the SigLIP2 example issues." + "reason": "Doc-example errors for SigLIP2 are unrelated to the generic GLM5 issue report." }, { "left": "issue:43425", "right": "issue:43957", "accept": false, - "reason": "Torch version incompatibility is not the same as a `meta`-device loading bug." + "reason": "Torch version incompatibility is not the same as meta-device loading failures on specific models." }, { - "left": "issue:43873", - "right": "issue:44387", + "left": "issue:42947", + "right": "issue:45704", "accept": false, - "reason": "Related broadly to quantization/offloading, but not the same concrete regression." + "reason": "Gradient checkpointing with PEFT LoRA and apex RMSNorm memory leak are unrelated subsystems." }, { - "left": "issue:42947", - "right": "issue:45704", + "left": "issue:43873", + "right": "issue:44387", "accept": false, - "reason": "Different bugs: PEFT gradient checkpointing behavior vs an apex RMSNorm memory leak." + "reason": "Both involve memory pressure, but one is offloading not working and the other is increased reserved memory under int4 quantization." }, { "left": "issue:43335", "right": "issue:45663", "accept": false, - "reason": "Different models and causes: SwitchTransformers sparse-layer construction vs Gemma-4 FSDP2 shared state rebuild." + "reason": "SwitchTransformers sparse-layer config bug is unrelated to Gemma4 FSDP2 shared_kv_states behavior." }, { "left": "issue:45663", "right": "issue:45759", "accept": false, - "reason": "Different issues in composite-model handling; no shared failure mode." + "reason": "Different model families and failures: Gemma4 FSDP2 KeyError vs Qwen multimodal config unwrapping." }, { "left": "issue:45663", "right": "issue:45859", "accept": false, - "reason": "Different code paths: FSDP2/shared state vs CPU-offload incompatibility in Qwen3_5Moe rotary embedding." + "reason": "Gemma4 FSDP2 shared_kv_states bug and Qwen3_5MoeTextRotaryEmbedding CPU-offload issue are separate code paths." }, { "left": "issue:44291", "right": "issue:44756", "accept": false, - "reason": "Different problems: `init_empty_weights` initialization error vs mmap-related OOM." + "reason": "init_empty_weights argument handling is unrelated to disabling mmap for Strix Halo OOM." }, { "left": "issue:44155", "right": "issue:45823", "accept": false, - "reason": "Different model-specific bugs; one is token/embedding leakage, the other is a Gemma4 device mismatch." + "reason": "Audio track leakage in batched inference is unrelated to Gemma4 device mismatch during forward." }, { "left": "issue:44387", "right": "issue:44756", "accept": false, - "reason": "Different memory/OOM scenarios with no clear common underlying bug." + "reason": "One is quantization-related CUDA memory growth; the other is a platform-specific mmap OOM workaround." }, { "left": "issue:43854", "right": "issue:45216", "accept": false, - "reason": "Different regressions: GLM-4.7-Flash loading vs Qwen3.5 save_pretrained checkpoint correctness." + "reason": "Different regression surfaces: GLM-4.7-Flash loading in tests vs Qwen3.5 save_pretrained checkpoint correctness." }, { "left": "issue:44292", "right": "issue:44912", "accept": false, - "reason": "Both are quantization loading issues, but they involve different models and formats (NVFP4 vs MXFP4) and likely different fixes." + "reason": "Both are quantized-model loading issues, but they concern different models, formats, and failure outcomes." }, { "left": "issue:44743", "right": "issue:44960", "accept": false, - "reason": "Unrelated: recurrent state reset bug vs a general GLM5 loading issue." + "reason": "Qwen3.5 recurrent-state cache bug is unrelated to the generic GLM5 issue." }, { "left": "issue:43957", "right": "issue:44756", "accept": false, - "reason": "Different load-path issues: `torch.device('meta')` compatibility vs mmap/OOM on Strix Halo." - }, - { - "left": "issue:45276", - "right": "issue:45292", - "accept": true, - "reason": "Same underlying bug class: `resize_token_embeddings` not updating the expected output embedding path, with one report being the generic form of the other." + "reason": "Meta-device loading failures and Strix Halo mmap OOM are different problem classes." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The only duplicate family that is well supported by the titles and soft-pair evidence is the `resize_token_embeddings` cluster around issues 45276/45335/45292, and the other pairwise rejections avoid overclaiming shared root cause. No mergeability concerns apply here since this is not a PR pair." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -2206,8 +2216,8 @@ "issue:44912", "issue:44918", "issue:44928", - "issue:45005", "issue:45084", + "issue:45276", "issue:45292", "issue:45335", "issue:45588", @@ -2216,14 +2226,13 @@ "issue:45698", "issue:45701", "issue:45759", - "issue:45823", - "issue:45902" + "issue:45823" ], "soft_pairs": [ "issue:45684|issue:45698", + "issue:45276|issue:45292", "issue:42947|issue:43381", "issue:43381|issue:44928", - "issue:45663|issue:45902", "issue:44164|issue:44756", "issue:44479|issue:44877", "issue:43646|issue:43708", @@ -2239,8 +2248,8 @@ "issue:44683|issue:45823", "issue:43931|issue:44863", "issue:43906|issue:44530", - "issue:43611|issue:43645", "issue:43425|issue:45823", + "issue:43611|issue:43645", "issue:44530|issue:45084", "issue:43425|issue:43716", "issue:43901|issue:44509", @@ -2250,252 +2259,245 @@ "issue:44479|issue:44805", "issue:43994|issue:44079", "issue:39404|issue:44671", - "issue:45005|issue:45902", "issue:44589|issue:45084", "issue:43425|issue:44918", "issue:44164|issue:45759" ], - "prepared_review_unit_hash": "a886f5bdfd74e9223700f8eb7558789702e872bda25aa75775732a79025a991e", + "prepared_review_unit_hash": "579d0df0b5fa4accc9c8297106d96f7d1bb5b01b2a82880bb22b48e2280ca080", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 34, - "serialized_chars": 37931, - "estimated_input_tokens": 9483, - "estimated_eval_tokens": 19222 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37052, + "estimated_input_tokens": 9263, + "estimated_eval_tokens": 18782 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is still mostly heterogeneous. 45292 remains the only clear duplicate anchor, with 45335 as a near-exact match; the Qwen3/Qwen3-VL pair should not be merged as a duplicate on the available evidence.", - "confidence": 0.78, - "canonical_issue_reason": "45292 is the best anchor because it has the clearest true duplicate in 45335 and states the broad underlying resize_token_embeddings propagation bug.", + "summary": "This cluster is mostly a grab bag of unrelated Transformers regressions and compatibility issues. I found one plausible duplicate pair around `resize_token_embeddings` not updating output embeddings; the rest look like distinct bugs or model-specific failures.", + "confidence": 0.64, + "canonical_issue_reason": "issue:43611 is the clearest broad regression-style issue in the set: it has a specific broken code path, a version boundary (Transformers 5.0.0), and a load-time compatibility failure that makes it a reasonable anchor for the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "45292 is the strongest issue to keep as the dedupe target because it is concrete, broadly phrased, and maps cleanly to 45335.", + "best_issue_reason": "issue:43611 is the most suitable representative issue because it is explicit, actionable, and broadly about Transformers 5 model-loading breakage rather than an incidental symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45684", "right": "issue:45698", "accept": false, - "reason": "Both touch custom-model save/load, but one is permissions propagation on save_pretrained and the other is wrong module resolution on load; different bugs." + "reason": "Both involve custom-model save/load, but one is about permissions on source files and the other about loading the wrong custom module; different bugs." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same underlying `resize_token_embeddings` bug: output embeddings are not updated; 45276 is just a Gemma4-specific manifestation with extra layer details." }, { "left": "issue:42947", "right": "issue:43381", "accept": false, - "reason": "Same broad gradient checkpointing area, but LoRA/PEFT ineffectiveness and eval-mode prohibition are different failures with different fixes." + "reason": "Both mention gradient checkpointing, but one is PEFT/LoRA ineffectiveness and the other is eval-mode restriction; not the same failure." }, { "left": "issue:43381", "right": "issue:44928", "accept": false, - "reason": "Eval-mode gradient checkpointing is unrelated to Qwen3.5 RLHF NaNs from 3D position_ids and SDPA fallback." - }, - { - "left": "issue:45663", - "right": "issue:45902", - "accept": false, - "reason": "Both are model-specific state/init problems, but Gemma4 FSDP2 KeyError and Qwen3.5 MoE inv_freq materialization are different code paths." + "reason": "Unrelated issues: eval-mode gradient checkpointing vs RLHF NaNs from Qwen3.5 3D position IDs and SDPA fallback." }, { "left": "issue:44164", "right": "issue:44756", "accept": false, - "reason": "Extra_state save/load handling is unrelated to the Strix Halo mmap OOM issue." + "reason": "`extra_state` serialization failure and Strix Halo mmap OOM are different subsystems and different bugs." }, { "left": "issue:44479", "right": "issue:44877", "accept": false, - "reason": "Different regressions: Qwen VL video-input behavior versus strict config loading." + "reason": "Video-input regression for Qwen VL models vs strict-config loading failure for granite_speech; not the same code path." }, { "left": "issue:43646", "right": "issue:43708", "accept": false, - "reason": "Custom model initialization breakage is unrelated to resume_from_checkpoint/max_steps calculation." + "reason": "Custom model initialization in notebooks is unrelated to `resume_from_checkpoint` max-step calculation." }, { "left": "issue:45588", "right": "issue:45663", "accept": false, - "reason": "Flash-attention sinkless-model crash and Gemma4 shared_kv_states KeyError are different bugs." + "reason": "Flash-attention `s_aux=None` crash and Gemma4 FSDP2 shared-key/value rebuild error are different model-specific forward-path bugs." }, { "left": "issue:45292", "right": "issue:45335", - "accept": true, - "reason": "Same underlying resize_token_embeddings bug; 45335 is a specific t5gemma manifestation of the broader output-embedding propagation failure." + "accept": false, + "reason": "Both are `resize_token_embeddings`-related, but one targets generic output embeddings while the other is about `decoder.embed_tokens`; too model-specific to treat as the same bug." }, { "left": "issue:43257", "right": "issue:44863", "accept": false, - "reason": "Both are checkpoint-loading complaints, but Qwen3 MOE conversion under accelerate+deepspeed is not the same as NemotronH checkpoint incompatibility." + "reason": "Qwen3 MOE weight conversion under accelerate/deepspeed and NemotronH checkpoint loading are distinct loading issues." }, { "left": "issue:44402", "right": "issue:44451", "accept": false, - "reason": "Tokenizer vocab-size mismatch and a broader load failure for a different model are not the same bug." + "reason": "Tokenizer vocab-size mismatch and inability to load ScandiBERT are not the same underlying defect." }, { "left": "issue:43645", "right": "issue:44164", "accept": false, - "reason": "Notebook/custom-model init breakage is different from extra_state serialization/deserialization." + "reason": "Jupyter custom model initialization and `extra_state` save/from_pretrained handling are separate problems." }, { "left": "issue:43828", "right": "issue:43957", "accept": false, - "reason": "Different dtype bugs: autocast mismatch in Phi-tiny-MoE versus meta-device loading breakage in other models." + "reason": "Autocast dtype mismatch for Phi-tiny-MoE and meta-device loading failures are different runtime environments and failure modes." }, { "left": "issue:43425", "right": "issue:43828", "accept": false, - "reason": "Generic Torch 2.10 incompatibility does not match the model-specific autocast dtype mismatch." + "reason": "Torch 2.10 compatibility is a version/platform issue; the other is a specific autocast dtype mismatch in a model forward path." }, { "left": "issue:39404", "right": "issue:45701", "accept": false, - "reason": "Whisper pipeline return_language regression is unrelated to tokenizer behavior changes across versions." + "reason": "Whisper `return_language` pipeline breakage and tokenizer version-dependent behavior are unrelated." }, { "left": "issue:44589", "right": "issue:44756", "accept": false, - "reason": "Float8 storage lookup failure and Strix Halo mmap OOM are unrelated." + "reason": "Float8 storage lookup failure and mmap-related OOM are unrelated runtime/storage issues." }, { "left": "issue:44683", "right": "issue:45823", "accept": false, - "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 device_map forward mismatch are different issues." + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 PLE device mismatch are different subsystems." }, { "left": "issue:43931", "right": "issue:44863", "accept": false, - "reason": "Different model families and failure modes: Qwen3-VL weight-shape mismatch versus NemotronH checkpoint loading." + "reason": "Qwen3-VL shape mismatch and NemotronH checkpoint loading are different model-loading bugs." }, { "left": "issue:43906", "right": "issue:44530", "accept": false, - "reason": "The isolated reproduction reference is too opaque to equate with the specific PagedAttentionCache linear_attention crash." + "reason": "A reproduction of an earlier issue and a PagedAttentionCache `linear_attention` crash are not the same bug." }, { - "left": "issue:43611", - "right": "issue:43645", + "left": "issue:43425", + "right": "issue:45823", "accept": false, - "reason": "Both are v5 custom-model regressions, but one is base_model_prefix loading and the other is notebook-based initialization." + "reason": "General Torch 2.10 incompatibility is not the same as Gemma4 device-map forward mismatch." }, { - "left": "issue:43425", - "right": "issue:45823", + "left": "issue:43611", + "right": "issue:43645", "accept": false, - "reason": "Torch compatibility and Gemma4 PLE device mismatch are unrelated." + "reason": "Both are Transformers 5 loading regressions, but one is about `base_model_prefix` and the other about notebook-defined custom models; different code paths." }, { "left": "issue:44530", "right": "issue:45084", "accept": false, - "reason": "PagedAttentionCache linear_attention crash is unrelated to non-template-node compilation errors." + "reason": "PagedAttentionCache group-type crash and non-template node compilation error are unrelated." }, { "left": "issue:43425", "right": "issue:43716", "accept": false, - "reason": "Generic Torch 2.10 incompatibility does not match the Mistral-3 image-preprocessor dtype mismatch." + "reason": "Torch version incompatibility and Mistral-3 image-preprocessor dtype mismatch are different bugs." }, { "left": "issue:43901", "right": "issue:44509", "accept": false, - "reason": "Both are docs-related, but they concern different pipeline-task deprecations." + "reason": "These are both docs issues, but they cover different removed/changed pipeline behaviors." }, { "left": "issue:39404", "right": "issue:44912", "accept": false, - "reason": "Whisper pipeline regression and MXFP4 quantization fallback are unrelated." + "reason": "Whisper pipeline behavior and MXFP4 quantization fallback are unrelated." }, { "left": "issue:44265", "right": "issue:44898", "accept": false, - "reason": "torch.export with torch_compilable_check is a different export-path bug from Perceiver interpolation/resolution failure." + "reason": "Torch export failure via `torch_compilable_check` is unrelated to Perceiver non-default-resolution failures." }, { "left": "issue:43299", "right": "issue:43931", "accept": false, - "reason": "Shared Qwen3/Qwen3-VL loading context is not enough; one is MOE weight conversion with accelerate+deepspeed, the other is a Qwen3-VL weight-shape mismatch, so the evidence is too weak to call them the same bug." + "reason": "Both are Qwen3-VL loading regressions, but one is a generic v5 breakage and the other is a specific checkpoint shape mismatch; not clearly the same bug." }, { "left": "issue:44479", "right": "issue:44805", "accept": false, - "reason": "Qwen video-input regression and a tensor mask shape IndexError are not the same bug." + "reason": "Qwen video-input regression and a mask/tensor shape IndexError are different failure modes." }, { "left": "issue:43994", "right": "issue:44079", "accept": false, - "reason": "Nonsensical SigLIP2 outputs and ModelOutput key assignment are unrelated." + "reason": "Siglip2 wrong outputs and `ModelOutput` key assignment behavior are unrelated." }, { "left": "issue:39404", "right": "issue:44671", "accept": false, - "reason": "Whisper pipeline regression and CamemBERT masked LM predictions are unrelated." - }, - { - "left": "issue:45005", - "right": "issue:45902", - "accept": false, - "reason": "Tied weights on translation models and Qwen3.5 MoE rotary embedding materialization are different issues." + "reason": "Whisper pipeline language return regression and CamemBERT masked-LM prediction regression are unrelated model bugs." }, { "left": "issue:44589", "right": "issue:45084", "accept": false, - "reason": "Float8 storage lookup and template-node compilation errors are unrelated." + "reason": "Float8 storage lookup and non-template node compilation failures are unrelated." }, { "left": "issue:43425", "right": "issue:44918", "accept": false, - "reason": "Torch version incompatibility is not the same as Qwen3.5 input-embedding unpacking in TRL SFT trainer." + "reason": "Torch version incompatibility and Qwen3.5 embedding unpacking with TRL are different issues." }, { "left": "issue:44164", "right": "issue:45759", "accept": false, - "reason": "Extra_state save/load handling and text_config unwrapping for composite configs are different loading paths." + "reason": "`extra_state` serialization and `from_config` text-config unwrapping for Qwen multimodal models are different code paths." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The 45292/45335 duplicate call is well grounded, but the accepted 43299/43931 soft-edge looks too strong for the packet: one issue is about Qwen3 MOE weight conversion with accelerate+deepspeed, while the other is a Qwen3-VL weight-shape mismatch. The titles indicate related model family/loading context, but not enough to conservatively call them the same underlying bug. The rest of the verdicts are mostly conservative, but this one overstates the evidence." + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The representative issue choice is defensible, and the only accepted soft-edge pair (issue:45276 / issue:45292) is supported by the shared `resize_token_embeddings` / `output_embeddings` wording in the titles. The rejected pairs are generally framed with caution and do not overstate duplicate evidence." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -2563,8 +2565,8 @@ "issue:45859" ], "soft_pairs": [ - "issue:44671|issue:44991", "issue:36032|issue:43452", + "issue:44671|issue:44991", "issue:43596|issue:44683", "issue:43856|issue:44756", "issue:32090|issue:39404", @@ -2592,246 +2594,246 @@ "issue:44849|issue:45325", "issue:43317|issue:43482", "issue:36032|issue:43645", - "issue:43404|issue:44466", "issue:44987|issue:44991", + "issue:43404|issue:44466", "issue:43883|issue:44991", "issue:43262|issue:43377" ], - "prepared_review_unit_hash": "4de6142263f09682895a3a5b8c4bd275828d77d25ad1447b2a6c7b53348f3b69", + "prepared_review_unit_hash": "7d2e2c866b202148be9a01a1cb43344f1843651457ab01603a1d08b08e138524", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37688, - "estimated_input_tokens": 9422, - "estimated_eval_tokens": 19100 + "serialized_chars": 37690, + "estimated_input_tokens": 9423, + "estimated_eval_tokens": 19102 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous regression reports rather than true duplicates. The only plausible duplicate pair is the GGUF loading regression; the rest differ by model, subsystem, or failure mode. No pull requests are present.", - "confidence": 0.41, - "canonical_issue_reason": "issue:43452 is the broadest and most central description of the GGUF loading regression, covering the shared from_pretrained/gguf_file code path and subsuming the more specific model report.", + "summary": "Mostly conservative clustering. I keep the audio feature-extraction duplicate, but I reject the 43644/43646 soft edge: both are v5 regressions, yet the packet does not show they are the same bug or even the same failure mode.", + "confidence": 0.83, + "canonical_issue_reason": "issue:43646 is the best umbrella issue for the custom-model-initialization regression described in this packet; it is broader and more representative than the more specific notebook-related report nearby.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43452 is the best representative because it names the generic loader breakage rather than a single model instance, making it the best umbrella issue for the GGUF failure.", + "best_issue_reason": "issue:43646 is the strongest general issue candidate because its title is broad, user-facing, and likely to absorb nearby custom-model-init reports without overclaiming cross-duplication.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44671", - "right": "issue:44991", + "left": "issue:36032", + "right": "issue:43452", "accept": false, - "reason": "Both are tokenizer-related, but one is a CamemBERT masked-LM prediction issue and the other is a tokenizer loading failure for a different model; not the same bug." + "reason": "Different failure modes: T5 tokenizer special-token conflict vs GGUF loader breakage." }, { - "left": "issue:36032", - "right": "issue:43452", + "left": "issue:44671", + "right": "issue:44991", "accept": false, - "reason": "T5 tokenizer class conflict during load vs GGUF/from_pretrained loading breakage are different code paths and failure modes." + "reason": "Both are v5 regressions, but one is CamemBERT MLM behavior and the other is tokenizer loading for a different model." }, { "left": "issue:43596", "right": "issue:44683", "accept": false, - "reason": "Deepspeed ZeRO-3/BertModel index error is unrelated to compiled flex_attention on torch>=2.9." + "reason": "Deepspeed zero3/BertModel init error is unrelated to compiled flex_attention on torch >= 2.9." }, { "left": "issue:43856", "right": "issue:44756", "accept": false, - "reason": "Both mention memory, but one is Qwen3 MoE training inefficiency and the other is mmap OOM on Strix Halo; different root causes." + "reason": "Qwen3 MoE memory usage and Strix Halo mmap OOM are unrelated resource issues." }, { "left": "issue:32090", "right": "issue:39404", "accept": false, - "reason": "Trainer GPU broadcast NoneType error and Whisper return_language pipeline regression are unrelated." + "reason": "Trainer GPU broadcast NoneType error is unrelated to Whisper return_language pipeline behavior." }, { "left": "issue:44991", "right": "issue:45081", "accept": false, - "reason": "Different tokenizer failures on different models; one is a general load regression, the other is a Mistral regex patch crash." + "reason": "Both involve loading/tokenizer paths, but they target different models and different bugs." }, { "left": "issue:44737", "right": "issue:45859", "accept": false, - "reason": "Both are device/offload related, but one is XLNet CPU placement in relative positional encoding and the other is Qwen3.5-MoE rotary embedding offload incompatibility." + "reason": "XLNet CPU arange inefficiency is unrelated to Qwen3.5 MoE CPU offload compatibility." }, { "left": "issue:44164", "right": "issue:44479", "accept": false, - "reason": "save/from_pretrained extra_state handling is unrelated to the Qwen2.5-VL video input regression." + "reason": "Serialization of extra_state is unrelated to the Qwen video-input regression." }, { "left": "issue:43582", "right": "issue:43632", "accept": false, - "reason": "AppleSilicon warmup TypeError and the _is_hf_initialized regression are different breakages." + "reason": "Apple Silicon allocator warmup TypeError is unrelated to the _is_hf_initialized flag regression." }, { "left": "issue:44488", "right": "issue:45081", "accept": false, - "reason": "Both concern tokenizers, but the model families and failure mechanisms differ." + "reason": "Both are model-loading reports, but the affected models and tokenizer code paths differ." }, { "left": "issue:44661", "right": "issue:44683", "accept": false, - "reason": "add-new-model-like mapping logic and compiled flex_attention are unrelated." + "reason": "TOKENIZER_MAPPING_NAMES handling and flex_attention compilation are unrelated subsystems." }, { "left": "issue:40990", "right": "issue:43421", "accept": false, - "reason": "High perplexity on a benchmark and runtime post-processor updates for special tokens are not the same bug." + "reason": "Perplexity regression on gpt-oss is unrelated to runtime special-token post-processing updates." }, { "left": "issue:43611", "right": "issue:43644", "accept": false, - "reason": "Loading models with base_model_prefix and non-persistent buffers being filled with junk are separate v5 regressions." + "reason": "Base-model-prefix loading failure and non-persistent-buffer junk are different v5 regressions." }, { "left": "issue:44488", "right": "issue:44987", "accept": false, - "reason": "Both are loading failures, but they affect different models and likely different underlying causes." + "reason": "Both are loading failures, but there is no evidence they share the same underlying bug." }, { "left": "issue:44206", "right": "issue:44220", - "accept": false, - "reason": "Audio feature-extractor crash due to an unsupported arg is not clearly the same as the generic _torch_extract_fbank_features issue." + "accept": true, + "reason": "Both point to the same audio feature-extraction bug around unsupported center handling / _torch_extract_fbank_features()." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Image preprocessor dtype mismatch and a template compilation TypeError are unrelated." + "reason": "Mistral-3 image-preprocessor dtype mismatch is unrelated to template compilation errors." }, { "left": "issue:43643", "right": "issue:44530", "accept": false, - "reason": "AutoConfig trust_remote_code field loss and PagedAttentionCache linear_attention crash are different subsystems." + "reason": "Missing fields from AutoConfig with trust_remote_code is unrelated to Qwen3.5 PagedAttention cache crashes." }, { "left": "issue:43883", "right": "issue:44488", "accept": false, - "reason": "Missing all_tied_weights_keys and inability to load a specific BERT model are unrelated." + "reason": "Missing all_tied_weights_keys and model-loading failure for a different checkpoint are not the same bug." }, { "left": "issue:34567", "right": "issue:40444", "accept": false, - "reason": "TrainerState token-count tracking and multi-image IterableDataset finetuning failure are unrelated." + "reason": "Trainer token-count tracking and multi-image IterableDataset finetuning failure are unrelated." }, { "left": "issue:44811", "right": "issue:45584", "accept": false, - "reason": "Both are Whisper special-token issues, but one is batch_decode ignoring skip_special_tokens while the other is empty-transcription generation failure; too different to merge." + "reason": "Whisper decode skipping special tokens and empty-transcription generation failure are different behaviors." }, { "left": "issue:36032", "right": "issue:42915", "accept": false, - "reason": "Tokenizer load conflict for T5 vs FineGrainedFP8Config failure for Qwen3Moe are unrelated." + "reason": "T5 tokenizer load conflict is unrelated to Qwen3 MoE FineGrainedFP8 failure." }, { "left": "issue:43644", "right": "issue:43646", "accept": false, - "reason": "Non-persistent buffer junk and custom model initialization breakage are different v5 regressions." + "reason": "Both are v5 regressions, but the packet does not establish the same underlying bug or code path." }, { "left": "issue:43452", "right": "issue:43482", - "accept": true, - "reason": "Both point to the same GGUF/from_pretrained loading path breaking in v5; the model-specific report is plausibly the same underlying loader bug." + "accept": false, + "reason": "gguf_file loading breakage and Qwen2.5-GGUF loading failure are separate loader issues." }, { "left": "issue:43541", "right": "issue:45084", "accept": false, - "reason": "torch dynamo tracing failure in grouped_mm is not the same as a non-template-node compilation error." + "reason": "Torch dynamo grouped_mm tracing error is unrelated to template compilation failure." }, { "left": "issue:44451", "right": "issue:44987", "accept": false, - "reason": "Different model loading failures on different checkpoints; no shared concrete bug is evident." + "reason": "Loading vesteinn/ScandiBERT and loading physical-intelligence/fast are different model-specific regressions." }, { "left": "issue:42915", "right": "issue:44530", "accept": false, - "reason": "FineGrainedFP8Config failure and PagedAttentionCache invalid group type are different MoE-related issues." + "reason": "Qwen3 MoE FP8 training and Qwen3.5 PagedAttention linear_attention crash are different code paths." }, { "left": "issue:44849", "right": "issue:45325", "accept": false, - "reason": "Hidden-state output bug and still-image rope index scaling bug affect different Qwen model paths." + "reason": "Qwen3.5 hidden-states output bug and Qwen2.5-VL rope-index scaling bug are unrelated." }, { "left": "issue:43317", "right": "issue:43482", "accept": false, - "reason": "device_map/offload dequantized-model loading and GGUF loading are separate loading problems." + "reason": "Dequantized model offload failure and Qwen2.5-GGUF loading failure are not the same bug." }, { "left": "issue:36032", "right": "issue:43645", "accept": false, - "reason": "Tokenizer special-token conflict and Jupyter custom-model initialization breakage are unrelated." + "reason": "Tokenizer special-token conflict is unrelated to custom model initialization in notebooks." }, { - "left": "issue:43404", - "right": "issue:44466", + "left": "issue:44987", + "right": "issue:44991", "accept": false, - "reason": "lm_head tie missing in Mistral3 and inconsistent serialization of lm_head.weight are related thematically but not the same concrete bug." + "reason": "Both fail on model loading in v5.x, but they are different checkpoints and different failure modes." }, { - "left": "issue:44987", - "right": "issue:44991", + "left": "issue:43404", + "right": "issue:44466", "accept": false, - "reason": "Two different model-loading failures with different triggers; not the same underlying issue." + "reason": "Mistral3 lm_head tying and inconsistent lm_head serialization are related to weights, but not the same concrete bug from the titles." }, { "left": "issue:43883", "right": "issue:44991", "accept": false, - "reason": "all_tied_weights_keys attribute error and tokenizer loading failure for a specific model are unrelated." + "reason": "Missing all_tied_weights_keys and tokenizer-loading failure are unrelated." }, { "left": "issue:43262", "right": "issue:43377", "accept": false, - "reason": "Audio processor sampling-rate default issue and MIMI batched-vs-single padding-mask bug are different." + "reason": "Audio processor chat-template sampling-rate default and MIMI padding-mask batching are different audio bugs." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is well grounded in the packet and stays conservative. The GGUF pair (issue:43452 / issue:43482) is the only clearly plausible duplicate, and the rest of the soft-edge rejections are appropriately cautious given the differing titles, subsystems, or model families. No PR pairs are present, so there is nothing mergeability-related to flag." + "accept": false, + "feedback": "The overall clustering is mostly conservative, but one accepted soft-edge is too strong for the packet. `issue:43644` (\u201cfills non-persistent buffers with junk\u201d) and `issue:43646` (\u201cbreaks custom model initialization\u201d) are both v5 regressions, but the packet does not ground them as the same bug or even clearly the same failure mode; the rationale overstates evidence by calling them the same underlying change. Because the acceptability of a soft-edge verdict must stay conservative, I can\u2019t accept this summary as-is." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -2923,8 +2925,8 @@ "issue:42915|issue:44589", "issue:43611|issue:44534", "issue:43531|issue:44877", - "issue:40990|issue:44811", "issue:36032|issue:41720", + "issue:40990|issue:44811", "issue:43425|issue:44403", "issue:44079|issue:45663", "issue:41553|issue:43582", @@ -2932,236 +2934,233 @@ "issue:42175|issue:45542", "issue:43818|issue:45357" ], - "prepared_review_unit_hash": "6ae41cf5e649f457975dd0b4f5c3125ca7653c98837374a964a5592244e7af42", + "prepared_review_unit_hash": "22ec130f16e76344a898e5938bf0593e1140e0345c3f214267bf21608f471de5", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37208, - "estimated_input_tokens": 9302, - "estimated_eval_tokens": 18860 + "serialized_chars": 37215, + "estimated_input_tokens": 9304, + "estimated_eval_tokens": 18864 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a loose collection of unrelated Transformers issues with a few true duplicate-like pairs. The clearest duplicate families are Qwen3.5 save_pretrained regressions, Qwen2.5-VL video position-id regression, and TensorFlow backend install packaging.", - "confidence": 0.73, - "canonical_issue_reason": "issue:45216 is the cleanest umbrella for a concrete regression with a clear follow-up duplicate (issue:45357) and a specific loading/saving failure mode.", + "summary": "This cluster is mostly a loose mix of unrelated transformer/model bugs, with several broad Transformers v5 regression reports but no clear duplicate set. I found no soft-edge pair that looks like the same underlying bug.", + "confidence": 0.89, + "canonical_issue_reason": "issue:43646 is the broadest representative of the Transformers v5 initialization regressions in this set, and it is easier to treat as the umbrella issue than the more narrowly scoped siblings.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45216 is the strongest canonical issue candidate in this set because it describes a concrete, version-linked regression rather than a vague symptom.", + "best_issue_reason": "issue:43646 is the strongest single issue candidate because it is a clear, general regression report about custom model initialization, which is a recurring theme across the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45084", "right": "issue:45507", "accept": false, - "reason": "Different bugs and code paths: template compilation failure vs GraniteMoEHybrid invalid method call." + "reason": "Different failures: template-node compilation vs an invalid method call in GraniteMoEHybrid." }, { "left": "issue:43994", "right": "issue:44610", "accept": false, - "reason": "Both are vision-model complaints, but they describe different failures: bad SigLIP2 outputs vs OmDet-Turbo size mismatch." + "reason": "Different models and symptoms: nonsensical outputs in SigLIP2 vs a processor/model input-size mismatch in OmDet-Turbo." }, { "left": "issue:43646", "right": "issue:44534", "accept": false, - "reason": "Unrelated regressions: custom model initialization vs non-persistent buffer initialization." + "reason": "Both are Transformers v5 regressions, but one is custom model init while the other is non-persistent buffer corruption; different bugs." }, { "left": "issue:43322", "right": "issue:43540", "accept": false, - "reason": "Different multimodal models and failures: Llava Next loading segfault vs Qwen3OmniMoe video ValueError." + "reason": "Different multimodal code paths: Llava Next loading segfault vs Qwen3OmniMoe video input validation failure." }, { "left": "issue:43645", "right": "issue:44534", "accept": false, - "reason": "Different underlying issue; one is notebook/custom model init, the other is buffer junk in v5." + "reason": "Jupyter/custom-model initialization breakage is not the same bug as junk-filled non-persistent buffers." }, { "left": "issue:44403", "right": "issue:45081", "accept": false, - "reason": "Logging/noise during loading is not the same as a tokenizer patch crash." + "reason": "Unnecessary loading noise is unrelated to the Mistral tokenizer regex patch crash." }, { "left": "issue:43377", "right": "issue:43819", "accept": false, - "reason": "Different audio-model bugs: padding-mask batching mismatch vs DAC latent/STE mismatch." + "reason": "Different audio/model implementations and failures: batching/padding-mask mismatch in MIMI vs DAC STE mismatch." }, { "left": "issue:42915", "right": "issue:45084", "accept": false, - "reason": "No shared code path; FP8 config failure is unrelated to template compilation." + "reason": "FineGrainedFP8Config failure in Qwen3Moe is unrelated to the non-template-node compilation error." }, { "left": "issue:44265", "right": "issue:44610", "accept": false, - "reason": "torch.export/torch_compilable_check failure is unrelated to OmDet-Turbo preprocessing size mismatch." + "reason": "torch.export with torch_compilable_check is a different export-time issue from a processor size mismatch." }, { "left": "issue:45216", "right": "issue:45357", - "accept": true, - "reason": "Same Qwen3.5 save_pretrained regression; 45357 looks like a narrower follow-up about the same incorrect checkpoint contents." + "accept": false, + "reason": "Both concern Qwen3.5 save_pretrained, but one is a general checkpoint regression and the other is specifically incorrect visual encoder keys; not clearly the same bug." }, { "left": "issue:44451", "right": "issue:45081", "accept": false, - "reason": "Different tokenization/loading failures; no evidence of the same bug." + "reason": "Different model-loading failures: ScandiBERT cannot load, while Mistral regex patching hits a tokenizer backend attribute error." }, { "left": "issue:44220", "right": "issue:44683", "accept": false, - "reason": "Different subsystems entirely: fbank feature extraction vs compiled flex_attention." + "reason": "Unrelated subsystems: audio feature extraction vs compiled flex_attention on newer torch." }, { "left": "issue:44987", "right": "issue:45081", "accept": false, - "reason": "Different model-loading failures; only superficial similarity in being tokenizer/model load issues." + "reason": "Both are loading-related, but one is a model-specific failure on physical-intelligence/fast and the other is a Mistral tokenizer patch crash." }, { "left": "issue:43404", "right": "issue:45127", "accept": false, - "reason": "Both touch tied embeddings, but one is a missing weight tie in a model class and the other is LoRA merge collapse after vocab extension." + "reason": "Both touch tied embeddings/lm_head behavior, but one is a missing weight tie in Mistral3 while the other is LoRA merge collapse with extended vocab." }, { "left": "issue:43632", "right": "issue:44756", "accept": false, - "reason": "Transformer v5 flag regression is unrelated to Strix Halo mmap OOM." + "reason": "Different regressions: _is_hf_initialized handling vs mmap/OOM behavior on Strix Halo." }, { "left": "issue:42915", "right": "issue:44291", "accept": false, - "reason": "Different failure modes: Qwen3Moe FP8 config vs init_empty_weights/_is_hf_initialized argument handling." + "reason": "Qwen3Moe FP8 config failure is not the same as init_empty_weights rejecting an unexpected _is_hf_initialized argument." }, { "left": "issue:44062", "right": "issue:44987", "accept": false, - "reason": "AddedToken special-arg conflict is unrelated to the physical-intelligence/fast loading failure." + "reason": "Tokenizer AddedToken argument duplication is unrelated to a model-loading failure for physical-intelligence/fast." }, { "left": "issue:44479", "right": "issue:45325", - "accept": true, - "reason": "Same Qwen2.5-VL video/temporal-position regression; 45325 appears to pinpoint the root cause behind the broader video-input regression." + "accept": false, + "reason": "Both are Qwen multimodal regressions, but one is a video-input regression and the other is still-image temporal position-id scaling; different code paths." }, { "left": "issue:36032", "right": "issue:43742", "accept": false, - "reason": "Different tokenizer/model-loading bugs with no shared concrete code path." + "reason": "T5 tokenizer method conflict and MobileLLM key error are different loading bugs." }, { "left": "issue:43296", "right": "issue:43531", "accept": false, - "reason": "PaddleOCR-VL load failure is unrelated to Qwen3-MoE sliding_window behavior." + "reason": "PaddleOCR-VL loading in vLLM is unrelated to Qwen3-MoE sliding_window behavior." }, { "left": "issue:40444", "right": "issue:45325", "accept": false, - "reason": "Different Qwen2.5-VL failures: iterable-dataset multi-image finetuning vs temporal position-id scaling." + "reason": "Multiple-image iterable training failure is different from Qwen2.5-VL still-image rope-index scaling." }, { "left": "issue:42915", "right": "issue:44589", "accept": false, - "reason": "Both mention FP8/floating-point8, but one is a config failure and the other is a missing storage type; not the same bug." + "reason": "Different failure modes: FP8 config handling vs missing Float8 storage class." }, { "left": "issue:43611", "right": "issue:44534", "accept": false, - "reason": "base_model_prefix loading regression is unrelated to non-persistent buffer corruption." + "reason": "base_model_prefix loading breakage is unrelated to junk-filled non-persistent buffers." }, { "left": "issue:43531", "right": "issue:44877", "accept": false, - "reason": "Different models and mechanisms: Qwen3-MoE sliding_window vs granite_speech config strictness." + "reason": "Sliding-window handling in Qwen3-MoE is different from strict config rejection for granite_speech." }, { - "left": "issue:40990", - "right": "issue:44811", + "left": "issue:36032", + "right": "issue:41720", "accept": false, - "reason": "Perplexity issue is unrelated to Whisper batch_decode skip_special_tokens handling." + "reason": "Tokenizer load conflict and Qwen3 device-mapping CUDA assert are unrelated." }, { - "left": "issue:36032", - "right": "issue:41720", + "left": "issue:40990", + "right": "issue:44811", "accept": false, - "reason": "Tokenizer add_special_tokens conflict is unrelated to Qwen3 auto device mapping cuda assert." + "reason": "High perplexity on WikiText and Whisper batch_decode skip_special_tokens are unrelated bugs." }, { "left": "issue:43425", "right": "issue:44403", "accept": false, - "reason": "Torch version incompatibility is not the same as a noisy load warning bug." + "reason": "Torch 2.10 compatibility and transformer loading noise are not the same problem." }, { "left": "issue:44079", "right": "issue:45663", "accept": false, - "reason": "ModelOutput key assignment and Gemma4 FSDP2 shared_kv_states are different bugs." + "reason": "ModelOutput key assignment and Gemma4 FSDP2 shared_kv_states rebuilding are different issues." }, { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "Bad AutoTokenizer error message is unrelated to Apple Silicon caching_allocator_warmup TypeError." + "reason": "Bad AutoTokenizer error messaging for Voxtral is unrelated to the Apple Silicon caching_allocator_warmup TypeError." }, { "left": "issue:43632", "right": "issue:44534", "accept": false, - "reason": "The _is_hf_initialized regression is unrelated to non-persistent buffer junk values." + "reason": "_is_hf_initialized regression and junk non-persistent buffers are different Transformers v5 bugs." }, { "left": "issue:42175", "right": "issue:45542", - "accept": true, - "reason": "Same packaging/backend problem: TensorFlow is missing from the installed backend path and triggers TF-backend errors." + "accept": false, + "reason": "Both involve TensorFlow/backend packaging, but one is missing TF in the torch extra and the other is a runtime undefined-backend error when only tensorboard is installed." }, { "left": "issue:43818", "right": "issue:45357", "accept": false, - "reason": "Video-LLaVA temporal-attention/weight-sharing bug is unrelated to Qwen3.5 save_pretrained key issues." + "reason": "Video-LLaVA tower/temporal-attention bug is unrelated to Qwen3.5 save_pretrained visual-key serialization." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is mostly grounded in the packet, and the rejected soft pairs are conservative. The accepted Qwen3.5 save_pretrained and TensorFlow-backend pairs are well supported by the titles; the Qwen2.5-VL video/position-id pair is somewhat inferential but still plausibly related rather than overstated. Overall this looks acceptable." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -3233,46 +3232,46 @@ "issue:44462|issue:45812", "issue:33666|issue:40444", "issue:43792|issue:44589", - "issue:44829|issue:45479", "issue:44683|issue:45020", + "issue:44829|issue:45479", "issue:44403|issue:44991", "issue:43381|issue:45923", "issue:43482|issue:45538", "issue:35707|issue:40444", "issue:44079|issue:44534", - "issue:43482|issue:43525", "issue:44291|issue:44589", + "issue:43482|issue:43525", "issue:43377|issue:45491", "issue:43749|issue:44079", - "issue:43644|issue:44079", "issue:40444|issue:45381", + "issue:43644|issue:44079", "issue:43531|issue:44291", "issue:41720|issue:45081", "issue:44265|issue:45072", "issue:43632|issue:45072", "issue:45325|issue:45381", - "issue:43611|issue:43646", "issue:43994|issue:44898", + "issue:43611|issue:43646", "issue:42175|issue:43296", "issue:43454|issue:44466", "issue:45325|issue:45834", "issue:43792|issue:43957", "issue:42915|issue:43482" ], - "prepared_review_unit_hash": "5c1d7744cef7c84d42f208d29d7279fa26c0d5f216709601a725b6779d5f0fb3", + "prepared_review_unit_hash": "3ba9e645c1e1ab0f5115603cefa6b13e76de31999cd0a56d6135056f84d3e84f", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36832, + "serialized_chars": 36830, "estimated_input_tokens": 9208, "estimated_eval_tokens": 18672 }, @@ -3281,198 +3280,198 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose set of transformer regression issues across loading, tokenizers, Qwen/VL, attention, and model-init behavior. I see one strong near-duplicate candidate around Qwen2.5-VL position/rope indexing, but the rest are distinct bugs.", - "confidence": 0.79, - "canonical_issue_reason": "Issue 43611 is the broadest representative of the shared theme: Transformers v5 model-loading/init regressions. It is more central to the cluster than the model-specific tokenizer, vision, or training bugs.", + "summary": "This set is dominated by Transformers v5/model-loading regression reports, but most pairs are model-specific or subsystem-specific rather than true duplicates. Only the Qwen2.5-VL position-id pair looks like the same underlying bug.", + "confidence": 0.78, + "canonical_issue_reason": "issue:43646 is the broadest representative of the cluster\u2019s main theme: Transformers v5 breaking model initialization/loading paths. Most other items are narrower variants or unrelated bugs.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43611 is the best single issue to anchor this cluster because it captures the common v5 loading/init breakage pattern that recurs across several items.", + "best_issue_reason": "issue:43646 is the best single issue to represent the cluster because it is broad, central, and closest to the recurring v5 initialization/loading regression pattern across the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43317", "right": "issue:43856", "accept": false, - "reason": "Both involve memory/device behavior, but one is gpu/cpu offload loading and the other is Qwen3 MoE training memory usage; not the same bug." + "reason": "Different failures: device offload/dequantized loading vs Qwen3 MoE training memory usage." }, { "left": "issue:43421", "right": "issue:44568", "accept": false, - "reason": "Both concern tokenizer special-token handling, but one is runtime post-processor updating and the other is add_special_tokens behavior for a specific tokenizer." + "reason": "Both involve tokenizer special tokens, but one is runtime post-processor updates and the other is add_special_tokens/BOS-EOS insertion for a specific tokenizer." }, { "left": "issue:44462", "right": "issue:45812", "accept": false, - "reason": "Both are AutoTokenizer regressions, but they describe different failures: ignoring tokenizer.json vs wrong token IDs for Granite models." + "reason": "Both are tokenizer-loading regressions, but they affect different behaviors and models; not the same concrete bug." }, { "left": "issue:33666", "right": "issue:40444", "accept": false, - "reason": "Different problem scopes: multi-GPU training for Qwen2-VL vs iterable-dataset failures with multiple images per prompt." + "reason": "Different Qwen-VL training problems: multi-GPU training vs iterable dataset with multiple images per prompt." }, { "left": "issue:43792", "right": "issue:44589", "accept": false, - "reason": "One is a model runtime failure, the other is a float8 storage lookup/type error; no clear shared code-path." + "reason": "Different root causes: Whisper model runtime failure vs missing Float8 storage type." }, { - "left": "issue:44829", - "right": "issue:45479", + "left": "issue:44683", + "right": "issue:45020", "accept": false, - "reason": "Both affect training, but one is flash_attention_3 degeneracy and the other is a classification loss bug for num_labels=1." + "reason": "Unrelated: compiled flex_attention on torch>=2.9 vs remote_code model breakage." }, { - "left": "issue:44683", - "right": "issue:45020", + "left": "issue:44829", + "right": "issue:45479", "accept": false, - "reason": "Torch flex_attention compilation failure and remote_code loading regressions are unrelated issues." + "reason": "Both mention degenerate training, but one is flash_attention_3 behavior and the other is a classification-loss setup bug." }, { "left": "issue:44403", "right": "issue:44991", "accept": false, - "reason": "Generic loading noise vs tokenizer loading regression for one specific model family; not the same bug." + "reason": "Different issues: noisy loading logs vs tokenizer loading failure for a specific model." }, { "left": "issue:43381", "right": "issue:45923", "accept": false, - "reason": "Gradient checkpointing in eval mode and a missing supports_gradient_checkpointing flag are different issues." + "reason": "One is gradient checkpointing in eval mode; the other is a missing supports_gradient_checkpointing flag on a remote-code model." }, { "left": "issue:43482", "right": "issue:45538", "accept": false, - "reason": "GGUF loading failure and CLIPTokenizer max-length behavior are unrelated." + "reason": "Different tokenizer/model-loading regressions with no clear shared code-path." }, { "left": "issue:35707", "right": "issue:40444", "accept": false, - "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to iterable dataset multi-image finetuning failures." + "reason": "Different generation/training contexts and different multimodal models." }, { "left": "issue:44079", "right": "issue:44534", "accept": false, - "reason": "ModelOutput key assignment and non-persistent buffer junk are different internal behaviors." + "reason": "ModelOutput key assignment vs non-persistent buffer initialization are different bugs." }, { - "left": "issue:43482", - "right": "issue:43525", + "left": "issue:44291", + "right": "issue:44589", "accept": false, - "reason": "GGUF loading failure and missing pad_token_id on Llama4Config are different model/config regressions." + "reason": "Different loading failures: unexpected _is_hf_initialized arg vs Float8 storage lookup." }, { - "left": "issue:44291", - "right": "issue:44589", + "left": "issue:43482", + "right": "issue:43525", "accept": false, - "reason": "Unexpected _is_hf_initialized argument during init_empty_weights is not the same as the float8 storage lookup failure." + "reason": "Different v5 loading failures; one is GGUF loading, the other is missing pad_token_id on Llama4Config." }, { "left": "issue:43377", "right": "issue:45491", "accept": false, - "reason": "Different bugs: MIMI batching/padding-mask support vs Gemma3 NaN embeddings with sliding-window attention." + "reason": "Both are input-shape/masking issues, but one is MIMI encoder padding masks and the other is Gemma3 sliding-window all-padding NaNs." }, { "left": "issue:43749", "right": "issue:44079", "accept": false, - "reason": "FSDP CPU RAM efficient loading and ModelOutput key assignment are unrelated." + "reason": "FSDP CPU-efficient loading is unrelated to ModelOutput key handling." }, { - "left": "issue:43644", - "right": "issue:44079", + "left": "issue:40444", + "right": "issue:45381", "accept": false, - "reason": "Non-persistent buffer junk and ModelOutput key assignment are separate regressions." + "reason": "Different Qwen2.5-VL problems: finetuning with multiple images vs wrong video position ids." }, { - "left": "issue:40444", - "right": "issue:45381", + "left": "issue:43644", + "right": "issue:44079", "accept": false, - "reason": "Multi-image finetuning failure is unrelated to Qwen2.5-VL video position-id computation." + "reason": "Non-persistent buffer junk vs ModelOutput key assignment are unrelated bugs." }, { "left": "issue:43531", "right": "issue:44291", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and init_empty_weights argument handling are unrelated." + "reason": "Qwen3-MoE sliding_window bug and init_empty_weights/_is_hf_initialized bug are different failure modes." }, { "left": "issue:41720", "right": "issue:45081", "accept": false, - "reason": "Qwen3 auto device-mapping cuda assert and Mistral tokenizer regex crash are different code paths." + "reason": "Qwen3 device-mapping cuda assert is unrelated to Mistral tokenizer regex patching." }, { "left": "issue:44265", "right": "issue:45072", "accept": false, - "reason": "torch.export/torch_compilable_check failure and dtype mismatch inference bugs are unrelated." + "reason": "torch.export with torch_compilable_check is unrelated to dtype mismatches in inference." }, { "left": "issue:43632", "right": "issue:45072", "accept": false, - "reason": "_is_hf_initialized breakage and bfloat16 dtype mismatches are different v5 regressions." + "reason": "_is_hf_initialized regression is a loading/init issue, not the same as dtype mismatch in bfloat16 inference." }, { "left": "issue:45325", "right": "issue:45381", "accept": true, - "reason": "Both point to the same Qwen2.5-VL position-id/rope-index computation in v5.3.0, affecting still-image and video paths in the same model family." + "reason": "Both describe wrong Qwen2.5-VL position-id generation in transformers 5.3.0; they likely share the same get_rope_index/rope-index code path." }, { - "left": "issue:43611", - "right": "issue:43646", + "left": "issue:43994", + "right": "issue:44898", "accept": false, - "reason": "Both are v5 loading/init regressions, but one is specifically about base_model_prefix loading and the other about custom model initialization; not clearly the same bug." + "reason": "Different vision model issues: nonsensical SigLIP2 outputs vs Perceiver non-default resolution failure." }, { - "left": "issue:43994", - "right": "issue:44898", + "left": "issue:43611", + "right": "issue:43646", "accept": false, - "reason": "SigLIP2 nonsensical outputs and Perceiver non-default resolution failure are distinct model-specific bugs." + "reason": "Both are v5 loading regressions, but the specific problems differ: base_model_prefix handling vs custom model initialization." }, { "left": "issue:42175", "right": "issue:43296", "accept": false, - "reason": "TensorFlow backend packaging and PaddleOCR-VL loading are unrelated." + "reason": "Packaging/backend dependency issue vs a model load failure in vLLM are unrelated." }, { "left": "issue:43454", "right": "issue:44466", "accept": false, - "reason": "Both involve lm_head/tied weights, but one is a specific model not tying weights and the other is inconsistent serialization behavior; too different to merge as one bug." + "reason": "One is tied lm_head weights in AyaVision; the other is device-dependent lm_head serialization in v5." }, { "left": "issue:45325", "right": "issue:45834", "accept": false, - "reason": "Qwen2.5-VL rope/position-id bug and Kosmos2.5 OCR long-input index error are different issues." + "reason": "Wrong Qwen2.5-VL position ids is unrelated to Kosmos2.5 OCR long-input index errors." }, { "left": "issue:43792", "right": "issue:43957", "accept": false, - "reason": "Whisper model runtime failure and meta-device loading regressions are not the same underlying bug." + "reason": "Both involve model loading on transformers 5+, but the concrete failures and affected code paths differ." }, { "left": "issue:42915", "right": "issue:43482", "accept": false, - "reason": "Qwen3MoE FP8 config failure and GGUF loading failure are unrelated." + "reason": "Different model families and different version-specific failures: FineGrainedFP8Config vs GGUF loading." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded and conservative overall. The summary correctly treats the cluster as mostly heterogeneous, and the single accepted soft pair (45325/45381) is a plausible near-duplicate based on the shared Qwen2.5-VL position-id/rope-index theme. The rest of the pair rejections are appropriately cautious." + "feedback": "The summary is grounded in the packet: many items are indeed v5/model-loading regressions, and the only accepted duplicate pair (45325/45381) is a conservative match. The rejected soft-edge verdicts generally stay within the evidence from titles and avoid overclaiming sameness." }, "evaluator_used": true, "retried": false, @@ -3550,8 +3549,8 @@ "issue:43881|issue:44488", "issue:43262|issue:43746", "issue:44291|issue:44877", - "issue:44568|issue:45920", "issue:43606|issue:44265", + "issue:44568|issue:45920", "issue:40990|issue:45584", "issue:34634|issue:35707", "issue:43873|issue:45701", @@ -3562,219 +3561,219 @@ "issue:43454|issue:43550", "issue:41720|issue:45823", "issue:43377|issue:44610", + "issue:39692|issue:40444", "issue:43454|issue:45072", "issue:43377|issue:44898", - "issue:39692|issue:40444", - "issue:43572|issue:45759", "issue:43257|issue:45923", + "issue:43572|issue:45759", "issue:40990|issue:42947", "issue:43526|issue:43883", - "issue:43749|issue:44534", "issue:44466|issue:45357", + "issue:43749|issue:44534", "issue:43454|issue:43720" ], - "prepared_review_unit_hash": "0e9a59f2d901a84ce7938c90e628991f9a9f2781711b7a177650286122d4159f", + "prepared_review_unit_hash": "69b2aa48d5f5c75d7f719fccec54f8edaf87002f65c1a7e82c629fdaddaf43ed", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36587, - "estimated_input_tokens": 9147, - "estimated_eval_tokens": 18550 + "serialized_chars": 36584, + "estimated_input_tokens": 9146, + "estimated_eval_tokens": 18548 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are distinct regressions around loading, offloading, tokenization, config handling, or model-specific multimodal behavior. The soft pairs share broad subsystem vocabulary but not the same concrete bug, so none should be merged as duplicates.", - "confidence": 0.72, - "canonical_issue_reason": "issue:43873 is the broadest representative of the offloading/quantization subgroup and is open, but it is not a true duplicate umbrella for the whole cluster.", + "summary": "The soft pairs are mostly superficially similar but not true duplicates: they span unrelated model-specific loading, tokenizer, offloading, multimodal, and serialization bugs. I would not merge any of the proposed soft pairs.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 43873 is the broadest representative of the offloading/quantization subtheme, but the overall set is heterogeneous and does not form a single duplicate cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the best single representative if one issue must be chosen because it covers the most central repeated theme here: offloading/device-map behavior under quantization.", + "best_issue_reason": "Issue 43873 is the closest thing to a cluster representative because it is the most generic, broadly worded bug in the set, centered on offloading/quantization behavior.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44387", "right": "issue:44617", "accept": false, - "reason": "Both mention CUDA/OOM, but one is a Transformers 5.x int4 quantization memory regression and the other is a model-specific Sam3Video OOM. Different root causes." + "reason": "Both mention OOM, but one is int4 quantization memory growth and the other is a Sam3Video inference OOM; different failures and code paths." }, { "left": "issue:43296", "right": "issue:45800", "accept": false, - "reason": "PaddleOCR-VL loading in vLLM/Transformers 4.53 is unrelated to a torch 2.4.1 vs Transformers 5.8 incompatibility. Same area, different bug." + "reason": "One is a PaddleOCR-VL model load failure in vLLM, the other is a torch/transformers version incompatibility; not the same bug." }, { "left": "issue:43701", "right": "issue:45325", "accept": false, - "reason": "Checkpoint resume key mismatch and Qwen2.5-VL rope position scaling are unrelated code paths." + "reason": "Checkpoint resume key mismatch and Qwen2.5-VL rope/position-id scaling are unrelated issues." }, { "left": "issue:44479", "right": "issue:45823", "accept": false, - "reason": "Both are multimodal/device issues, but one is a video-input regression in Qwen variants and the other is a Gemma4 PLE device mismatch during forward. Not the same bug." + "reason": "Both are multimodal regressions involving device/multimodal handling, but they affect different models and different failure modes." }, { "left": "issue:43881", "right": "issue:44488", "accept": false, - "reason": "GLM-4V-9B loading failure and failure to load cjvt/sleng-bert are unrelated model-loading problems." + "reason": "Both are model loading problems, but they concern different models and distinct root causes." }, { "left": "issue:43262", "right": "issue:43746", "accept": false, - "reason": "Audio chat-template sampling-rate default mismatch is unrelated to GraniteSpeech PEFT local-checkpoint loading." + "reason": "Audio chat-template sampling-rate defaulting and GraniteSpeech PEFT checkpoint loading are unrelated." }, { "left": "issue:44291", "right": "issue:44877", "accept": false, - "reason": "Unexpected _is_hf_initialized in init_empty_weights and strict config rejecting granite_speech are both config-related, but they are different failure modes." + "reason": "Both involve loading/config behavior in v5, but one is an init_empty_weights argument mismatch and the other is strict config validation; not the same bug." }, { - "left": "issue:44568", - "right": "issue:45920", + "left": "issue:43606", + "right": "issue:44265", "accept": false, - "reason": "A specific add_special_tokens tokenizer regression is not the same as a broader AutoTokenizer wrong-token-ids regression across many models." + "reason": "CPU offload device mismatch and torch.export failing on torch_compilable_check are different code paths." }, { - "left": "issue:43606", - "right": "issue:44265", + "left": "issue:44568", + "right": "issue:45920", "accept": false, - "reason": "CPU offload device mismatch for BarkSmall is unrelated to torch.export.export failures with torch_compilable_check." + "reason": "One is a specific add_special_tokens regression for a tokenizer, the other is broad wrong-token-id behavior across several models." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "High perplexity on gpt-oss-20b and Whisper empty-transcription failure after align_special_tokens are unrelated." + "reason": "High perplexity on gpt-oss and Whisper empty-transcription generation failure are unrelated." }, { "left": "issue:34634", "right": "issue:35707", "accept": false, - "reason": "BarkProcessor voice_preset and progressive generation with inputs_embeds/past_key_values are unrelated." + "reason": "Bark voice_preset and progressive generation with inputs_embeds/past_key_values are different bugs." }, { "left": "issue:43873", "right": "issue:45701", "accept": false, - "reason": "Offloading with quantization and tokenization changing across versions are different bugs." + "reason": "Offloading/quantization issues are not the same as tokenization changes across versions." }, { "left": "issue:44186", "right": "issue:44265", "accept": false, - "reason": "LayoutLMv2 tokenizer crashes on NER/padding are unrelated to torch.export and torch_compilable_check." + "reason": "Tokenizer crashes on NER/padding versus export-time torch_compilable_check failure; no shared underlying bug." }, { "left": "issue:43317", "right": "issue:43873", "accept": false, - "reason": "Both involve offloading/quantization, but one is a dequantized-model load failure on gpu+cpu offload and the other is a broader quantized offloading issue. Too different to merge." + "reason": "Both involve offload/quantization language, but one is dequantized model loading with gpu+cpu offload and the other is a generic offloading complaint; too broad to merge." }, { "left": "issue:35707", "right": "issue:41720", "accept": false, - "reason": "inputs_embeds/past_key_values generation and Qwen3 auto device mapping cudaErrorAssert are unrelated." + "reason": "Progressive generation with cached embeddings and a Qwen3 cudaErrorAssert under auto device mapping are unrelated." }, { "left": "issue:43606", "right": "issue:43819", "accept": false, - "reason": "Bark CPU-offload device mismatch and DAC.from_latents/forward STE mismatch are unrelated model bugs." + "reason": "Bark CPU offload device mismatch and DAC from_latents vs forward mismatch are different model behaviors." }, { "left": "issue:43454", "right": "issue:43550", "accept": false, - "reason": "lm_head weight tying failure in AyaVisionConfig is unrelated to a Bamba torch.compile SDPA failure." + "reason": "lm_head tying/garbage generation and torch.compile SDPA failure are unrelated." }, { "left": "issue:41720", "right": "issue:45823", "accept": false, - "reason": "Both mention device_map/forward failures, but the models and concrete failure modes differ completely." + "reason": "Both mention device mapping, but they affect different models and likely different backend faults." }, { "left": "issue:43377", "right": "issue:44610", "accept": false, - "reason": "MIMI padding-mask batching bug and OmDet-Turbo processor/model image-size mismatch are unrelated." + "reason": "Missing padding-mask support in MIMI and OmDet-Turbo processor/model size mismatch are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "Doc-example model/processor mismatch plus quantization failure is not the same as Qwen2.5-VL multi-image finetuning failure." }, { "left": "issue:43454", "right": "issue:45072", "accept": false, - "reason": "Tied-weights serialization and bf16 dtype mismatches in SwitchTransformers/TimmWrapperModel are different issues." + "reason": "Weight-tying serialization and bfloat16 dtype mismatch are different problems." }, { "left": "issue:43377", "right": "issue:44898", "accept": false, - "reason": "Missing padding mask support in MIMI encoder is unrelated to Perceiver interpolation/position-encoding behavior." + "reason": "Both are shape/batching-related vision issues, but one is padding-mask support and the other is resolution/interpolation behavior." }, { - "left": "issue:39692", - "right": "issue:40444", + "left": "issue:43257", + "right": "issue:45923", "accept": false, - "reason": "SigLIP2 doc example errors and Qwen2.5-VL iterable-dataset multi-image finetuning failures are distinct multimodal bugs." + "reason": "MOE weight conversion during accelerate/deepspeed loading and a missing supports_gradient_checkpointing flag are unrelated." }, { "left": "issue:43572", "right": "issue:45759", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig and failure to unwrap text_config in composite Qwen multimodal configs are not the same bug." - }, - { - "left": "issue:43257", - "right": "issue:45923", - "accept": false, - "reason": "Qwen3 MOE loading with accelerate+deepspeed and missing supports_gradient_checkpointing on a trust_remote_code variant are unrelated." + "reason": "StableLm pad_token_idx config regression and composite Qwen config unwrapping are different config-loading bugs." }, { "left": "issue:40990", "right": "issue:42947", "accept": false, - "reason": "High perplexity on a language model benchmark and gradient checkpointing ineffective with PEFT LoRA are unrelated." + "reason": "Perplexity regression and gradient checkpointing ineffectiveness are unrelated." }, { "left": "issue:43526", "right": "issue:43883", "accept": false, - "reason": "BeitImageProcessorFast reduce_labels bug and missing all_tied_weights_keys on MolmoForCausalLM are unrelated." + "reason": "BeitImageProcessorFast label reduction and missing all_tied_weights_keys are unrelated." }, { - "left": "issue:43749", - "right": "issue:44534", + "left": "issue:44466", + "right": "issue:45357", "accept": false, - "reason": "FSDP CPU RAM efficient loading and non-persistent buffer initialization in v5 are different loading/runtime issues." + "reason": "Both touch serialization, but one is lm_head weight behavior and the other is incorrect visual encoder keys; not the same issue." }, { - "left": "issue:44466", - "right": "issue:45357", + "left": "issue:43749", + "right": "issue:44534", "accept": false, - "reason": "Inconsistent lm_head.weight serialization and incorrect visual encoder keys in Qwen3.5 save_pretrained are both serialization bugs, but not the same one." + "reason": "FSDP CPU RAM-efficient loading and non-persistent buffer initialization are different v5 regressions." }, { "left": "issue:43454", "right": "issue:43720", "accept": false, - "reason": "AyaVision tied-weights bug and BitNet packed-weight unpacking during accelerate loading are unrelated." + "reason": "Weight tying in AyaVision and packed-weight unpacking in BitNet loading are unrelated." } ] }, @@ -3853,15 +3852,15 @@ "issue:43296|issue:44877", "issue:41720|issue:44977", "issue:44008|issue:45823", - "issue:40990|issue:45701", - "issue:43404|issue:45357", "issue:45632|issue:45684", - "issue:44186|issue:45072", + "issue:43404|issue:45357", + "issue:40990|issue:45701", "issue:42617|issue:42915", + "issue:44186|issue:45072", "issue:43873|issue:44756", "issue:43577|issue:43708", - "issue:44423|issue:45406", "issue:44387|issue:45704", + "issue:44423|issue:45406", "issue:36331|issue:42617", "issue:44877|issue:44991", "issue:43550|issue:43720", @@ -3875,232 +3874,232 @@ "issue:43316|issue:44913", "issue:42915|issue:43317", "issue:43295|issue:45701", - "issue:44671|issue:44743", "issue:43526|issue:43881", + "issue:44671|issue:44743", "issue:45081|issue:45812", "issue:41720|issue:45161", "issue:43531|issue:43572" ], - "prepared_review_unit_hash": "1500bd448ba6d87adf9076077151f09c1895c230dd3693b645229f6ce631a45f", + "prepared_review_unit_hash": "1b1c244f137251e30d2079577daaaaf0da1ea5a4e61fc8eb570daa568bf9d06a", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37424, - "estimated_input_tokens": 9356, - "estimated_eval_tokens": 18968 + "serialized_chars": 37427, + "estimated_input_tokens": 9357, + "estimated_eval_tokens": 18970 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The items are mostly unrelated standalone bugs across different models, processors, and trainer paths. None of the soft pairs look like the same underlying defect, so I would keep them separate. If forced to pick one representative issue, 45812 is the broadest active regression in the set.", - "confidence": 0.93, - "canonical_issue_reason": null, + "summary": "The cluster is loosely centered on Transformers regressions around loading, offloading, quantization, tokenizer/config handling, and multimodal processors, but the soft pairs mostly describe different concrete failures rather than true duplicates.", + "confidence": 0.58, + "canonical_issue_reason": "issue:43873 is the broadest and most central report in the set: it is still open, has the most engagement/inbound references, and sits closest to the recurring offloading/quantization theme shared by several nearby issues.", "canonical_pr_reason": null, - "best_issue_reason": "45812 is the broadest active regression here: it is open, has inbound references, and affects tokenization across all Granite models.", + "best_issue_reason": "issue:43873 is the best representative issue for this cluster because it is the most general, active, and thematically central; the other issues are more model- or API-specific edge cases.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43881", "right": "issue:44991", "accept": false, - "reason": "Different model families and failure modes: GLM-4V loading vs EST-RoBERTa tokenizer loading." + "reason": "Different models and symptoms: a model loading failure vs tokenizer loading regression." }, { "left": "issue:41720", "right": "issue:43322", "accept": false, - "reason": "Both are load/runtime crashes, but they involve different models and distinct code paths (device mapping assert vs Llava Next segfault)." + "reason": "Both are loading failures, but they affect different models and different crash modes." }, { "left": "issue:43296", "right": "issue:44877", "accept": false, - "reason": "PaddleOCR-VL loading in vLLM and Granite speech strict-config loading are separate configuration/model issues." + "reason": "Both involve loading/config handling, but the models and underlying failure paths are unrelated." }, { "left": "issue:41720", "right": "issue:44977", "accept": false, - "reason": "Qwen3 auto device-mapping assert is not the same bug as Qwen3.5 flash-attention generation failure." + "reason": "Same family name, but one is a device-map assert during loading and the other is a flash-attention generation issue." }, { "left": "issue:44008", "right": "issue:45823", "accept": false, - "reason": "Different Gemma generations and different failures: variable-name collision vs device mismatch in forward." + "reason": "Different Gemma generations and different bug classes: attribute collision vs device mismatch." }, { - "left": "issue:40990", - "right": "issue:45701", + "left": "issue:45632", + "right": "issue:45684", "accept": false, - "reason": "High perplexity on GPT-OSS and tokenization changes across versions are related only loosely; not the same concrete defect." + "reason": "Both are path/file-handling related, but one is cache collision and the other is permission propagation on save." }, { "left": "issue:43404", "right": "issue:45357", "accept": false, - "reason": "Mistral3 lm_head tying and Qwen3.5 save_pretrained key serialization are unrelated bugs." + "reason": "Different models and different regressions: tied weights vs save_pretrained key mapping." }, { - "left": "issue:45632", - "right": "issue:45684", + "left": "issue:40990", + "right": "issue:45701", "accept": false, - "reason": "Cache-path collisions for trust_remote_code and permission propagation in save_pretrained are different filesystem behaviors." + "reason": "Perplexity regression and tokenization change are not the same concrete bug." }, { - "left": "issue:44186", - "right": "issue:45072", + "left": "issue:42617", + "right": "issue:42915", "accept": false, - "reason": "Tokenizer crash on NER/batched padding is unrelated to dtype mismatches in bfloat16 inference." + "reason": "Both are model execution failures, but they concern different scripts, models, and code paths." }, { - "left": "issue:42617", - "right": "issue:42915", + "left": "issue:44186", + "right": "issue:45072", "accept": false, - "reason": "3d_parallel.py execution failure and Qwen3Moe FineGrainedFP8Config failure are different model/runtime problems." + "reason": "Tokenizer/padding crash is unrelated to dtype mismatch in inference." }, { "left": "issue:43873", "right": "issue:44756", "accept": false, - "reason": "Quantization offloading issues and Strix Halo mmap OOM are both memory-related but not the same code-path bug." + "reason": "Both touch memory/offload behavior, but one is quantization offloading and the other is mmap/OOM on Strix Halo." }, { "left": "issue:43577", "right": "issue:43708", "accept": false, - "reason": "BLIP2 dtype loading bug and Trainer checkpoint-resume max_steps bug are unrelated." + "reason": "Different problem domains: dtype propagation in BLIP-2 vs Trainer checkpoint-resume math." }, { - "left": "issue:44423", - "right": "issue:45406", + "left": "issue:44387", + "right": "issue:45704", "accept": false, - "reason": "Both hit serving paths, but one is a multimodal batching crash and the other is a missing _tokenizer attribute in Gemma4Processor." + "reason": "Both mention memory pressure, but one is int4 CUDA reserved memory and the other is an apex RMSNorm leak." }, { - "left": "issue:44387", - "right": "issue:45704", + "left": "issue:44423", + "right": "issue:45406", "accept": false, - "reason": "Int4 quantization reserved-memory regression and T5 apex RMSNorm memory leak are different memory issues with different root causes." + "reason": "Both crash in serving, but the processor errors are different and likely separate code paths." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "CustomTrainer compute_loss signature breakage is unrelated to 3d_parallel.py startup failure." + "reason": "Trainer API incompatibility is unrelated to a 3d_parallel execution failure." }, { "left": "issue:44877", "right": "issue:44991", "accept": false, - "reason": "Strict config loading for granite_speech and tokenizer loading for EST-RoBERTa are separate issues." + "reason": "Config strictness and tokenizer loading regression are different failure sources." }, { "left": "issue:43550", "right": "issue:43720", "accept": false, - "reason": "torch.compile with SDPA on Bamba and BitNet packed-weight loading are unrelated model/runtime bugs." + "reason": "Both involve advanced execution settings, but one is torch.compile+SDPA and the other is accelerate loading/unpacking." }, { "left": "issue:36331", "right": "issue:42915", "accept": false, - "reason": "Trainer API signature regression does not match a Qwen3Moe FP8 loading failure." + "reason": "A Trainer signature mismatch is unrelated to a FineGrainedFP8 model load failure." }, { "left": "issue:43450", "right": "issue:45701", "accept": false, - "reason": "Batched video-shape handling and tokenization changes are different subsystems." + "reason": "Video processor shape bug does not match a tokenizer behavior regression." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "Both involve image inputs, but OmDet-Turbo uses the wrong processor size while Perceiver fails on non-default resolution handling." + "reason": "Both concern image sizes, but the models and expected resolutions differ." }, { "left": "issue:44186", "right": "issue:44610", "accept": false, - "reason": "Tokenizer padding/truncation crashes and image processor resolution mismatch are not the same defect." + "reason": "Tokenizer batching crash and image resolution mismatch are different subsystems and bugs." }, { "left": "issue:44479", "right": "issue:44918", "accept": false, - "reason": "Video-input regression across Qwen VL models and Qwen3.5 SFT embedding unpacking are different failures." + "reason": "Video input regression and embedding unpacking failure are not the same underlying issue." }, { "left": "issue:43550", "right": "issue:43606", "accept": false, - "reason": "Different models and mechanisms: SDPA/compile failure vs CPU-offload device mismatch." + "reason": "Different failure modes: compile/SDPA bug vs CPU offload device mismatch." }, { "left": "issue:41720", "right": "issue:42617", "accept": false, - "reason": "Qwen3 auto device mapping and 3d_parallel.py failure are unrelated." + "reason": "Auto device-map CUDA assert and 3d_parallel runtime failure are not the same bug." }, { "left": "issue:43316", "right": "issue:44913", "accept": false, - "reason": "Gemma3TextConfig API discrepancy and GPTNeoX rotary_pct reload bug are separate configuration issues." + "reason": "Both are config-related, but one is API inconsistency and the other is config value not persisting." }, { "left": "issue:42915", "right": "issue:43317", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure and dequantized model offload loading failure do not share the same root cause." + "reason": "Both are load/offload problems, but they target different quantization/dequantization scenarios." }, { "left": "issue:43295", "right": "issue:45701", "accept": false, - "reason": "Processor/tokenizer regression in v4.57.5 and version-dependent tokenization changes are not the same concrete bug." + "reason": "Processor/tokenizer API regression is not the same as a general tokenization change report." }, { - "left": "issue:44671", - "right": "issue:44743", + "left": "issue:43526", + "right": "issue:43881", "accept": false, - "reason": "CamemBERT masked-LM predictions and Qwen3.5 recurrent-state reset are unrelated model behaviors." + "reason": "BEiT label reduction bug is unrelated to a model loading failure." }, { - "left": "issue:43526", - "right": "issue:43881", + "left": "issue:44671", + "right": "issue:44743", "accept": false, - "reason": "BEiT label reduction bug and GLM-4V loading failure are unrelated." + "reason": "CamemBERT masked LM output bug and Qwen recurrent-state bug affect different model families and code paths." }, { "left": "issue:45081", "right": "issue:45812", "accept": false, - "reason": "Mistral regex patch crashing during tokenizer load and Granite token-ID regression are different tokenizer bugs." + "reason": "Tokenizer patch crash and Granite token-ID regression are different tokenizer bugs." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "Qwen3 auto device mapping and GPT-OSS MoE tensor-parallel-only failure are different distributed-loading issues." + "reason": "Both mention GPT/Qwen-style model work, but they are different models and different execution modes." }, { "left": "issue:43531", "right": "issue:43572", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and StableLm pad_token_idx config regression are unrelated." + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to a StableLm config missing pad_token_idx." } ] }, @@ -4181,8 +4180,8 @@ "issue:44186|issue:44361", "issue:43388|issue:45059", "issue:45406|issue:45464", - "issue:44492|issue:44509", "issue:33357|issue:35707", + "issue:44492|issue:44509", "issue:43421|issue:45701", "issue:41720|issue:43746", "issue:43606|issue:43720", @@ -4194,224 +4193,224 @@ "issue:43606|issue:43653", "issue:43531|issue:45759", "issue:34634|issue:41720", - "issue:43421|issue:43927", "issue:34567|issue:35707", + "issue:43421|issue:43927", "issue:44368|issue:45357", "issue:43782|issue:45381", "issue:44898|issue:45072", "issue:44451|issue:44991", "issue:44077|issue:44625", - "issue:44987|issue:45042", - "issue:45759|issue:45859" + "issue:45759|issue:45859", + "issue:44987|issue:45042" ], - "prepared_review_unit_hash": "d20d476bd29e5e45a83d7e8af63ccab9f99fbad6b3c59f5b4798fefce7ee4169", + "prepared_review_unit_hash": "2a21a3b4327dee18bff08e3cd8ce45de3c3f5e85db4d1a84a31d299d2c1370d5", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36471, - "estimated_input_tokens": 9118, - "estimated_eval_tokens": 18492 + "serialized_chars": 36464, + "estimated_input_tokens": 9116, + "estimated_eval_tokens": 18488 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of unrelated Transformers bug reports. The only clear duplicate-style pair is the two Qwen2_5_VLProcessor apply_chat_template crashes on batched input with padding=False; the rest share broad themes (tokenizers, Qwen, offloading, multimodal) but not the same concrete bug.", - "confidence": 0.83, - "canonical_issue_reason": "issue:44514 is the clearest canonical issue: it is the earlier report of the exact batched-input/padding=False crash, and issue:44545 appears to be the same bug phrased almost identically.", + "summary": "Most pairs are unrelated despite superficial subsystem overlap. The only clear duplicate is the two identical Qwen2_5_VLProcessor batched-input crash reports; the rest should stay separate issues.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 44545 is the best canonical representative for the Qwen2_5_VLProcessor padding=False crash: it matches 44514 exactly on title/symptom and has slightly more discussion/inbound references.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44514 is the best representative issue because it captures the only strong duplicate pair in the set and has the cleanest, most specific description of the underlying failure.", + "best_issue_reason": "Issue 44545 is the strongest issue to keep as the main reference for this duplicate bug because it is the clearest, most central report of the exact same crash.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43540", "right": "issue:43782", "accept": false, - "reason": "Both involve Qwen multimodal models, but one is a video-processing ValueError and the other is a weight_only loading error for a different model path; not the same bug." + "reason": "Different models and failure modes: video input ValueError vs from_pretrained weight_only load error." }, { "left": "issue:36010", "right": "issue:42617", "accept": false, - "reason": "GenerationMixin import failure and inability to run 3d_parallel.py are unrelated issues with no shared concrete code path." + "reason": "Unrelated problems: GenerationMixin import error vs 3d_parallel.py execution failure." }, { "left": "issue:43421", "right": "issue:45812", "accept": false, - "reason": "Both mention tokenizer behavior, but one is about runtime post-processor updates and the other is a silent token-ID regression for Granite models; different failures and fixes." + "reason": "Both concern tokenization internals, but one is runtime post-processor updates and the other is incorrect Granite token IDs; not the same bug." }, { "left": "issue:42915", "right": "issue:43322", "accept": false, - "reason": "Qwen3Moe FP8 config failure and Llava Next segmentation fault are different model-loading bugs." + "reason": "Different code paths and symptoms: Qwen3Moe FP8 config failure vs Llava Next segmentation fault." }, { "left": "issue:44186", "right": "issue:44361", "accept": false, - "reason": "Both are tokenizer/task crashes, but they affect different tokenizers and different failure modes." + "reason": "Both tokenizer-related, but different models and errors (NER/batching crash vs AttributeError on MLuke tasks)." }, { "left": "issue:43388", "right": "issue:45059", "accept": false, - "reason": "One is a metrics-gathering label truncation bug; the other is a SAM3 text/bounding-box behavior issue. Not the same underlying defect." + "reason": "Different features and failures: gather_for_metrics label handling vs SAM3 text/bbox behavior." }, { "left": "issue:45406", "right": "issue:45464", "accept": false, - "reason": "Gemma4Processor missing _tokenizer in transformers serve is unrelated to Qwen3.5 streaming chat/completions failure." - }, - { - "left": "issue:44492", - "right": "issue:44509", - "accept": false, - "reason": "A typo in cache strategies and stale docs mentioning removed pipeline tasks are separate documentation issues." + "reason": "Unrelated API crashes in different processors/models." }, { "left": "issue:33357", "right": "issue:35707", "accept": false, - "reason": "MacOS bus error with a CLIP model and progressive generation with inputs_embeds/past_key_values are unrelated." + "reason": "Completely different bugs: MacOS bus error with CLIP vs generation with inputs_embeds/past_key_values." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache strategy docs is unrelated to removed pipeline-task docs." }, { "left": "issue:43421", "right": "issue:45701", "accept": false, - "reason": "Both touch tokenization, but one is a runtime post-processor update bug and the other is a broad version-dependent tokenization change report; not the same concrete issue." + "reason": "Generic tokenization regression is too broad; this is not the same specific runtime special-token bug." }, { "left": "issue:41720", "right": "issue:43746", "accept": false, - "reason": "A800 auto-device-map cudaErrorAssert and GraniteSpeech PEFT local checkpoint loading are different loading/runtime problems." + "reason": "Different model-specific issues: device-mapping cudaErrorAssert vs PEFT local checkpoint loading." }, { "left": "issue:43606", "right": "issue:43720", "accept": false, - "reason": "CPU offload device mismatch for suno/bark-small and BitNet packed-weight unpacking during accelerate loading are both load-related, but they are distinct bugs." + "reason": "Both involve loading/offload, but one is CPU offload device mismatch and the other is packed-weight unpacking during accelerate loading." }, { "left": "issue:30064", "right": "issue:34634", "accept": false, - "reason": "Void segmentation-map processing and BarkProcessor voice_preset handling are unrelated processor bugs." + "reason": "Unrelated processors: void segmentation map handling vs Bark voice_preset." }, { "left": "issue:43550", "right": "issue:43653", "accept": false, - "reason": "torch.compile SDPA failure in Bamba-9B-v2 and BigBirdTokenizer special-token registration are unrelated." + "reason": "Different components and symptoms: torch.compile/SDPA failure vs tokenizer special-token registration." }, { "left": "issue:43873", "right": "issue:45704", "accept": false, - "reason": "Quantization/offloading behavior and a T5 apex memory leak are different issues." + "reason": "Quantization offloading issue and apex RMSNorm memory leak are unrelated." }, { "left": "issue:44514", "right": "issue:44545", "accept": true, - "reason": "Same processor, same batched-input case, same padding=False crash; the titles and described failure align closely enough to be duplicates." + "reason": "Exact same Qwen2_5_VLProcessor.apply_chat_template batched-input crash with padding=False; these are duplicates of the same bug." }, { "left": "issue:44625", "right": "issue:45479", "accept": false, - "reason": "Both concern classification configuration, but one is Qwen3.5 num_labels propagation and the other is a generic zero-loss problem with num_labels=1; different root causes." + "reason": "Different classification bugs: config propagation in Qwen3.5 vs degenerate zero-loss behavior across sequence-classification models." }, { "left": "issue:43606", "right": "issue:43653", "accept": false, - "reason": "The CPU offload device mismatch and the BigBirdTokenizer special-token/decode issue are entirely different bugs." + "reason": "Device mismatch on CPU offload is unrelated to BigBirdTokenizer special-token decode behavior." }, { "left": "issue:43531", "right": "issue:45759", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE and from_config not unwrapping text_config in composite Qwen multimodal configs are not the same bug." + "reason": "Both mention Qwen3-family configs, but one is sliding_window behavior and the other is text_config unwrapping in from_config." }, { "left": "issue:34634", "right": "issue:41720", "accept": false, - "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping cudaErrorAssert are unrelated." + "reason": "No shared underlying bug: BarkProcessor voice_preset vs Qwen3 auto device mapping CUDA assert." }, { - "left": "issue:43421", - "right": "issue:43927", + "left": "issue:34567", + "right": "issue:35707", "accept": false, - "reason": "Special-token post-processor updates and DiaConfig losing custom token IDs after save/load are separate tokenizer/config issues." + "reason": "TrainerState token counting and progressive generation are unrelated." }, { - "left": "issue:34567", - "right": "issue:35707", + "left": "issue:43421", + "right": "issue:43927", "accept": false, - "reason": "TrainerState token counting and progressive generation with past_key_values are unrelated." + "reason": "Both involve token IDs/special tokens, but one is runtime post-processing and the other is config save/load corruption." }, { "left": "issue:44368", "right": "issue:45357", "accept": false, - "reason": "A warning about tie_word_embeddings during LoRA fine-tuning and incorrect visual encoder keys saved by Qwen3.5 are different problems." + "reason": "A warning about tie_word_embeddings is unrelated to incorrect visual encoder keys on save_pretrained." }, { "left": "issue:43782", "right": "issue:45381", "accept": false, - "reason": "Qwen3VL weight_only loading failure and Qwen2.5-VL video vision_position_ids are both VL-related but not the same code-path bug." + "reason": "Different Qwen VL bugs: weight_only load failure vs wrong video vision_position_ids." }, { "left": "issue:44898", "right": "issue:45072", "accept": false, - "reason": "Perceiver interpolation failure and bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel are unrelated." + "reason": "Different model classes and failures: Perceiver interpolation bug vs bfloat16 dtype mismatch." }, { "left": "issue:44451", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading regressions, but they affect different models and likely different failure mechanisms." + "reason": "Both are loading failures, but they affect different models and likely different tokenizer code paths; not enough evidence of the same bug." }, { "left": "issue:44077", "right": "issue:44625", "accept": false, - "reason": "PatchTSMixer post_init policy and Qwen3.5 num_labels propagation are unrelated." + "reason": "Optional post_init policy and num_labels propagation are unrelated." }, { - "left": "issue:44987", - "right": "issue:45042", + "left": "issue:45759", + "right": "issue:45859", "accept": false, - "reason": "Loading physical-intelligence/fast and PIL processors requiring torchvision are separate regressions." + "reason": "Different Qwen3.5 multimodal issues: config unwrapping vs CPU-offload rotary embedding compatibility." }, { - "left": "issue:45759", - "right": "issue:45859", + "left": "issue:44987", + "right": "issue:45042", "accept": false, - "reason": "Composite config unwrapping and Qwen3_5MoeTextRotaryEmbedding CPU-offload incompatibility are different issues." + "reason": "Unrelated regressions: model loading failure vs PIL image processors requiring torchvision." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded and conservative overall. The summary correctly centers the only clearly duplicate-looking pair in the packet, and the soft-edge rejections stay appropriately cautious without overstating evidence. The canonical issue choice is also supported by the earlier timestamp and near-identical title." + "feedback": "The summary is grounded in the packet: 44514 and 44545 are the same Qwen2_5_VLProcessor.apply_chat_template batched-input/padding=False crash, and 44545 does have slightly more discussion/inbound references. The other soft-edge rejections are cautious and consistent with the issue titles; no pair is overstated as a duplicate without clear evidence." }, "evaluator_used": true, "retried": false, @@ -4487,15 +4486,15 @@ "issue:44918|issue:44936", "issue:44062|issue:44451", "issue:45663|issue:45923", - "issue:43421|issue:43653", "issue:39692|issue:44610", + "issue:43421|issue:43653", "issue:45588|issue:45759", "issue:42886|issue:44843", "issue:43381|issue:43844", "issue:43421|issue:44186", "issue:43994|issue:44671", - "issue:43525|issue:43531", "issue:45704|issue:45859", + "issue:43525|issue:43531", "issue:45081|issue:45356", "issue:39692|issue:43873", "issue:45356|issue:45907", @@ -4513,221 +4512,224 @@ "issue:43335|issue:43756", "issue:43295|issue:45834" ], - "prepared_review_unit_hash": "8fa040c68357dbf4ab15dd57a2af7090a999e9abf41cc6c7cb3d430755d737ec", + "prepared_review_unit_hash": "f44e7b8298df2b14fc752c6b63af296954196354218b8a0461b8f0bb2556289c", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 37147, - "estimated_input_tokens": 9287, - "estimated_eval_tokens": 18830 + "serialized_chars": 37150, + "estimated_input_tokens": 9288, + "estimated_eval_tokens": 18832 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: the soft pairs mostly share only broad themes like tokenizers, multimodal models, or offline loading, but they do not describe the same concrete bug or change. I would not merge any of the candidate pairs.", - "confidence": 0.94, - "canonical_issue_reason": null, + "summary": "This cluster is mostly a loose semantic grouping of unrelated Hugging Face/Transformers issues. I found one plausible duplicate pair around offline tokenizer loading; the rest are similar only at the subsystem level and should stay separate.", + "confidence": 0.78, + "canonical_issue_reason": "Issue 44843 is the most concrete root-cause report in the only plausible duplicate pair: it identifies the exact offline failure path in `_patch_mistral_regex` / `model_info()`, whereas 42886 is the broader symptom report.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44843 is the clearest single report: it names a specific failing path (`_patch_mistral_regex` / `model_info()`), gives a focused offline-mode regression, and is more actionable than the rest of the mixed bag.", + "best_issue_reason": "44843 is the best representative issue because it pins the bug to a specific code path and failure mode, making it the most useful canonical anchor for triage.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45704", "right": "issue:45923", "accept": false, - "reason": "Different failures: T5 RMSNorm memory leak vs missing `supports_gradient_checkpointing` on Nemotron-3-Nano-Omni." + "reason": "Different areas: T5 RMSNorm/apex memory leak vs missing gradient_checkpointing flag on a Nemotron remote-code variant." }, { "left": "issue:43531", "right": "issue:44589", "accept": false, - "reason": "Unrelated bugs: Qwen3-MoE `sliding_window` behavior vs a Float8 storage import/type error." + "reason": "Unrelated failures: Qwen3-MoE sliding-window logic vs a Float8 storage deserialization error." }, { "left": "issue:44918", "right": "issue:44936", "accept": false, - "reason": "Different trainer failures: Qwen3.5 embedding unpacking in SFT vs `trainer.evaluate()` after `trainer.train()`." + "reason": "Both are trainer-related, but the concrete bugs differ: SFT embedding unpacking vs evaluate() failing after train()." }, { "left": "issue:44062", "right": "issue:44451", "accept": false, - "reason": "`AddedToken` constructor argument collision is unrelated to ScandiBERT model loading failure." + "reason": "Tokenizer construction bug vs model loading failure for a specific checkpoint; no shared underlying code path is evident." }, { "left": "issue:45663", "right": "issue:45923", "accept": false, - "reason": "Gemma-4 FSDP2 `shared_kv_states` KeyError is unrelated to Nemotron gradient-checkpointing flag." + "reason": "Different model families and mechanisms: Gemma-4 FSDP2 shared_kv_states bug vs Nemotron gradient_checkpointing metadata." }, { - "left": "issue:43421", - "right": "issue:43653", + "left": "issue:39692", + "right": "issue:44610", "accept": false, - "reason": "Both concern tokenizers, but one is runtime post-processor updates and the other is BigBird mask-token registration/empty decode." + "reason": "Documentation/example mismatch vs processor/model image-size mismatch; same broad vision space, but not the same bug." }, { - "left": "issue:39692", - "right": "issue:44610", + "left": "issue:43421", + "right": "issue:43653", "accept": false, - "reason": "Different multimodal problems: SigLIP2 doc example/quantization issues vs OmDet-Turbo processor size mismatch." + "reason": "Both involve tokenizer special tokens, but one is runtime post-processor updates and the other is BigBird mask-token registration/decoding." }, { "left": "issue:45588", "right": "issue:45759", "accept": false, - "reason": "Flash-attention `s_aux=None` crash is unrelated to Qwen multimodal `text_config` unwrapping in `from_config`." + "reason": "Flash-attention sinkless-model crash and composite config unwrapping are unrelated code paths." }, { "left": "issue:42886", "right": "issue:44843", - "accept": false, - "reason": "Both mention offline/tokenizer loading, but 44843 is a specific `_patch_mistral_regex`/`model_info()` bug while 42886 is a broader cache-load failure." + "accept": true, + "reason": "Both describe the same offline tokenizer-loading failure; 44843 names the concrete root cause in `_patch_mistral_regex`/`model_info()`." }, { "left": "issue:43381", "right": "issue:43844", "accept": false, - "reason": "Gradient checkpointing eval-mode restriction is unrelated to ZeRO-3 training gradient blow-up." + "reason": "Gradient checkpointing in eval mode and ZeRO-3 gradient growth are different training-time issues." }, { "left": "issue:43421", "right": "issue:44186", "accept": false, - "reason": "Runtime special-token/post-processor update is not the same as LayoutLMv2 NER padding/truncation crashes." + "reason": "Both are tokenizer-related, but the bugs are different: runtime special-token updates vs LayoutLMv2 NER/padding crashes." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "Different model-specific output quality regressions: SigLIP2 inference vs CamemBERT masked LM predictions." + "reason": "Different model behaviors: SigLIP2 AutoModel/pipeline mismatch vs CamemBERT masked-LM prediction regression." }, { - "left": "issue:43525", - "right": "issue:43531", + "left": "issue:45704", + "right": "issue:45859", "accept": false, - "reason": "Missing `pad_token_id` on `Llama4Config` is unrelated to Qwen3-MoE `sliding_window` behavior." + "reason": "T5 apex RMSNorm memory leak and Qwen3_Moe CPU-offload incompatibility are unrelated." }, { - "left": "issue:45704", - "right": "issue:45859", + "left": "issue:43525", + "right": "issue:43531", "accept": false, - "reason": "T5 Apex RMSNorm leak and Qwen3_5Moe CPU-offload incompatibility are different code paths." + "reason": "Llama4 pad_token_id attribute error is unrelated to Qwen3-MoE sliding_window handling." }, { "left": "issue:45081", "right": "issue:45356", "accept": false, - "reason": "Related tokenizer-regex area, but one is an AttributeError on `backend_tokenizer` and the other is a codec-handling regression/warning issue." + "reason": "Both touch tokenizer regex handling, but one is a hard crash on Mistral tokenizer load and the other is a Kimi-K2.5 codec/regression warning issue." }, { "left": "issue:39692", "right": "issue:43873", "accept": false, - "reason": "SigLIP2 doc/quantization errors are unrelated to general offloading-not-working-with-quantization behavior." + "reason": "SigLIP2 docs/example problems are not the same as the quantization/offloading bug in the other issue." }, { "left": "issue:45356", "right": "issue:45907", "accept": false, - "reason": "Tokenizer regression in Kimi-K2.5 is unrelated to `list_repo_files` failing to forward `revision`." + "reason": "Kimi tokenizer regression and list_repo_files revision forwarding are unrelated." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "Adapter loading from local checkpoints is unrelated to quantization/offloading behavior." + "reason": "PEFT/local-checkpoint loading for GraniteSpeech does not match the quantization offloading problem." }, { "left": "issue:40444", "right": "issue:43746", "accept": false, - "reason": "Qwen2.5-VL iterable-dataset multi-image failure is unrelated to GraniteSpeech PEFT adapter loading." + "reason": "Qwen2.5-VL IterableDataset multi-image finetuning failure is unrelated to GraniteSpeech PEFT local-load issues." }, { "left": "issue:43596", "right": "issue:44220", "accept": false, - "reason": "DeepSpeed ZeRO-3 init on BertModel is unrelated to `_torch_extract_fbank_features()`." + "reason": "Deepspeed ZeRO-3/BertModel index error and `_torch_extract_fbank_features()` are different failure modes." }, { "left": "issue:43746", "right": "issue:45381", "accept": false, - "reason": "GraniteSpeech adapter loading and Qwen2.5-VL video `vision_position_ids` are different bugs." + "reason": "PEFT checkpoint loading and Qwen2.5-VL video vision_position_ids are separate multimodal bugs." }, { "left": "issue:43540", "right": "issue:44008", "accept": false, - "reason": "Qwen3OmniMoe video-input ValueError is unrelated to Gemma 3n `audio_mel_mask` variable collision." + "reason": "Qwen3OmniMoe video-processing ValueError and Gemma 3n audio_mel_mask name collision are unrelated." }, { "left": "issue:43720", "right": "issue:43819", "accept": false, - "reason": "BitNet packed-weight loading and DAC latent/STE mismatch are unrelated." + "reason": "BitNet packed-weight loading and DAC.from_latents STE mismatch are different models and different code paths." }, { "left": "issue:43618", "right": "issue:43697", "accept": false, - "reason": "CLIPOutput attentions assignment regression is unrelated to RTDetrV2 output differences." + "reason": "CLIPOutput attentions regression is unrelated to RTDetrV2 output drift in Transformers v5." }, { "left": "issue:45183", "right": "issue:45479", "accept": false, - "reason": "Input validation XOR messaging bug is unrelated to the single-label classification zero-loss regression." + "reason": "Input validation messaging across models is not the same bug as the single-label classification zero-loss issue." }, { "left": "issue:44220", "right": "issue:44661", "accept": false, - "reason": "Audio fbank extraction failure is unrelated to `add-new-model-like` failing in `TOKENIZER_MAPPING_NAMES`." + "reason": "Audio feature extraction and add-new-model-like/TOKENIZER_MAPPING_NAMES are unrelated tokenizer/model registry issues." }, { "left": "issue:43653", "right": "issue:43720", "accept": false, - "reason": "BigBird special-token registration bug is unrelated to BitNet accelerate-loading unpacking." + "reason": "BigBird special-token decode bug is unrelated to BitNet accelerate loading of packed weights." }, { "left": "issue:30064", "right": "issue:36010", "accept": false, - "reason": "Void segmentation-map processing and `GenerationMixin` import failure are unrelated." + "reason": "Void segmentation-map processing and a GenerationMixin import error are completely unrelated." }, { "left": "issue:43335", "right": "issue:43756", "accept": false, - "reason": "SwitchTransformers sparse-layer creation and Smollm3 RoPE-layer drop are different model-configuration bugs." + "reason": "SwitchTransformers sparse-layer config and Smollm3 RoPE-layer mismatch are different architecture/configuration bugs." }, { "left": "issue:43295", "right": "issue:45834", "accept": false, - "reason": "Custom processor/tokenizer image regression is unrelated to Kosmos2.5 long OCR index errors." + "reason": "Custom processor.tokenizer regression in v4.57.5 is not the same as Kosmos2.5 long OCR index errors." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is conservative and grounded in the packet. The only accepted pair, 42886/44843, is plausibly the same offline tokenizer-loading bug, while the other pairwise rejections are adequately supported by differing titles and failure modes. The summary avoids overclaiming and stays within the evidence available." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -4803,232 +4805,229 @@ "issue:44843|issue:45081", "issue:43577|issue:43818", "issue:43531|issue:43643", - "issue:43421|issue:44811", "issue:43653|issue:43819", "issue:43643|issue:44877", + "issue:43421|issue:44811", "issue:43540|issue:45381", "issue:43782|issue:44977", "issue:44448|issue:44671", "issue:45584|issue:45834", "issue:43825|issue:43994", "issue:44561|issue:45020", - "issue:44485|issue:44509", "issue:39692|issue:45834", + "issue:44485|issue:44509", "issue:43644|issue:43749", "issue:44821|issue:45084", "issue:30064|issue:36331", "issue:41720|issue:44991", - "issue:44186|issue:44442", "issue:43825|issue:43827", - "issue:43577|issue:43582", - "issue:44479|issue:45834" + "issue:44186|issue:44442", + "issue:44479|issue:45834", + "issue:43577|issue:43582" ], - "prepared_review_unit_hash": "ba4ac4b2191cc6bb97936822564f64be7e57dcef8f5efd8c82896361f730151e", + "prepared_review_unit_hash": "22fa7d1edcd72870d1a8f0afa0c32ed56a5d443358ed7ad664d341dfbcc40fd4", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36173, - "estimated_input_tokens": 9044, - "estimated_eval_tokens": 18344 + "serialized_chars": 36172, + "estimated_input_tokens": 9043, + "estimated_eval_tokens": 18342 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a loose collection of unrelated Transformers issues. Only two soft pairs look like true duplicates/near-duplicates: the v5 pipeline-removal docs/error-message fallout, and the remote_code breakage caused by removal of `is_torch_fx_available`.", - "confidence": 0.78, - "canonical_issue_reason": "issue:45020 is the broadest open umbrella in the set, centered on recent `remote_code` regressions and able to absorb the more specific breakage reports.", + "summary": "Most candidate pairs are clearly unrelated. The only close theme is the v5 pipeline-removal/docs cluster, but the paired issues there still describe different artifacts (runtime error text vs documentation cleanup), so I rejected all soft edges. No PRs were provided.", + "confidence": 0.67, + "canonical_issue_reason": "issue:43825 is the most concrete, user-facing item in the nearest thematic cluster (v5 pipeline-removal behavior), but it is still not a clean duplicate anchor for the rest of the set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the best global representative because it captures a cross-cutting compatibility regression rather than a narrow model-specific symptom.", + "best_issue_reason": "issue:43825 is the best single issue to anchor triage among the near-match set because it describes a specific runtime regression rather than a broad or documentation-only report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43716", "right": "issue:44008", "accept": false, - "reason": "Different bugs in different code paths: image preprocessor dtype mismatch vs a Gemma3n variable-name collision in forward()." + "reason": "Different models and different failure modes: image preprocessor dtype mismatch vs Gemma 3n variable-name collision." }, { "left": "issue:43550", "right": "issue:45072", "accept": false, - "reason": "Unrelated model failures: Bamba torch.compile/SDPA issue vs dtype mismatch in SwitchTransformers/TimmWrapperModel inference." + "reason": "Both involve model execution regressions, but one is Bamba + torch.compile/SDPA and the other is SwitchTransformers/TimmWrapper dtype mismatch; not the same bug." }, { "left": "issue:43550", "right": "issue:43819", "accept": false, - "reason": "Different models and failures; no shared concrete bug path." + "reason": "Unrelated code paths: torch.compile/SDPA failure versus DAC latent/forward mismatch." }, { "left": "issue:43322", "right": "issue:45685", "accept": false, - "reason": "A Llava Next loading segfault and an MPS histogram kernel limitation are not the same bug." + "reason": "Different subsystems and symptoms: Llava Next loading segfault vs MPS histogram Int error." }, { "left": "issue:44062", "right": "issue:44488", "accept": false, - "reason": "Both are tokenizer-loading themed, but the failures are different and not clearly the same underlying defect." + "reason": "One is an AddedToken constructor argument error; the other is a tokenizer/model load failure for a specific repo." }, { "left": "issue:44186", "right": "issue:44898", "accept": false, - "reason": "Completely different model/component paths: LayoutLMv2 tokenizer crash vs Perceiver image classification resolution issue." + "reason": "Different tokenization/image-classification failures with no shared concrete code path." }, { "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 an attribute error on tokenizer internals; too distinct to merge as duplicates." + "reason": "Both touch _patch_mistral_regex, but they fail for different reasons: offline model_info call vs missing backend_tokenizer attribute." }, { "left": "issue:43577", "right": "issue:43818", "accept": false, - "reason": "BLIP2 dtype propagation and Video-LLaVA temporal-attention/weight-sharing are unrelated bugs." + "reason": "Different models and defects: BLIP2 dtype propagation vs Video-LLaVA missing temporal attention/shared weights." }, { "left": "issue:43531", "right": "issue:43643", "accept": false, - "reason": "Qwen3-MoE sliding-window behavior and `trust_remote_code` config field loss are separate issues." - }, - { - "left": "issue:43421", - "right": "issue:44811", - "accept": false, - "reason": "Both involve token handling, but one is runtime special-token updates and the other is Whisper decode behavior; not the same bug." + "reason": "Qwen3-MoE sliding-window behavior is unrelated to trust_remote_code config field loss." }, { "left": "issue:43653", "right": "issue:43819", "accept": false, - "reason": "BigBird special-token registration and DAC latent/forward mismatch are unrelated." + "reason": "Tokenizer special-token registration bug is unrelated to DAC from_latents/forward mismatch." }, { "left": "issue:43643", "right": "issue:44877", "accept": false, - "reason": "Different config-loading problems with no shared concrete failure path." + "reason": "Both concern config loading, but one is missing fields with trust_remote_code and the other is strict-config rejection of granite_speech." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Both mention tokenizer behavior, but one is runtime post-processor updates and the other is Whisper batch_decode skip_special_tokens handling." }, { "left": "issue:43540", "right": "issue:45381", "accept": false, - "reason": "Both are video-related Qwen issues, but they affect different models and different failures." + "reason": "Related model family, but different regressions and different affected video-processing paths." }, { "left": "issue:43782", "right": "issue:44977", "accept": false, - "reason": "Different Qwen3-family issues: weight-only loading error vs flash-attention generation failure." + "reason": "Different Qwen variants and different failure modes: weight_only load error vs flash-attention generation bug." }, { "left": "issue:44448", "right": "issue:44671", "accept": false, - "reason": "Separate model regressions with different symptoms and code paths." + "reason": "Both are output regressions, but they affect different models and distinct inference paths." }, { "left": "issue:45584", "right": "issue:45834", "accept": false, - "reason": "Whisper empty-transcription behavior and Kosmos2.5 long-OCR indexing are unrelated." + "reason": "Whisper empty-transcription generation and Kosmos2.5 long-OCR index error are unrelated." }, { "left": "issue:43825", "right": "issue:43994", "accept": false, - "reason": "Both mention pipeline/model output problems, but one is a v5 pipeline deprecation message and the other is a SigLIP2 model-output regression." + "reason": "Both are about docs/model usage, but one is pipeline error-message text and the other is AutoModel/pipeline output quality for SigLIP2." }, { "left": "issue:44561", "right": "issue:45020", - "accept": true, - "reason": "`issue:44561` is a specific `remote_code` breakage caused by removing `is_torch_fx_available`; `issue:45020` is the broader umbrella for the same class of recent `remote_code` regressions." - }, - { - "left": "issue:44485", - "right": "issue:44509", "accept": false, - "reason": "RoPE implementation discussion and pipeline-doc references are unrelated." + "reason": "The former is a specific API-removal breakage; the latter is a broad umbrella report about remote_code breakage." }, { "left": "issue:39692", "right": "issue:45834", "accept": false, - "reason": "SigLIP2 documentation/model mismatch and Kosmos2.5 OCR indexing are different problems." + "reason": "SigLIP2 docs/model mismatch and Kosmos2.5 OCR index error are unrelated." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "GLM-5 RoPE discussion is unrelated to v5 pipeline-task documentation cleanup." }, { "left": "issue:43644", "right": "issue:43749", "accept": false, - "reason": "Non-persistent buffer initialization and FSDP CPU RAM efficient loading are distinct issues." + "reason": "Non-persistent buffer initialization bug is unrelated to FSDP CPU RAM-efficient loading." }, { "left": "issue:44821", "right": "issue:45084", "accept": false, - "reason": "AutoImageProcessor URL loading and a Jinja/template compilation error are unrelated." + "reason": "AutoImageProcessor URL loading and template-node compilation errors are different failures." }, { "left": "issue:30064", "right": "issue:36331", "accept": false, - "reason": "ImageProcessor void segmentation maps and CustomTrainer compute_loss kwargs are different bugs." + "reason": "Image segmentation-map processing and CustomTrainer compute_loss signature mismatch are unrelated." }, { "left": "issue:41720", "right": "issue:44991", "accept": false, - "reason": "Qwen device-mapping crash and tokenizer-loading failure for EMBEDDIA/est-roberta are unrelated." - }, - { - "left": "issue:44186", - "right": "issue:44442", - "accept": false, - "reason": "LayoutLMv2 tokenizer padding/NER crash and FastSpeech2Conformer AutoTokenizer loading are different failures." + "reason": "Qwen3 auto device-mapping CUDA assert and est-roberta tokenizer loading are different issues." }, { "left": "issue:43825", "right": "issue:43827", - "accept": true, - "reason": "Both are fallout from v5 pipeline removals: one is incorrect error text, the other is stale docs still referencing removed pipeline tasks." + "accept": false, + "reason": "Both relate to v5 pipeline removals, but one reports a runtime error-message regression and the other is a documentation cleanup issue; not the same concrete bug." }, { - "left": "issue:43577", - "right": "issue:43582", + "left": "issue:44186", + "right": "issue:44442", "accept": false, - "reason": "BLIP2 dtype propagation and Apple Silicon allocator warmup TypeError do not share the same bug." + "reason": "LayoutLMv2 tokenizer NER/padding crash and FastSpeech2Conformer AutoTokenizer loading failure are unrelated." }, { "left": "issue:44479", "right": "issue:45834", "accept": false, - "reason": "Different model families and failure modes: Qwen video-input regression vs Kosmos2.5 long OCR input indexing." + "reason": "Qwen video-input regression and Kosmos2.5 long OCR index error are different model paths and symptoms." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "BLIP2 dtype propagation and Apple Silicon caching_allocator_warmup TypeError are unrelated." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays fairly conservative. The choice of issue:45020 as the broad umbrella for recent `remote_code` regressions is well supported by the title, and the two accepted soft pairs are plausibly same-bug/near-duplicate umbrella cases. The rejected pairs mostly track clearly different models, components, or failure modes, so I don\u2019t see obvious overstatement." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -5096,8 +5095,8 @@ ], "soft_pairs": [ "issue:44805|issue:44936", - "issue:44671|issue:45701", "issue:43606|issue:45072", + "issue:44671|issue:45701", "issue:34634|issue:36010", "issue:44610|issue:45072", "issue:43782|issue:45325", @@ -5114,8 +5113,8 @@ "issue:45459|issue:45803", "issue:34634|issue:36331", "issue:45084|issue:45538", - "issue:43653|issue:44912", "issue:43329|issue:45600", + "issue:43653|issue:44912", "issue:43742|issue:45020", "issue:43638|issue:44936", "issue:43746|issue:44811", @@ -5124,20 +5123,20 @@ "issue:43257|issue:43696", "issue:44589|issue:45042" ], - "prepared_review_unit_hash": "79929f7b3c6599a2c5a15f076c0ebd168267f88e06d86834a959b7d330e90d12", + "prepared_review_unit_hash": "b5589a7e7dc6d6a9bb85b23a8522e4c73b7c581b7eb2c5ee818bbb7b4b6c0e02", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36441, + "serialized_chars": 36442, "estimated_input_tokens": 9111, "estimated_eval_tokens": 18478 }, @@ -5146,180 +5145,180 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly distinct Transformers bug reports that only look similar at a high level (loading, tokenization, dtype, processor, or trainer issues). I do not see any pair that should be merged as true duplicates or same-code-path fixes.", - "confidence": 0.94, - "canonical_issue_reason": "Issue 45020 is the broadest umbrella here: it describes recent Transformers versions breaking remote_code models, which overlaps with several loading/tokenizer regressions, but it is still too general to be a strong duplicate canonical.", + "summary": "This cluster is mostly heterogeneous: a mix of unrelated model-loading, processor, tokenizer, trainer, and backend-specific regressions. I did not find any soft pair that looks like the same underlying bug or change.", + "confidence": 0.82, + "canonical_issue_reason": "issue:45020 is the broadest and most representative report in the set: a general regression where newer Transformers versions break remote-code models, which matches the cluster\u2019s overall loading/regression theme better than the narrower model-specific issues.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45020 is the best single representative for the cluster because it is open, broad, and captures the widest class of version-regression loading failures among the set.", + "best_issue_reason": "issue:45020 is the best standalone issue here because it is broad, clearly described, and likely to serve as the most useful umbrella reference for related version-regression reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44805", "right": "issue:44936", "accept": false, - "reason": "Different failures: mask/index shape mismatch vs evaluate-after-train behavior." + "reason": "Different failure modes: a mask/index shape mismatch versus an evaluation-after-training regression. No evidence they share the same code path." }, { - "left": "issue:44671", - "right": "issue:45701", + "left": "issue:43606", + "right": "issue:45072", "accept": false, - "reason": "CamemBERT masked-LM predictions bug is not the same as a generic tokenization-version change report." + "reason": "Different models and different bug classes: CPU offload device mismatch vs bfloat16 dtype mismatch." }, { - "left": "issue:43606", - "right": "issue:45072", + "left": "issue:44671", + "right": "issue:45701", "accept": false, - "reason": "CPU offload device mismatch and bfloat16 dtype mismatch are different code paths." + "reason": "CamemBERT masked-LM prediction regression is not the same as a broad tokenization-change report." }, { "left": "issue:34634", "right": "issue:36010", "accept": false, - "reason": "Bark voice_preset malfunction is unrelated to the GenerationMixin import error." + "reason": "BarkProcessor voice_preset failure and GenerationMixin import error are unrelated." }, { "left": "issue:44610", "right": "issue:45072", "accept": false, - "reason": "Processor image-size mismatch vs inference dtype mismatch; not the same bug." + "reason": "Processor output size mismatch and dtype mismatch are distinct problems." }, { "left": "issue:43782", "right": "issue:45325", "accept": false, - "reason": "Qwen3VL weight_only load error and rope-index scaling bug are different issues." + "reason": "Different Qwen variants and different bugs: weight_only loading error vs rope-index scaling." }, { "left": "issue:43720", "right": "issue:45072", "accept": false, - "reason": "Packed-weight unpacking in BitNet is separate from bfloat16 dtype mismatch." + "reason": "Packed-weight unpacking during accelerate loading is unrelated to generic dtype mismatch behavior." }, { "left": "issue:42175", "right": "issue:45081", "accept": false, - "reason": "Missing TensorFlow backend in torch install is unrelated to a Mistral tokenizer regex crash." + "reason": "Package/backend dependency issue vs tokenizer regex crash; no shared underlying bug." }, { "left": "issue:45081", "right": "issue:45406", "accept": false, - "reason": "Both are tokenizer/processor attribute errors, but they affect different models and different missing attributes." + "reason": "Both are tokenizer/processor loading problems, but they affect different components and failure causes." }, { "left": "issue:43749", "right": "issue:44291", "accept": false, - "reason": "FSDP CPU RAM efficient loading and init_empty_weights _is_hf_initialized are different loading bugs." + "reason": "Both involve model loading, but one is FSDP RAM-efficient loading and the other is init_empty_weights argument handling; not the same bug." }, { "left": "issue:44060", "right": "issue:44368", "accept": false, - "reason": "Similar warning theme, but different models and different tie-weight warning behavior." + "reason": "These are different tied-weights warnings with different triggers and configurations." }, { "left": "issue:43650", "right": "issue:43906", "accept": false, - "reason": "A non-technical placeholder issue is not a duplicate of a reproduction for issue 38071." + "reason": "One is non-technical placeholder content; the other is a reproduction of a specific upstream issue." }, { "left": "issue:43582", "right": "issue:44857", "accept": false, - "reason": "Apple Silicon allocator TypeError and CUDA AMP loss crash are unrelated." + "reason": "AppleSilicon allocator warmup TypeError and CUDA AMP loss crash are unrelated." }, { "left": "issue:43262", "right": "issue:43819", "accept": false, - "reason": "Audio chat-template sampling-rate default is unrelated to DAC latent reconstruction mismatch." + "reason": "Audio chat-template sampling-rate defaulting and DAC latent conversion mismatch are separate issues." }, { "left": "issue:43526", "right": "issue:43697", "accept": false, - "reason": "BEiT reduce_labels bug and RTDetrV2 output drift are different model behaviors." + "reason": "BeitImageProcessorFast label reduction bug does not match an RTDetrV2 output-difference regression." }, { "left": "issue:43526", "right": "issue:44554", "accept": false, - "reason": "No shared code path: BEiT label reduction vs MPS attention correctness." + "reason": "Label reduction bug and MPS attention correctness issue are unrelated." }, { "left": "issue:45459", "right": "issue:45803", "accept": false, - "reason": "Both are bad exception handling, but they are different functions and different swallowed errors." + "reason": "Both concern broad exception handling, but they are different code paths and different observable bugs." }, { "left": "issue:34634", "right": "issue:36331", "accept": false, - "reason": "BarkProcessor preset handling is unrelated to CustomTrainer.compute_loss kwargs." + "reason": "BarkProcessor preset behavior and CustomTrainer compute_loss signature mismatch are unrelated." }, { "left": "issue:45084", "right": "issue:45538", "accept": false, - "reason": "Template compilation error and CLIPTokenizer max_length bug are unrelated." + "reason": "Template-compilation TypeError and tokenizer model_max_length are different tokenizer-layer issues." }, { - "left": "issue:43653", - "right": "issue:44912", + "left": "issue:43329", + "right": "issue:45600", "accept": false, - "reason": "BigBird special-token registration and MXFP4 quantization fallback are separate issues." + "reason": "Undefined multimodal-token video branch symbols and removed auto-mapping configs are unrelated CI/codegen problems." }, { - "left": "issue:43329", - "right": "issue:45600", + "left": "issue:43653", + "right": "issue:44912", "accept": false, - "reason": "Undefined multimodal-token variables and stale auto_mappings config references are different CI/code bugs." + "reason": "BigBird special-token decoding bug and MXFP4 quantization fallback bug are distinct." }, { "left": "issue:43742", "right": "issue:45020", "accept": false, - "reason": "MobileLLM key error may be version-related, but it is not the same concrete bug as remote_code breakage in general." + "reason": "A key error for a specific model is not the same as the broader remote_code regression." }, { "left": "issue:43638", "right": "issue:44936", "accept": false, - "reason": "Zero3/Bert index error and trainer.evaluate() failure after train are distinct trainer/runtime problems." + "reason": "Zero3 mask indexing failure and train-then-evaluate failure are different runtime issues." }, { "left": "issue:43746", "right": "issue:44811", "accept": false, - "reason": "PEFT local checkpoint loading and Whisper batch_decode skip_special_tokens are unrelated." + "reason": "PEFT local checkpoint loading and Whisper batch_decode skip_special_tokens behavior do not share a concrete code path." }, { "left": "issue:43257", "right": "issue:43329", "accept": false, - "reason": "Qwen3 MOE weight conversion/loading issue and multimodal token counting bug do not share a fix." + "reason": "MOE weight conversion during accelerate/deepspeed loading is unrelated to multimodal token counting." }, { "left": "issue:43452", "right": "issue:44991", "accept": false, - "reason": "Tokenizer/model loading with gguf_file and EMBEDDIA tokenizer loading regression are different failure modes." + "reason": "Both involve from_pretrained/tokenizer loading, but they target different APIs and failure mechanisms." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Weight conversion/loading bug is not an out-of-memory issue." + "reason": "Loading/conversion bug and CUDA OOM are clearly unrelated." }, { "left": "issue:44589", "right": "issue:45042", "accept": false, - "reason": "Float8 storage lookup failure and torchvision requirement regression are unrelated." + "reason": "Float8 storage lookup failure and PIL/torchvision dependency regression are different classes of issues." } ] }, @@ -5393,23 +5392,23 @@ "issue:45859" ], "soft_pairs": [ - "issue:38175|issue:44610", "issue:43854|issue:44977", + "issue:38175|issue:44610", "issue:39692|issue:44912", "issue:43819|issue:43927", "issue:36010|issue:43645", - "issue:43454|issue:43577", "issue:43720|issue:44610", + "issue:43454|issue:43577", "issue:43550|issue:44912", "issue:43262|issue:44811", "issue:43329|issue:44479", "issue:43531|issue:45042", - "issue:43749|issue:44756", - "issue:44849|issue:45701", "issue:36010|issue:36331", + "issue:43749|issue:44756", "issue:44568|issue:45356", "issue:43643|issue:45759", "issue:44568|issue:45488", + "issue:44849|issue:45701", "issue:34567|issue:40990", "issue:43329|issue:43450", "issue:43257|issue:43931", @@ -5417,214 +5416,217 @@ "issue:43257|issue:43756", "issue:43697|issue:45357", "issue:43452|issue:43582", - "issue:44987|issue:45538", "issue:44291|issue:45042", + "issue:44987|issue:45538", "issue:44479|issue:45381", "issue:42915|issue:43742" ], - "prepared_review_unit_hash": "9dd1b3a2f9452fa9b2dc028957d1462187cdf7dcf3825ab877e5bd390311a5d0", + "prepared_review_unit_hash": "cc46032d771ace898aa070f2da3becd97461984e766e308f7e4fd16805687f4c", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36350, - "estimated_input_tokens": 9088, - "estimated_eval_tokens": 18432 + "serialized_chars": 36345, + "estimated_input_tokens": 9087, + "estimated_eval_tokens": 18430 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These pairs are mostly loose topical matches across transformers/model-loading regressions, not true duplicates. The issues usually involve different models, different failure modes, or different code paths, so none of the soft edges should be merged. No PRs were provided.", - "confidence": 0.34, - "canonical_issue_reason": "Issue 45701 is the broadest umbrella item here: it describes version-dependent tokenization changes and has multiple references, making it the best representative of the only partially shared theme in this set. The other issues are more specific and not duplicates of it.", + "summary": "Mostly a heterogeneous set of transformer bug reports, not one duplicate family. I only see one strong duplicate-style overlap: the Qwen2.5-VL/Qwen3.5 video-input regression pair.", + "confidence": 0.44, + "canonical_issue_reason": "issue:44479 is the best canonical anchor for the one clear duplicate family here because it captures the broader Qwen video-input regression that also covers the specific qwen2.5-vl symptom in issue:45381.", "canonical_pr_reason": null, - "best_issue_reason": "45701 is the most general, central issue among the set and best serves as a representative anchor for the tokenization/regression subgroup. It still should not absorb the more specific model-, processor-, or loading-related bugs.", + "best_issue_reason": "issue:45701 is the broadest open issue, but the set is too mixed to make it a strong global duplicate canonical; issue:44479 is the best local match to the only accepted overlap.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:38175", - "right": "issue:44610", + "left": "issue:43854", + "right": "issue:44977", "accept": false, - "reason": "Both are vision-model bugs, but one is unexpected zero probabilities in SigLIP2 and the other is a processor/model input-size mismatch in OmDet-Turbo. Different models and different failure modes." + "reason": "Different models and failure modes: GLM-4.7-Flash loading/test issue vs Qwen3.5 flash-attention generation behavior." }, { - "left": "issue:43854", - "right": "issue:44977", + "left": "issue:38175", + "right": "issue:44610", "accept": false, - "reason": "Both mention Qwen-related loading/generation problems, but one is a unit-test model-loading issue and the other is a flash-attention generation regression. Not the same code-path bug." + "reason": "Both are vision-related, but one is a SigLIP2 probability bug and the other is an OmDet-Turbo processor size mismatch." }, { "left": "issue:39692", "right": "issue:44912", "accept": false, - "reason": "SigLIP2 docs example errors are about example/model-processor mismatch and quantization failure; git-oss-20b MXFP4 is a quantization fallback/loading issue. Distinct bugs." + "reason": "Both mention quantization, but they affect different models and different code paths (docs example/processor vs MXFP4 fallback)." }, { "left": "issue:43819", "right": "issue:43927", "accept": false, - "reason": "DAC.from_latents vs forward mismatch is a codec/quantization consistency bug, while DiaConfig loses custom token IDs on save/load and breaks generation. Unrelated subsystems." + "reason": "Unrelated bugs: DAC latent conversion mismatch vs DiaConfig losing token IDs on save/load." }, { "left": "issue:36010", "right": "issue:43645", "accept": false, - "reason": "GenerationMixin import failure and notebook custom model initialization problems are both Transformers 5 issues, but they affect different APIs and causes." + "reason": "Both are import/config compatibility regressions, but they are distinct failures with different triggers and code paths." }, { - "left": "issue:43454", - "right": "issue:43577", + "left": "issue:43720", + "right": "issue:44610", "accept": false, - "reason": "AyaVision lm_head tying failure causing garbage text is a multimodal weight-tying bug; Blip2 dtype staying float32 is a loading/dtype propagation issue. Not duplicates." + "reason": "Packed-weight loading bug in BitNet is unrelated to an OmDet-Turbo image processor shape issue." }, { - "left": "issue:43720", - "right": "issue:44610", + "left": "issue:43454", + "right": "issue:43577", "accept": false, - "reason": "BitNet packed weights not unpacked during accelerate loading is a model-loading bug; OmDet-Turbo\u2019s 640px processor output is an image preprocessing mismatch. Different paths." + "reason": "Different models and symptoms: AyaVision lm_head tying vs BLIP2 dtype propagation." }, { "left": "issue:43550", "right": "issue:44912", "accept": false, - "reason": "Bamba-9B-v2 torch.compile+SDPA failure is a compilation/runtime issue, while git-oss-20b MXFP4 falling back to bf16 is a quantization loading issue." + "reason": "Different problem classes: SDPA/torch.compile failure in Bamba vs quantization fallback in git-oss-20b." }, { "left": "issue:43262", "right": "issue:44811", "accept": false, - "reason": "Audio apply_chat_template defaulting to 16kHz is a sampling-rate default bug; Whisper batch_decode ignoring skip_special_tokens is a decoding-parameter propagation bug." + "reason": "Both audio-related, but one is chat-template sampling rate defaulting and the other is batch_decode skip_special_tokens handling." }, { "left": "issue:43329", "right": "issue:44479", "accept": false, - "reason": "The video branch undefined-variable bug in _get_num_multimodal_tokens is a helper logic error; the v5.3 video regression spans multiple VLMs and position-id handling. Related area, different defect." + "reason": "Both involve video multimodal processing, but one is an undefined-helper bug in token counting and the other is a broader video-input regression." }, { "left": "issue:43531", "right": "issue:45042", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and PIL backend requiring torchvision are unrelated regressions in different subsystems." - }, - { - "left": "issue:43749", - "right": "issue:44756", - "accept": false, - "reason": "FSDP_CPU_RAM_EFFICIENT_LOADING is a distributed-loading regression; Disable mmap on Strix Halo is a platform-specific OOM workaround. Not the same issue." + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to PIL image processors requiring torchvision." }, { - "left": "issue:44849", - "right": "issue:45701", + "left": "issue:36010", + "right": "issue:36331", "accept": false, - "reason": "Qwen3.5 output_hidden_states bug is a model forward-path regression, while 45701 is a broad tokenization-change report. Different functionality." + "reason": "ImportError for GenerationMixin and compute_loss signature mismatch are separate API breakages." }, { - "left": "issue:36010", - "right": "issue:36331", + "left": "issue:43749", + "right": "issue:44756", "accept": false, - "reason": "GenerationMixin import error and CustomTrainer.compute_loss num_items_in_batch TypeError are both API regressions, but they involve unrelated modules and fixes." + "reason": "Different loading/memory issues: FSDP CPU RAM-efficient loading vs mmap OOM on Strix Halo." }, { "left": "issue:44568", "right": "issue:45356", "accept": false, - "reason": "mdeberta-v3-base add_special_tokens regression and Kimi-K2.5 codec handling warning/regression are tokenizer issues, but they affect different tokenizers and behavior." + "reason": "Both tokenizer regressions, but the BOS/EOS special-token bug is unrelated to Kimi-K2.5 codec handling." }, { "left": "issue:43643", "right": "issue:45759", "accept": false, - "reason": "AutoConfig missing fields with trust_remote_code and AutoModelForCausalLM.from_config not unwrapping text_config are both config-loading issues, but the API failures differ." + "reason": "Both are config-loading issues, but one is trust_remote_code field loss and the other is composite Qwen multimodal text_config unwrapping." }, { "left": "issue:44568", "right": "issue:45488", "accept": false, - "reason": "Both are tokenizer-related, but one is BOS/EOS insertion in mdeberta and the other is LlamaTokenizer overriding ByteLevel with Metaspace for DeepSeek models. Not duplicates." + "reason": "Tokenizer regressions in different tokenizers and different mechanisms; not the same underlying bug." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Qwen3.5 hidden-states output bug is unrelated to the broad tokenization-change report." }, { "left": "issue:34567", "right": "issue:40990", "accept": false, - "reason": "TrainerState token counting not updating and gpt-oss-20b perplexity are unrelated training/benchmarking problems." + "reason": "TrainerState token-count tracking and GPT-OSS perplexity are unrelated." }, { "left": "issue:43329", "right": "issue:43450", "accept": false, - "reason": "Undefined video-token helper variables and batched video processor shape are both video-path issues, but one is a missing-symbol bug and the other is an output-shape bug." + "reason": "Both are video-related, but one is token counting and the other is batched output shape." }, { "left": "issue:43257", "right": "issue:43931", "accept": false, - "reason": "Qwen3 MOE weights not converted under accelerate+deepspeed and Qwen3-VL-30B shape mismatch are both loading issues, but they target different model families and failure modes." + "reason": "Different Qwen3 loading failures: accelerate/deepspeed weight conversion vs VL weight-shape mismatch." }, { "left": "issue:43606", "right": "issue:45859", "accept": false, - "reason": "CPU offload device mismatch for bark-small and Qwen3_5MoeTextRotaryEmbedding incompatibility with CPU offload are both offload-related, but they are separate model-specific bugs." + "reason": "Both mention CPU offload, but they affect different models and different failure mechanisms." }, { "left": "issue:43257", "right": "issue:43756", "accept": false, - "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is a loading/conversion bug; Smollm3 RoPE layer dropping is a model-architecture mismatch. Different defects." + "reason": "Qwen3 MoE loading/accelerate issue and Smollm3 RoPE-layer dropout are unrelated." }, { "left": "issue:43697", "right": "issue:45357", "accept": false, - "reason": "RTDetrV2 different outputs in Transformers v5 and Qwen3.5 save_pretrained saving wrong visual encoder keys are both regressions, but one is inference behavior and the other is serialization." + "reason": "RTDetrV2 output regression and Qwen3.5 save_pretrained key mapping are distinct bugs." }, { "left": "issue:43452", "right": "issue:43582", "accept": false, - "reason": "gguf_file breaking from_pretrained and caching_allocator_warmup TypeError on Apple Silicon are unrelated loader/platform issues." + "reason": "gguf_file loading breakage and AppleSilicon warmup TypeError are unrelated." }, { - "left": "issue:44987", - "right": "issue:45538", + "left": "issue:44291", + "right": "issue:45042", "accept": false, - "reason": "Loading physical-intelligence/fast fails in v5.1.0, while CLIPTokenizer using 10**30 model_max_length is a tokenizer constant bug. No shared code-path." + "reason": "init_empty_weights/_is_hf_initialized incompatibility is unrelated to the PIL image processor torchvision regression." }, { - "left": "issue:44291", - "right": "issue:45042", + "left": "issue:44987", + "right": "issue:45538", "accept": false, - "reason": "init_empty_weights unexpected _is_hf_initialized argument and PIL backend requiring torchvision are entirely separate regressions." + "reason": "Different tokenizer issues: loading physical-intelligence/fast vs CLIPTokenizer max-length sentinel." }, { "left": "issue:44479", "right": "issue:45381", - "accept": false, - "reason": "v5.3 video input regression across Qwen VLMs and qwen2.5-vl video vision_position_ids bug overlap in subject area, but one is broad regression and the other is a specific positional-id error." + "accept": true, + "reason": "Same underlying video-input regression in the Qwen2.5-VL/Qwen3.5 family; one report is broad and the other is a specific symptom of the same code path." }, { "left": "issue:42915", "right": "issue:43742", "accept": false, - "reason": "Qwen3Moe failing with FineGrainedFP8Config and MobileLLM-125M key error during load are different model-loading problems with different causes." + "reason": "Different model-loading failures: FineGrainedFP8Config on Qwen3Moe vs key error loading MobileLLM-125M." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "Grounded and conservative overall. The only accepted overlap is the Qwen video-input regression family, and the summary does not overclaim broader duplication across the heterogeneous set." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -5691,253 +5693,265 @@ ], "soft_pairs": [ "issue:33357|issue:34567", - "issue:44987|issue:45907", "issue:41720|issue:43540", + "issue:44987|issue:45907", "issue:42915|issue:44821", "issue:43927|issue:44912", "issue:42915|issue:45538", "issue:43317|issue:45538", "issue:42175|issue:43824", - "issue:43577|issue:45072", "issue:43540|issue:44479", + "issue:43577|issue:45072", "issue:43540|issue:44560", "issue:42175|issue:44938", "issue:44062|issue:44991", "issue:38175|issue:44898", "issue:45137|issue:45405", "issue:42175|issue:44488", - "issue:44737|issue:45588", "issue:43906|issue:45381", "issue:28282|issue:34689", + "issue:44737|issue:45588", "issue:42915|issue:43749", - "issue:43526|issue:45479", "issue:43742|issue:44291", "issue:43582|issue:43782", + "issue:43526|issue:45479", "issue:28282|issue:29127", - "issue:43756|issue:45325", "issue:43540|issue:45823", + "issue:43756|issue:45325", "issue:43493|issue:43994", "issue:34567|issue:36331", "issue:44291|issue:44821", "issue:44062|issue:45081" ], - "prepared_review_unit_hash": "2cc850285e44b37ac9d16a04d3f658af64d773485952bd2ab3205251ae9fe7ce", + "prepared_review_unit_hash": "c81dbd0d05f05e52d103ec8225a26e221a6bb63cc9cfac2d8a96863e2becc8ab", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36329, - "estimated_input_tokens": 9083, - "estimated_eval_tokens": 18422 + "serialized_chars": 36326, + "estimated_input_tokens": 9082, + "estimated_eval_tokens": 18420 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous closed issues. The clearest duplicate-like cluster is the Qwen video-input regression area, but it should be handled conservatively: 44479 is the strongest anchor, 43540 looks adjacent/possibly related, and 44560 does not have enough concrete evidence to be merged with 43540. A second plausible cluster is the SigLIP2 discrepancy pair (43493/43994). No PRs are present.", - "confidence": 0.72, - "canonical_issue_reason": "Issue 44479 is the broadest, most root-cause-oriented Qwen video-input regression report, so it is the best anchor for the limited Qwen-related overlap in this packet.", + "summary": "This set is mostly a grab bag of unrelated issues, so only a few soft pairs look like true duplicates. The clearest overlap is around Qwen video/position-id regressions and a separate SigLIP2 discrepancy pair.", + "confidence": 0.62, + "canonical_issue_reason": "Issue 44479 is the broadest concrete regression report in the set, covering multiple Qwen VL/video-input variants and serving as the best umbrella for the accepted Qwen-related duplicates.", "canonical_pr_reason": null, - "best_issue_reason": "44479 is the strongest representative issue because it describes the shared Qwen video-input regression at the highest level; 43540 is the closest neighboring symptom, while 44560 is not supported strongly enough to join the cluster.", + "best_issue_reason": "44479 is the most representative issue overall: it is specific enough to be actionable, but broad enough to subsume the closest matching reports in the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:33357", "right": "issue:34567", "accept": false, - "reason": "Different failures in unrelated areas: CLIP/MacOS bus error vs TrainerState token counting." + "reason": "Different subsystems and failures: MacOS bus error in CLIP vs TrainerState token counter update." }, { - "left": "issue:44987", - "right": "issue:45907", + "left": "issue:41720", + "right": "issue:43540", "accept": false, - "reason": "Different bug classes: model loading failure vs repository file listing not forwarding revision." + "reason": "Both involve Qwen, but one is device-mapping CUDA assert and the other is a video-input ValueError for a different model path." }, { - "left": "issue:41720", - "right": "issue:43540", + "left": "issue:44987", + "right": "issue:45907", "accept": false, - "reason": "Both mention Qwen, but the concrete failures differ: auto device-map cuda assert vs video-input ValueError." + "reason": "Loading a model regression is unrelated to a repo-files API forwarding bug in CI." }, { "left": "issue:42915", "right": "issue:44821", "accept": false, - "reason": "Unrelated: FineGrainedFP8/Qwen3Moe config handling vs AutoImageProcessor URL loading." + "reason": "FineGrainedFP8/Qwen3Moe loading failure is unrelated to AutoImageProcessor URL loading." }, { "left": "issue:43927", "right": "issue:44912", "accept": false, - "reason": "Different code paths: DiaConfig save/load token IDs vs MXFP4 quantization fallback to bf16." + "reason": "Custom config save/load corruption is not the same as an MXFP4 quantization fallback issue." }, { "left": "issue:42915", "right": "issue:45538", "accept": false, - "reason": "Only broad loading overlap; one is Qwen3Moe FP8 config, the other is CLIPTokenizer max-length metadata." + "reason": "Qwen3Moe FP8 loading and CLIPTokenizer max_length are different bugs." }, { "left": "issue:43317", "right": "issue:45538", "accept": false, - "reason": "Different problems: device_map/offload dequantized loading vs tokenizer max-length metadata." + "reason": "Device-map/offload loading failure is unrelated to tokenizer max_length." }, { "left": "issue:42175", "right": "issue:43824", "accept": false, - "reason": "Packaging/backend install issue vs missing Qwen2.5-VL export/import symbol." + "reason": "Packaging/backend installation issue vs model import error for Qwen2.5-VL." }, { - "left": "issue:43577", - "right": "issue:45072", + "left": "issue:43540", + "right": "issue:44479", "accept": false, - "reason": "Both are dtype-related, but they affect different models and likely different initialization paths." + "reason": "Related domain, but different model families and symptoms; not clearly the same concrete bug." }, { - "left": "issue:43540", - "right": "issue:44479", - "accept": true, - "reason": "Same Qwen video-input regression family: 44479 describes the broader v5.3.0 breakage, and 43540 is a narrower symptom in that area." + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Both mention dtype, but they affect different architectures and code paths." }, { "left": "issue:43540", "right": "issue:44560", "accept": false, - "reason": "Only generic Qwen/video similarity is visible; there is no explicit shared target, code path, or other concrete evidence that these are the same bug." + "reason": "Same broad video-input area, but the titles do not establish the same underlying bug with enough confidence." }, { "left": "issue:42175", "right": "issue:44938", "accept": false, - "reason": "Packaging/backend install issue vs a Python 3.14 import/load failure; no shared underlying bug." + "reason": "TensorFlow backend packaging vs Python 3.14 import failure are unrelated." }, { "left": "issue:44062", "right": "issue:44991", "accept": false, - "reason": "Tokenizer construction bug vs tokenizer loading failure for a specific model; too different to merge." + "reason": "Tokenizer construction TypeError and a model-specific tokenizer load failure are too indirect to treat as duplicates." }, { "left": "issue:38175", "right": "issue:44898", "accept": false, - "reason": "Both are vision-model behavior issues, but they concern different models and mechanisms." + "reason": "Different models and different failure modes; both are vision issues only at a very broad level." }, { "left": "issue:45137", "right": "issue:45405", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error is unrelated to the PEFT version bump issue." + "reason": "DeepSpeed ZeRO3 deque error is unrelated to PEFT version pinning." }, { "left": "issue:42175", "right": "issue:44488", "accept": false, - "reason": "Different failures: missing TensorFlow backend in torch install vs loading a specific model checkpoint." - }, - { - "left": "issue:44737", - "right": "issue:45588", - "accept": false, - "reason": "XLNet CPU tensor creation bug vs flash-attention sinkless-model AttributeError are unrelated." + "reason": "Installation/backend packaging issue vs model load failure for a specific checkpoint." }, { "left": "issue:43906", "right": "issue:45381", "accept": false, - "reason": "The title only shows a reproduction of an older issue; there is no clear evidence it is the same bug as the Qwen2.5-VL video position-id regression." + "reason": "The reproduction issue has no clear evidence it matches the Qwen2.5-VL position-id bug." }, { "left": "issue:28282", "right": "issue:34689", "accept": false, - "reason": "Both are model-loading problems, but they involve different models and different error modes." + "reason": "Missing PyTorch dependency is unrelated to a model-loading regression for Llama 3.2 Vision." }, { - "left": "issue:42915", - "right": "issue:43749", + "left": "issue:44737", + "right": "issue:45588", "accept": false, - "reason": "FP8/Qwen3Moe configuration issue vs broken FSDP CPU RAM-efficient loading; different layers of the stack." + "reason": "XLNet CPU positional encoding overhead and flash_attention sink handling are different code paths." }, { - "left": "issue:43526", - "right": "issue:45479", + "left": "issue:42915", + "right": "issue:43749", "accept": false, - "reason": "Different label-handling bugs: image processor reduction vs degenerate zero loss in sequence classification." + "reason": "Qwen3Moe FP8 loading and FSDP CPU RAM efficient loading are separate training/loading mechanisms." }, { "left": "issue:43742", "right": "issue:44291", "accept": false, - "reason": "KeyError while loading one model vs init_empty_weights signature mismatch in a different code path." + "reason": "KeyError on a specific model load is not the same as the init_empty_weights API break." }, { "left": "issue:43582", "right": "issue:43782", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError vs Qwen3VL weight_only loading error are unrelated." + "reason": "Apple Silicon allocator warmup TypeError and Qwen3VL weight_only load error are unrelated." }, { - "left": "issue:28282", - "right": "issue:29127", + "left": "issue:43526", + "right": "issue:45479", "accept": false, - "reason": "ImportError for missing PyTorch vs LayoutLMv3 box-content error; no shared bug." + "reason": "Image processor label reduction bug is unrelated to sequence-classification loss degeneracy." }, { - "left": "issue:43756", - "right": "issue:45325", + "left": "issue:28282", + "right": "issue:29127", "accept": false, - "reason": "Different RoPE/position-id issues in different model families; too little evidence of a shared root cause." + "reason": "Missing PyTorch import error and LayoutLMv3 box validation message are distinct issues." }, { "left": "issue:43540", "right": "issue:45823", "accept": false, - "reason": "Qwen video-input processing bug vs Gemma4 PLE device-mismatch during forward are unrelated." + "reason": "Different models and different failure symptoms; not the same device-mapping bug." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "RoPE-layer mismatch in SmolLM3 is not the same as Qwen2.5-VL temporal position-id scaling." }, { "left": "issue:43493", "right": "issue:43994", "accept": true, - "reason": "Same SigLIP2 underlying discrepancy: one reports divergence from the original JAX implementation, the other reports nonsensical outputs from AutoModel/pipeline on the same model family." + "reason": "Both describe the same SigLIP2 implementation discrepancy, one as HF-vs-JAX mismatch and the other as nonsensical outputs in AutoModel/pipeline." }, { "left": "issue:34567", "right": "issue:36331", "accept": false, - "reason": "TrainerState token counter update bug vs CustomTrainer compute_loss signature breakage; different trainer issues." + "reason": "TrainerState token counting and CustomTrainer compute_loss API mismatch are different trainer bugs." }, { "left": "issue:44291", "right": "issue:44821", "accept": false, - "reason": "Different load-time failures: init_empty_weights argument mismatch vs loading AutoImageProcessor from URL." + "reason": "Init-empty-weights loading break is unrelated to AutoImageProcessor URL loading." }, { "left": "issue:44062", "right": "issue:45081", "accept": false, - "reason": "Tokenizer AddedToken duplication bug vs Mistral regex patch crash; both tokenizer-related but not the same defect." + "reason": "Both are tokenizer-related, but the specific failure mechanisms and affected code paths are different." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": true, + "reason": "Same Qwen video-input regression family in transformers 5.3.0; the specific StopIteration report fits the broader video-input breakage." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": true, + "reason": "Both point to the same Qwen2.5-VL position-id/rope-index regression affecting still-image and video inputs." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The summary is broadly grounded, but one soft-edge accept is too strong for the evidence provided: issue:43540 vs issue:44560 is only supported by generic title similarity (both involve Qwen/video), while the packet gives no explicit links, shared targets, or other concrete evidence that they are the same bug or mergeable into one PR. The Qwen video-input cluster should be described more conservatively. The other false verdicts look fine." + "accept": true, + "feedback": "The cluster summary is grounded in the packet and stays conservative. The accepted overlaps are the ones with clear title-level alignment (SigLIP2 pair, Qwen video/position-id regressions), and the rejected pairs avoid over-claiming duplicates when the evidence is only broad topical similarity." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -5957,6 +5971,7 @@ "nodes": [ "issue:24643", "issue:29127", + "issue:38175", "issue:41762", "issue:43295", "issue:43452", @@ -5999,7 +6014,6 @@ "issue:45375", "issue:45381", "issue:45405", - "issue:45440", "issue:45600", "issue:45823", "issue:45920" @@ -6015,214 +6029,226 @@ "issue:45405|issue:45600", "issue:43697|issue:45216", "issue:43688|issue:43749", - "issue:44484|issue:44704", "issue:44514|issue:45381", "issue:43720|issue:44898", + "issue:44484|issue:44704", "issue:41762|issue:44936", - "issue:43452|issue:43577", "issue:24643|issue:29127", + "issue:43452|issue:43577", "issue:44625|issue:44877", "issue:44514|issue:45290", "issue:43452|issue:44843", "issue:44466|issue:45005", - "issue:43643|issue:44561", "issue:43819|issue:44811", "issue:43295|issue:45356", - "issue:43653|issue:44568", + "issue:43643|issue:44561", "issue:44464|issue:44743", "issue:44936|issue:45823", - "issue:43688|issue:45440" + "issue:43653|issue:44568", + "issue:38175|issue:43295" ], - "prepared_review_unit_hash": "fc3beb990c5cd60f86ff86b8ba32d656f9d15cf735e71087d84dd59983e99c2b", + "prepared_review_unit_hash": "d9e9f681de5af9f56d610906072b6176b1422c544750a068fb370df35c27e2a7", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 27, - "serialized_chars": 36161, - "estimated_input_tokens": 9041, - "estimated_eval_tokens": 18338 + "serialized_chars": 36144, + "estimated_input_tokens": 9036, + "estimated_eval_tokens": 18328 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are not duplicates of one another; they are a grab-bag of unrelated HF/Transformers issues with only superficial title similarity. The soft pairs mostly share broad themes like tokenizer/config/model-loading regressions, but they point to different code paths and should stay separate.", - "confidence": 0.93, - "canonical_issue_reason": "No true canonical duplicate exists for this set because the issues are distinct. If one must be chosen as the representative, issue 45920 is the broadest open regression report and most umbrella-like.", + "summary": "This cluster is mostly a noisy mix of unrelated regression reports around tokenizers, configs, loading, and generation. The best representative issue is the broad AutoTokenizer token-ID regression, but none of the soft pairs look like the same concrete bug/change.", + "confidence": 0.44, + "canonical_issue_reason": "issue:45920 is the broadest and most central report: it describes a cross-model AutoTokenizer regression affecting multiple model families, which best matches the cluster\u2019s dominant tokenizer/config regression theme.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45920 is the best representative only in a loose clustering sense: it is broad, current, and covers a general tokenizer regression affecting multiple models. That said, it should not be treated as a duplicate of the others.", + "best_issue_reason": "issue:45920 is the strongest cluster representative because it is open, broad in scope, and explicitly aggregates failures across several model families rather than a single narrow edge case.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45357", "right": "issue:45920", "accept": false, - "reason": "Different bugs: Qwen3.5 save_pretrained visual-key serialization vs general AutoTokenizer token-id regression across several models." + "reason": "Both concern Qwen-related regressions, but one is about incorrect save_pretrained visual encoder keys while the other is about wrong token IDs from AutoTokenizer; different failure modes and code paths." }, { "left": "issue:43643", "right": "issue:45375", "accept": false, - "reason": "Both involve config fields, but one is trust_remote_code AutoConfig object loss and the other is a missing Qwen3_5MoeVisionConfig field." + "reason": "Both involve config-field handling, but one is trust_remote_code returning missing fields and the other is strict dropping a specific Qwen vision field; not the same bug." }, { "left": "issue:44792", "right": "issue:44964", "accept": false, - "reason": "Different models and failures: janus test failure vs Phi-4-multimodal load failure." + "reason": "Different models and symptoms: a janus test failure versus loading Phi-4 multimodal; no clear shared underlying code-path bug." }, { "left": "issue:43452", "right": "issue:44488", "accept": false, - "reason": "Both are loading-related, but one is gguf_file/AutoTokenizer+AutoModel and the other is a specific model load failure for cjvt/sleng-bert." + "reason": "Both are model loading issues, but one is gguf_file handling for AutoTokenizer/AutoModel and the other is a specific model load failure; too broad and not the same concrete defect." }, { "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "Both touch image/tokenizer internals, but one is PIL backend torchvision dependency handling and the other is a Mistral regex crash." + "reason": "Unrelated subsystems: PIL image processor torchvision dependency versus Mistral regex patch crash." }, { "left": "issue:43742", "right": "issue:44589", "accept": false, - "reason": "Different load-path failures: MobileLLM key error vs Float8 storage lookup during deserialization." + "reason": "Different failure classes: key error during model loading versus missing Float8 storage object." }, { "left": "issue:43531", "right": "issue:44821", "accept": false, - "reason": "Qwen3-MoE sliding window behavior is unrelated to AutoImageProcessor URL loading." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to AutoImageProcessor URL loading." }, { "left": "issue:45405", "right": "issue:45600", "accept": false, - "reason": "One is a PEFT version pin problem; the other is stale auto_mappings entries for removed Sam3LiteText configs." + "reason": "One is a version pin bump for PEFT; the other is a stale auto_mappings reference causing CI breakage." }, { "left": "issue:43697", "right": "issue:45216", "accept": false, - "reason": "Different issues: RTDetrV2 output mismatch in v5 vs Qwen3.5 checkpoint save regression." + "reason": "Different regressions in different areas: RTDetrV2 output changes versus Qwen3.5 save_pretrained serialization." }, { "left": "issue:43688", "right": "issue:43749", "accept": false, - "reason": "Both mention loading/training infrastructure, but auxiliary-loss normalization is unrelated to FSDP CPU RAM efficient loading." - }, - { - "left": "issue:44484", - "right": "issue:44704", - "accept": false, - "reason": "save_pretrained shard-size question is unrelated to AutoProcessor forwarding kwargs to cached_file." + "reason": "Auxiliary-loss normalization and FSDP CPU RAM efficient loading are unrelated issues." }, { "left": "issue:44514", "right": "issue:45381", "accept": false, - "reason": "Both involve chat/template or vision inputs, but one is batching crash in apply_chat_template and the other is wrong vision_position_ids for Qwen2.5-VL video." + "reason": "Both are Qwen/VL-related, but one is chat template batching with padding=False and the other is wrong vision_position_ids for video input; different bugs." }, { "left": "issue:43720", "right": "issue:44898", "accept": false, - "reason": "BitNet packed-weight loading bug is unrelated to Perceiver image classification on non-default resolution." + "reason": "BitNet packed-weight loading and Perceiver interpolation behavior are unrelated code paths." }, { - "left": "issue:41762", - "right": "issue:44936", + "left": "issue:44484", + "right": "issue:44704", "accept": false, - "reason": "Gemma3 ZeRO-3 load index error is unrelated to trainer.evaluate() failing after training." + "reason": "A save_pretrained shard-size question is not the same as AutoProcessor failing to forward kwargs to cached_file." }, { - "left": "issue:43452", - "right": "issue:43577", + "left": "issue:41762", + "right": "issue:44936", "accept": false, - "reason": "Both concern model/tokenizer loading, but gguf_file breakage is unrelated to Blip2 dtype propagation." + "reason": "DeepSpeed ZeRO-3 model loading error versus trainer.evaluate failure after train; different layers of the stack and different symptoms." }, { "left": "issue:24643", "right": "issue:29127", "accept": false, - "reason": "DeepSpeed weight dimensionality error and LayoutLMv3 box validation/error-message clarity are different bugs." + "reason": "DeepSpeed weight-dimension runtime error and LayoutLMv3 box-info error-message clarity are unrelated." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "gguf_file loading breakage and Blip2 dtype propagation are distinct model-loading regressions." }, { "left": "issue:44625", "right": "issue:44877", "accept": false, - "reason": "Qwen3.5 num_labels propagation and strict config blocking granite_speech config are different config bugs." + "reason": "Both touch config handling, but one is Qwen3.5 num_labels propagation and the other is granite_speech strict config loading; not the same bug." }, { "left": "issue:44514", "right": "issue:45290", "accept": false, - "reason": "Both are chat-template related, but one is batched padding behavior and the other is assistant tool-call handling." + "reason": "Both are chat-template crashes, but the triggering conditions differ materially: batched input with padding=False versus assistant tool-call messages with no content." }, { "left": "issue:43452", "right": "issue:44843", "accept": false, - "reason": "gguf_file loading breakage is unrelated to the mistral regex offline model_info bug." + "reason": "AutoTokenizer gguf_file loading and mistral-regex offline-mode behavior are unrelated." }, { "left": "issue:44466", "right": "issue:45005", "accept": false, - "reason": "Serialization of tied lm_head weights is distinct from tied-weight issues in translation models, even if both mention weights." - }, - { - "left": "issue:43643", - "right": "issue:44561", - "accept": false, - "reason": "Different root causes: missing fields from trust_remote_code AutoConfig versus removal of is_torch_fx_available breaking remote-code models." + "reason": "Both mention tied weights, but one is device-dependent serialization of lm_head.weight while the other is a broader translation-model issue; insufficient evidence they share one concrete bug." }, { "left": "issue:43819", "right": "issue:44811", "accept": false, - "reason": "DAC from_latents mismatch is unrelated to Whisper processor.batch_decode ignoring skip_special_tokens." + "reason": "DAC latent decoding mismatch and Whisper batch_decode skip_special_tokens handling are unrelated." }, { "left": "issue:43295", "right": "issue:45356", "accept": false, - "reason": "Processor/tokenizer API regression is unrelated to Kimi-K2.5 tokenizer codec handling and warning text." + "reason": "Both are regressions around processors/tokenizers, but one is processor.tokenizer access and image passing in custom code, while the other is Kimi-K2.5 codec handling and regex warnings." }, { - "left": "issue:43653", - "right": "issue:44568", + "left": "issue:43643", + "right": "issue:44561", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug is unrelated to mdeberta-v3 add_special_tokens not adding BOS/EOS in v5." + "reason": "Both involve trust_remote_code, but one is missing config fields and the other is removal of is_torch_fx_available breaking remote code; different root causes." }, { "left": "issue:44464", "right": "issue:44743", "accept": false, - "reason": "Chunked generation inconsistency with compiled forward is unrelated to Qwen3.5 recurrent-state reset in modular_qwen3_5.py." + "reason": "Both mention generation/cache behavior, but one is compiled-forward chunked generation inconsistency and the other is recurrent states resetting in modular_qwen3_5; not the same bug." }, { "left": "issue:44936", "right": "issue:45823", "accept": false, - "reason": "Trainer lifecycle behavior and Gemma4 PLE device mismatch are separate problems." + "reason": "Trainer lifecycle failure and Gemma4 PLE device-map mismatch are unrelated." }, { - "left": "issue:43688", - "right": "issue:45440", + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Both concern tokenizer special tokens, but one is BigBird mask-token registration and the other is mdeberta BOS/EOS addition; different tokenizers and defects." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Siglip2 zero probabilities and processor.tokenizer regression are unrelated." + }, + { + "left": "issue:45375", + "right": "issue:45920", + "accept": false, + "reason": "Both are Qwen3.5-related, but one is a missing vision-config field dropped by strict and the other is wrong token IDs from AutoTokenizer across many models; different layers and symptoms." + }, + { + "left": "issue:45381", + "right": "issue:44792", "accept": false, - "reason": "Auxiliary loss normalization in MoE models is unrelated to DeepseekV3MoE divergence from remote implementation." + "reason": "Qwen2.5-VL video position-id bug and janus test failure are not the same issue." } ] }, @@ -6254,11 +6280,9 @@ "issue:34567", "issue:35141", "issue:36010", - "issue:38175", "issue:41720", "issue:43122", "issue:43278", - "issue:43295", "issue:43329", "issue:43452", "issue:43493", @@ -6266,6 +6290,7 @@ "issue:43531", "issue:43550", "issue:43577", + "issue:43606", "issue:43646", "issue:43653", "issue:43688", @@ -6291,285 +6316,252 @@ "issue:44991", "issue:45325", "issue:45357", + "issue:45440", "issue:45588", "issue:45701", "issue:45704" ], "soft_pairs": [ - "issue:38175|issue:43295", + "issue:43688|issue:45440", + "issue:43653|issue:44361", "issue:44361|issue:44442", + "issue:41720|issue:43906", "issue:43688|issue:45704", - "issue:43653|issue:44361", "issue:30064|issue:33666", - "issue:41720|issue:43906", - "issue:43577|issue:43697", "issue:28282|issue:32090", + "issue:43577|issue:43697", "issue:28282|issue:33357", - "issue:43688|issue:45701", "issue:28282|issue:35141", + "issue:43688|issue:45701", "issue:43646|issue:43749", "issue:44843|issue:44991", "issue:43653|issue:43927", "issue:43531|issue:43742", "issue:43550|issue:43927", "issue:44464|issue:44977", - "issue:43577|issue:45588", "issue:43329|issue:45325", + "issue:43577|issue:45588", "issue:43577|issue:44554", "issue:44479|issue:44625", "issue:43550|issue:44361", "issue:43122|issue:43906", "issue:32090|issue:34567", - "issue:43504|issue:43550", "issue:41720|issue:44464", + "issue:43504|issue:43550", "issue:43504|issue:43720", - "issue:44849|issue:45357", "issue:43504|issue:44610", "issue:32090|issue:33666", + "issue:44849|issue:45357", "issue:43452|issue:44451", "issue:43278|issue:43688", "issue:43742|issue:44821", + "issue:33666|issue:36010", "issue:43493|issue:43825", - "issue:33666|issue:36010" + "issue:43606|issue:43927", + "issue:43720|issue:43927" ], - "prepared_review_unit_hash": "888c1bbec8358cea0ea6bd8814a647c5451f25ac6994f110282640c863c32ad4", + "prepared_review_unit_hash": "db154134b66dd838e029b544a663f40c12cc90116d2c9a79079e04331324a05e", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 35, - "serialized_chars": 38233, - "estimated_input_tokens": 9559, - "estimated_eval_tokens": 19374 + "soft_pair_count": 37, + "serialized_chars": 38809, + "estimated_input_tokens": 9703, + "estimated_eval_tokens": 19662 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is very heterogeneous: most items are distinct model-loading, training, tokenizer, or multimodal regressions rather than true duplicates. I found no soft-edge pair that clearly meets the same-underlying-bug bar.", - "confidence": 0.79, - "canonical_issue_reason": "issue:45701 is the broadest and most umbrella-like title in the set, so it is the least-bad representative bucket for version-dependent tokenizer behavior; however, the cluster is too mixed for a strong canonical choice.", + "summary": "The cluster is highly heterogeneous: it mixes unrelated tokenizer, vision, training, generation, loading, and dtype regressions. The soft-similarity links are mostly superficial, so none of the candidate pairs look like true duplicates or mergeable PR-equivalents.", + "confidence": 0.93, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45701 is the most general open issue here and the best single representative for a loose cluster centered on version-related tokenizer changes, though it does not cover many of the other items well.", + "best_issue_reason": "issue:43646 is the broadest regression-style report and the closest thing to an anchor, but the set is too mixed to treat it as a real duplicate hub.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:38175", - "right": "issue:43295", + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Different model families and different bugs: auxiliary-loss normalization vs DeepSeekV3 implementation divergence." + }, + { + "left": "issue:43653", + "right": "issue:44361", "accept": false, - "reason": "Different subsystems and failures: zero-probability output on SigLIP2 vs a regression around processor.tokenizer/image passing." + "reason": "Both involve tokenizers, but they are different tokenizer classes with different failure modes and fixes." }, { "left": "issue:44361", "right": "issue:44442", "accept": false, - "reason": "Different tokenizer classes and distinct load-time errors; not the same bug." + "reason": "Both are tokenizer-loading issues, but the affected tokenizer and error path are unrelated." }, { - "left": "issue:43688", - "right": "issue:45704", + "left": "issue:41720", + "right": "issue:43906", "accept": false, - "reason": "Auxiliary-loss normalization math bug vs a T5/apex RMSNorm memory leak; unrelated code paths." + "reason": "No evidence they describe the same CUDA/device-mapping bug; 43906 is just an isolated repro reference." }, { - "left": "issue:43653", - "right": "issue:44361", + "left": "issue:43688", + "right": "issue:45704", "accept": false, - "reason": "Special-token registration bug in BigBirdTokenizer vs MLukeTokenizer task attribute failure." + "reason": "Unrelated subsystems: expert-loss normalization vs T5 RMSNorm memory leak." }, { "left": "issue:30064", "right": "issue:33666", "accept": false, - "reason": "Segmentation-map processing failure vs Qwen2-VL multi-GPU training request; different problems." + "reason": "Image segmentation map processing and multi-GPU training are different problems." }, { - "left": "issue:41720", - "right": "issue:43906", + "left": "issue:28282", + "right": "issue:32090", "accept": false, - "reason": "One is a concrete Qwen3 cudaErrorAssert report; the other is only an isolated reproduction of another issue, with no evidence they are the same bug." + "reason": "Import-time PyTorch detection error vs trainer broadcast NoneType error; not the same bug." }, { "left": "issue:43577", "right": "issue:43697", "accept": false, - "reason": "BLIP2 dtype loading issue vs RTDetrV2 output divergence in v5; unrelated symptoms." + "reason": "Different model families and different issues: dtype propagation vs incorrect outputs in RTDetrV2." }, { "left": "issue:28282", - "right": "issue:32090", + "right": "issue:33357", "accept": false, - "reason": "Missing PyTorch import error vs Trainer distributed broadcast TypeError." + "reason": "PyTorch missing import vs MacOS bus error on CLIP are unrelated." }, { "left": "issue:28282", - "right": "issue:33357", + "right": "issue:35141", "accept": false, - "reason": "ImportError for AutoModel/PyTorch missing vs MacOS bus error loading a CLIP model." + "reason": "ImportError and embedding reinitialization are distinct code paths and bugs." }, { "left": "issue:43688", "right": "issue:45701", "accept": false, - "reason": "Aux-loss normalization bug vs a broad tokenization-version regression; different functionality." - }, - { - "left": "issue:28282", - "right": "issue:35141", - "accept": false, - "reason": "Environment import failure vs embedding reinitialization after resize; not the same issue." + "reason": "Aux-loss normalization is unrelated to a tokenization-version regression." }, { "left": "issue:43646", "right": "issue:43749", "accept": false, - "reason": "Custom model initialization breakage vs FSDP CPU RAM-efficient loading breakage." + "reason": "Custom model initialization breakage and FSDP CPU RAM efficient loading are different failures." }, { "left": "issue:44843", "right": "issue:44991", "accept": false, - "reason": "Offline-mode failure from unconditional model_info() call vs tokenizer loading failure for a specific model." + "reason": "Both are tokenizer-related, but one is an offline hub lookup bug and the other is a specific tokenizer load failure." }, { "left": "issue:43653", "right": "issue:43927", "accept": false, - "reason": "Tokenizer special-token registration bug vs config save/load losing custom token IDs." + "reason": "Special-token registration bug vs saved custom token IDs being lost; different artifacts and causes." }, { "left": "issue:43531", "right": "issue:43742", "accept": false, - "reason": "Qwen3-MoE sliding_window issue vs key error loading MobileLLM-125M." + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to a key error loading MobileLLM-125M." }, { "left": "issue:43550", "right": "issue:43927", "accept": false, - "reason": "torch.compile/SDPA failure in Bamba vs DiaConfig token ID persistence bug." + "reason": "torch.compile/SDPA failure and config save/load token-ID loss are unrelated." }, { "left": "issue:44464", "right": "issue:44977", "accept": false, - "reason": "Compiled-forward chunked generation mismatch vs flash-attention generation failure; different triggers." + "reason": "Compiled-forward chunked generation and flash-attention generation failures are different code paths." }, { - "left": "issue:43577", - "right": "issue:45588", + "left": "issue:43329", + "right": "issue:45325", "accept": false, - "reason": "BLIP2 dtype-loading bug vs flash_attention integration crashing on s_aux=None." + "reason": "Video-token counting bug and still-image rope-index scaling bug are distinct multimodal issues." }, { - "left": "issue:43329", - "right": "issue:45325", + "left": "issue:43577", + "right": "issue:45588", "accept": false, - "reason": "Undefined helper in video token counting vs still-image temporal position scaling bug." + "reason": "BLIP2 dtype propagation and flash_attention s_aux=None crash are unrelated." }, { "left": "issue:43577", "right": "issue:44554", "accept": false, - "reason": "BLIP2 dtype inconsistency vs MPS attention correctness issue with differing head dims." + "reason": "Dtype loading for BLIP2 does not match an MPS attention correctness issue." }, { "left": "issue:44479", "right": "issue:44625", "accept": false, - "reason": "Video-input regression across Qwen VL models vs num_labels propagation bug in Qwen3.5." - }, - { - "left": "issue:43550", - "right": "issue:44361", - "accept": false, - "reason": "torch.compile/SDPA model failure vs tokenizer AttributeError on tasks." - }, - { - "left": "issue:43122", - "right": "issue:43906", - "accept": false, - "reason": "Tokenizer output changes across versions vs an isolated reproduction of another issue; no direct same-bug evidence." - }, - { - "left": "issue:32090", - "right": "issue:34567", - "accept": false, - "reason": "Different Trainer-related problems: NoneType broadcast vs num_input_tokens_seen not updating." - }, - { - "left": "issue:43504", - "right": "issue:43550", - "accept": false, - "reason": "Legacy-field model loading failure vs Bamba compile/SDPA runtime failure." - }, - { - "left": "issue:41720", - "right": "issue:44464", - "accept": false, - "reason": "Qwen3 auto device mapping cuda assert vs chunked generation inconsistency under compiled forward." - }, - { - "left": "issue:43504", - "right": "issue:43720", - "accept": false, - "reason": "Beit legacy-field loading bug vs BitNet packed-weight unpacking during accelerate loading." + "reason": "Video-input regression in Qwen VL models is unrelated to num_labels propagation in Qwen3.5." }, { "left": "issue:44849", "right": "issue:45357", "accept": false, - "reason": "Hidden-states output bug vs save_pretrained writing incorrect visual encoder keys." + "reason": "Different bugs: hidden-states output mismatch vs save_pretrained writing wrong visual keys." }, { "left": "issue:43452", "right": "issue:44451", "accept": false, - "reason": "gguf_file breaks from_pretrained paths vs a specific model failing to load; not the same concrete defect." + "reason": "General gguf tokenizer/model loading breakage and a specific model tokenizer load failure are not the same underlying defect." }, { "left": "issue:43278", "right": "issue:43688", "accept": false, - "reason": "Embedding dtype drift between train/eval vs auxiliary-loss normalization; unrelated behaviors." + "reason": "Embedding dtype mismatch during eval is unrelated to auxiliary-loss normalization." }, { "left": "issue:43742", "right": "issue:44821", "accept": false, - "reason": "Model loading key error vs AutoImageProcessor-from-URL failure." + "reason": "Loading a text model with a key error is unrelated to loading an image processor from a URL." }, { - "left": "issue:43493", - "right": "issue:43825", + "left": "issue:33666", + "right": "issue:36010", "accept": false, - "reason": "SigLIP2 numerical discrepancy vs an incorrect pipeline error message about translation support." + "reason": "Multi-GPU training and a GenerationMixin import error are unrelated." }, { - "left": "issue:33666", - "right": "issue:36010", + "left": "issue:43493", + "right": "issue:43825", "accept": false, - "reason": "Qwen2-VL multi-GPU training topic vs GenerationMixin import error; unrelated." + "reason": "SigLIP2 implementation discrepancy and pipeline translation-task messaging are different issues." }, { - "left": "issue:43504", - "right": "issue:44610", + "left": "issue:43606", + "right": "issue:43927", "accept": false, - "reason": "Beit legacy-field loading bug vs OmDet-Turbo processor/model input-size mismatch." + "reason": "CPU offload device mismatch has nothing to do with DiaConfig token-ID persistence." }, { - "left": "issue:32090", - "right": "issue:33666", + "left": "issue:43720", + "right": "issue:43927", "accept": false, - "reason": "Trainer broadcast NoneType error vs Qwen2-VL multi-GPU training request; not the same bug." + "reason": "Packed-weight unpacking during accelerate loading is unrelated to DiaConfig save/load token-ID loss." } ] }, @@ -6642,31 +6634,29 @@ "issue:45823" ], "soft_pairs": [ + "issue:43262|issue:43295", "issue:44568|issue:45812", - "issue:43606|issue:43927", - "issue:43720|issue:43927", "issue:43756|issue:44913", - "issue:43262|issue:43295", "issue:43525|issue:43572", - "issue:43493|issue:43844", "issue:28218|issue:33357", "issue:43720|issue:44361", - "issue:39692|issue:43493", "issue:33666|issue:36331", + "issue:39692|issue:43493", + "issue:43493|issue:43844", "issue:43540|issue:44977", "issue:43278|issue:43844", "issue:44964|issue:45405", - "issue:43334|issue:43572", "issue:44625|issue:45584", + "issue:43334|issue:43572", "issue:43257|issue:43299", "issue:43756|issue:45083", "issue:45200|issue:45229", "issue:44488|issue:44843", "issue:43334|issue:45759", "issue:43646|issue:44466", - "issue:43824|issue:44589", "issue:39692|issue:43295", "issue:45229|issue:45823", + "issue:43824|issue:44589", "issue:33357|issue:33666", "issue:44589|issue:45520", "issue:43504|issue:43606", @@ -6676,232 +6666,220 @@ "issue:43504|issue:43653", "issue:44464|issue:45584" ], - "prepared_review_unit_hash": "aa45d98263cea0fcbc901d24376bba6bfea227c437ae5457e594346c33953a80", + "prepared_review_unit_hash": "d4f3abd4771b6d1eef72afec65cb8f4792ae5aa9977fd0c44655e99c95a3678f", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 33, - "serialized_chars": 37367, - "estimated_input_tokens": 9342, - "estimated_eval_tokens": 18940 + "soft_pair_count": 31, + "serialized_chars": 36769, + "estimated_input_tokens": 9193, + "estimated_eval_tokens": 18642 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is heterogeneous: tokenizer regressions, multimodal model loading failures, trainer/config issues, and CI/device mismatches are all mixed together. The soft-similarity pairs look superficially related by keywords, but they do not describe the same underlying bug or change, so none should be merged as duplicates.", - "confidence": 0.19, - "canonical_issue_reason": "issue:45812 is the broadest tokenizer regression report in the set: it affects all Granite models, has the strongest cross-reference signal, and is clearer as a canonical anchor than the more model-specific tokenizer bugs.", + "summary": "The cluster is heterogeneous: most pairs share only broad subsystem or model-family similarity, not the same underlying bug. I would not merge any of the soft-edge pairs.", + "confidence": 0.92, + "canonical_issue_reason": "No single canonical issue stands out for this mixed set; if one must be chosen, issue 45812 is the clearest broad-impact regression with a concrete tokenizer failure.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45812 is the best representative issue because it is concrete, broadly scoped, and central to the tokenizer-regression subset; the rest of the issues are too diverse to serve as a single duplicate cluster.", + "best_issue_reason": "Issue 45812 is the strongest standalone candidate because it describes a silent, model-wide regression affecting Granite tokenization and is clearly reproducible.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44568", - "right": "issue:45812", - "accept": false, - "reason": "Both are tokenizer bugs, but one is mdeberta BOS/EOS insertion and the other is Granite token-ID regression; different code paths and likely different fixes." - }, - { - "left": "issue:43606", - "right": "issue:43927", + "left": "issue:43262", + "right": "issue:43295", "accept": false, - "reason": "CPU offload device mismatch is unrelated to DiaConfig token-ID persistence; no shared bug mechanism." + "reason": "Both involve processor/tokenizer behavior, but one is an audio sampling-rate default and the other is a multimodal regression around processor.tokenizer/images; different code paths." }, { - "left": "issue:43720", - "right": "issue:43927", + "left": "issue:44568", + "right": "issue:45812", "accept": false, - "reason": "Packed-weight unpacking during accelerate loading is a model-loading path issue, not the same as lost custom token IDs in config." + "reason": "Both are tokenizer regressions, but they affect different models and fail differently: BOS/EOS handling for mdeberta-v3 vs wrong token IDs for Granite." }, { "left": "issue:43756", "right": "issue:44913", "accept": false, - "reason": "RoPE layer-count mismatch in Smollm3 is not the same as GPTNeoX rotary_pct resetting on reload." - }, - { - "left": "issue:43262", - "right": "issue:43295", - "accept": false, - "reason": "Audio chat-template sampling-rate defaulting and processor.tokenizer regression are unrelated multimodal API changes." + "reason": "Both mention RoPE-related config behavior, but one is about Smollm3 layer dropping and the other is a GPTNeoX config value not persisting on reload." }, { "left": "issue:43525", "right": "issue:43572", "accept": false, - "reason": "Both mention missing pad-token fields, but they are different configs/models with different failure surfaces." - }, - { - "left": "issue:43493", - "right": "issue:43844", - "accept": false, - "reason": "SigLIP2 implementation drift and ZeRO-3 gradient growth are different numerical/model-parallelism problems." + "reason": "Both are missing token-id fields in configs, but they concern different model families and different attributes (pad_token_id vs pad_token_idx)." }, { "left": "issue:28218", "right": "issue:33357", "accept": false, - "reason": "Tokenizer added-space behavior is unrelated to the MacOS bus error on a community CLIP model." + "reason": "Tokenizer spacing bug vs MacOS bus error in a CLIP model; no shared underlying defect." }, { "left": "issue:43720", "right": "issue:44361", "accept": false, - "reason": "These are distinct tokenizer/model-loading failures; one is packed weights with accelerate, the other is MLukeTokenizer task errors." + "reason": "BitNet accelerate-loading unpacking bug vs tokenizer AttributeError on tasks; unrelated." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Multi-GPU training request for Qwen2-VL vs a CustomTrainer signature break; different problems." }, { "left": "issue:39692", "right": "issue:43493", "accept": false, - "reason": "SigLIP2 docs/example errors are not the same as the HF-vs-JAX implementation discrepancy." + "reason": "Both touch SigLIP2, but one is a documentation/example mismatch and the other is an implementation discrepancy with JAX; not the same bug." }, { - "left": "issue:33666", - "right": "issue:36331", + "left": "issue:43493", + "right": "issue:43844", "accept": false, - "reason": "Qwen2-VL multi-GPU training and CustomTrainer compute_loss signature mismatch are different training-time issues." + "reason": "SigLIP2 implementation mismatch vs a gradient blow-up with ZeRO-3; unrelated." }, { "left": "issue:43540", "right": "issue:44977", "accept": false, - "reason": "Video-input processing in Qwen3OmniMoe is separate from Qwen3.5 flash-attention generation behavior." + "reason": "Qwen3 OmniMoe video-input validation vs Qwen3.5 flash-attention generation failure; different code paths." }, { "left": "issue:43278", "right": "issue:43844", "accept": false, - "reason": "Embedding dtype drift in evaluate is not the same as gradient explosion under HfDeepSpeedConfig + ZeRO-3." + "reason": "Embedding dtype drift in eval vs abnormal gradients under DeepSpeed; no direct match." }, { "left": "issue:44964", "right": "issue:45405", "accept": false, - "reason": "Model loading failure for Phi-4 multimodal is unrelated to the PEFT version bump issue." + "reason": "Phi-4 multimodal loading failure vs an unreleased PEFT version bump; unrelated." }, { - "left": "issue:43334", - "right": "issue:43572", + "left": "issue:44625", + "right": "issue:45584", "accept": false, - "reason": "Qwen3-VL pad_token_id absence and StableLm pad_token_idx absence are similar in theme but are separate config regressions." + "reason": "Qwen3.5 config propagation bug vs Whisper empty-transcription generation failure; different models and failures." }, { - "left": "issue:44625", - "right": "issue:45584", + "left": "issue:43334", + "right": "issue:43572", "accept": false, - "reason": "num_labels propagation and Whisper empty-transcription generation after align_special_tokens are unrelated bugs." + "reason": "Both are missing pad-token metadata in configs, but they affect different model classes and arise in different loading paths." }, { "left": "issue:43257", "right": "issue:43299", "accept": false, - "reason": "Both concern Qwen3 MOE loading, but one is accelerate+deepspeed weight conversion and the other is generic transformers dev loading; not clearly the same defect." + "reason": "Both involve Qwen3 MOE loading, but one is accelerate/deepspeed weight conversion and the other is a dev-version multimodal config break; not the same issue." }, { "left": "issue:43756", "right": "issue:45083", "accept": false, - "reason": "Smollm3 RoPE-layer loss and qwen3_omni_moe feature-extract length behavior do not share a concrete code path." + "reason": "Both are Qwen3 OmniMoe-related, but one is RoPE-layer selection and the other is a feature-length helper behaving unexpectedly." }, { "left": "issue:45200", "right": "issue:45229", "accept": false, - "reason": "Text-only mm_token_type_ids defaults and Gemini 4 multi-GPU CUDA OOM are different multimodal inference/training problems." + "reason": "Text-only token-type-id defaulting vs multi-GPU inference OOM; unrelated." }, { "left": "issue:44488", "right": "issue:44843", "accept": false, - "reason": "Loading cjvt/sleng-bert and offline model_info() calls in AutoTokenizer are separate tokenizer-loading failures." + "reason": "Model-loading failure for a specific checkpoint vs offline-mode break in AutoTokenizer patching; different causes." }, { "left": "issue:43334", "right": "issue:45759", "accept": false, - "reason": "Both involve composite Qwen configs, but one is missing pad_token_id and the other is from_config not unwrapping text_config; different failure points." + "reason": "Both relate to composite Qwen multimodal configs, but one is a missing pad_token_id attribute and the other is from_config not unwrapping text_config." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "Custom model initialization breaks under v5, while lm_head.weight serialization depends on device/tied-weights behavior; not the same bug." - }, - { - "left": "issue:43824", - "right": "issue:44589", - "accept": false, - "reason": "Missing import/export of Qwen2_5_VL is unrelated to Float8 storage lookup failure." + "reason": "Both are v5 regressions, but one is custom model initialization and the other is tied-weight serialization; unrelated code paths." }, { "left": "issue:39692", "right": "issue:43295", "accept": false, - "reason": "Documentation/example mistakes are not the same as the regression around processor.tokenizer and image passing." + "reason": "Documentation/example issues vs a processor/tokenizer regression; not the same underlying bug." }, { "left": "issue:45229", "right": "issue:45823", "accept": false, - "reason": "Both are Gemma4-related, but one is inference OOM and the other is device mismatch in PLE; different runtime issues." + "reason": "Both involve Gemma4 and distributed/device behavior, but one is CUDA OOM during multi-GPU inference and the other is a PLE device mismatch on forward." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "Import error for a Qwen2.5-VL class vs a missing Float8 storage type during load; unrelated." }, { "left": "issue:33357", "right": "issue:33666", "accept": false, - "reason": "A MacOS bus error on CLIP is unrelated to Qwen2-VL multi-GPU training." + "reason": "MacOS bus error in CLIP vs Qwen2-VL multi-GPU training; no shared bug." }, { "left": "issue:44589", "right": "issue:45520", "accept": false, - "reason": "Float8 storage deserialization and flash_attn import_utils KeyError are unrelated environment/runtime issues." + "reason": "Different import/load failures: missing Float8 storage type vs flash_attn KeyError on Python 3.13." }, { "left": "issue:43504", "right": "issue:43606", "accept": false, - "reason": "Legacy-field load failure for Beit semantic segmentation is unrelated to CPU offload device mismatch." + "reason": "Legacy-field model loading bug vs CPU offload device mismatch; unrelated." }, { "left": "issue:43572", "right": "issue:44291", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig and unexpected _is_hf_initialized argument during init_empty_weights are different config/init regressions." + "reason": "Missing pad_token_idx in StableLmConfig vs init_empty_weights rejecting _is_hf_initialized; different compatibility issues." }, { "left": "issue:44423", "right": "issue:44734", "accept": false, - "reason": "Continuous-batching multimodal crash and KV-cache continuation indexing bug are both serve-side, but they are distinct code paths and fixes." + "reason": "Both are serving crashes, but one is multimodal continuous-batching and the other is KV-cache continuation indexing; not the same root cause." }, { "left": "issue:43927", "right": "issue:44361", "accept": false, - "reason": "DiaConfig token-ID persistence and MLukeTokenizer task AttributeError are unrelated model/tokenizer bugs." + "reason": "DiaConfig save/load token-ID loss vs MLukeTokenizer AttributeError; unrelated." }, { "left": "issue:43504", "right": "issue:43653", "accept": false, - "reason": "Legacy-field load failure in Beit is not the same as BigBirdTokenizer mask-token registration/decode issue." + "reason": "Beit legacy-field loading issue vs BigBird tokenizer special-token registration; different components and failures." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Chunked generation inconsistency with compiled forward and Whisper empty transcription after align_special_tokens are different generation regressions." + "reason": "Compiled-forward chunked generation inconsistency vs Whisper empty-transcription handling; unrelated." } ] }, @@ -6981,20 +6959,20 @@ "issue:44792|issue:45405", "issue:29942|issue:30333", "issue:44521|issue:45290", - "issue:43232|issue:44849", "issue:43844|issue:44964", - "issue:44554|issue:45446", "issue:28218|issue:34567", "issue:43582|issue:44977", + "issue:44554|issue:45446", "issue:45084|issue:45520", + "issue:43232|issue:44849", "issue:15354|issue:38175", "issue:43540|issue:45200", "issue:45072|issue:45198", - "issue:44315|issue:44849", "issue:45200|issue:45823", "issue:37428|issue:43957", - "issue:44554|issue:45356", + "issue:44315|issue:44849", "issue:43493|issue:43701", + "issue:44554|issue:45356", "issue:43295|issue:44610", "issue:43873|issue:44464", "issue:44466|issue:45127", @@ -7002,196 +6980,196 @@ "issue:44792|issue:45563", "issue:45200|issue:45325" ], - "prepared_review_unit_hash": "03a5e53b571e5224d9dd83f99ea05e1722fd60f372025d9aec2a4ebecf38f5a1", + "prepared_review_unit_hash": "5b0a397b57349e1108c6ee4006be66b6636fc4fe9159e4b6db54ebf75ca5485b", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 27, - "serialized_chars": 35281, - "estimated_input_tokens": 8821, - "estimated_eval_tokens": 17898 + "serialized_chars": 35280, + "estimated_input_tokens": 8820, + "estimated_eval_tokens": 17896 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is not a true duplicate cluster: the items span many unrelated bugs, with only superficial topical overlap in a few model/failure areas. I would reject all soft-edge merges.", - "confidence": 0.91, - "canonical_issue_reason": "issue:43994 is the most representative concrete user-facing regression in the set: it has a clear model-specific symptom, a likely reproducible failure mode, and it is one of the few items that is semantically adjacent to another issue in the cluster.", + "summary": "The items are mostly heterogeneous, and the soft-similarity pairs do not look like true duplicates. They share broad themes or model families, but not the same concrete bug or fix path.", + "confidence": 0.83, + "canonical_issue_reason": "No single canonical issue stands out for the whole set; the cluster is too mixed to represent one underlying bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43994 is the best single issue to anchor this cluster because it is a clear, actionable model-behavior bug rather than a narrow environment/test/reporting problem.", + "best_issue_reason": "If one issue must represent the set, issue 44521 is the most broadly impactful open regression with active discussion, but it is not a duplicate target for the other items.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "Same model family, but different bugs: one reports nonsensical outputs in AutoModel/pipeline, the other an unexpected `position id` key issue. Similar area, not the same underlying defect." + "reason": "Both mention position-id/output regressions, but one is a SigLIP2 inference quality bug and the other is a generic unexpected-key issue; different failures." }, { "left": "issue:43606", "right": "issue:44112", "accept": false, - "reason": "Both are CI/device-mismatch flavored, but they involve different models and different failure causes: CPU offload on bark-small versus a stale device override test in GraniteSpeech." + "reason": "Both are CI/device-mismatch failures, but they affect different models and code paths." }, { "left": "issue:43575", "right": "issue:45405", "accept": false, - "reason": "Completely different problems: Qwen2.5 TP OOM versus an unreleased PEFT version requirement." + "reason": "OOM during tensor-parallel loading is unrelated to a PEFT version pin regression." }, { "left": "issue:43278", "right": "issue:43701", "accept": false, - "reason": "Different code paths and symptoms: dtype change in evaluation versus resume-from-checkpoint key mismatch." + "reason": "A dtype shift in eval is not the same bug as a resume-from-checkpoint key mismatch." }, { "left": "issue:44792", "right": "issue:45405", "accept": false, - "reason": "A janus image-generation test failure is unrelated to the PEFT version pinning issue." + "reason": "Janus test failure and MIN_PEFT_VERSION bump are unrelated." }, { "left": "issue:29942", "right": "issue:30333", "accept": false, - "reason": "One is a Flash Attention 2 test failure, the other is MLflow reporting not being updated after training failure; no shared underlying bug." + "reason": "Flash Attention 2 test failures do not match the MLflow reporting issue." }, { "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "Both touch chat templating, but one is all-zero assistant masks for multimodal inputs while the other crashes on tool-call assistant messages with no content; distinct failures." - }, - { - "left": "issue:43232", - "right": "issue:44849", - "accept": false, - "reason": "Generation kwargs after `sync_gpus` and Qwen3.5 `output_hidden_states=True` are separate bugs in different generation paths." + "reason": "Both involve chat templating, but one is all-zero assistant masks for multimodal inputs and the other is a crash on tool-call assistant messages." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "ZeRO-3 gradient behavior and a multimodal model load failure are unrelated issues." - }, - { - "left": "issue:44554", - "right": "issue:45446", - "accept": false, - "reason": "An MPS attention correctness bug is not the same as a PyTorch version check/import issue for AuxRequest." + "reason": "ZeRO-3 gradient behavior and latest-transformers model loading are different bugs." }, { "left": "issue:28218", "right": "issue:34567", "accept": false, - "reason": "Tokenizer spacing after added tokens and `num_input_tokens_seen` not updating are different tokenizer/trainer-state problems." + "reason": "Tokenizer spacing after added tokens is unrelated to TrainerState token-count updates." }, { "left": "issue:43582", "right": "issue:44977", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError and Qwen3.5 flash-attention generation issues do not share the same root cause." + "reason": "Apple Silicon allocator warmup TypeError and Qwen3.5 flash-attention generation bug are unrelated." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "An MPS attention correctness issue is not the same as a PyTorch version-check import bug." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "Compile-time template-node errors and a Python 3.13 `flash_attn` import KeyError are unrelated." + "reason": "Compilation failure for non-template nodes and a Python 3.13 flash_attn import KeyError are separate issues." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation state update after sync_gpus is unrelated to the hidden-states output bug." }, { "left": "issue:15354", "right": "issue:38175", "accept": false, - "reason": "TorchScript generator-expression export failure and SigLIP2 zero-probability outputs are different bugs with no shared code-path evidence." + "reason": "TorchScript GeneratorExp support and SigLIP2 zero-probabilities are unrelated." }, { "left": "issue:43540", "right": "issue:45200", "accept": false, - "reason": "Video-processing failure in Qwen3OmniMoe and Gemma4 `mm_token_type_ids` defaults are separate multimodal input bugs." + "reason": "Video processing failure in Qwen3OmniMoe is not the same as Gemma 4 token-type defaults." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "A bfloat16 dtype mismatch in inference and Wav2Vec2 save/tokenization failures are unrelated." - }, - { - "left": "issue:44315", - "right": "issue:44849", - "accept": false, - "reason": "Liger Kernel not being applied with `model_init` is unrelated to the Qwen3.5 hidden-states regression." + "reason": "A dtype mismatch in CI and a Wav2Vec2 save/tokenization failure are different concrete bugs." }, { "left": "issue:45200", "right": "issue:45823", "accept": false, - "reason": "Both concern Gemma4, but one is missing/default `mm_token_type_ids` while the other is a PLE device mismatch under `device_map=\"auto\"`; different defects." + "reason": "Same model family, but one is missing mm_token_type_ids defaults and the other is a PLE device mismatch; not one fix." }, { "left": "issue:37428", "right": "issue:43957", "accept": false, - "reason": "Importing `_flash_supports_window_size` and meta-device model loading failures are different flash-attention/model-loading issues." + "reason": "flash_supports_window_size import failure and meta-device model loading regressions are unrelated." }, { - "left": "issue:44554", - "right": "issue:45356", + "left": "issue:44315", + "right": "issue:44849", "accept": false, - "reason": "MPS attention correctness and Kimi-K2.5 tokenizer regression are unrelated." + "reason": "Liger Kernel application with model_init is unrelated to the Qwen3.5 hidden-states bug." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "A SigLIP2 HF-vs-JAX discrepancy is not the same as a resume-from-checkpoint key mismatch." + "reason": "SigLIP2 JAX discrepancy and checkpoint key mismatch do not share a code path." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "MPS attention dim mismatch and tokenizer codec/warning regression are different subsystems." }, { "left": "issue:43295", "right": "issue:44610", "accept": false, - "reason": "Processor/tokenizer regression and OmDet-Turbo image-size mismatch are different model-integration bugs." + "reason": "Processor/tokenizer regression and OmDet-Turbo image-size mismatch are separate bugs." }, { "left": "issue:43873", "right": "issue:44464", "accept": false, - "reason": "Quantization/offloading behavior and chunked generation with compiled forward are distinct problems." + "reason": "Quantization offloading behavior and compiled-forward chunked generation are not the same issue." }, { "left": "issue:44466", "right": "issue:45127", "accept": false, - "reason": "Serialization of tied weights on different devices is unrelated to LoRA merging with extended vocabulary collapse." + "reason": "Both touch weights/serialization, but one is inconsistent lm_head serialization and the other is LoRA merge collapse with extended vocab." }, { "left": "issue:44898", "right": "issue:45412", "accept": false, - "reason": "Perceiver positional interpolation failure and RT-DETR memory leak are not the same underlying bug." + "reason": "Perceiver interpolation failure and RT-DETR memory release are unrelated." }, { "left": "issue:44792", "right": "issue:45563", "accept": false, - "reason": "A failing Janus image-generation test and a stale paged-generate warning are different issues." + "reason": "Janus image-generation test failure and a stale paged-generate warning are different problems." }, { "left": "issue:45200", "right": "issue:45325", "accept": false, - "reason": "Gemma4 multimodal token-type defaults and Qwen2.5-VL rope-index scaling are different model-specific bugs." + "reason": "Gemma 4 token-type defaults and Qwen2.5-VL rope-index scaling are unrelated." } ] }, @@ -7226,7 +7204,6 @@ "issue:43334", "issue:43335", "issue:43388", - "issue:43425", "issue:43450", "issue:43493", "issue:43525", @@ -7272,245 +7249,238 @@ "issue:43295|issue:43450", "issue:43334|issue:43525", "issue:43824|issue:43975", - "issue:43388|issue:43526", "issue:43295|issue:43873", + "issue:43388|issue:43526", "issue:38175|issue:45412", "issue:42915|issue:43824", "issue:38175|issue:43493", "issue:28218|issue:35141", - "issue:44442|issue:45198", "issue:44514|issue:44521", + "issue:44442|issue:45198", "issue:43278|issue:43856", "issue:43653|issue:44112", "issue:43827|issue:43901", "issue:43450|issue:43575", - "issue:44206|issue:45584", "issue:43450|issue:43493", + "issue:44206|issue:45584", "issue:45127|issue:45245", "issue:43819|issue:45083", "issue:43262|issue:43450", "issue:43819|issue:44112", "issue:44610|issue:44857", - "issue:44360|issue:44561", "issue:43278|issue:43825", "issue:43606|issue:45083", - "issue:37428|issue:43425", + "issue:44360|issue:44561", "issue:43335|issue:43550" ], - "prepared_review_unit_hash": "42a6e415fa75c9cab80bcff86f1026216a9674c7b54606a46de74344245612b0", + "prepared_review_unit_hash": "a7e47e411776c1f8a67130787f0f5338df0a3c08412da6b34a8966f5b51eb669", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 31, - "serialized_chars": 37091, - "estimated_input_tokens": 9273, - "estimated_eval_tokens": 18802 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36271, + "estimated_input_tokens": 9068, + "estimated_eval_tokens": 18392 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pair looks like a true duplicate or same underlying change; the candidates mostly share broad subsystems or similar regression wording, but each points to a different concrete failure mode. The strongest issue candidate overall is the multimodal `apply_chat_template` bug in issue 44521.", - "confidence": 0.88, - "canonical_issue_reason": "issue 44521 is the clearest anchor: it describes a specific, reproducible multimodal `apply_chat_template` failure with a well-defined wrong output (`assistant_masks` all zeros) and active discussion.", + "summary": "The set is heterogeneous: the soft-similarity pairs mostly share only broad themes (tokenizers, configs, processors, loading, training) but not the same concrete bug or fix path. I would not merge any of the proposed pairs as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue 44521 is the most suitable global representative because it is narrowly scoped, actionable, and tied to a concrete API behavior rather than a vague subsystem complaint.", + "best_issue_reason": "Issue 44521 is the strongest standalone issue here: it is active, has multiple inbound references, and describes a concrete, reproducible multimodal apply_chat_template bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43572", "right": "issue:44987", "accept": false, - "reason": "Both involve loading/runtime breakage, but one is a config field regression and the other is a model-load failure in a different path; not the same bug." + "reason": "Both involve model loading regressions, but one is a specific missing config field and the other is a broad load failure on a different model; not the same underlying bug." }, { "left": "issue:37428", "right": "issue:43824", "accept": false, - "reason": "Both are ImportErrors, but they reference different missing symbols from different modules and model families." + "reason": "These are unrelated ImportErrors for different missing symbols in different code paths." }, { "left": "issue:43844", "right": "issue:44936", "accept": false, - "reason": "These are unrelated training/evaluation issues: one is gradient growth with ZeRO-3, the other is `trainer.evaluate()` after training." + "reason": "One is a gradient explosion under ZeRO-3, the other is an evaluate-after-train failure; different symptoms and fixes." }, { "left": "issue:43572", "right": "issue:43957", "accept": false, - "reason": "Both are version-related loading problems, but one is a missing config attribute and the other is meta-device model loading failures across specific models." + "reason": "Both are loading issues, but they point to different failure modes and code paths: missing pad_token_idx vs meta-device loading breakage." }, { "left": "issue:43295", "right": "issue:43450", "accept": false, - "reason": "One is a processor/tokenizer regression for multimodal custom code; the other is a batched video processor shape bug." + "reason": "Custom model processor/tokenizer regression and batched video-shape bug are different processor problems." }, { "left": "issue:43334", "right": "issue:43525", "accept": false, - "reason": "Both are `pad_token_id`-style attribute errors, but they affect different configs/models and likely different code paths." + "reason": "Both mention missing pad_token_id, but they are model-specific config bugs for different config classes and likely separate fixes." }, { "left": "issue:43824", "right": "issue:43975", "accept": false, - "reason": "These are different import/detokenization problems affecting different models and failure surfaces." + "reason": "An import/export failure is not the same as a detokenization regression." }, { - "left": "issue:43388", - "right": "issue:43526", + "left": "issue:43295", + "right": "issue:43873", "accept": false, - "reason": "Both concern label handling in processors, but one is a metric-gathering bug and the other is `reduce_labels` returning a truncated array." + "reason": "Processor API regression and offloading-with-quantization behavior are unrelated." }, { - "left": "issue:43295", - "right": "issue:43873", + "left": "issue:43388", + "right": "issue:43526", "accept": false, - "reason": "One is a processor API regression; the other is an offloading problem with quantization. Same broad area, different bug." + "reason": "Both concern labels, but one is metrics gathering in Trainer and the other is BeitImageProcessorFast label reduction." }, { "left": "issue:38175", "right": "issue:45412", "accept": false, - "reason": "Both mention model memory/behavior issues, but one is zero probabilities in SigLIP2 and the other is memory not being released for RT-DETR." + "reason": "SigLIP2 probability issues and RT-DETR memory release problems are unrelated." }, { "left": "issue:42915", "right": "issue:43824", "accept": false, - "reason": "Different failures: Qwen3Moe FP8 config error versus a missing top-level import for a VL model class." + "reason": "Qwen3Moe FP8 configuration failure is unrelated to a Qwen2.5-VL import error." }, { "left": "issue:38175", "right": "issue:43493", "accept": false, - "reason": "Same model family, but one is zero probabilities in HF inference and the other is a discrepancy versus the original JAX implementation." + "reason": "Both mention SigLIP2, but one is zero probabilities and the other is an implementation discrepancy versus JAX; not clearly the same bug." }, { "left": "issue:28218", "right": "issue:35141", "accept": false, - "reason": "Tokenizer spacing after added tokens is unrelated to embedding reinitialization during resize when weights are untied." + "reason": "Tokenizer spacing after added tokens and embedding reinitialization after resize are different tokenizer/model initialization bugs." }, { - "left": "issue:44442", - "right": "issue:45198", + "left": "issue:44514", + "right": "issue:44521", "accept": false, - "reason": "Both are tokenizer-related, but one is AutoTokenizer loading and the other is save/tokenization failure for a wav2vec2 model." + "reason": "Both involve apply_chat_template on multimodal inputs, but one is a batched-padding crash and the other is incorrect assistant masks; similar area, not the same concrete bug." }, { - "left": "issue:44514", - "right": "issue:44521", + "left": "issue:44442", + "right": "issue:45198", "accept": false, - "reason": "Both hit `apply_chat_template`, but one crashes on batched padded input while the other returns wrong assistant masks for multimodal input." + "reason": "Different tokenizer/model failures on different architectures; no shared code-path evidence." }, { "left": "issue:43278", "right": "issue:43856", "accept": false, - "reason": "Both concern training/memory or dtype behavior, but one is BF16-to-FP32 embedding dtype drift and the other is MoE memory inefficiency." + "reason": "dtype drift in evaluation and MoE training memory inefficiency are unrelated." }, { "left": "issue:43653", "right": "issue:44112", "accept": false, - "reason": "Both are CI-related, but one is tokenizer special-token registration and the other is a stale device override test." + "reason": "Tokenizer special-token registration and a stale CI device-override test are unrelated." }, { "left": "issue:43827", "right": "issue:43901", "accept": false, - "reason": "Both are docs/API cleanup around pipelines, but they refer to different tasks and different deprecated arguments." + "reason": "Both are docs around pipelines, but they address different API/documentation mismatches." }, { "left": "issue:43450", - "right": "issue:43575", - "accept": false, - "reason": "A batched video processor shape bug is unrelated to a tensor-parallel OOM during model loading." - }, - { - "left": "issue:44206", - "right": "issue:45584", + "right": "issue:43575", "accept": false, - "reason": "Different audio-related regressions: one is a feature extractor center-argument crash, the other is Whisper empty-transcription failure after token alignment." + "reason": "A video processor shape bug and a large-model OOM loading issue are unrelated." }, { "left": "issue:43450", "right": "issue:43493", "accept": false, - "reason": "Same general vision area, but one is batched processor output shape and the other is implementation discrepancy against JAX." + "reason": "Video batching shape and SigLIP2 implementation discrepancy are different problems." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Feature-extractor arg validation and Whisper empty-transcription handling are distinct code paths." }, { "left": "issue:45127", "right": "issue:45245", "accept": false, - "reason": "LoRA merge collapse with extended vocabulary is unrelated to the category-count limit runtime error." + "reason": "LoRA/vocab collapse and category-limit runtime error are unrelated." }, { "left": "issue:43819", "right": "issue:45083", "accept": false, - "reason": "Both are internal helper-function problems, but they affect different DAC/qwen3_omni_moe code paths and behaviors." + "reason": "DAC latent conversion mismatch and qwen3_omni_moe feature-length helper behavior are unrelated." }, { "left": "issue:43262", "right": "issue:43450", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting is unrelated to video processor batch-shape handling." + "reason": "Audio chat-template sample-rate defaulting and batched video output shape are different processor bugs." }, { "left": "issue:43819", "right": "issue:44112", "accept": false, - "reason": "No shared underlying bug: DAC latent decoding mismatch versus a stale CI device override test." + "reason": "DAC forward mismatch and a stale CI device override test are unrelated." }, { "left": "issue:44610", "right": "issue:44857", "accept": false, - "reason": "Different models and failure modes: processor/model input-size mismatch versus AMP/CUDA loss crash." - }, - { - "left": "issue:44360", - "right": "issue:44561", - "accept": false, - "reason": "A discussion about a DSA indexer activation choice is unrelated to `is_torch_fx_available` removal breaking remote-code models." + "reason": "Wrong processor output size and AMP/CUDA loss crash are unrelated." }, { "left": "issue:43278", "right": "issue:43825", "accept": false, - "reason": "Embedding dtype drift during evaluation is unrelated to an incorrect translation-task pipeline error message." + "reason": "Embedding dtype regression and pipeline error-message wording are unrelated." }, { "left": "issue:43606", "right": "issue:45083", "accept": false, - "reason": "CPU offload device mismatch for Bark is unrelated to a qwen3_omni_moe helper-length calculation issue." + "reason": "CPU offload device mismatch and a qwen3_omni_moe helper-function issue are unrelated." }, { - "left": "issue:37428", - "right": "issue:43425", + "left": "issue:44360", + "right": "issue:44561", "accept": false, - "reason": "An import symbol regression is not the same as general Torch 2.10 incompatibility." + "reason": "A model-architecture discussion and a trust_remote_code breakage from removed helper API are not the same bug." }, { "left": "issue:43335", "right": "issue:43550", "accept": false, - "reason": "Both concern model configuration/runtime behavior, but one is sparse-layer creation and the other is torch.compile with SDPA on a different model." + "reason": "Sparse-layer creation and torch.compile/SDPA failure are different model/config regressions." } ] }, @@ -7544,6 +7514,7 @@ "issue:39692", "issue:40444", "issue:43335", + "issue:43425", "issue:43504", "issue:43526", "issue:43550", @@ -7584,230 +7555,285 @@ ], "soft_pairs": [ "issue:45290|issue:45381", + "issue:37428|issue:43425", "issue:44448|issue:44743", "issue:44857|issue:45593", "issue:45200|issue:45588", "issue:29942|issue:32090", - "issue:37428|issue:44351", "issue:43611|issue:43931", + "issue:37428|issue:44351", "issue:39692|issue:43526", "issue:44206|issue:45823", "issue:43335|issue:43653", "issue:43819|issue:44493", - "issue:44462|issue:44561", - "issue:44561|issue:44779", "issue:43550|issue:44112", "issue:43335|issue:43504", - "issue:45310|issue:45478", - "issue:44855|issue:45083", "issue:29127|issue:32090", + "issue:44855|issue:45083", + "issue:45310|issue:45478", "issue:43335|issue:45083", "issue:45588|issue:45859", "issue:39692|issue:44493", + "issue:44462|issue:44561", + "issue:44561|issue:44779", "issue:44292|issue:45520", "issue:38175|issue:44493", - "issue:44448|issue:44779", "issue:43335|issue:45072", + "issue:44448|issue:44779", "issue:43335|issue:45593", "issue:43720|issue:44112", "issue:44977|issue:45229", "issue:40444|issue:43526", - "issue:43723|issue:44554", "issue:34689|issue:35141", + "issue:43723|issue:44554", "issue:44448|issue:44977", "issue:43606|issue:44857", "issue:44857|issue:45072" ], - "prepared_review_unit_hash": "e8f2209aead4a1deca6a2247a52f80bfc5dbf41d83049670f7638201c19472e8", + "prepared_review_unit_hash": "e34c6b721a9979253b9557110240d67301454c973c1d4ac721f89d28b3581c3c", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 34, - "serialized_chars": 37415, - "estimated_input_tokens": 9354, - "estimated_eval_tokens": 18964 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38238, + "estimated_input_tokens": 9560, + "estimated_eval_tokens": 19376 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issue reports; only one soft pair is a true duplicate: the two identical Qwen3.5 MoE from_pretrained regression reports.", - "confidence": 0.96, - "canonical_issue_reason": "issue:45310 is the earlier exact-match report of the Qwen3.5 MoE from_pretrained regression, so it is the best canonical issue over issue:45478.", + "summary": "This cluster is mostly heterogeneous: nearly all soft pairs are false duplicates across different models/subsystems. The only clear duplicate is the Qwen3.5 MoE from_pretrained regression pair (issues 45310 and 45478).", + "confidence": 0.94, + "canonical_issue_reason": "Issue 45310 is the earliest and most complete report of the only true duplicate theme in this set, and 45478 repeats the same Qwen3.5 MoE from_pretrained failure on transformers>=5.4.0.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the most representative issue here because it is the earliest and clearest duplicate target for the only confirmed overlap.", + "best_issue_reason": "45310 is the best representative issue because it has the clearest description of the shared regression and anchors the one real duplicate pair in the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45290", "right": "issue:45381", "accept": false, - "reason": "Different bugs: chat-template crash with tool calls vs video vision_position_ids in Qwen2.5-VL." + "reason": "Different bugs: chat template tool-call handling vs Qwen2.5-VL video vision_position_ids." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Flash-attention import error vs general Torch 2.10 incompatibility; not the same issue." }, { "left": "issue:44448", "right": "issue:44743", "accept": false, - "reason": "Different models and code paths: Pegasus output drift vs Qwen recurrent-state reset with cache." + "reason": "Pegasus v4/v5 output drift and Qwen recurrent-state reset are unrelated model behaviors." }, { "left": "issue:44857", "right": "issue:45593", "accept": false, - "reason": "Unrelated failures in different models: AMP/CUDA crash vs missing auxiliary losses in D-FINE." + "reason": "LwDetr AMP/CUDA crash and D-FINE auxiliary-loss behavior are different code paths." }, { "left": "issue:45200", "right": "issue:45588", "accept": false, - "reason": "Different Gemma4 text-only token-id behavior vs flash_attention.py sink-less-model crash." + "reason": "Gemma4 text-only mm_token_type_ids problem vs flash_attention sink-less model crash." }, { "left": "issue:29942", "right": "issue:32090", "accept": false, - "reason": "Flash Attention 2 test failures are not the same as the Trainer _gpu_broadcast_one NoneType error." + "reason": "Flash Attention 2 test failures vs Trainer _gpu_broadcast_one NoneType error are unrelated." }, { - "left": "issue:37428", - "right": "issue:44351", + "left": "issue:43611", + "right": "issue:43931", "accept": false, - "reason": "Both are import errors, but for different missing symbols and unrelated failure modes." + "reason": "base_model_prefix loading regression vs Qwen3-VL weight-shape mismatch." }, { - "left": "issue:43611", - "right": "issue:43931", + "left": "issue:37428", + "right": "issue:44351", "accept": false, - "reason": "Different model-loading regressions: base_model_prefix handling vs Qwen3-VL weight-shape mismatch." + "reason": "Different missing symbols: _flash_supports_window_size vs HybridCache." }, { "left": "issue:39692", "right": "issue:43526", "accept": false, - "reason": "Documentation/example errors are unrelated to BeitImageProcessorFast reduce_labels behavior." + "reason": "SigLIP2 docs/quantization issues vs BeitImageProcessorFast reduce_labels bug." }, { "left": "issue:44206", "right": "issue:45823", "accept": false, - "reason": "Different regressions: unsupported center arg in feature extractor vs Gemma4 PLE device mismatch." + "reason": "Feature-extractor center arg regression vs Gemma4 PLE device mismatch." }, { "left": "issue:43335", "right": "issue:43653", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug and BigBirdTokenizer special-token bug are unrelated." + "reason": "SwitchTransformers sparse-layer config bug vs BigBirdTokenizer special-token registration bug." }, { "left": "issue:43819", "right": "issue:44493", "accept": false, - "reason": "DAC from_latents mismatch is unrelated to the position-id key warnings in many models." - }, - { - "left": "issue:44462", - "right": "issue:44561", - "accept": false, - "reason": "AutoTokenizer tokenizer.json handling is unrelated to removal of is_torch_fx_available." - }, - { - "left": "issue:44561", - "right": "issue:44779", - "accept": false, - "reason": "Trust-remote-code breakage from API removal is unrelated to Deepseek tokenizer output errors." + "reason": "DAC.from_latents STE mismatch vs unexpected position-id key issues; different failures." }, { "left": "issue:43550", "right": "issue:44112", "accept": false, - "reason": "Different issues: Bamba torch.compile SDPA failure vs GraniteSpeech stale device override test." + "reason": "Bamba torch.compile/SDPA issue vs GraniteSpeech stale device override CI failure." }, { "left": "issue:43335", "right": "issue:43504", "accept": false, - "reason": "Both touch different BEiT/SwitchTransformers areas with different root causes." + "reason": "SwitchTransformers config bug and Beit pretrained legacy-field loading bug are unrelated." }, { - "left": "issue:45310", - "right": "issue:45478", - "accept": true, - "reason": "Same underlying Qwen3.5 MoE from_pretrained regression; titles and failure description match closely." + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "LayoutLMv3 error-message clarity vs Trainer broadcast TypeError." }, { "left": "issue:44855", "right": "issue:45083", "accept": false, - "reason": "Python 3.13 IndentationError in DebertaV2Model is unrelated to qwen3_omni_moe feat-extract length behavior." + "reason": "DebertaV2 Python 3.13 IndentationError vs qwen3_omni_moe helper-length behavior." }, { - "left": "issue:29127", - "right": "issue:32090", - "accept": false, - "reason": "LayoutLMv3 error-message clarity issue is unrelated to Trainer broadcast NoneType failure." + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Same reported regression: transformers>=5.4.0 breaks Qwen3.5 MoE from_pretrained loading." }, { "left": "issue:43335", "right": "issue:45083", "accept": false, - "reason": "SwitchTransformers sparse-layer config and qwen3_omni_moe length helper behavior are unrelated." + "reason": "SwitchTransformers sparse-layer creation and qwen3_omni_moe feature-length logic are unrelated." }, { "left": "issue:45588", "right": "issue:45859", "accept": false, - "reason": "Both involve flash-attention-adjacent code, but they are different bugs in different functions." + "reason": "Both involve Qwen-related attention/code, but one is sink-less flash_attention aux handling and the other is CPU-offload rotary embedding incompatibility." }, { "left": "issue:39692", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 doc/example issues are unrelated to the generic position-id key warnings." + "reason": "SigLIP2 doc example errors vs generic position-id key warnings are not the same bug." + }, + { + "left": "issue:44462", + "right": "issue:44561", + "accept": false, + "reason": "AutoTokenizer tokenizer.json precedence issue vs removal of is_torch_fx_available breaking trust_remote_code models." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer/loading regressions affecting different mechanisms and models." }, { "left": "issue:44292", "right": "issue:45520", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime error is unrelated to the Python 3.13 flash_attn import KeyError." + "reason": "Qwen-3-8B-NVFP4 runtime failure vs flash_attn import KeyError on Python 3.13." }, { "left": "issue:38175", "right": "issue:44493", "accept": false, - "reason": "Unexpected zero probabilities in SigLIP2 is unrelated to position-id key warnings." + "reason": "SIGLIP2 zero-probabilities issue vs unexpected key with position id; not the same defect." + }, + { + "left": "issue:43335", + "right": "issue:45072", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug vs dtype mismatch in bfloat16 inference." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Pegasus output regression and Deepseek tokenizer regression are unrelated." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "SwitchTransformers config bug vs D-FINE auxiliary-loss omission are different bugs." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "BitNet packed-weight loading issue vs GraniteSpeech device override CI failure." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug vs Gemma4 multi-GPU OOM are unrelated." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "Qwen2.5-VL iterable dataset/multi-image fine-tuning failure vs BeitImageProcessorFast label reduction bug." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Llama 3.2 Vision model loading regression vs post_init embedding reinitialization bug; different code paths." + }, + { + "left": "issue:43723", + "right": "issue:44554", + "accept": false, + "reason": "AutoTokenizer.from_pretrained v5 loading issue vs MPS attention correctness bug." }, { "left": "issue:44448", "right": "issue:44977", "accept": false, - "reason": "Pegasus output difference and Qwen3.5 flash-attention generation issue are different model-specific regressions." + "reason": "Pegasus output drift and Qwen3.5 flash-attention generation failure are unrelated." }, { "left": "issue:43606", "right": "issue:44857", "accept": false, - "reason": "CPU-offload device mismatch for Bark is unrelated to LwDetrImageLoss AMP/CUDA crash." + "reason": "CPU offload device mismatch vs float16 AMP image-loss crash; different subsystems." }, { "left": "issue:44857", "right": "issue:45072", "accept": false, - "reason": "Different failures: LwDetrImageLoss AMP crash vs dtype mismatch in SwitchTransformers/TimmWrapperModel." + "reason": "Image-loss AMP crash and dtype mismatch in SwitchTransformers/TimmWrapperModel are not duplicates." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded and conservative overall. The duplicate call on issue:45310 / issue:45478 is supported by the identical titles and close descriptions, and the rest of the soft-edge rejections stay within title-level evidence without overclaiming. The canonical/best-issue choice is also reasonable because 45310 is the earlier report." + "feedback": "The cluster summary is well grounded in the packet. The only accepted duplicate pair is supported by identical titles and matching regression wording for issues 45310 and 45478, while the other soft-edge rejections stay conservative and model/subsystem-specific." }, "evaluator_used": true, "retried": false, @@ -7879,20 +7905,20 @@ ], "soft_pairs": [ "issue:44568|issue:45491", - "issue:45198|issue:45803", "issue:28218|issue:32090", - "issue:43927|issue:45698", "issue:44977|issue:45563", + "issue:43927|issue:45698", "issue:43701|issue:45083", + "issue:45198|issue:45803", "issue:43550|issue:45083", "issue:29942|issue:35141", - "issue:45459|issue:45712", "issue:45372|issue:45588", "issue:43873|issue:44493", + "issue:45459|issue:45712", "issue:32090|issue:33357", "issue:37428|issue:44987", - "issue:43412|issue:45588", "issue:43873|issue:45229", + "issue:43412|issue:45588", "issue:45245|issue:45563", "issue:43278|issue:43329", "issue:41720|issue:45229", @@ -7905,225 +7931,225 @@ "issue:44857|issue:45588", "issue:43065|issue:45412", "issue:44448|issue:44466", - "issue:43066|issue:44462", "issue:43278|issue:43381", "issue:43653|issue:44743", - "issue:45588|issue:45823" + "issue:45588|issue:45823", + "issue:43066|issue:44462" ], - "prepared_review_unit_hash": "91f608b87f81fcccbb12acefd605d899c9da92c75bda0708b298eef6f83cc92d", + "prepared_review_unit_hash": "4031cbe16c5e8621a4aa40589ea46299347d103b2d678760a247dd5a43d00042", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 37199, - "estimated_input_tokens": 9300, - "estimated_eval_tokens": 18856 + "serialized_chars": 37193, + "estimated_input_tokens": 9299, + "estimated_eval_tokens": 18854 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is highly heterogeneous: tokenizer, multimodal, training, export, flash-attention, and device-mapping issues are mixed together. None of the soft pairs look like the same underlying bug or change closely enough to deduplicate safely.", - "confidence": 0.97, - "canonical_issue_reason": "No single canonical issue fits this cluster; the items span unrelated problems across many subsystems and time periods.", + "summary": "These items are a heterogeneous grab-bag of unrelated Transformers bugs. The soft pairs mostly share superficial subsystem keywords (tokenizers, Gemma, flash attention, save/load) but do not describe the same underlying defect or mergeable fix.", + "confidence": 0.92, + "canonical_issue_reason": "No strong canonical duplicate anchor exists because the cluster is not coherent. If one issue must be chosen as a representative, issue:45372 is the most discussion-rich and currently active, but it is not a true umbrella for the rest.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44462 is the clearest self-contained bug report here: it names a concrete regression in AutoTokenizer loading tokenizer.json from the repo and is likely the most actionable reference issue.", + "best_issue_reason": "Issue:45372 is the best single representative only in the sense of being recent, user-facing, and relatively high-signal; however, it should not be treated as a duplicate target for the other issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44568", "right": "issue:45491", "accept": false, - "reason": "Different problems: missing BOS/EOS with add_special_tokens vs NaN embeddings from sliding-window attention and all-padding windows." + "reason": "Both concern tokenizer/model output behavior, but one is about special tokens not being added and the other is a Gemma3 NaN embedding bug from mixed-length batching. Different code paths and fixes." }, { - "left": "issue:45198", - "right": "issue:45803", + "left": "issue:28218", + "right": "issue:32090", "accept": false, - "reason": "Both involve tokenization/tensor conversion broadly, but one is a Wav2Vec2 save/tokenization failure and the other is a bare-except bug in Fuyu tensor conversion." + "reason": "Tokenizer spacing after added tokens is unrelated to Trainer distributed broadcast NoneType errors." }, { - "left": "issue:28218", - "right": "issue:32090", + "left": "issue:44977", + "right": "issue:45563", "accept": false, - "reason": "Unrelated symptoms and code paths: extra space after added token vs Trainer NoneType broadcast error." + "reason": "Qwen3.5 flash-attention generation failure and a stale generate() warning are different bugs with different symptoms and fixes." }, { "left": "issue:43927", "right": "issue:45698", "accept": false, - "reason": "Both touch save/load behavior, but one is custom token IDs being lost and the other is loading the wrong custom module after save_pretrained." + "reason": "Both involve save/load behavior, but one is config token IDs being lost and the other is loading the wrong custom module. Not the same defect." }, { - "left": "issue:44977", - "right": "issue:45563", + "left": "issue:43701", + "right": "issue:45083", "accept": false, - "reason": "Flash-attention generation failure is not the same as a stale warning for num_return_sequences." + "reason": "Resume-from-checkpoint key mismatch is unrelated to a Qwen3 Omni MoE helper-function bug." }, { - "left": "issue:43701", - "right": "issue:45083", + "left": "issue:45198", + "right": "issue:45803", "accept": false, - "reason": "Checkpoint key mismatch and a feature-extraction helper returning unexpected lengths are unrelated." + "reason": "Wav2Vec2 save_pretrained/tokenization failure and a bare-except swallowing errors in Fuyu conversion are distinct issues." }, { "left": "issue:43550", "right": "issue:45083", "accept": false, - "reason": "torch.compile + SDPA failure in Bamba is a different code path from qwen3_omni_moe feature-length logic." + "reason": "A torch.compile/SDPA failure in Bamba is unrelated to a feature-extraction length helper bug in qwen3_omni_moe." }, { "left": "issue:29942", "right": "issue:35141", "accept": false, - "reason": "Failing Flash Attention 2 tests and embedding reinitialization after resize are distinct issues." - }, - { - "left": "issue:45459", - "right": "issue:45712", - "accept": false, - "reason": "Tokenizer error masking with missing protobuf has nothing to do with leftover dummy classes or dir(transformers) leakage." + "reason": "Failing Flash Attention 2 tests and embedding reinitialization after resizing token embeddings are different regressions." }, { "left": "issue:45372", "right": "issue:45588", "accept": false, - "reason": "Gemma4 processor import failure and a flash_attention.py crash on s_aux=None are different bugs." + "reason": "Gemma4 processor import failure due to ReasoningEffort is a separate problem from flash_attention.py crashing on s_aux=None." }, { "left": "issue:43873", "right": "issue:44493", "accept": false, - "reason": "Quantization/offloading behavior and unexpected position-id keys are separate problems." + "reason": "Quantization/offloading behavior and unexpected position-id keys are unrelated runtime/configuration bugs." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "A broad tokenizer exception-handling bug and leftover dummy classes leaking into imports are not the same underlying change." }, { "left": "issue:32090", "right": "issue:33357", "accept": false, - "reason": "Trainer broadcast NoneType error and Mac bus error with a CLIP model are unrelated failures." + "reason": "Trainer GPU broadcast NoneType error and MacOS bus error with a CLIP model have no shared concrete root cause." }, { "left": "issue:37428", "right": "issue:44987", "accept": false, - "reason": "Both are import/loading failures, but they concern different modules and failure modes; not the same concrete bug." + "reason": "Both mention loading failures, but one is a missing flash-attention utility import and the other is a model-specific transformers>=5.1.0 load failure; too little evidence they are the same bug." }, { - "left": "issue:43412", - "right": "issue:45588", + "left": "issue:45372", + "right": "issue:45406", "accept": false, - "reason": "Executorch export guard/support issues are not the same as a flash-attention crash on sink-less models." + "reason": "Both are Gemma4-related, but one is processor loading import failure and the other is serve crashing because Gemma4Processor lacks _tokenizer. Different failure points." }, { - "left": "issue:43873", - "right": "issue:45229", + "left": "issue:43412", + "right": "issue:45588", "accept": false, - "reason": "Quantization/offloading bug and multi-GPU CUDA OOM are different concerns." + "reason": "Executorch export failures for Gemma3n and flash-attention crashes on sink-less models are unrelated." }, { "left": "issue:45245", "right": "issue:45563", "accept": false, - "reason": "Category-count runtime limit and a stale generation warning are unrelated." + "reason": "Category-count runtime error and stale num_return_sequences warning are not the same defect." }, { "left": "issue:43278", "right": "issue:43329", "accept": false, - "reason": "Embedding dtype drift in eval and undefined video-branch variables are different bugs." + "reason": "Embedding dtype drift in eval and undefined video-token helper names are distinct issues." }, { "left": "issue:41720", "right": "issue:45229", "accept": false, - "reason": "Auto device mapping cudaErrorAssert on A800 is not the same as a multi-GPU OOM." + "reason": "Qwen3 auto device-map cudaErrorAssert and Gemma4 multi-GPU OOM are different runtime problems." }, { "left": "issue:10000", "right": "issue:15354", "accept": false, - "reason": "German DistilBertModel issue and torch.jit.script GeneratorExp export failure have no shared underlying bug." + "reason": "German DistilBertModel issue and torch.jit GeneratorExp export failure are unrelated model/export bugs." }, { "left": "issue:43976", "right": "issue:45440", "accept": false, - "reason": "Python version compatibility and DeepSeekV3MoE divergence are unrelated." + "reason": "Python-version compatibility and DeepSeekV3MoE divergence are unrelated." }, { "left": "issue:45372", "right": "issue:45406", "accept": false, - "reason": "Gemma4 processor loading import error and Gemma4Processor missing _tokenizer on serve are distinct failures." + "reason": "Repeated pair: same rationale as above; import-time processor loading bug vs serve-time missing attribute bug are not duplicates." }, { "left": "issue:28218", "right": "issue:29127", "accept": false, - "reason": "Tokenizer spacing after added tokens and layoutlmv3 error-message clarity are not the same bug." + "reason": "Tokenizer added-space behavior and LayoutLMv3 error-message clarity are unrelated." }, { "left": "issue:43065", "right": "issue:43994", "accept": false, - "reason": "A dummy Conv2d placeholder in Sam3PixelDecoder is unrelated to SigLIP2 producing nonsensical outputs." + "reason": "Dummy nn.Conv2d in Sam3PixelDecoder and SigLIP2 nonsensical outputs are different model bugs." }, { "left": "issue:45372", "right": "issue:45459", "accept": false, - "reason": "Gemma4 processor import breakage and tokenizer protobuf error masking are different code paths." + "reason": "ReasoningEffort import failure in Gemma4 processor loading and protobuf error handling in tokenizer decode are unrelated." }, { "left": "issue:44857", "right": "issue:45588", "accept": false, - "reason": "AMP/CUDA crash in LwDetrImageLoss is not the same as flash_attention.py crashing on s_aux=None." + "reason": "AMP crash in LwDetrImageLoss and flash_attention s_aux=None are different code paths." }, { "left": "issue:43065", "right": "issue:45412", "accept": false, - "reason": "Sam3PixelDecoder dummy layer and RT-DETR memory not being released are unrelated." + "reason": "Dummy layer definition in Sam3PixelDecoder and RT-DETR memory release issues are not the same underlying bug." }, { "left": "issue:44448", "right": "issue:44466", "accept": false, - "reason": "Different issues: Pegasus output changes across versions vs inconsistent lm_head.weight serialization tied to device." - }, - { - "left": "issue:43066", - "right": "issue:44462", - "accept": false, - "reason": "Wrong tokenizer decoder type in v5 and AutoTokenizer ignoring tokenizer.json are related to tokenizers, but they are not the same concrete bug." + "reason": "Different model output in v4 vs v5 and inconsistent lm_head serialization are separate versioning/serialization problems." }, { "left": "issue:43278", "right": "issue:43381", "accept": false, - "reason": "Embedding dtype mismatch in eval and gradient checkpointing in eval mode are separate issues." + "reason": "Eval/training state problems, but one is dtype regression and the other is gradient checkpointing restriction in eval; not the same fix." }, { "left": "issue:43653", "right": "issue:44743", "accept": false, - "reason": "BigBirdTokenizer special-token registration and Qwen3 recurrent-state reset are unrelated." + "reason": "BigBirdTokenizer special-token registration and Qwen3.5 recurrent-state reset are unrelated." }, { "left": "issue:45588", "right": "issue:45823", "accept": false, - "reason": "A flash-attention crash on sink-less models and Gemma4 PLE device mismatch are different problems." + "reason": "Both are Gemma4-adjacent, but flash_attention.py s_aux handling and PLE device mismatch under device_map='auto' are different concrete bugs." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Tokenizer decoder-type selection and AutoTokenizer ignoring tokenizer.json are distinct tokenizer-loading issues." } ] }, @@ -8155,6 +8181,7 @@ "issue:33290", "issue:33357", "issue:34567", + "issue:35141", "issue:36010", "issue:39692", "issue:42915", @@ -8179,7 +8206,6 @@ "issue:44361", "issue:44442", "issue:44451", - "issue:44464", "issue:44479", "issue:44610", "issue:44871", @@ -8198,8 +8224,8 @@ ], "soft_pairs": [ "issue:25251|issue:32090", - "issue:44186|issue:45198", "issue:43994|issue:45412", + "issue:44186|issue:45198", "issue:44206|issue:44479", "issue:44451|issue:45520", "issue:44361|issue:45198", @@ -8212,230 +8238,230 @@ "issue:43257|issue:43278", "issue:33357|issue:39692", "issue:29127|issue:34567", - "issue:44610|issue:45412", "issue:32090|issue:33290", + "issue:44610|issue:45412", "issue:45103|issue:45372", "issue:45440|issue:45593", "issue:43824|issue:44279", "issue:43653|issue:44442", - "issue:43761|issue:45859", "issue:43381|issue:43701", "issue:43526|issue:45834", + "issue:43761|issue:45859", "issue:29942|issue:36010", "issue:43526|issue:44186", "issue:29127|issue:33357", "issue:43334|issue:43531", - "issue:44977|issue:45245", - "issue:44464|issue:45491" + "issue:29127|issue:35141", + "issue:44977|issue:45245" ], - "prepared_review_unit_hash": "d871fdca64d5118c75b662b4f8bc42784d19c24f4da02aa5a97bd1b48a095231", + "prepared_review_unit_hash": "f6e29ee813731e0344287f0f77e4db6c31bfddf18ad95e72997196a0b1e0eedf", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36655, - "estimated_input_tokens": 9164, - "estimated_eval_tokens": 18584 + "serialized_chars": 36694, + "estimated_input_tokens": 9174, + "estimated_eval_tokens": 18604 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: it contains many unrelated tokenizer, processor, model-loading, quantization, and runtime issues. I found no strong duplicate groups among the provided soft pairs, so all soft edges are rejected.", - "confidence": 0.14, - "canonical_issue_reason": "No clear duplicate cluster stands out; if one issue must anchor the set, issue 43994 is the most self-contained and reproducible runtime failure among otherwise mixed reports.", + "summary": "This cluster is highly heterogeneous: the soft pairs mostly share only broad subsystem or model-family vocabulary, not the same concrete bug. I would not merge any of the suggested soft edges. If a single representative is needed, the broadest issue is the Qwen multi-model video regression report.", + "confidence": 0.88, + "canonical_issue_reason": "No pair here looks like a true duplicate set; if one issue must represent the cluster, issue 44479 is the broadest and clearest regression report because it covers multiple Qwen variants with one concrete video-input failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43994 is the best representative of the cluster only in the loose sense that it is a concrete, actionable bug report; however, the overall set is not a cohesive duplicate cluster.", + "best_issue_reason": "Issue 44479 is the best overall representative because it is specific enough to summarize a concrete regression while still spanning several affected models, making it the most cluster-like of the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:25251", "right": "issue:32090", "accept": false, - "reason": "Different subsystems and failure modes: pipeline output shape vs Trainer GPU broadcast NoneType error." + "reason": "Unrelated failures: pipeline output shape vs Trainer broadcast TypeError." }, { - "left": "issue:44186", - "right": "issue:45198", + "left": "issue:43994", + "right": "issue:45412", "accept": false, - "reason": "Different model/tokenizer paths and different bugs: LayoutLMv2 NER padding crash vs Wav2Vec2 save/tokenization failure." + "reason": "Different components and symptoms: SigLIP2 doc/example mismatch vs RT-DETR memory leak." }, { - "left": "issue:43994", - "right": "issue:45412", + "left": "issue:44186", + "right": "issue:45198", "accept": false, - "reason": "Unrelated problems: SigLIP2 bad outputs vs RT-DETR memory not released." + "reason": "Tokenizer padding/NER crash vs Wav2Vec2 save/tokenization failure; no shared code-path." }, { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Different code paths: feature extractor center-arg crash vs Qwen video-input regression." + "reason": "Feature extractor center-arg crash is unrelated to Qwen video-input regression." }, { "left": "issue:44451", "right": "issue:45520", "accept": false, - "reason": "Different issue classes: model loading failure for ScandiBERT vs Python 3.13 flash_attn import key error." + "reason": "Model loading failure vs flash_attn import handling on Python 3.13; different bug classes." }, { "left": "issue:44361", "right": "issue:45198", "accept": false, - "reason": "Both involve tokenization, but the concrete failures and models are unrelated." + "reason": "Different tokenizers and failure modes; no evidence of one underlying bug." }, { "left": "issue:42915", "right": "issue:42994", "accept": false, - "reason": "Both mention quantization, but one is a FineGrainedFP8 model failure and the other is a save_pretrained issue; not the same bug." + "reason": "Both involve quantization, but one is model/runtime failure and the other is saving quantized models." }, { "left": "issue:44871", "right": "issue:44977", "accept": false, - "reason": "Different models and different symptoms: eos_token_id config mismatch vs flash-attention generation regression." + "reason": "Eos-token config mismatch vs flash-attention generation regression; unrelated." }, { "left": "issue:44265", "right": "issue:45198", "accept": false, - "reason": "Different mechanisms: torch.export with torch_compilable_check vs Wav2Vec2 tokenization/save failure." + "reason": "torch.export exportability issue is not the same as a tokenizer/save_pretrained problem." }, { "left": "issue:43756", "right": "issue:45440", "accept": false, - "reason": "Different architectures and bug types: Smollm3 RoPE layer mismatch vs DeepseekV3MoE divergence from remote implementation." + "reason": "Both are model-implementation mismatches, but for different models and different defects." }, { "left": "issue:43421", "right": "issue:45491", "accept": false, - "reason": "Tokenizer special-token runtime updates are unrelated to Gemma3 sliding-window NaN embeddings." + "reason": "Runtime special-token/post-processor update bug vs Gemma3 NaN embedding issue; unrelated." }, { "left": "issue:29942", "right": "issue:30064", "accept": false, - "reason": "Flash Attention 2 test failures are unrelated to void segmentation map processing." + "reason": "FlashAttention test failures are unrelated to void segmentation-map processing." }, { "left": "issue:43257", "right": "issue:43278", "accept": false, - "reason": "Qwen3 MOE weight conversion/loading bug is unrelated to embedding dtype drift between train and eval." + "reason": "Accelerate/deepspeed weight conversion bug vs training/eval dtype change; different root causes." }, { "left": "issue:33357", "right": "issue:39692", "accept": false, - "reason": "Both mention CLIP/SigLIP-family workflows, but one is a MacOS bus error and the other is a documentation example with multiple mistakes." + "reason": "Both mention CLIP-like vision models, but one is a bus error and the other is a documentation/example error bundle." }, { "left": "issue:29127", "right": "issue:34567", "accept": false, - "reason": "LayoutLMv3 error-message clarity issue vs TrainerState token counter not updating; unrelated." + "reason": "LayoutLMv3 error-message clarity and TrainerState token-count updates are unrelated." }, { - "left": "issue:44610", - "right": "issue:45412", + "left": "issue:32090", + "right": "issue:33290", "accept": false, - "reason": "Processor resolution mismatch vs model memory leak are different bugs and code paths." + "reason": "Trainer broadcast NoneType error and DeepSpeed OOM with Adafactor are different issues." }, { - "left": "issue:32090", - "right": "issue:33290", + "left": "issue:44610", + "right": "issue:45412", "accept": false, - "reason": "Trainer broadcast NoneType error is unrelated to DeepSpeed Adafactor OOM." + "reason": "Processor image-size mismatch is unrelated to RT-DETR memory reclamation." }, { "left": "issue:45103", "right": "issue:45372", "accept": false, - "reason": "Auto-docstring annotation parsing crash is unrelated to Gemma 4 processor import failure." + "reason": "auto_docstring annotations crash and mistral_common import error are unrelated." }, { "left": "issue:45440", "right": "issue:45593", "accept": false, - "reason": "DeepSeekV3MoE implementation divergence is unrelated to D-FINE auxiliary-loss behavior." + "reason": "Different model families and different behaviors: implementation divergence vs loss computation." }, { "left": "issue:43824", "right": "issue:44279", "accept": false, - "reason": "Importing a missing Qwen2.5-VL class is unrelated to a generic dependency issue." + "reason": "Specific Qwen2.5-VL import failure is not the same as a generic dependency issue." }, { "left": "issue:43653", "right": "issue:44442", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug is unrelated to loading FastSpeech2ConformerTokenizer." - }, - { - "left": "issue:43761", - "right": "issue:45859", - "accept": false, - "reason": "CLIPVision hidden_states regression and Qwen3.5 MoE CPU-offload incompatibility are different model-path bugs." + "reason": "Tokenizer special-token registration bug vs AutoTokenizer loading failure for a different tokenizer." }, { "left": "issue:43381", "right": "issue:43701", "accept": false, - "reason": "Gradient checkpointing in eval mode vs resume_from_checkpoint key mismatch are unrelated trainer issues." + "reason": "Gradient checkpointing in eval mode and checkpoint key mismatch are unrelated." }, { "left": "issue:43526", "right": "issue:45834", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug is unrelated to Kosmos2.5 OCR long-input indexing." + "reason": "BEiT label reduction bug is unrelated to Kosmos2.5 OCR index errors." + }, + { + "left": "issue:43761", + "right": "issue:45859", + "accept": false, + "reason": "CLIPVision hidden_states regression is unrelated to CPU-offload incompatibility in Qwen3_5Moe rotary embeddings." }, { "left": "issue:29942", "right": "issue:36010", "accept": false, - "reason": "Flash Attention tests and GenerationMixin import error do not share the same underlying bug." + "reason": "FlashAttention test failures are not the same as a GenerationMixin import regression." }, { "left": "issue:43526", "right": "issue:44186", "accept": false, - "reason": "Different processors and different failure modes; not the same bug." + "reason": "ImageProcessor label handling and LayoutLMv2 tokenizer crashes are different code paths." }, { "left": "issue:29127", "right": "issue:33357", "accept": false, - "reason": "LayoutLMv3 error messaging and CLIP MacOS bus error are unrelated." + "reason": "LayoutLMv3 error messaging and CLIP Mac bus error share no concrete failure mode." }, { "left": "issue:43334", "right": "issue:43531", "accept": false, - "reason": "Qwen3-VL pad_token_id import-time AttributeError and Qwen3-MoE sliding_window behavior are different issues." + "reason": "Both are Qwen-family issues, but one is a missing config attribute and the other is a sliding-window behavior bug." }, { - "left": "issue:44977", - "right": "issue:45245", + "left": "issue:29127", + "right": "issue:35141", "accept": false, - "reason": "Flash-attention generation regression is unrelated to category-count runtime limits." + "reason": "LayoutLMv3 box-validation messaging vs token-embedding reinit in post_init are unrelated." }, { - "left": "issue:44464", - "right": "issue:45491", + "left": "issue:44977", + "right": "issue:45245", "accept": false, - "reason": "Chunked generation with compiled forward and Gemma3 NaN embeddings are different bugs." + "reason": "Flash-attention generation regression and category-cardinality runtime error are not the same bug." } ] }, @@ -8459,7 +8485,6 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-399", "nodes": [ - "issue:29127", "issue:33290", "issue:34567", "issue:34689", @@ -8490,6 +8515,7 @@ "issue:44206", "issue:44387", "issue:44442", + "issue:44464", "issue:44479", "issue:44743", "issue:44811", @@ -8508,9 +8534,9 @@ "issue:45907" ], "soft_pairs": [ - "issue:29127|issue:35141", - "issue:44442|issue:45072", + "issue:44464|issue:45491", "issue:42915|issue:43931", + "issue:44442|issue:45072", "issue:43262|issue:44206", "issue:43526|issue:44857", "issue:43232|issue:43388", @@ -8519,9 +8545,9 @@ "issue:44062|issue:45907", "issue:43575|issue:43701", "issue:43646|issue:44077", + "issue:43575|issue:44387", "issue:43761|issue:44079", "issue:45003|issue:45440", - "issue:43575|issue:44387", "issue:43723|issue:45479", "issue:44442|issue:45803", "issue:43761|issue:44188", @@ -8537,202 +8563,202 @@ "issue:43381|issue:43873", "issue:43931|issue:45216" ], - "prepared_review_unit_hash": "b0791181acff28eac01b0e766bf2d95f1a30f5aab90c38a561653c7976395bdf", + "prepared_review_unit_hash": "efe8f8078e5e823f5b38202962c764b6464ab3046a368b8d44223996f0347cdf", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 35665, - "estimated_input_tokens": 8917, - "estimated_eval_tokens": 18090 + "serialized_chars": 35639, + "estimated_input_tokens": 8910, + "estimated_eval_tokens": 18076 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: it mixes unrelated loading, generation, quantization, processor, and training bugs. None of the soft pairs look like true duplicates, so all candidate edges are rejected. The broadest representative issue is the Transformers 5.0.0 custom model initialization regression (#43646), but only as a loose thematic anchor.", - "confidence": 0.66, - "canonical_issue_reason": "#43646 is the broadest, most representative issue in the set: it describes a general Transformers 5.0.0 model-initialization regression rather than a narrow model-specific symptom.", + "summary": "The soft pairs are mostly superficial similarity matches; they span unrelated regressions across loading, generation, quantization, tokenizers, and model configs, so none look like true duplicates. No PRs are present.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43873 is the strongest anchor here: it is open, has the most discussion and inbound references, and describes a broad quantization/offloading runtime regression that seems more central than the other isolated reports.", "canonical_pr_reason": null, - "best_issue_reason": "#43646 is the best single issue to stand in for the cluster\u2019s main v5-regression theme, though the set is still too mixed to treat as one duplicate family.", + "best_issue_reason": "issue:43873 is the best global representative because it is the most widely discussed and externally referenced issue in the set, making it the most useful canonical issue if one must be chosen.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:29127", - "right": "issue:35141", + "left": "issue:44464", + "right": "issue:45491", "accept": false, - "reason": "Different problems: LayoutLMv3 error messaging vs token embedding reinitialization after resize." + "reason": "Both are generation-related, but one is about compiled chunked generation inconsistencies while the other is a Gemma3 NaN embedding bug from sliding-window attention; different failure modes." }, { - "left": "issue:44442", - "right": "issue:45072", + "left": "issue:42915", + "right": "issue:43931", "accept": false, - "reason": "Tokenizer loading failure vs bfloat16 dtype mismatch in inference; not the same code path." + "reason": "Both involve Qwen loading, but one is a FineGrainedFP8Config failure and the other is a model weight-shape mismatch; not the same underlying bug." }, { - "left": "issue:42915", - "right": "issue:43931", + "left": "issue:44442", + "right": "issue:45072", "accept": false, - "reason": "FineGrainedFP8Config failure vs model weight-shape mismatch on load; distinct bugs." + "reason": "Tokenizer loading failure and bfloat16 dtype mismatch are unrelated code paths, despite both affecting model startup/inference." }, { "left": "issue:43262", "right": "issue:44206", "accept": false, - "reason": "Chat-template audio sample-rate default bug vs unsupported `center` arg crash in a feature extractor." + "reason": "Audio processor default sampling-rate handling is different from an unsupported `center` argument crash in a feature extractor." }, { "left": "issue:43526", "right": "issue:44857", "accept": false, - "reason": "Label-reduction bug in BeitImageProcessorFast vs AMP/CUDA loss crash in LwDetrImageLoss." + "reason": "One is a label-reduction logic bug, the other is an AMP/CUDA loss crash; they do not share the same root cause." }, { "left": "issue:43232", "right": "issue:43388", "accept": false, - "reason": "Generation kwargs update bug vs metric gathering dropping tuple labels; unrelated behavior." + "reason": "`_update_model_kwargs_for_generation` after GPU sync and `gather_for_metrics` truncating tuple labels are separate bugs in different subsystems." }, { "left": "issue:43441", "right": "issue:45588", "accept": false, - "reason": "Both mention FlashAttention, but one is a model-specific regression and the other is an `s_aux=None` crash in integration code." + "reason": "Both mention flash attention, but one is a model/regression issue for Ministral-3 and the other is a sink-less `s_aux=None` crash in the integration layer." }, { "left": "issue:44743", "right": "issue:44811", "accept": false, - "reason": "Qwen recurrent-state reset bug vs Whisper batch_decode skip_special_tokens handling; different subsystems." + "reason": "Recurrent state reset with cache and Whisper batch decode skipping special-token handling are unrelated." }, { "left": "issue:44062", "right": "issue:45907", "accept": false, - "reason": "AddedToken constructor duplication vs list_repo_files revision forwarding; unrelated." + "reason": "`AddedToken(..., special=...)` argument duplication and `list_repo_files` not forwarding `revision` are unrelated API bugs." }, { "left": "issue:43575", "right": "issue:43701", "accept": false, - "reason": "TP load OOM for Qwen2-57B vs resume-from-checkpoint key mismatch; not the same issue." + "reason": "Tensor-parallel OOM on load and resume-from-checkpoint key mismatch are different problems." }, { "left": "issue:43646", "right": "issue:44077", "accept": false, - "reason": "Custom model initialization regression vs patchtsmixer post_init policy change; different concrete failures." + "reason": "Custom model initialization breakage is not the same as the `patchtsmixer` `post_init` policy issue." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both mention OOM, but one is TP model loading OOM and the other is increased reserved memory under int4 quantization; different root causes." }, { "left": "issue:43761", "right": "issue:44079", "accept": false, - "reason": "CLIPVisionModel hidden_states regression vs ModelOutput key assignment bug; distinct code paths." + "reason": "Missing `hidden_states` in CLIPVisionModel and incorrect `ModelOutput` key assignment are related only broadly; not the same concrete bug." }, { "left": "issue:45003", "right": "issue:45440", "accept": false, - "reason": "Unsafe sys.modules access vs DeepSeekV3MoE divergence from remote implementation." - }, - { - "left": "issue:43575", - "right": "issue:44387", - "accept": false, - "reason": "Specific Qwen TP OOM vs broader int4 reserved-memory/OOM regression; same symptom, different bug." + "reason": "Unsafe `sys.modules` access in `modeling_utils` and DeepseekV3MoE divergence are unrelated." }, { "left": "issue:43723", "right": "issue:45479", "accept": false, - "reason": "Tokenizer loading issue vs zero-loss classification bug; unrelated." + "reason": "Tokenizer-loading regression and zero-loss sequence-classification behavior are separate issues." }, { "left": "issue:44442", "right": "issue:45803", "accept": false, - "reason": "AutoTokenizer load failure vs bare-except in tensor conversion; different failures." + "reason": "AutoTokenizer failing to load a tokenizer and a bare `except:` swallowing errors are unrelated bugs." }, { "left": "issue:43761", "right": "issue:44188", "accept": false, - "reason": "Hidden-states regression vs attention-kernel divergence under `torch.compile`; unrelated." + "reason": "CLIP hidden-state regression and attention-kernel divergence under `torch.compile` do not share the same code path." }, { "left": "issue:44479", "right": "issue:45290", "accept": false, - "reason": "Video-input regression for Qwen variants vs chat-template crash with tool calls; different paths." + "reason": "Video-input regression for Qwen VL models and chat-template crashes with tool calls are different feature paths." }, { "left": "issue:43316", "right": "issue:43525", "accept": false, - "reason": "Different config API regressions: Gemma3TextConfig discrepancy vs missing pad_token_id on Llama4Config." + "reason": "API discrepancy in `Gemma3TextConfig` and missing `pad_token_id` on `Llama4Config` are distinct config issues." }, { "left": "issue:42907", "right": "issue:45698", "accept": false, - "reason": "Save-time dequantization bug vs wrong custom module loaded after save_pretrained; related area but not the same bug." + "reason": "Both are save/load related, but one is about saving dequantized models and the other about loading the wrong custom module after save/load; not the same bug." }, { "left": "issue:44442", "right": "issue:44898", "accept": false, - "reason": "Tokenizer load failure vs Perceiver image-classification failure; no shared concrete cause." + "reason": "Tokenizer load failure and Perceiver non-default-resolution image classification failure are unrelated." }, { "left": "issue:33290", "right": "issue:34567", "accept": false, - "reason": "Deepspeed/Adafactor OOM vs TrainerState token counter not updating; unrelated." + "reason": "OOM with Adafactor/DeepSpeed and `num_input_tokens_seen` not updating are unrelated trainer issues." }, { "left": "issue:34689", "right": "issue:43611", "accept": false, - "reason": "Llama 3.2 Vision load break vs base_model_prefix loading regression; both loading-related but distinct." + "reason": "Both are model-loading regressions, but one is specific to Llama 3.2 Vision while the other is a `base_model_prefix` loading issue; different causes." }, { "left": "issue:39692", "right": "issue:45491", "accept": false, - "reason": "Documentation example errors vs Gemma3 NaN embeddings; different bug classes." + "reason": "SigLIP2 documentation example errors and Gemma3 NaN embeddings are not the same underlying bug." }, { "left": "issue:33290", "right": "issue:35141", "accept": false, - "reason": "OOM in deepspeed/Adafactor vs embedding reinitialization on resize; unrelated." + "reason": "OOM in training versus output embedding reinitialization during `post_init` are unrelated." }, { "left": "issue:33290", "right": "issue:34689", "accept": false, - "reason": "OOM during training vs model-loading regression; not duplicate." + "reason": "Training OOM with DeepSpeed/Adafactor and model loading failure for Llama 3.2 Vision are different failures." }, { "left": "issue:43381", "right": "issue:43873", "accept": false, - "reason": "Gradient checkpointing in eval-mode bug vs quantization offloading issue; different behavior." + "reason": "Gradient checkpointing being disallowed in eval mode is unrelated to quantization/offloading behavior." }, { "left": "issue:43931", "right": "issue:45216", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch vs Qwen3.5 save_pretrained checkpoint regression; same family only at a very broad level." + "reason": "Model weight-shape mismatch on load and a corrupted checkpoint saved by `save_pretrained` are distinct persistence bugs." } ] }, @@ -8815,8 +8841,8 @@ "issue:43749|issue:43931", "issue:33290|issue:36331", "issue:34689|issue:36010", - "issue:44190|issue:44496", "issue:43873|issue:44871", + "issue:44190|issue:44496", "issue:43381|issue:43856", "issue:44855|issue:45137", "issue:43540|issue:44514", @@ -8827,188 +8853,188 @@ "issue:44162|issue:44792", "issue:43065|issue:44617", "issue:25251|issue:28282", - "issue:43388|issue:43688", "issue:43295|issue:43825", + "issue:43388|issue:43688", "issue:16998|issue:22355" ], - "prepared_review_unit_hash": "5a12409eeb1125645da8ea7513cf965b00a1da83be9227bbf367043a8c621f63", + "prepared_review_unit_hash": "e202d11bf4c5a117f5ba32f6f4075bb455e5e3d27c8027a7782c62390b1269e5", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 25, - "serialized_chars": 34612, - "estimated_input_tokens": 8653, - "estimated_eval_tokens": 17562 + "serialized_chars": 34607, + "estimated_input_tokens": 8652, + "estimated_eval_tokens": 17560 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a grab-bag of mostly unrelated Transformers issues spanning model loading, pipelines, quantization/offloading, training, and multimodal regressions. No soft-edge pair looks like the same underlying bug/change closely enough to merge.", - "confidence": 0.93, - "canonical_issue_reason": "issue:43749 is the broadest, most central-looking runtime regression in the set, with high discussion and multiple inbound references around a core loading path breakage.", + "summary": "This cluster is a set of largely unrelated Transformers issues. The soft-similarity edges are mostly superficial (shared words like import/config/video/model loading) rather than the same underlying bug, so all candidate duplicate links should be rejected.", + "confidence": 0.94, + "canonical_issue_reason": "No single canonical issue fits this set: the items cover distinct failures across model loading, processors, training, config, and docs, with no shared concrete code path.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43749 is the strongest representative issue because it affects a foundational feature, has significant thread activity, and is phrased as a clear actionable regression.", + "best_issue_reason": "If one issue must be picked as the most representative, issue 43873 is the best fit because it is active, user-facing, and broad in impact; but it is not a true duplicate anchor for the rest of the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43493", "right": "issue:43827", "accept": false, - "reason": "Both mention Gemma 3/translation-adjacent docs, but one is a model implementation discrepancy and the other is outdated docs referencing pipeline() after v5 removals." + "reason": "One is a SigLIP2 implementation mismatch; the other is stale docs still referencing pipeline(). Different problem types and code paths." }, { "left": "issue:43550", "right": "issue:44743", "accept": false, - "reason": "One is a torch.compile + SDPA failure for Bamba-9B-v2; the other is a Qwen3.5 recurrent-state cache bug. Different models and code paths." + "reason": "Torch.compile/SDPA failure in Bamba vs recurrent state reset in modular_qwen3_5.py. Different models and bugs." }, { "left": "issue:44077", "right": "issue:44479", "accept": false, - "reason": "Patchtsmixer post_init validation is unrelated to the Qwen video-input regression; no shared underlying bug." + "reason": "patchtsmixer post_init policy change is unrelated to Qwen video-input regression." }, { "left": "issue:43720", "right": "issue:44442", "accept": false, - "reason": "BitNet packed-weight loading under accelerate and AutoTokenizer loading FastSpeech2ConformerTokenizer are unrelated import/loading failures." + "reason": "BitNet packed-weight loading issue is unrelated to AutoTokenizer failing for FastSpeech2ConformerTokenizer." }, { "left": "issue:43316", "right": "issue:45759", "accept": false, - "reason": "Gemma3TextConfig API mismatch and Qwen 3.5/3.6 composite config unwrapping are different multimodal config issues." + "reason": "Gemma3TextConfig API discrepancy is a config-class consistency issue; Qwen 3.5/3.6 bug is about unwrapping text_config in AutoModelForCausalLM.from_config." }, { "left": "issue:41628", "right": "issue:45372", "accept": false, - "reason": "AutoImageProcessor import failure and ReasoningEffort/mistral_common processor-loading breakage are different import dependencies." + "reason": "AutoImageProcessor import failure and ReasoningEffort import failure are separate import regressions affecting different modules." }, { "left": "issue:43643", "right": "issue:44913", "accept": false, - "reason": "Missing fields with trust_remote_code=True and GPTNeoX rotary_pct not persisting are separate config serialization problems." + "reason": "trust_remote_code returning missing fields is not the same as GPTNeoXConfig rotary_pct not persisting on reload." }, { "left": "issue:43749", "right": "issue:43931", "accept": false, - "reason": "FSDP CPU-efficient loading and Qwen3-VL weight shape mismatch are both loading-related but clearly different failure modes and subsystems." + "reason": "FSDP CPU RAM efficient loading broken vs Qwen3-VL weight-shape mismatch; different loading paths and symptoms." }, { "left": "issue:33290", "right": "issue:36331", "accept": false, - "reason": "Adafactor+DeepSpeed OOM and CustomTrainer.compute_loss signature mismatch are different training bugs." + "reason": "AdaFactor+DeepSpeed OOM is a memory issue; CustomTrainer.compute_loss signature break is an API compatibility issue." }, { "left": "issue:34689", "right": "issue:36010", "accept": false, - "reason": "Llama 3.2 Vision load breakage in 4.46.2 and GenerationMixin import failure are unrelated regressions." + "reason": "Llama 3.2 Vision model loading regression and GenerationMixin import error are distinct regressions." }, { - "left": "issue:44190", - "right": "issue:44496", + "left": "issue:43873", + "right": "issue:44871", "accept": false, - "reason": "Local dataset loading in a training example and unrecognized OLMo-Hybrid config are unrelated issues." + "reason": "Quantization offloading behavior and Gemma-3 eos_token_id mismatch are unrelated." }, { - "left": "issue:43873", - "right": "issue:44871", + "left": "issue:44190", + "right": "issue:44496", "accept": false, - "reason": "Quantization/offloading behavior and Gemma-3 eos_token_id inconsistency are separate model/runtime bugs." + "reason": "Local dataset loading in a script is unrelated to an unrecognized model/config.json model_type error." }, { "left": "issue:43381", "right": "issue:43856", "accept": false, - "reason": "Gradient checkpointing in eval mode and MoE memory inefficiency are different training/runtime concerns." + "reason": "Gradient checkpointing eval-mode limitation is not the same as MoE memory inefficiency during training." }, { "left": "issue:44855", "right": "issue:45137", "accept": false, - "reason": "PyTorch 3.13 indentation/import parsing issue and DeepSpeed ZeRO3 deque underflow are unrelated." + "reason": "Python 3.13 IndentationError in DebertaV2 import is unrelated to DeepSpeed ZeRO3 deque pop error." }, { "left": "issue:43540", "right": "issue:44514", "accept": false, - "reason": "Qwen3OmniMoe video ValueError and Qwen2_5_VL batched chat-template crash are both multimodal but distinct code paths." + "reason": "Qwen3OmniMoe video processing ValueError and Qwen2_5_VL batched chat-template crash are different processors and failure modes." }, { "left": "issue:43316", "right": "issue:44877", "accept": false, - "reason": "Gemma3TextConfig API discrepancy and strict config blocking granite_speech loading are different config-compatibility bugs." + "reason": "Gemma3TextConfig API mismatch and strict config blocking granite_speech loading are separate config-handling issues." }, { "left": "issue:43673", "right": "issue:44743", "accept": false, - "reason": "GenerationMixin cache missing in v5 chunked_prefill and Qwen3.5 recurrent states resetting are unrelated generation bugs." + "reason": "GenerationMixin cache missing in v5.0.0 during chunked_prefill is unrelated to Qwen recurrent state reset in modular_qwen3_5.py." }, { "left": "issue:42175", "right": "issue:43316", "accept": false, - "reason": "Missing TensorFlow backend in torch-only install and Gemma3TextConfig API discrepancy are unrelated packaging vs config issues." + "reason": "TensorFlow missing from a torch-only install and Gemma3TextConfig inconsistency are unrelated packaging/config issues." }, { "left": "issue:43404", "right": "issue:43441", "accept": false, - "reason": "Weight tying bug in Mistral3ForConditionalGeneration and Ministral-3 FlashAttention failure are different model-specific regressions." + "reason": "Mistral3 lm_head weight tying bug and Ministral-3 FlashAttention failure are different model-specific regressions." }, { "left": "issue:44162", "right": "issue:44792", "accept": false, - "reason": "ESM2 breakage and janus generate-images test failure are not the same defect." + "reason": "ESM2 broken workflow issue and janus image-generation test failure do not share the same concrete bug." }, { "left": "issue:43065", "right": "issue:44617", "accept": false, - "reason": "Sam3PixelDecoder dummy Conv2d and Sam3Video CUDA OOM are separate SAM3 issues with different causes." + "reason": "Dummy Conv2d in Sam3PixelDecoder is a design/implementation question, while Sam3Video CUDA OOM is a runtime memory problem." }, { "left": "issue:25251", "right": "issue:28282", "accept": false, - "reason": "Pipeline top_k nesting behavior and AutoModel PyTorch import error are unrelated." + "reason": "Pipeline top_k output nesting change and AutoModel missing PyTorch import error are unrelated API behaviors." }, { - "left": "issue:43388", - "right": "issue:43688", + "left": "issue:43295", + "right": "issue:43825", "accept": false, - "reason": "gather_for_metrics dropping tuple labels and auxiliary-loss normalization in MoE models affect different training internals." + "reason": "Processor.tokenizer regression and pipeline translation error-message regression are different v5 behavior breaks." }, { - "left": "issue:43295", - "right": "issue:43825", + "left": "issue:43388", + "right": "issue:43688", "accept": false, - "reason": "Processor.tokenizer regression in v4.57.5 and pipeline translation support messaging in v5 are different API regressions." + "reason": "gather_for_metrics dropping tuple labels and auxiliary loss normalization are separate training bugs." }, { "left": "issue:16998", "right": "issue:22355", "accept": false, - "reason": "DeBERTa-V3 model_max_length question and transformers.onnx import error are unrelated." + "reason": "DeBERTa model_max_length question and missing transformers.onnx module are unrelated." } ] }, @@ -9034,6 +9060,7 @@ "nodes": [ "issue:32090", "issue:35141", + "issue:41720", "issue:41950", "issue:42617", "issue:43065", @@ -9068,13 +9095,12 @@ "issue:44792", "issue:44871", "issue:44913", - "issue:44933", "issue:44964", "issue:44977", "issue:45127", "issue:45200", + "issue:45245", "issue:45290", - "issue:45362", "issue:45381", "issue:45406", "issue:45563", @@ -9085,249 +9111,249 @@ "issue:41950|issue:43475", "issue:44077|issue:44534", "issue:44871|issue:45584", - "issue:45127|issue:45200", "issue:43550|issue:43701", + "issue:45127|issue:45200", "issue:43761|issue:44479", "issue:43606|issue:43701", "issue:43295|issue:44871", "issue:42617|issue:43575", - "issue:43452|issue:44496", "issue:43575|issue:43742", + "issue:43452|issue:44496", "issue:44734|issue:45406", "issue:43646|issue:44913", "issue:43504|issue:45698", - "issue:43931|issue:44060", "issue:43575|issue:43873", - "issue:43761|issue:44242", "issue:44617|issue:45584", - "issue:45563|issue:45584", + "issue:43761|issue:44242", + "issue:43931|issue:44060", "issue:32090|issue:35141", + "issue:45563|issue:45584", "issue:44464|issue:44871", - "issue:44479|issue:44734", "issue:44792|issue:44977", "issue:43065|issue:44610", + "issue:44479|issue:44734", "issue:44521|issue:45381", - "issue:43906|issue:45290", "issue:43827|issue:43994", + "issue:43906|issue:45290", "issue:44464|issue:45563", "issue:44534|issue:44913", "issue:44190|issue:44964", - "issue:44933|issue:45362" + "issue:41720|issue:45245" ], - "prepared_review_unit_hash": "baf49edef4a3cb92871b7d3d48ed51a7e724ce18e48710a8eb7bec22d7b300b8", + "prepared_review_unit_hash": "0e5afe1b5555a4f7d965b68abbb36abe5f9d4c98af485689c95c03a7f1a78ddb", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36917, - "estimated_input_tokens": 9230, - "estimated_eval_tokens": 18716 + "serialized_chars": 36957, + "estimated_input_tokens": 9240, + "estimated_eval_tokens": 18736 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a loose semantic grouping of mostly unrelated Transformers issues. None of the soft pairs look like true duplicates or mergeable into one fix; there are no PRs in the cluster.", - "confidence": 0.89, - "canonical_issue_reason": "issue:43646 is the broadest representative of the loading/initialization regression subset, but the cluster is not a single underlying bug.", + "summary": "This cluster is highly heterogeneous: it mixes unrelated bugs in loading, multimodal processing, generation, offloading, and config handling. None of the soft-edge pairs appear to be true duplicates of the same underlying bug/change.", + "confidence": 0.97, + "canonical_issue_reason": "No single issue is a clear duplicate hub for this mixed cluster; if one must be chosen, issue 43761 is the broadest, most clearly scoped regression report.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43646 is the closest thing to a cluster anchor because it is relatively general and overlaps the most with other loading/init-related reports.", + "best_issue_reason": "Issue 43761 is the cleanest representative bug report in the set: concrete regression framing, specific failure mode, and broadly relevant model behavior.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41950", "right": "issue:43475", "accept": false, - "reason": "Different failures: pipeline processor lookup vs missing vision encoder field in SAM3 video." + "reason": "Both touch multimodal/video paths, but one is a processor-selection bug and the other is a missing SAM3 video field; different code paths and symptoms." }, { "left": "issue:44077", "right": "issue:44534", "accept": false, - "reason": "Unrelated model internals: optional post_init constraint vs non-persistent buffer initialization junk." + "reason": "One is about optional post_init in patchtsmixer, the other about non-persistent buffers being corrupted in v5; unrelated bugs." }, { "left": "issue:44871", "right": "issue:45584", "accept": false, - "reason": "Different symptom and code path: EOS config mismatch vs Whisper empty-transcription failure after align_special_tokens." + "reason": "Both involve token/eos handling, but one is a Gemma config mismatch and the other is Whisper empty-transcription generation after align_special_tokens; not the same defect." }, { - "left": "issue:45127", - "right": "issue:45200", + "left": "issue:43550", + "right": "issue:43701", "accept": false, - "reason": "Different model behaviors: LoRA merge with tied embeddings vs Gemma 4 mm_token_type_ids defaulting." + "reason": "torch.compile+SDPA on Bamba and resume_from_checkpoint key mismatch are unrelated failures in different subsystems." }, { - "left": "issue:43550", - "right": "issue:43701", + "left": "issue:45127", + "right": "issue:45200", "accept": false, - "reason": "torch.compile/SDPA OOM on Bamba is unrelated to resume_from_checkpoint key mismatches." + "reason": "Both relate to model configuration, but one is LoRA/vocabulary merging collapse and the other is Gemma 4 multimodal token type defaults; different underlying issues." }, { "left": "issue:43761", "right": "issue:44479", "accept": false, - "reason": "CLIPVision hidden_states regression is not the same as Qwen video-input regression." + "reason": "Both are v5 regressions, but one is CLIPVisionModel hidden_states being dropped and the other is Qwen VL video input behavior; different model families and code paths." }, { "left": "issue:43606", "right": "issue:43701", "accept": false, - "reason": "CPU offload device mismatch is a different bug from checkpoint key mismatch on resume." + "reason": "CPU offload device mismatch for bark-small is unrelated to checkpoint resume key mismatches." }, { "left": "issue:43295", "right": "issue:44871", "accept": false, - "reason": "Processor.tokenizer regression is unrelated to eos_token_id configuration inconsistency." + "reason": "Processor.tokenizer API breakage is a different bug from Gemma eos_token_id config inconsistency." }, { "left": "issue:42617", "right": "issue:43575", "accept": false, - "reason": "Running 3d_parallel.py and TP OOM on Qwen2-57B are different runtime/resource problems." + "reason": "A parallel script failure and a tp-induced OOM when loading a large Qwen model are not the same bug." }, { - "left": "issue:43452", - "right": "issue:44496", + "left": "issue:43575", + "right": "issue:43742", "accept": false, - "reason": "gguf_file loading breakage is not the same as missing model_type config for custom model loading." + "reason": "Qwen2-57B tensor-parallel OOM and MobileLLM key error on load are distinct loading failures." }, { - "left": "issue:43575", - "right": "issue:43742", + "left": "issue:43452", + "right": "issue:44496", "accept": false, - "reason": "Tensor-parallel OOM and MobileLLM key error are distinct load-time issues." + "reason": "gguf_file breaking auto classes is different from a model config missing model_type." }, { "left": "issue:44734", "right": "issue:45406", "accept": false, - "reason": "KV-cache continuation indexing bug is unrelated to Gemma4Processor missing _tokenizer." + "reason": "Both affect transformers serve, but one crashes in KV-cache continuation indexing and the other lacks _tokenizer on Gemma4Processor." }, { "left": "issue:43646", "right": "issue:44913", "accept": false, - "reason": "Generic v5 custom model init breakage and GPTNeoX rotary_pct reload bug are different code paths." + "reason": "Custom model initialization breakage in v5 and GPTNeoX rotary_pct reload reset are unrelated configuration bugs." }, { "left": "issue:43504", "right": "issue:45698", "accept": false, - "reason": "Both involve loading, but one is a legacy preset field and the other is wrong custom-module resolution after save/load." - }, - { - "left": "issue:43931", - "right": "issue:44060", - "accept": false, - "reason": "Qwen3-VL weight-shape mismatch is unrelated to the tied-weights warning in Qwen3-Next." + "reason": "Legacy field loading for BEiT semantic segmentation is not the same as loading the wrong custom module after save_pretrained." }, { "left": "issue:43575", "right": "issue:43873", "accept": false, - "reason": "OOM during tensor-parallel load is not the same as quantization offloading behavior." + "reason": "Both concern memory/loading, but tp OOM on a specific model and offloading-with-quantization behavior are different problems." }, { - "left": "issue:43761", - "right": "issue:44242", + "left": "issue:44617", + "right": "issue:45584", "accept": false, - "reason": "Vision hidden_states regression and MoE load-balancing loss omission do not share a concrete fix." + "reason": "Sam3Video CUDA OOM and Whisper empty transcription generation are unrelated runtime failures." }, { - "left": "issue:44617", - "right": "issue:45584", + "left": "issue:43761", + "right": "issue:44242", "accept": false, - "reason": "Sam3Video CUDA OOM is unrelated to Whisper empty transcription after token alignment." + "reason": "A CLIPVisionModel forward regression is not the same as missing load-balancing loss when output_router_logits=False." }, { - "left": "issue:45563", - "right": "issue:45584", + "left": "issue:43931", + "right": "issue:44060", "accept": false, - "reason": "A stale generate warning and a real Whisper generation crash are different bugs." + "reason": "Qwen3-VL weight-shape mismatch and Qwen3-Next tied-weights warning are separate model/config issues." }, { "left": "issue:32090", "right": "issue:35141", "accept": false, - "reason": "Trainer GPU broadcast TypeError and embedding reinit after resize are different mechanisms." + "reason": "Trainer GPU broadcast TypeError and embedding reinitialization after resizing are unrelated bugs." }, { - "left": "issue:44464", - "right": "issue:44871", + "left": "issue:45563", + "right": "issue:45584", "accept": false, - "reason": "Compiled chunked generation inconsistency is not the same as EOS config mismatch." + "reason": "A stale warning in paged generate and Whisper failure on empty transcription are different generation-path issues." }, { - "left": "issue:44479", - "right": "issue:44734", + "left": "issue:44464", + "right": "issue:44871", "accept": false, - "reason": "Video-input regression in Qwen variants is unrelated to serve KV-cache tensor indexing." + "reason": "Compiled chunked generation inconsistency and Gemma eos_token_id mismatch are not the same bug." }, { "left": "issue:44792", "right": "issue:44977", "accept": false, - "reason": "Janus image-generation test failure is a different issue than Qwen3.5 flash-attention generation." + "reason": "Janus image-generation test failure and Qwen3.5 flash-attention generation bug are different model-specific issues." }, { "left": "issue:43065", "right": "issue:44610", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder and OmDet-Turbo processor size mismatch are unrelated model-specific bugs." + "reason": "A dummy Conv2d in Sam3PixelDecoder and OmDet-Turbo's 640px/224px mismatch are unrelated architecture/preprocessing problems." }, { - "left": "issue:44521", - "right": "issue:45381", + "left": "issue:44479", + "right": "issue:44734", "accept": false, - "reason": "All-zero assistant masks in multimodal chat templating and wrong vision_position_ids are separate multimodal bugs." + "reason": "Both are server or multimodal regressions, but one is video-input behavior in Qwen VL models and the other is serve KV-cache continuation indexing." }, { - "left": "issue:43906", - "right": "issue:45290", + "left": "issue:44521", + "right": "issue:45381", "accept": false, - "reason": "The isolated reproduction issue and apply_chat_template crashing on tool-call messages are not the same failure." + "reason": "All-zero assistant masks in apply_chat_template and incorrect vision_position_ids in Qwen2.5-VL are different multimodal template bugs." }, { "left": "issue:43827", "right": "issue:43994", "accept": false, - "reason": "Docs still referencing pipeline() is unrelated to Siglip2 producing bad outputs with AutoModel/pipeline." + "reason": "Docs still referencing pipeline() and SigLIP2 producing nonsense via AutoModel/pipeline are not the same underlying issue." + }, + { + "left": "issue:43906", + "right": "issue:45290", + "accept": false, + "reason": "An isolated reproduction issue and apply_chat_template crashing on tool-call messages are unrelated." }, { "left": "issue:44464", "right": "issue:45563", "accept": false, - "reason": "Compiled forward inconsistency and stale warning text are different severity and code paths." + "reason": "Compiled-forward inconsistency and a stale paged-generate warning do not share the same defect." }, { "left": "issue:44534", "right": "issue:44913", "accept": false, - "reason": "Non-persistent buffer junk and rotary_pct resetting on reload are distinct config/runtime problems." + "reason": "Non-persistent buffer junk in v5 and GPTNeoX rotary_pct reset on reload are different configuration/runtime problems." }, { "left": "issue:44190", "right": "issue:44964", "accept": false, - "reason": "Local dataset loading in a training script is unrelated to a model loading error for Phi-4 multimodal." + "reason": "Local dataset loading failure in a training script and a latest-transformers model load failure for Phi-4 multimodal are separate issues." }, { - "left": "issue:44933", - "right": "issue:45362", + "left": "issue:41720", + "right": "issue:45245", "accept": false, - "reason": "Missing import from image_utils is unrelated to Qwen3.5 chat crashes." + "reason": "Qwen3 auto device mapping cudaErrorAssert and category cardinality overflow are unrelated errors." } ] }, @@ -9356,7 +9382,6 @@ "issue:38617", "issue:39692", "issue:41628", - "issue:41720", "issue:41950", "issue:43065", "issue:43295", @@ -9365,6 +9390,7 @@ "issue:43479", "issue:43540", "issue:43550", + "issue:43582", "issue:43644", "issue:43650", "issue:43976", @@ -9384,275 +9410,269 @@ "issue:44792", "issue:44855", "issue:44913", + "issue:44933", "issue:44938", "issue:45127", "issue:45245", "issue:45276", "issue:45357", + "issue:45362", "issue:45375", "issue:45446", "issue:45542", "issue:45584", "issue:45588", "issue:45600", - "issue:45632", "issue:45803", "issue:45907" ], "soft_pairs": [ - "issue:41720|issue:45245", + "issue:44933|issue:45362", "issue:44792|issue:45600", "issue:44855|issue:45588", "issue:44737|issue:45803", - "issue:43388|issue:44079", + "issue:43388|issue:43582", "issue:45127|issue:45357", "issue:44462|issue:45907", "issue:43335|issue:44855", + "issue:44077|issue:44079", "issue:44938|issue:45542", "issue:43976|issue:44855", - "issue:44077|issue:44079", "issue:43550|issue:44190", "issue:43644|issue:44913", "issue:43540|issue:44734", "issue:44112|issue:45600", "issue:44496|issue:45803", - "issue:45276|issue:45584", "issue:43335|issue:44190", "issue:43650|issue:44220", + "issue:45276|issue:45584", "issue:39692|issue:44190", "issue:45245|issue:45584", - "issue:43295|issue:45276", "issue:43479|issue:45375", + "issue:43295|issue:45276", "issue:41950|issue:44016", "issue:36246|issue:41628", "issue:36246|issue:38617", "issue:44186|issue:44734", - "issue:44683|issue:45446", "issue:38175|issue:43065", + "issue:44683|issue:45446", "issue:44016|issue:44220", - "issue:44734|issue:45632", "issue:44466|issue:44913" ], - "prepared_review_unit_hash": "b2d44c9e058adc74073d17a29ad5d79fe313a6e4397af7f021738d196f838135", + "prepared_review_unit_hash": "7c05a5fef0e6bc7d733d19b0018f5e8ccd2b1858f0a288e6104d7478909575d2", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 32, - "serialized_chars": 36989, - "estimated_input_tokens": 9248, - "estimated_eval_tokens": 18752 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37184, + "estimated_input_tokens": 9296, + "estimated_eval_tokens": 18848 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous: import/export regressions, model-specific runtime bugs, config/serialization issues, and a few CI/test failures. None of the soft pairs look like the same concrete underlying bug or change, so I would not merge any of them.", - "confidence": 0.91, - "canonical_issue_reason": "issue:41628 is a clear, broadly relevant import regression with a simple symptom and strong duplicate-search value.", + "summary": "This cluster is noisy: most items are unrelated regressions, import/export issues, model-specific bugs, or CI/notebook problems. The soft-similar pairs mostly share broad vocabulary only, not the same concrete code path or fix.", + "confidence": 0.88, + "canonical_issue_reason": "issue:41628 is the broadest, most user-facing import/export regression in the set and best fits the loose theme of missing public symbols.", "canonical_pr_reason": null, - "best_issue_reason": "issue:41628 is the most generally useful representative bug report in this set because it is well-scoped, user-facing, and easy to recognize as a duplicate.", + "best_issue_reason": "issue:41628 is the clearest representative of the cluster\u2019s import-regression subset and is more general than the model-specific or highly niche reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:41720", - "right": "issue:45245", + "left": "issue:44933", + "right": "issue:45362", "accept": false, - "reason": "Different failures: Qwen3 device-mapping CUDA assert vs a category-count runtime limit." + "reason": "One is a missing import from image_utils; the other is a Qwen3.5 chat crash. Different failure modes and code paths." }, { "left": "issue:44792", "right": "issue:45600", "accept": false, - "reason": "Janus image-generation test failure is unrelated to Sam3LiteText auto-mapping references." + "reason": "Janus image generation test failure vs. auto_mappings referencing removed Sam3LiteText configs. Unrelated subsystems." }, { "left": "issue:44855", "right": "issue:45588", "accept": false, - "reason": "Python 3.13 import parsing issue is unrelated to flash_attention handling of s_aux=None." + "reason": "Python 3.13 import parse error in DebertaV2 vs. flash_attention AttributeError on s_aux=None. No shared bug." }, { "left": "issue:44737", "right": "issue:45803", "accept": false, - "reason": "XLNet device placement bug and a bare-except in Fuyu tensor conversion are different code paths." + "reason": "XLNet device-placement bug vs. FuyuBatchFeature bare-except handling. Different models and fixes." }, { "left": "issue:43388", - "right": "issue:44079", + "right": "issue:43582", "accept": false, - "reason": "Metric-gathering label truncation and ModelOutput key assignment are unrelated bugs." + "reason": "Metric-gathering label truncation vs. Apple Silicon allocator TypeError. No overlap in code path." }, { "left": "issue:45127", "right": "issue:45357", "accept": false, - "reason": "LoRA merge collapse and incorrect visual encoder save keys are both multimodal, but not the same defect." + "reason": "LoRA merge/tied-weights collapse vs. Qwen3.5 save_pretrained key serialization. Related only by broad model saving behavior." }, { "left": "issue:44462", "right": "issue:45907", "accept": false, - "reason": "Tokenizer loading ignores tokenizer.json, while the other issue is missing a revision passthrough in file listing." + "reason": "Tokenizer file selection vs. list_repo_files revision forwarding. Both repo-related, but not the same bug." }, { "left": "issue:43335", "right": "issue:44855", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug is unrelated to the DebertaV2 import parsing failure." + "reason": "SwitchTransformers config layer creation bug vs. DebertaV2 import parse failure. Unrelated." + }, + { + "left": "issue:44077", + "right": "issue:44079", + "accept": false, + "reason": "Optional post_init validation in patchtsmixer vs. ModelOutput key assignment semantics. Different objects and behavior." }, { "left": "issue:44938", "right": "issue:45542", "accept": false, - "reason": "Python 3.14 import/load failure and tensorboard-without-TF backend error are different issues." + "reason": "Python 3.14 load failure vs. tensorboard-only backend detection. Not the same regression." }, { "left": "issue:43976", "right": "issue:44855", "accept": false, - "reason": "Version-compatibility problem in Transformers 5.1.0 is unrelated to the DebertaV2 import regression." - }, - { - "left": "issue:44077", - "right": "issue:44079", - "accept": false, - "reason": "PatchTST post_init validation and ModelOutput key handling are separate framework bugs." + "reason": "Python version compatibility issue vs. a JIT parsing/import problem. Different layers of the stack." }, { "left": "issue:43550", "right": "issue:44190", "accept": false, - "reason": "Bamba torch.compile SDPA failure is unrelated to the local-dataset loading problem in the example script." + "reason": "torch.compile SDPA failure in Bamba vs. local dataset loading in an example script. Unrelated." }, { "left": "issue:43644", "right": "issue:44913", "accept": false, - "reason": "Non-persistent buffer serialization junk and GPTNeoX rotary_pct reload behavior do not share the same code path." + "reason": "Non-persistent buffer initialization vs. GPTNeoX config reload of rotary_pct. Different components." }, { "left": "issue:43540", "right": "issue:44734", "accept": false, - "reason": "Qwen3OmniMoe video-input validation and serve KV-cache continuation indexing are unrelated." + "reason": "Qwen3OmniMoe video input validation vs. serve KV-cache indexing. Distinct failures." }, { "left": "issue:44112", "right": "issue:45600", "accept": false, - "reason": "GraniteSpeech CI stale-device test and Sam3LiteText auto-mapping breakage are both CI-related but not the same bug." + "reason": "GraniteSpeech CI test flake vs. removed Sam3LiteText auto-mapping references. Both CI-related, but not the same bug." }, { "left": "issue:44496", "right": "issue:45803", "accept": false, - "reason": "Unrecognized model loading and Fuyu bare-except tensor conversion are unrelated." - }, - { - "left": "issue:45276", - "right": "issue:45584", - "accept": false, - "reason": "Embedding resize propagation bug and Whisper empty-transcription generation failure are different issues." + "reason": "Unrecognized model/config issue vs. bare except swallowing errors. No shared root cause." }, { "left": "issue:43335", "right": "issue:44190", "accept": false, - "reason": "SwitchTransformers config creation and local dataset loading are unrelated." + "reason": "SwitchTransformers sparse-layer config bug vs. dataset loading example failure. Unrelated." }, { "left": "issue:43650", "right": "issue:44220", "accept": false, - "reason": "A placeholder data issue and an fbank feature extraction bug are unrelated." + "reason": "Placeholder issue title vs. fbank feature extraction bug. No meaningful match." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Gemma4 embedding resize issue vs. Whisper empty-transcription generation failure. Different models and code paths." }, { "left": "issue:39692", "right": "issue:44190", "accept": false, - "reason": "SigLIP2 docs/example errors do not match the local dataset loading failure." + "reason": "SigLIP2 documentation example errors vs. local dataset loading in a training script. Separate issues." }, { "left": "issue:45245", "right": "issue:45584", "accept": false, - "reason": "Category-count limit error is unrelated to Whisper generation after special-token alignment." + "reason": "Category cardinality runtime error vs. Whisper generation bug. Unrelated." }, { - "left": "issue:43295", - "right": "issue:45276", + "left": "issue:43479", + "right": "issue:45375", "accept": false, - "reason": "Processor/tokenizer regression and embedding-resize propagation are different multimodal API bugs." + "reason": "Phi4Multimodal default config initialization vs. Qwen3_5MoeVisionConfig field dropping. Both config issues, but different models and semantics." }, { - "left": "issue:43479", - "right": "issue:45375", + "left": "issue:43295", + "right": "issue:45276", "accept": false, - "reason": "Default config initialization with None is not the same as missing deepstack_visual_indexes serialization." + "reason": "Processor/tokenizer regression vs. resize_token_embeddings not updating Gemma4 outputs. Different mechanisms." }, { "left": "issue:41950", "right": "issue:44016", "accept": false, - "reason": "Video-classification pipeline processor mismatch and notebook syntax error are unrelated." + "reason": "Video-classification pipeline processor lookup vs. notebook syntax error. No shared underlying bug." }, { "left": "issue:36246", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but one is a missing model-specific class export and the other is a missing top-level AutoImageProcessor import." + "reason": "Both are import errors, but they involve different missing symbols and different modules; not the same concrete fix." }, { "left": "issue:36246", "right": "issue:38617", "accept": false, - "reason": "Different missing symbols from different modules; no shared underlying bug." + "reason": "Qwen2_5_VLImageProcessor export missing vs. layer_type_validation import missing from configuration_utils. Different symbols and files." }, { "left": "issue:44186", "right": "issue:44734", "accept": false, - "reason": "LayoutLMv2 tokenizer padding/NER crash is unrelated to serve KV-cache indexing." - }, - { - "left": "issue:44683", - "right": "issue:45446", - "accept": false, - "reason": "Both touch flex_attention, but one is a compile failure on torch>=2.9 and the other is a version-check/import gate issue." + "reason": "Tokenizer crashes on NER padding vs. server KV-cache continuation indexing. Different tokenizer/model-server paths." }, { "left": "issue:38175", "right": "issue:43065", "accept": false, - "reason": "SigLIP2 zero-probability behavior and Dummy Conv2d in Sam3PixelDecoder are unrelated." + "reason": "Zero probabilities in SigLIP2 vs. dummy Conv2d in Sam3PixelDecoder. No common bug." }, { - "left": "issue:44016", - "right": "issue:44220", + "left": "issue:44683", + "right": "issue:45446", "accept": false, - "reason": "Notebook syntax error and audio feature extraction bug are unrelated." + "reason": "Compiled flex_attention fails on torch>=2.9 vs. incorrect PyTorch version check for AuxRequest import. Related to attention imports only at a very high level." }, { - "left": "issue:44734", - "right": "issue:45632", + "left": "issue:44016", + "right": "issue:44220", "accept": false, - "reason": "KV-cache continuation crash and trust_remote_code cache-path collision are different problems." + "reason": "Notebook syntax error vs. audio feature extraction issue. Unrelated." }, { "left": "issue:44466", "right": "issue:44913", "accept": false, - "reason": "Tokenizer serialization/loading and GPTNeoX config reload are unrelated." + "reason": "Serialization of tied lm_head weights vs. GPTNeoX rotary_pct reload behavior. Different config/serialization bugs." } ] }, @@ -9705,6 +9725,7 @@ "issue:44496", "issue:44521", "issue:44561", + "issue:44734", "issue:44933", "issue:44964", "issue:44987", @@ -9717,7 +9738,6 @@ "issue:45325", "issue:45335", "issue:45357", - "issue:45405", "issue:45406", "issue:45584", "issue:45593", @@ -9726,24 +9746,25 @@ "issue:45759" ], "soft_pairs": [ + "issue:44734|issue:45632", "issue:42175|issue:44279", "issue:43404|issue:45216", "issue:44561|issue:45632", "issue:44246|issue:45003", "issue:43643|issue:44496", - "issue:44964|issue:45600", "issue:45406|issue:45632", + "issue:44964|issue:45600", "issue:44112|issue:44964", "issue:42907|issue:43441", "issue:43854|issue:44964", - "issue:43295|issue:45335", "issue:44016|issue:44062", + "issue:43295|issue:45335", "issue:43976|issue:44987", "issue:43673|issue:44336", "issue:43531|issue:45600", - "issue:43526|issue:43901", "issue:43824|issue:44933", "issue:44496|issue:45042", + "issue:43526|issue:43901", "issue:43299|issue:45216", "issue:44075|issue:44368", "issue:45335|issue:45584", @@ -9751,227 +9772,226 @@ "issue:43901|issue:44496", "issue:43381|issue:45593", "issue:43299|issue:45357", - "issue:44351|issue:45278", "issue:44496|issue:45759", + "issue:44351|issue:45278", "issue:44521|issue:45325", "issue:44242|issue:45200", - "issue:43976|issue:45405", "issue:41950|issue:45020" ], - "prepared_review_unit_hash": "a1cef573d538b396d26cb6aba00c40c0e58a0b366fb9635d999d94c4c9895cc4", + "prepared_review_unit_hash": "6183d0bee9e580044c980e52bf1a321a5c0b6b4a9a7021d633f3604696d3bea3", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36934, - "estimated_input_tokens": 9234, - "estimated_eval_tokens": 18724 + "serialized_chars": 36957, + "estimated_input_tokens": 9240, + "estimated_eval_tokens": 18736 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a grab-bag of distinct Transformers regressions around model loading, config/export, and multimodal processing. None of the suggested soft pairs look like true duplicates; they mostly share only broad subsystem or version-regression similarity.", - "confidence": 0.77, - "canonical_issue_reason": "Issue 45020 is the broadest open umbrella for the recurring remote_code/model-loading regressions in this set, so it is the best canonical anchor among the issues.", + "summary": "The items are mostly heterogeneous bug reports with only loose lexical overlap around loading/imports/remote_code/multimodal behavior; none of the soft pairs look like true duplicates or mergeable PR-equivalents.", + "confidence": 0.68, + "canonical_issue_reason": "issue:45020 is the broadest open umbrella here: it covers recent-version breakage for remote_code models, which is the closest shared theme across the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45020 is the most representative high-level issue here because it captures the general class of recent-version breakage affecting custom/remote-code models, which is the closest shared theme across the cluster.", + "best_issue_reason": "issue:45020 is the best single representative issue because it is broad, user-facing, and likely to subsume several of the version-regression/model-loading reports in this set.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44734", + "right": "issue:45632", + "accept": false, + "reason": "Serve KV-cache tensor indexing crash and trust_remote_code cache-path collision are different code paths and symptoms." + }, { "left": "issue:42175", "right": "issue:44279", "accept": false, - "reason": "One is a pip extra/backend packaging complaint; the other is a generic dependency issue. Same ecosystem, but not the same bug." + "reason": "One is missing TensorFlow in an extra install; the other is a generic dependency issue. Not the same bug." }, { "left": "issue:43404", "right": "issue:45216", "accept": false, - "reason": "Different models and different failures: untied lm_head weights vs incorrect checkpoint saving." + "reason": "Both relate to model saving/state, but they affect different models and failure modes: untied weights vs incorrect checkpoint keys." }, { "left": "issue:44561", "right": "issue:45632", "accept": false, - "reason": "Both mention trust_remote_code, but one is a removed helper causing breakage and the other is a cache-path collision." + "reason": "Both involve trust_remote_code, but one is a missing helper API and the other is a local cache-path collision." }, { "left": "issue:44246", "right": "issue:45003", "accept": false, - "reason": "Slow import timing and unsafe sys.modules access are not the same concrete defect." + "reason": "Import slowness and unsafe sys.modules access are related at most by module import, not the same concrete defect." }, { "left": "issue:43643", "right": "issue:44496", "accept": false, - "reason": "One is missing fields from AutoConfig with trust_remote_code; the other is an unrecognized-model/config issue." + "reason": "Missing fields from AutoConfig with trust_remote_code is not the same issue as failing to recognize a model type in config.json." }, { - "left": "issue:44964", - "right": "issue:45600", + "left": "issue:45406", + "right": "issue:45632", "accept": false, - "reason": "Different targets and root causes: Phi-4 multimodal load failure vs stale auto_mappings references breaking CI." + "reason": "Gemma4Processor missing _tokenizer in serve is unrelated to remote-code cache path collisions." }, { - "left": "issue:45406", - "right": "issue:45632", + "left": "issue:44964", + "right": "issue:45600", "accept": false, - "reason": "Processor _tokenizer attribute crash is unrelated to a local-model cache path collision." + "reason": "These are different model-loading regressions affecting different components and fixes." }, { "left": "issue:44112", "right": "issue:44964", "accept": false, - "reason": "A stale CI test failure in GraniteSpeech is unrelated to a model-load regression for Phi-4 multimodal." + "reason": "A stale CI test failure in GraniteSpeech is not the same as a model-loading failure for Phi-4 multimodal." }, { "left": "issue:42907", "right": "issue:43441", "accept": false, - "reason": "Both are Ministral-related, but one is save/dequantization and the other is FlashAttention loading." + "reason": "Both mention Ministral, but saving dequantized weights and FlashAttention loading are distinct bugs." }, { "left": "issue:43854", "right": "issue:44964", "accept": false, - "reason": "Different model families and different load failures; not the same underlying bug." + "reason": "Different models and failure points; one is a GLM import/load issue, the other is Phi-4 multimodal loading." }, { - "left": "issue:43295", - "right": "issue:45335", + "left": "issue:44016", + "right": "issue:44062", "accept": false, - "reason": "These are separate regressions in different model classes and code paths." + "reason": "A notebook syntax error and an AddedToken argument TypeError are unrelated." }, { - "left": "issue:44016", - "right": "issue:44062", + "left": "issue:43295", + "right": "issue:45335", "accept": false, - "reason": "Notebook syntax error and AddedToken argument handling are unrelated." + "reason": "These affect different model families and different APIs: processor/tokenizer handling vs embedding resizing." }, { "left": "issue:43976", "right": "issue:44987", "accept": false, - "reason": "Both are version-related breakages, but one is Python compatibility and the other is a specific model load failure." + "reason": "Python-version compatibility and a model load failure are not the same underlying bug." }, { "left": "issue:43673", "right": "issue:44336", "accept": false, - "reason": "Generation cache regression and ANSI color output are unrelated." + "reason": "Generation cache behavior during chunked prefill and ANSI formatting in loading_report are unrelated." }, { "left": "issue:43531", "right": "issue:45600", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to stale config references in auto_mappings." - }, - { - "left": "issue:43526", - "right": "issue:43901", - "accept": false, - "reason": "ImageProcessor reduce_labels bug and docs mismatch for text classification are different issues." + "reason": "Qwen3-MoE sliding_window behavior and removed auto_mappings configs are different subsystems and failures." }, { "left": "issue:43824", "right": "issue:44933", "accept": false, - "reason": "Both are import errors, but they refer to different missing symbols in different modules." + "reason": "Both are import-related, but they concern different missing symbols in different modules." }, { "left": "issue:44496", "right": "issue:45042", "accept": false, - "reason": "Unrecognized model config and PIL backend requiring torchvision are separate problems." + "reason": "Unrecognized model config and PIL backend image-processor torchvision dependency are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "A BeitImageProcessorFast label bug and a documentation/behavior mismatch in TextClassificationPipeline are not duplicates." }, { "left": "issue:43299", "right": "issue:45216", "accept": false, - "reason": "Both involve Qwen regressions, but one is loading a MoE model and the other is broken save_pretrained output." + "reason": "Both are Qwen3.5 regressions, but one is model loading and the other is incorrect save_pretrained output." }, { "left": "issue:44075", "right": "issue:44368", "accept": false, - "reason": "Unused SGD args and a tie_word_embeddings warning are not the same bug." + "reason": "Unused SGD arguments and a tie_word_embeddings warning concern different components and symptoms." }, { "left": "issue:45335", "right": "issue:45584", "accept": false, - "reason": "These affect different models and different APIs: embedding resize vs Whisper generation after token alignment." + "reason": "Embedding resizing in t5gemma and Whisper empty-transcription generation are unrelated bugs." }, { "left": "issue:43976", "right": "issue:44279", "accept": false, - "reason": "Python-version incompatibility and generic dependency trouble do not point to one concrete defect." + "reason": "Python compatibility and a generic dependency issue do not describe the same defect." }, { "left": "issue:43901", "right": "issue:44496", "accept": false, - "reason": "Documentation drift is unrelated to model config loading failure." + "reason": "Docs mismatch for text classification is not the same as a config/model-type loading error." }, { "left": "issue:43381", "right": "issue:45593", "accept": false, - "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary-loss omission are unrelated training bugs." + "reason": "Gradient checkpointing eval-mode behavior and D-FINE aux-loss handling are different training bugs." }, { "left": "issue:43299", "right": "issue:45357", "accept": false, - "reason": "Both are Qwen3.5 regressions, but one is model loading and the other is incorrect visual-encoder keys on save." + "reason": "Related model family, but one is load failure and the other is incorrect visual-encoder keys on save." }, { - "left": "issue:44351", - "right": "issue:45278", + "left": "issue:44496", + "right": "issue:45759", "accept": false, - "reason": "A specific missing HybridCache import is not the same concrete issue as a broad report of many import errors." + "reason": "An unrecognized model config and composite-config unwrapping in AutoModelForCausalLM are different code paths." }, { - "left": "issue:44496", - "right": "issue:45759", + "left": "issue:44351", + "right": "issue:45278", "accept": false, - "reason": "Unrecognized model config and missing text_config unwrapping are different failure modes." + "reason": "A missing HybridCache export is just one symptom; the other is a broader import-error report, not the same bug." }, { "left": "issue:44521", "right": "issue:45325", "accept": false, - "reason": "Assistant-mask generation and rope-index scaling are separate multimodal processing bugs." + "reason": "Assistant mask generation for multimodal chat templates and rope-index scaling are separate multimodal bugs." }, { "left": "issue:44242", "right": "issue:45200", "accept": false, - "reason": "Missing load-balancing loss and defaulting mm_token_type_ids are different training-time behaviors." - }, - { - "left": "issue:43976", - "right": "issue:45405", - "accept": false, - "reason": "Python support regression and an unreleased peft minimum version are unrelated release issues." + "reason": "Router-logit load-balancing loss and Gemma4 mm_token_type_ids defaults are unrelated training issues." }, { "left": "issue:41950", "right": "issue:45020", "accept": false, - "reason": "Video-classification processor selection and remote_code model regressions are different code paths." + "reason": "A video-classification pipeline processor lookup bug is unrelated to remote_code model breakage." } ] }, @@ -10013,6 +10033,7 @@ "issue:43723", "issue:43854", "issue:43901", + "issue:43976", "issue:44016", "issue:44162", "issue:44261", @@ -10044,14 +10065,15 @@ "issue:45907" ], "soft_pairs": [ + "issue:43976|issue:45405", "issue:43479|issue:43525", "issue:44779|issue:45812", "issue:41950|issue:45823", "issue:43577|issue:45216", - "issue:44683|issue:45542", - "issue:39692|issue:43901", "issue:38175|issue:44617", + "issue:44683|issue:45542", "issue:43299|issue:43854", + "issue:39692|issue:43901", "issue:44279|issue:45405", "issue:43295|issue:44779", "issue:41950|issue:43650", @@ -10067,208 +10089,214 @@ "issue:42994|issue:43317", "issue:44466|issue:45276", "issue:44261|issue:44263", - "issue:43441|issue:44683", "issue:43650|issue:45290", + "issue:43441|issue:44683", "issue:44623|issue:45397", "issue:44623|issue:44749", "issue:43723|issue:44162" ], - "prepared_review_unit_hash": "adb073f25f58e30e9c5212466cb00fd7202106a3872134d38938abed56a465ca", + "prepared_review_unit_hash": "be6284bc14fbfa19ba5f38516efc0779780349f6374f88399ecba77678efeef2", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 28, - "serialized_chars": 35496, - "estimated_input_tokens": 8874, - "estimated_eval_tokens": 18004 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36357, + "estimated_input_tokens": 9090, + "estimated_eval_tokens": 18436 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is mostly heterogeneous Transformers bugs across unrelated subsystems. None of the suggested soft pairs look like true duplicates under a strict triage standard; the strongest standalone issue is the Granite AutoTokenizer v5 regression.", - "confidence": 0.71, - "canonical_issue_reason": "Issue 45812 is the most concrete high-impact bug report: it describes a silent v4\u2192v5 regression, has broad model impact, and includes clear symptom framing.", + "summary": "No soft pairs look like true duplicates or the same fix; most are only loosely related by broad subsystems like tokenization, loading, or multimodal configs. I would keep them separate. The strongest issue to anchor triage is the open Granite token-ID regression.", + "confidence": 0.9, + "canonical_issue_reason": "issue:45812 is the strongest canonical issue candidate: it is open, clearly scoped, describes a concrete silent regression, and has multiple inbound references, making it the best anchor for dedupe/triage.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45812 is the best representative issue because it is specific, user-facing, and likely to attract the right fix work quickly.", + "best_issue_reason": "issue:45812 is the most actionable and broadly impactful issue in this set, with a specific reproducible tokenizer regression and ongoing attention from other references.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Unrelated problems: Python version compatibility vs an unreleased dependency version bump." + }, { "left": "issue:43479", "right": "issue:43525", "accept": false, - "reason": "Different configs and failures: Phi4Multimodal default sub-config initialization vs Llama4 missing pad_token_id." + "reason": "Both are config-related, but one is multimodal default initialization and the other is a missing pad_token_id attribute; different bugs." }, { "left": "issue:44779", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and likely different code paths; not the same bug." + "reason": "Both involve tokenization regressions, but they affect different model families and likely different code paths/root causes." }, { "left": "issue:41950", "right": "issue:45823", "accept": false, - "reason": "Unrelated problems: video pipeline processor lookup vs Gemma4 device mismatch during forward." + "reason": "Different subsystems and failures: video-classification processor lookup vs Gemma4 PLE device mismatch." }, { "left": "issue:43577", "right": "issue:45216", "accept": false, - "reason": "Different lifecycle stages and symptoms: dtype propagation on load vs incorrect checkpoint saving." - }, - { - "left": "issue:44683", - "right": "issue:45542", - "accept": false, - "reason": "Completely different subsystems: compiled flex_attention vs TensorFlow/tensorboard dependency handling." + "reason": "Both concern model loading/saving behavior, but the observed failures and affected code paths are not the same." }, { - "left": "issue:39692", - "right": "issue:43901", + "left": "issue:38175", + "right": "issue:44617", "accept": false, - "reason": "Both mention docs, but one is a broken SigLIP2 example and the other is stale pipeline documentation; not the same issue." + "reason": "Completely different symptoms and components: zero probabilities in SigLIP2 vs CUDA OOM in Sam3Video." }, { - "left": "issue:38175", - "right": "issue:44617", + "left": "issue:44683", + "right": "issue:45542", "accept": false, - "reason": "Different models and failure modes: zero probabilities in SigLIP2 vs CUDA OOM in Sam3Video." + "reason": "Attention/torch-related on one side, backend installation issue on the other; no shared concrete bug." }, { "left": "issue:43299", "right": "issue:43854", "accept": false, - "reason": "Both are model-loading failures, but for different models and likely unrelated regressions." + "reason": "Both are model-loading failures in v5, but they concern different models and distinct failure modes." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "Both are documentation-related, but one is a broken SigLIP2 example and the other is stale pipeline docs." }, { "left": "issue:44279", "right": "issue:45405", "accept": false, - "reason": "One is a generic dependency issue; the other is a specific minimum-version bump to an unreleased PEFT version." + "reason": "Dependency complaints, but not the same dependency or failure mechanism." }, { "left": "issue:43295", "right": "issue:44779", "accept": false, - "reason": "Both involve tokenizer behavior regressions, but they are different symptoms and code paths." + "reason": "Both are regressions around tokenizer/processor behavior, but the concrete bug is different: processor.tokenizer access vs wrong tokenization results." }, { "left": "issue:41950", "right": "issue:43650", "accept": false, - "reason": "No substantive overlap: video-classification processor lookup vs an unrelated placeholder issue." + "reason": "No shared bug: video pipeline processor lookup vs unrelated placeholder issue." }, { "left": "issue:44462", "right": "issue:45081", "accept": false, - "reason": "Different tokenizer-loading bugs: ignoring tokenizer.json vs a Mistral regex patch crash." + "reason": "Different tokenizer-loading failures: repository tokenizer.json being ignored vs a crash in Mistral regex patching." }, { "left": "issue:33290", "right": "issue:45656", "accept": false, - "reason": "Both mention deepspeed optimizers, but one is OOM with Adafactor and the other is double-stepping the optimizer." + "reason": "Both mention deepspeed/optimizer behavior, but one is OOM with Adafactor and the other is double optimizer stepping." }, { "left": "issue:44297", "right": "issue:44462", "accept": false, - "reason": "Related area, but different bugs: tokenizer_class serialization mismatch vs AutoTokenizer preferring repository tokenizer files." + "reason": "Both touch tokenizer serialization/loading, but one is save_pretrained metadata mismatch and the other is AutoTokenizer ignoring tokenizer.json." }, { "left": "issue:45081", "right": "issue:45907", "accept": false, - "reason": "Unrelated: tokenizer regex patch crash vs list_repo_files revision forwarding in CI." + "reason": "Tokenizer patch crash vs list_repo_files not forwarding revision; unrelated APIs." }, { "left": "issue:44016", "right": "issue:44291", "accept": false, - "reason": "Notebook syntax error and init_empty_weights TypeError are unrelated." + "reason": "Notebook syntax error vs init_empty_weights TypeError; unrelated." }, { "left": "issue:45600", "right": "issue:45712", "accept": false, - "reason": "Both are repo hygiene/CI issues, but they concern different stale references and different files." + "reason": "Both are repo maintenance/CI issues, but they target different files and problems." }, { "left": "issue:43441", "right": "issue:45834", "accept": false, - "reason": "Different model-specific failures: FlashAttention incompatibility vs long-ocr input index error." + "reason": "Both mention model/runtime failures, but one is FlashAttention on Ministral-3 and the other is an OCR index error in Kosmos2.5." }, { "left": "issue:43450", "right": "issue:44279", "accept": false, - "reason": "Different categories: batched video processor shape bug vs generic dependency issue." + "reason": "Video processor shape bug vs generic dependency issue; no common underlying bug." }, { "left": "issue:43065", "right": "issue:44898", "accept": false, - "reason": "Distinct model/component bugs: dummy Conv2d in Sam3PixelDecoder vs Perceiver non-default-resolution failure." + "reason": "Different vision-model bugs: a dummy Conv2d in Sam3PixelDecoder vs Perceiver resize/interpolation failure." }, { "left": "issue:42994", "right": "issue:43317", "accept": false, - "reason": "Both touch quantization/offload, but one is saving a quantized model and the other is loading a dequantized model with device_map=auto." + "reason": "Both involve quantization/loading, but saving a quantized model and loading a dequantized model with offload are different issues." }, { "left": "issue:44466", "right": "issue:45276", "accept": false, - "reason": "Different serialization/weight-management bugs with no clear shared failure path." + "reason": "Both relate to embeddings/serialization, but one is device-dependent lm_head serialization and the other is resize_token_embeddings not propagating to Gemma4 components." }, { "left": "issue:44261", "right": "issue:44263", "accept": false, - "reason": "Unrelated defects: RMS norm epsilon precision issue vs torch.split return handling." + "reason": "One is a missing rms_norm_eps config precision issue; the other is torch.split return handling in GlmMoeDsaIndexer." }, { - "left": "issue:43441", - "right": "issue:44683", + "left": "issue:43650", + "right": "issue:45290", "accept": false, - "reason": "Different attention-stack failures: FlashAttention model support vs compiled flex_attention on newer torch." + "reason": "Placeholder issue vs chat template crash on tool-call assistant messages; unrelated." }, { - "left": "issue:43650", - "right": "issue:45290", + "left": "issue:43441", + "right": "issue:44683", "accept": false, - "reason": "No overlap: placeholder issue vs chat template crash with tool-call assistant messages." + "reason": "Both are attention-related, but one is model-specific FlashAttention failure and the other is compiled flex_attention on newer torch." }, { "left": "issue:44623", "right": "issue:45397", "accept": false, - "reason": "Different loading/saving paths: processor.save_pretrained missing files vs gemma-4 zero3 from_pretrained." + "reason": "Both are save/load related, but they affect different APIs and failure modes." }, { "left": "issue:44623", "right": "issue:44749", "accept": false, - "reason": "Different performance/serialization problems; the Chinese issue is about slow filtering after upgrade, not processor saving." + "reason": "Processor save_pretrained file omission vs a performance regression when filtering data; unrelated." }, { "left": "issue:43723", "right": "issue:44162", "accept": false, - "reason": "Generic tokenizer-loading issue vs a model-specific ESM2 breakage; no evidence of the same root cause." + "reason": "Tokenizer loading bug vs a separate model breakage in ESM2; no same underlying defect." } ] }, @@ -10344,9 +10372,9 @@ "issue:43653|issue:45593", "issue:43262|issue:44521", "issue:43673|issue:43761", - "issue:43784|issue:45542", - "issue:44361|issue:45593", "issue:44016|issue:45823", + "issue:44361|issue:45593", + "issue:43784|issue:45542", "issue:42175|issue:43976", "issue:44671|issue:44779", "issue:43901|issue:43994", @@ -10359,9 +10387,9 @@ "issue:43525|issue:43643", "issue:43450|issue:44617", "issue:42491|issue:43299", - "issue:45561|issue:45563", "issue:44589|issue:45362", "issue:41628|issue:43976", + "issue:45561|issue:45563", "issue:43901|issue:44246", "issue:44336|issue:44373", "issue:44242|issue:45593", @@ -10371,20 +10399,20 @@ "issue:43352|issue:45250", "issue:43531|issue:45362" ], - "prepared_review_unit_hash": "a6d9cc3484531fa8c6fabd37bd34fc08f98ad405e6231a79d81972a684bece0a", + "prepared_review_unit_hash": "bd3825e2701f6905fc75a2257b3e0df9e763be5df3c1054498b6a0a27fb658be", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 35401, + "serialized_chars": 35403, "estimated_input_tokens": 8851, "estimated_eval_tokens": 17958 }, @@ -10393,186 +10421,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is heterogeneous: most items are distinct transformer bugs spanning tokenizers, multimodal processors, model loading, packaging, docs, and performance. The soft pairs only share broad subsystem similarity, not the same concrete underlying bug or change, so none should be merged as duplicates.", - "confidence": 0.94, - "canonical_issue_reason": null, + "summary": "This cluster is very heterogeneous: most items are unrelated regressions across tokenizers, vision/audio processors, model loading, packaging, and docs. I do not see any true duplicate pair among the soft edges.", + "confidence": 0.9, + "canonical_issue_reason": "issue:43122 is the broadest, most central user-facing regression in the set (tokenization changed between v4.57.3 and v5.0), so it is the best anchor if one issue must represent this mixed cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44521 is the strongest standalone bug report here: it is open, detailed, and has clear inbound interest, but it is not a duplicate of the other issues in this cluster.", + "best_issue_reason": "issue:43122 is the most general and broadly applicable regression report here; it has the clearest duplicate potential compared with the more model-specific or documentation-only reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43653", "right": "issue:45593", "accept": false, - "reason": "Different areas and symptoms: BigBirdTokenizer special-token decoding vs D-FINE auxiliary-loss behavior." + "reason": "BigBirdTokenizer special-token decoding vs D-FINE auxiliary-loss behavior; different models and code paths." }, { "left": "issue:43262", "right": "issue:44521", "accept": false, - "reason": "Both involve chat template/processor code, but one is audio sampling-rate defaulting and the other is multimodal assistant-mask generation; different bugs." + "reason": "Both mention apply_chat_template, but one is audio sampling rate and the other is multimodal assistant masks; not the same bug." }, { "left": "issue:43673", "right": "issue:43761", "accept": false, - "reason": "Generation cache/chunked_prefill regression vs CLIP vision hidden_states output; unrelated code paths." + "reason": "Generation cache/chunked_prefill regression vs CLIPVisionModel hidden_states bug; unrelated subsystems." }, { - "left": "issue:43784", - "right": "issue:45542", + "left": "issue:44016", + "right": "issue:45823", "accept": false, - "reason": "Separate import-time dependency issues: sentence-transformers NameError vs missing TensorFlow backend wiring." + "reason": "Notebook syntax error vs Gemma4 device-map forward mismatch; unrelated." }, { "left": "issue:44361", "right": "issue:45593", "accept": false, - "reason": "MLukeTokenizer task-time AttributeError is unrelated to D-FINE auxiliary losses." + "reason": "MLukeTokenizer task AttributeError vs D-FINE auxiliary losses; unrelated." }, { - "left": "issue:44016", - "right": "issue:45823", + "left": "issue:43784", + "right": "issue:45542", "accept": false, - "reason": "Notebook syntax/docs issue vs Gemma4 forward device-map runtime bug." + "reason": "sentence-transformers import NameError vs missing TensorFlow backend packaging; both are import-time issues but not the same one." }, { "left": "issue:42175", "right": "issue:43976", "accept": false, - "reason": "Packaging/backend install mismatch vs Python-version compatibility; not the same failure." + "reason": "Package extras/backends issue vs Python version compatibility regression; different failure modes." }, { "left": "issue:44671", "right": "issue:44779", "accept": false, - "reason": "Different model families and likely different root causes: CamemBERT masked-LM outputs vs Deepseek tokenizer regression." + "reason": "CamemBERT masked-LM predictions vs Deepseek tokenizer regression; both are model/tokenizer bugs but not the same code path or model." }, { "left": "issue:43901", "right": "issue:43994", "accept": false, - "reason": "Docs mismatch for return_all_scores vs model/pipeline inference producing bad outputs; different issues." + "reason": "Docs mismatch for return_all_scores vs incorrect SigLIP2 outputs; one is documentation, the other is model behavior." }, { "left": "issue:43122", "right": "issue:43650", "accept": false, - "reason": "Tokenizer behavior regression vs unrelated placeholder issue titled 'ADD THE DATA'." + "reason": "Tokenizer regression across versions vs an unrelated low-information issue; no shared bug." }, { "left": "issue:44987", "right": "issue:45542", "accept": false, - "reason": "Model loading failure for a specific repo vs TensorFlow backend installation problem; unrelated." + "reason": "Loading a specific model on >=5.1.0 vs TensorFlow backend installation error; unrelated." }, { "left": "issue:43824", "right": "issue:43976", "accept": false, - "reason": "Specific missing export/import symbol vs Python 3.9/3.10 compatibility; not the same bug." + "reason": "Missing Qwen2_5_VL import vs Python 3.9/3.10 compatibility problem; unrelated." }, { "left": "issue:44297", "right": "issue:45397", "accept": false, - "reason": "Tokenizer save metadata mismatch vs Gemma4 zero3 loading failure; different code paths." + "reason": "Tokenizer save_pretrained metadata mismatch vs Gemma4 ZeRO3 from_pretrained failure; unrelated." }, { "left": "issue:43525", "right": "issue:43881", "accept": false, - "reason": "Llama4Config missing pad_token_id is a config-field bug, but glm-4v loading failed is a separate model-specific loading issue." + "reason": "Llama4Config missing pad_token_id vs glm-4v loading failure; different models and causes." }, { "left": "issue:44188", "right": "issue:44336", "accept": false, - "reason": "Attention-kernel divergence under torch.compile vs ANSI color codes in loading_report; unrelated." + "reason": "torch.compile attention-kernel divergence vs ANSI codes in loading_report; unrelated." }, { "left": "issue:43525", "right": "issue:43643", "accept": false, - "reason": "Both mention config fields, but one is a missing pad_token_id on Llama4Config and the other is trust_remote_code losing fields from AutoConfig; too different to merge." + "reason": "Both involve missing config fields, but one is Llama4 pad_token_id and the other is trust_remote_code AutoConfig fields; not the same bug." }, { "left": "issue:43450", "right": "issue:44617", "accept": false, - "reason": "Incorrect batched video-processor shape vs CUDA OOM in Sam3Video; different symptom and fix." + "reason": "Batched video processor shape bug vs Sam3Video CUDA OOM; unrelated." }, { "left": "issue:42491", "right": "issue:43299", "accept": false, - "reason": "Both touch Qwen3 MoE compatibility, but one is LoRA/model reuse across branches and the other is a Qwen3VL loading regression; not the same concrete bug." - }, - { - "left": "issue:45561", - "right": "issue:45563", - "accept": false, - "reason": "pytest-xdist file race vs stale generate() warning; unrelated." + "reason": "Qwen3 MoE LoRA compatibility across hf versions vs Qwen3VL loading break on v5; related family, but different concrete failures." }, { "left": "issue:44589", "right": "issue:45362", "accept": false, - "reason": "Storage-type import error vs Qwen3.5 chat crash; no shared code path." + "reason": "Float8 storage deserialization error vs Qwen3.5 chat crash; unrelated." }, { "left": "issue:41628", "right": "issue:43976", "accept": false, - "reason": "Missing AutoImageProcessor import is unrelated to Python version support." + "reason": "Cannot import AutoImageProcessor vs Python version support problem; unrelated." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "pytest-xdist file race vs stale generate() warning; different testing/runtime issues." }, { "left": "issue:43901", "right": "issue:44246", "accept": false, - "reason": "Docs wording issue vs import-time slowness; not the same bug." + "reason": "Documentation mismatch vs slow import behavior; unrelated." }, { "left": "issue:44336", "right": "issue:44373", "accept": false, - "reason": "Terminal ANSI formatting bug vs incorrect docstring for position_ids; unrelated." + "reason": "ANSI output in loading_report vs wrong docstring for position_ids; both docs/UI-ish, but not the same bug." }, { "left": "issue:44242", "right": "issue:45593", "accept": false, - "reason": "Both mention missing losses, but they are different losses in different models/flags and not one fixable PR." + "reason": "MoE load-balancing loss omission vs D-FINE auxiliary losses; both are loss-related but for different models and mechanisms." }, { "left": "issue:44295", "right": "issue:45397", "accept": false, - "reason": "Buffer access error for position_ids vs Gemma4 zero3 loading failure; different root causes." + "reason": "position_ids buffer access bug vs Gemma4 from_pretrained/ZeRO3 issue; unrelated." }, { "left": "issue:41950", "right": "issue:44220", "accept": false, - "reason": "Video-classification pipeline processor lookup vs audio fbank extraction issue; different modalities and bugs." + "reason": "Video-classification pipeline selecting image processors vs audio fbank extraction issue; different modalities and code paths." }, { "left": "issue:44246", "right": "issue:44496", "accept": false, - "reason": "Import performance regression vs unrecognized-model config error; unrelated." + "reason": "Import latency regression vs unrecognized model/config error; unrelated." }, { "left": "issue:43352", "right": "issue:45250", "accept": false, - "reason": "Specific model not supporting Flash Attention 2.0 yet vs a generic Flash Attention topic; too broad and not the same change." + "reason": "Model-specific Flash Attention 2 support error vs generic Flash Attention 2 discussion; not the same concrete bug." }, { "left": "issue:43531", "right": "issue:45362", "accept": false, - "reason": "Qwen3-MoE sliding_window bug vs Qwen3.5 chat crash; related model family but different failure modes and fixes." + "reason": "Qwen3-MoE sliding_window issue vs Qwen3.5-35B chat crash; different models and failure signatures." } ] }, @@ -10660,231 +10688,231 @@ "issue:44877|issue:45375", "issue:44779|issue:45292", "issue:45292|issue:45584", - "issue:45292|issue:45701", "issue:43901|issue:45003", + "issue:45292|issue:45701", "issue:43232|issue:44336", "issue:42886|issue:45229", - "issue:43065|issue:43408", "issue:44496|issue:45003", + "issue:43065|issue:43408", "issue:44987|issue:45459", "issue:45375|issue:45759", "issue:43257|issue:45137", "issue:43262|issue:44246", - "issue:44060|issue:44077", "issue:44995|issue:45200", + "issue:44060|issue:44077", "issue:43475|issue:43540", - "issue:43653|issue:44162", "issue:44466|issue:45335", + "issue:43653|issue:44162", "issue:44368|issue:45563" ], - "prepared_review_unit_hash": "543015ab137fe6a1c9c073c111cc5cc3f5f51e4e1ca042d98ff764661a1f26c5", + "prepared_review_unit_hash": "5711006eb8754be780c969999188723f74ea9a1daa1b16a7b688643c4b5161cc", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36866, - "estimated_input_tokens": 9217, - "estimated_eval_tokens": 18690 + "serialized_chars": 36863, + "estimated_input_tokens": 9216, + "estimated_eval_tokens": 18688 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is highly heterogeneous: import/API regressions, model-specific config bugs, generation issues, tokenizer behavior, and performance complaints. None of the soft pairs look like the same concrete bug/change, so all candidate duplicate edges are rejected.", - "confidence": 0.82, - "canonical_issue_reason": "issue:45003 is the best anchor because it is a broad core-library bug in modeling_utils, making it more central and cross-cutting than the other mostly model-specific tickets.", + "summary": "The cluster is heterogeneous: mostly unrelated issue reports spanning import errors, model/config mismatches, generation bugs, and model-specific regressions. The soft matches are all superficial overlaps rather than the same underlying bug or change.", + "confidence": 0.89, + "canonical_issue_reason": "Issue 45003 is the broadest, most cross-cutting internal bug in the set; the rest are mostly narrower model- or feature-specific regressions, so it is the best representative if one must be chosen.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45003 is the most suitable global representative: it targets a core utility path, has broad potential impact, and is closer to a root-cause issue than the more symptom-specific tickets.", + "best_issue_reason": "Issue 45003 is the most generally impactful and system-level item in an otherwise diverse cluster, making it the best single issue candidate.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36246", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but they involve different missing symbols from different packages and code paths." + "reason": "Both are import errors, but they involve different symbols and different packages; not the same defect." }, { "left": "issue:36246", "right": "issue:44351", "accept": false, - "reason": "Same broad symptom class, but different missing imports and unrelated modules." + "reason": "Separate missing-import problems for unrelated names/modules." }, { "left": "issue:43526", "right": "issue:45003", "accept": false, - "reason": "One is a BEiT image processor label bug; the other is an unsafe sys.modules access in modeling_utils." + "reason": "One is a BeitImageProcessorFast label-reduction bug; the other is unsafe sys.modules access in modeling_utils. Different code paths." }, { "left": "issue:44336", "right": "issue:44485", "accept": false, - "reason": "ANSI terminal output handling and GLM-5 RoPE are unrelated subsystems and failures." + "reason": "ANSI-output handling and GLM-5 RoPE implementation are unrelated issues." }, { "left": "issue:45003", "right": "issue:45561", "accept": false, - "reason": "A modeling_utils import/runtime bug and a pytest-xdist file race are not the same code-path problem." + "reason": "sys.modules access in modeling_utils is unrelated to a pytest-xdist race in testing utils." }, { "left": "issue:43673", "right": "issue:44561", "accept": false, - "reason": "Chunked-prefill cache handling and removal of is_torch_fx_available affect different generation/import surfaces." + "reason": "Generation cache behavior and removal of is_torch_fx_available are different regression classes." }, { "left": "issue:44246", "right": "issue:44871", "accept": false, - "reason": "Import latency and Gemma eos_token_id mismatch are unrelated issues." + "reason": "Load-balancing loss logic and Gemma-3 eos_token_id config mismatch are unrelated." }, { "left": "issue:44242", "right": "issue:44857", "accept": false, - "reason": "Both mention losses, but one is gating router loss computation and the other is an AMP/CUDA crash in a different loss module." + "reason": "Different model families and different failures: router-loss accounting vs AMP/CUDA crash." }, { "left": "issue:44162", "right": "issue:45200", "accept": false, - "reason": "ESM2 breakage and mm_token_type_ids defaults are different model/config problems." + "reason": "ESM2 breakage and Gemma 4 mm_token_type_ids defaulting are unrelated model-specific bugs." }, { "left": "issue:44295", "right": "issue:44749", "accept": false, - "reason": "A position_ids buffer bug and a tokenization slowdown/regression are not the same defect." + "reason": "A position_ids buffer error and a tokenizer-speed regression are not the same issue." }, { "left": "issue:36246", "right": "issue:43596", "accept": false, - "reason": "Different import targets and different DeepSpeed/BERT failure modes." + "reason": "ImportError for Qwen2.5-VL image processor vs DeepSpeed ZeRO3 BertModel index error; unrelated." }, { "left": "issue:44877", "right": "issue:45375", "accept": false, - "reason": "Both involve config strictness, but one is loading granite_speech and the other is a missing field being dropped in Qwen vision config." + "reason": "Both are config-related, but they concern different model families and different missing-fields/loading behaviors." }, { "left": "issue:44779", "right": "issue:45292", "accept": false, - "reason": "Tokenizer correctness and resize_token_embeddings/output_embeddings are unrelated behaviors." + "reason": "DeepSeek tokenization regression and resize_token_embeddings propagation are different bugs." }, { "left": "issue:45292", "right": "issue:45584", "accept": false, - "reason": "Embedding resize propagation and Whisper empty-transcription generation failure are different bugs." + "reason": "Embedding resize propagation and Whisper empty-transcription handling are unrelated." }, { - "left": "issue:45292", - "right": "issue:45701", + "left": "issue:43901", + "right": "issue:45003", "accept": false, - "reason": "Resize-token-embedding behavior is unrelated to a version-dependent tokenization change." + "reason": "Docs wording for TextClassificationPipeline is unrelated to sys.modules access in modeling_utils." }, { - "left": "issue:43901", - "right": "issue:45003", + "left": "issue:45292", + "right": "issue:45701", "accept": false, - "reason": "Documentation mismatch in a pipeline and an unsafe sys.modules access are unrelated." + "reason": "Embedding resize behavior and general tokenization changes are not the same defect." }, { "left": "issue:43232", "right": "issue:44336", "accept": false, - "reason": "Generation kwargs handling after sync_gpus and ANSI report formatting are unrelated." + "reason": "Generation kwargs update after sync_gpus and ANSI code emission are unrelated." }, { "left": "issue:42886", "right": "issue:45229", "accept": false, - "reason": "Offline cache loading and a multi-GPU OOM are not the same failure." + "reason": "Offline cache loading and Gemma4 multi-GPU OOM are distinct problems." }, { - "left": "issue:43065", - "right": "issue:43408", + "left": "issue:44496", + "right": "issue:45003", "accept": false, - "reason": "Sam3PixelDecoder dummy conv and sam3_video vs sam3_tracker warning are different SAM3 issues." + "reason": "Unrecognized model config and unsafe sys.modules access are different failure modes." }, { - "left": "issue:44496", - "right": "issue:45003", + "left": "issue:43065", + "right": "issue:43408", "accept": false, - "reason": "Unrecognized model config and sys.modules access are unrelated problems." + "reason": "Both mention SAM3, but one is a pixel-decoder implementation issue and the other is a model-type warning from config mismatch." }, { "left": "issue:44987", "right": "issue:45459", "accept": false, - "reason": "Model loading failure for a specific repo and protobuf error masking are different tokenizer/loading paths." + "reason": "A loading failure on transformers>=5.1.0 and hidden tokenizer errors from protobuf exception handling are unrelated." }, { "left": "issue:45375", "right": "issue:45759", "accept": false, - "reason": "Missing strict config field and AutoModelForCausalLM text_config unwrapping are distinct multimodal config issues." + "reason": "Missing Qwen3.5MoeVisionConfig field and AutoModelForCausalLM.from_config text_config unwrapping are different config-handling bugs." }, { "left": "issue:43257", "right": "issue:45137", "accept": false, - "reason": "Qwen3 MOE weight conversion under accelerate+deepspeed is unrelated to a ZeRO3 deque underflow." + "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed and a ZeRO3 deque index error are not the same code-path problem." }, { "left": "issue:43262", "right": "issue:44246", "accept": false, - "reason": "Audio chat-template sampling-rate defaults and import performance are unrelated." + "reason": "Audio sampling-rate defaults and MoE load-balancing loss are unrelated." }, { - "left": "issue:44060", - "right": "issue:44077", + "left": "issue:44995", + "right": "issue:45200", "accept": false, - "reason": "Incorrect tied-weights warning and patchtsmixer post_init policy are different model-specific changes." + "reason": "Stale cache on second forward pass vs Gemma 4 token-type defaults are different bugs." }, { - "left": "issue:44995", - "right": "issue:45200", + "left": "issue:44060", + "right": "issue:44077", "accept": false, - "reason": "GlmMoeDsa cache corruption and Gemma mm_token_type_ids defaults are unrelated." + "reason": "A tied-weights warning in Qwen3-Next and a patchtsmixer post_init policy change are unrelated." }, { "left": "issue:43475", "right": "issue:43540", "accept": false, - "reason": "A missing attribute on Sam3VisionEncoderOutput and a Qwen3OmniMoe video ValueError are different model failures." + "reason": "Both are multimodal/video-related, but they affect different models and different failure points." }, { - "left": "issue:43653", - "right": "issue:44162", + "left": "issue:44466", + "right": "issue:45335", "accept": false, - "reason": "BigBirdTokenizer special-token registration and ESM2 breakage are unrelated." + "reason": "Both involve embedding/tied-weight behavior, but one is serialization consistency and the other is resize propagation into decoder embeddings; not the same underlying bug." }, { - "left": "issue:44466", - "right": "issue:45335", + "left": "issue:43653", + "right": "issue:44162", "accept": false, - "reason": "Both mention embedding/weight handling, but one is device-dependent serialization and the other is resize_token_embeddings propagation." + "reason": "BigBirdTokenizer special-token registration and ESM2 breakage are unrelated." }, { "left": "issue:44368", "right": "issue:45563", "accept": false, - "reason": "A tie_word_embeddings warning and a stale num_return_sequences warning are not the same bug." + "reason": "A tied-embeddings warning and a stale num_return_sequences warning are unrelated." } ] }, @@ -10992,20 +11020,20 @@ "issue:45310|issue:45375", "issue:41628|issue:44351" ], - "prepared_review_unit_hash": "1b079788843463b9c4b084bc8eae1f15e8fd181b9a05414b6a8a52d938304e41", + "prepared_review_unit_hash": "3a82fd340f8e2c0f553641ab7a1c66894debb5e51293a973f4da26bb0c565e69", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37435, + "serialized_chars": 37434, "estimated_input_tokens": 9359, "estimated_eval_tokens": 18974 }, @@ -11014,210 +11042,216 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a grab-bag of unrelated Transformers issues, with one clear duplicate pair: the Qwen3ForCausalLM VRAM leak reports. I would anchor the cluster on the earlier, more detailed report and reject the rest as distinct bugs.", - "confidence": 0.83, - "canonical_issue_reason": "issue:42673 is the best canonical issue because issue:43704 is the same Qwen3ForCausalLM VRAM leak report with nearly identical title and scope, and 42673 has the fuller discussion/activity.", + "summary": "This cluster is mostly a mix of superficially similar Transformers bug reports, but the only clear duplicate edge is the repeated Qwen3ForCausalLM VRAM leak report. The rest differ by model, subsystem, or concrete failure mode despite overlapping keywords like v5, quantization, generation, or import errors.", + "confidence": 0.92, + "canonical_issue_reason": "issue:42673 is the best canonical issue because it is the earlier, more discussed report of the exact same Qwen3ForCausalLM VRAM leak in multi-dataloader-thread usage as issue:43704.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42673 is the strongest issue to keep as the representative because it is the clearest duplicate target and the most substantive report among the matching pair.", + "best_issue_reason": "issue:42673 is the strongest representative issue in this cluster: it is an exact duplicate candidate with clear repro context and a specific code-path bug, unlike the many broader or unrelated reports here.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44368", "right": "issue:44977", "accept": false, - "reason": "Both involve Qwen3.5, but one is a fine-tuning warning and the other is a flash-attention generation failure; different concrete bugs." + "reason": "Both mention Qwen3.5, but one is a tie_word_embeddings warning during LoRA fine-tuning while the other is a flash-attention generation failure. Different bugs and code paths." }, { "left": "issue:44995", "right": "issue:45588", "accept": false, - "reason": "Different code paths: stale indexer cache on second forward pass vs flash-attention crash on sink-less models." + "reason": "GlmMoeDsa stale indexer cache on second forward pass is unrelated to flash_attention.py crashing on s_aux=None for sink-less models." }, { "left": "issue:45397", "right": "issue:45656", "accept": false, - "reason": "Both mention DeepSpeed, but one is a zero3 from_pretrained issue and the other is an optimizer step-count bug." + "reason": "Gemma-4 Zero3 from_pretrained failure and a double optimizer-step bug with deepspeed are different failures with no shared concrete code-path." }, { "left": "issue:43673", "right": "issue:44734", "accept": false, - "reason": "Both are generation/cache related, but they describe different failures in different paths: chunked_prefill cache missing vs KV continuation tensor indexing." + "reason": "One is a GenerationMixin cache regression in chunked_prefill; the other is a transformers serve tensor-indexing crash in /v1/responses. Not the same bug." }, { "left": "issue:45137", "right": "issue:45600", "accept": false, - "reason": "Completely different bugs: ZeRO3 deque underflow vs a broken auto_mappings reference to removed Sam3LiteText configs." + "reason": "DeepSpeed ZeRO3 deque underflow and stale auto_mappings references to removed Sam3LiteText configs are unrelated." }, { "left": "issue:43931", "right": "issue:44368", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch is unrelated to the Qwen3.5 LoRA tie_word_embeddings warning." + "reason": "Qwen3-VL-30B weight-shape mismatch is a model-loading error; the other is a warning about tie_word_embeddings during fine-tuning." }, { "left": "issue:44933", "right": "issue:45459", "accept": false, - "reason": "Different import/tokenizer issues: missing image_utils export vs protobuf-decode-error handling hiding tokenizer errors." + "reason": "A missing import from image_utils and tokenizer error masking when protobuf is absent are separate import/exception-handling issues." }, { "left": "issue:44869", "right": "issue:45356", "accept": false, - "reason": "Whisper timestamp decoding crash and Kimi-K2.5 tokenizer regression are unrelated." + "reason": "Whisper timestamp decoding on a trailing replacement character and a Kimi-K2.5 tokenizer regression are different tokenization/decoding bugs." }, { "left": "issue:43232", "right": "issue:43673", "accept": false, - "reason": "Both are generation internals, but the reported failures are different and not clearly the same bug." + "reason": "sync_gpus generation kwargs handling and missing GenerationMixin cache during chunked_prefill are distinct generation regressions." }, { "left": "issue:44297", "right": "issue:44623", "accept": false, - "reason": "Both concern save_pretrained, but one is tokenizer metadata mismatch and the other is missing processor files." + "reason": "tokenizer.save_pretrained writing the wrong tokenizer_class and processor.save_pretrained missing files are related only at a high level; not the same concrete bug." }, { "left": "issue:43673", "right": "issue:45278", "accept": false, - "reason": "A specific generation cache bug is not the same as a broad set of import errors after upgrading versions." + "reason": "A generation cache regression and a broad report of many import errors after upgrading to v5.5.0 are not the same issue." }, { "left": "issue:43646", "right": "issue:44162", "accept": false, - "reason": "Trust-remote-code config field loss is a different issue from ESM2 being broken." + "reason": "Transformers 5 breaking custom model initialization is broader than the ESM2-specific breakage described here; different impacted code paths." }, { "left": "issue:44361", "right": "issue:44749", "accept": false, - "reason": "Different symptoms and domains: MLukeTokenizer task AttributeError vs slow data filtering after version upgrade." + "reason": "MLukeTokenizer attribute errors on tasks and a performance slowdown after upgrading to 5.3.0 are unrelated." }, { "left": "issue:44623", "right": "issue:45656", "accept": false, - "reason": "processor.save_pretrained file emission and a duplicated optimizer step are unrelated." + "reason": "Missing files from processor.save_pretrained and optimizer.step being called twice under deepspeed are completely different behaviors." }, { "left": "issue:44749", "right": "issue:45656", "accept": false, - "reason": "Performance regression after upgrade is not the same as optimizer step being called twice." + "reason": "A throughput regression after upgrade and duplicate optimizer stepping with deepspeed do not describe the same bug." }, { "left": "issue:44749", "right": "issue:45310", "accept": false, - "reason": "Data filtering slowdown and Qwen3.5 MoE from_pretrained failure are different bugs." + "reason": "Performance regression after upgrade is not the same as Qwen3.5 Moe from_pretrained failing." }, { "left": "issue:28282", "right": "issue:30990", "accept": false, - "reason": "Both are loading-related, but one is a missing PyTorch ImportError and the other is Sentence Transformers hanging while loading." + "reason": "AutoModel import error due to missing PyTorch and Sentence Transformers hanging while loading are unrelated installation/runtime problems." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are opposite directions and different bugs." + "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure with v5 are different model I/O problems." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF32 API-settings discussion is unrelated to InternVL2 meta-initialization incompatibility." + "reason": "TF32 API-setting guidance and old InternVL2 meta-initialization incompatibility are unrelated." }, { "left": "issue:43475", "right": "issue:45381", "accept": false, - "reason": "Different multimodal model bugs: missing SAM3 vision field vs wrong qwen2.5-vl video position ids." + "reason": "SAM 3 video missing fpn_position_embeddings and qwen2.5-vl video vision_position_ids mismatch are separate video-model bugs." }, { "left": "issue:43475", "right": "issue:43643", "accept": false, - "reason": "Missing SAM3 attribute and missing fields in AutoConfig from trust_remote_code are not the same underlying issue." + "reason": "A missing field on Sam3VisionEncoderOutput is unrelated to AutoConfig returning missing fields under trust_remote_code=True." }, { "left": "issue:45230", "right": "issue:45397", "accept": false, - "reason": "A placeholder bug report is not a duplicate of the Gemma-4 zero3 issue." + "reason": "A generic bug report and a specific Gemma-4 Zero3 loading failure are not duplicates." }, { "left": "issue:44038", "right": "issue:44623", "accept": false, - "reason": "Qwen3-VL-Moe Transformers 5.0 incompatibility and processor.save_pretrained missing files are unrelated." + "reason": "Qwen3-VL-Moe model bug and processor.save_pretrained missing files are unrelated." }, { "left": "issue:41628", "right": "issue:44933", "accept": false, - "reason": "Missing AutoImageProcessor import and missing image_utils import are similar in theme but not the same concrete missing symbol." + "reason": "Two import errors, but for different symbols and different subsystems: AutoImageProcessor versus image_utils." }, { "left": "issue:45081", "right": "issue:45103", "accept": false, - "reason": "Mistral regex/tokenizer backend crash and auto-docstring annotation crash are different code paths." + "reason": "A Mistral regex patch crash and an auto_docstring future-annotations crash are unrelated exceptions." }, { "left": "issue:43475", "right": "issue:44661", "accept": false, - "reason": "SAM3 vision output attribute error is unrelated to add-new-model-like failing on TOKENIZER_MAPPING_NAMES." + "reason": "SAM 3 video output attribute loss and add-new-model-like failing on TOKENIZER_MAPPING_NAMES are different failures." }, { "left": "issue:42994", "right": "issue:43525", "accept": false, - "reason": "Quantized saving failure and Llama4Config missing pad_token_id are different bugs." + "reason": "Quantized saving failure and Llama4Config missing pad_token_id are separate model/config issues." }, { "left": "issue:45375", "right": "issue:45478", "accept": false, - "reason": "The missing deepstack_visual_indexes field may contribute to Qwen3.5 Moe problems, but these are not clearly the same concrete bug as the from_pretrained error." + "reason": "Missing deepstack_visual_indexes in Qwen3_5MoeVisionConfig and a Qwen3.5 Moe from_pretrained error are related by model family only, not the same concrete bug." }, { "left": "issue:42673", "right": "issue:43704", "accept": true, - "reason": "Same issue: identical Qwen3ForCausalLM VRAM leak in multiple dataloader threads, with 43704 clearly duplicating 42673." + "reason": "Exact same title and same failure mode: Qwen3ForCausalLM leaks VRAM when used in multiple dataloader threads. This is a clear duplicate." }, { "left": "issue:43232", "right": "issue:44188", "accept": false, - "reason": "Generation kwargs handling after sync_gpus is unrelated to attention-kernel divergence under torch.compile." + "reason": "A generation kwargs update bug after sync_gpus and attention-kernel divergence under torch.compile are different issues." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Quantization/offloading behavior and pipeline-parallel incorrect output are different failures." + "reason": "Quantization/offloading behavior and incorrect LLM output under pipeline parallelism do not share the same root cause." }, { "left": "issue:45310", "right": "issue:45375", "accept": false, - "reason": "Both are Qwen3.5 MoE-related, but one is a from_pretrained failure and the other is a missing config field; not enough to treat as the same bug." + "reason": "One is a Qwen3.5 Moe from_pretrained error; the other is a missing config field silently dropped by strict serialization. Related ecosystem, different bugs." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "AutoImageProcessor missing import and HybridCache missing import are separate API export regressions." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded overall. The representative issue choice is supported by the packet: 42673 and 43704 have the same title and clearly match as duplicates, with 42673 being earlier and more active. The rest of the soft-edge verdicts stay conservative and are generally justified by distinct titles/symptoms rather than assuming deeper linkage." + "feedback": "The summary is well grounded in the packet: issue:42673 and issue:43704 are indeed identical titles with a clear duplicate pattern, and the rest of the cluster is mostly separated by model or subsystem. The soft-edge rejections are conservative overall and do not overclaim duplicate status." }, "evaluator_used": true, "retried": false, @@ -11263,7 +11297,6 @@ "issue:44038", "issue:44060", "issue:44079", - "issue:44164", "issue:44188", "issue:44222", "issue:44242", @@ -11282,6 +11315,7 @@ "issue:45137", "issue:45200", "issue:45310", + "issue:45356", "issue:45561", "issue:45712", "issue:45753", @@ -11294,14 +11328,14 @@ "issue:44222|issue:45310", "issue:36010|issue:41628", "issue:43673|issue:44188", - "issue:42994|issue:44164", "issue:42175|issue:45712", + "issue:42994|issue:43824", "issue:43232|issue:44792", "issue:43377|issue:44242", - "issue:43526|issue:45561", "issue:44038|issue:45310", - "issue:44496|issue:45561", + "issue:43526|issue:45561", "issue:43504|issue:45561", + "issue:44496|issue:45561", "issue:44964|issue:45137", "issue:42175|issue:44933", "issue:42898|issue:44568", @@ -11312,221 +11346,228 @@ "issue:43784|issue:43824", "issue:44188|issue:44373", "issue:44623|issue:45310", - "issue:42898|issue:45812", "issue:43278|issue:45137", + "issue:42898|issue:45812", "issue:44734|issue:45103", "issue:30333|issue:33290", "issue:43723|issue:44568", - "issue:44060|issue:45125" + "issue:44060|issue:45125", + "issue:42898|issue:45356" ], - "prepared_review_unit_hash": "4e5c1d3b7eb161c15dbd50c89984734a0cb0682b24af27e034a19f1309075557", + "prepared_review_unit_hash": "fa44a991afb32d81abb7feb027f8a9dd1da1d160227b175565e56a5be1f05365", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36545, - "estimated_input_tokens": 9137, - "estimated_eval_tokens": 18530 + "soft_pair_count": 31, + "serialized_chars": 36918, + "estimated_input_tokens": 9230, + "estimated_eval_tokens": 18716 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is mostly a grab bag of distinct Transformers regressions and import/runtime issues. None of the soft pairs look safe to collapse into duplicates; they usually differ by model family, failure mode, or code path.", - "confidence": 0.89, - "canonical_issue_reason": "Issue:45310 is the cleanest representative of the small Qwen3.5/MoE regression subtheme: it is concrete, versioned, and model-specific, but it still does not subsume the other issues.", + "summary": "These items are mostly unrelated transformers bugs that were grouped by loose title similarity rather than shared root cause. No soft pair looks like a safe duplicate merge candidate.", + "confidence": 0.24, + "canonical_issue_reason": "issue:43824 is the strongest single anchor only because it has the most concrete user-facing import regression signal and the highest reference activity; however, the cluster is too heterogeneous to treat it as a true canonical duplicate target.", "canonical_pr_reason": null, - "best_issue_reason": "Issue:45310 is the most actionable single anchor in this cluster because it states a specific from_pretrained failure for a named model/version combination and is closest to the other Qwen/MoE reports.", + "best_issue_reason": "issue:43824 is the most actionable representative of the set: it is a concrete import failure, has strong discussion/reference signals, and is self-contained enough for triage even though it does not unify the other issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44945", "right": "issue:45753", "accept": false, - "reason": "Different bugs: pipeline-parallel output corruption vs an infinite loop on a specific image." + "reason": "Both mention Qwen models, but one is incorrect LLM output under pipeline parallelism and the other is an infinite loop on a specific image; different failure modes and likely different code paths." }, { "left": "issue:42831", "right": "issue:44222", "accept": false, - "reason": "Both mention FP8, but one is an accuracy issue and the other is a save_pretrained failure." + "reason": "One is an accuracy regression in FineGrainedFP8, the other is a save_pretrained failure for FP8 moe models; related theme, but not the same bug." }, { "left": "issue:42222", "right": "issue:45200", "accept": false, - "reason": "Unrelated models and symptoms: VitPose breakage vs Gemma 4 token-type defaults." + "reason": "Vitpose model breakage and Gemma 4 mm_token_type_ids defaulting are unrelated model-specific issues." }, { "left": "issue:44222", "right": "issue:45310", "accept": false, - "reason": "Same broad MoE area, but different code paths: save_pretrained failure vs from_pretrained loading error." + "reason": "Both involve FP8/Qwen-related regressions, but one is moe save_pretrained and the other is a from_pretrained error on transformers>=5.4.0; not the same concrete bug." }, { "left": "issue:36010", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but for different symbols and different breakpoints in the package exports." + "reason": "Both are import errors, but they involve different symbols (GenerationMixin vs AutoImageProcessor) and different compatibility problems." }, { "left": "issue:43673", "right": "issue:44188", "accept": false, - "reason": "Generation cache/chunked-prefill issue is separate from attention-kernel divergence under torch.compile." + "reason": "Chunked prefill GenerationMixin cache issues and attention-kernel divergence under torch.compile are separate generation vs kernel behaviors." }, { - "left": "issue:42994", - "right": "issue:44164", + "left": "issue:42175", + "right": "issue:45712", "accept": false, - "reason": "Both are save/load related, but quantized model saving and extra_state handling are distinct bugs." + "reason": "Package backend dependency omission and leftover dummy classes leaking into dir(transformers) are unrelated maintenance issues." }, { - "left": "issue:42175", - "right": "issue:45712", + "left": "issue:42994", + "right": "issue:43824", "accept": false, - "reason": "Dependency/backend installation mismatch is unrelated to leftover dummy class cleanup." + "reason": "Quantized model saving and missing Qwen2.5-VL import are different subsystems and failure surfaces." }, { "left": "issue:43232", "right": "issue:44792", "accept": false, - "reason": "Generation kwargs after sync_gpus is unrelated to a Janus image-generation test failure." + "reason": "A generation kwargs update bug after sync_gpus is unrelated to a failed janus image-generation test." }, { "left": "issue:43377", "right": "issue:44242", "accept": false, - "reason": "Missing padding-mask support in MIMI encoder is unrelated to load-balancing loss omission." - }, - { - "left": "issue:43526", - "right": "issue:45561", - "accept": false, - "reason": "Label reduction bug in BeitImageProcessorFast is unrelated to the xdist captured_info.txt race." + "reason": "Missing padding-mask support in MIMI encoder and router-logits load-balancing loss are different model behaviors." }, { "left": "issue:44038", "right": "issue:45310", "accept": false, - "reason": "Related model family, but different models and failures: Qwen3-VL-Moe vs Qwen3.5 MoE from_pretrained." + "reason": "Qwen3-VL-Moe and Qwen3.5 Moe both concern Qwen, but the model variants and reported from_pretrained failures are not clearly the same bug." }, { - "left": "issue:44496", + "left": "issue:43526", "right": "issue:45561", "accept": false, - "reason": "Model config loading error is unrelated to a pytest-xdist file race." + "reason": "A BeitImageProcessorFast label-reduction bug and a pytest-xdist race on captured_info.txt are unrelated." }, { "left": "issue:43504", "right": "issue:45561", "accept": false, - "reason": "Legacy-field model loading bug is unrelated to the testing-utils race condition." + "reason": "Beit semantic-segmentation preset loading and an xdist file race are different issues with no shared code path." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Unrecognized model loading in from_pretrained is unrelated to the testing-utils captured_info.txt race." }, { "left": "issue:44964", "right": "issue:45137", "accept": false, - "reason": "Phi-4 multimodal loading and DeepSpeed ZeRO3 deque errors are different problems." + "reason": "Phi-4 multimodal loading failure and DeepSpeed ZeRO3 deque underflow are distinct runtime problems." }, { "left": "issue:42175", "right": "issue:44933", "accept": false, - "reason": "Package/backend dependency issue is unrelated to a missing image_utils import." + "reason": "Tensorflow backend packaging and a nonexistent import from image_utils are separate import/package issues." }, { "left": "issue:42898", "right": "issue:44568", "accept": false, - "reason": "Both involve tokenizer behavior in v5, but one is a global cleanup_spaces change and the other is missing BOS/EOS for a specific tokenizer." + "reason": "clean_up_tokenization_spaces behavior changes and add_special_tokens BOS/EOS regression are different tokenizer semantics." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Different subsystems: multimodal config defaults vs Whisper timestamp decoding crash." + "reason": "Default vision/audio config initialization and Whisper timestamp decode crash are unrelated multimodal vs speech bugs." }, { "left": "issue:30990", "right": "issue:35141", "accept": false, - "reason": "Sentence Transformers loading hang is unrelated to token-embedding resizing/reinitialization." + "reason": "Sentence Transformers loading hang and token-embedding reinitialization after resize are unrelated issues." }, { "left": "issue:43976", "right": "issue:44933", "accept": false, - "reason": "Python-version compatibility bug is unrelated to an import from image_utils." + "reason": "Python version compatibility for transformers 5.1.0 and a missing image_utils import are not the same defect." }, { "left": "issue:43479", "right": "issue:44079", "accept": false, - "reason": "Config initialization bug and ModelOutput key assignment bug are separate code paths." + "reason": "Incorrect multimodal config defaults and ModelOutput key assignment when previously None are different object-initialization bugs." }, { "left": "issue:43784", "right": "issue:43824", "accept": false, - "reason": "Different import failures in different packages/symbols; not the same underlying bug." + "reason": "Sentence-transformers import NameError and Qwen2.5-VL import failure are both import-related but affect different modules and causes." }, { "left": "issue:44188", "right": "issue:44373", "accept": false, - "reason": "Runtime attention-kernel divergence is unrelated to a docstring typo." + "reason": "Attention-kernel divergence under compile and a wrong docstring for position_ids are unrelated; one is a runtime bug, the other documentation." }, { "left": "issue:44623", "right": "issue:45310", "accept": false, - "reason": "Processor save_pretrained missing files is a different save-path bug than Qwen3.5 MoE loading." + "reason": "Missing processor.save_pretrained files and Qwen3.5 Moe from_pretrained regression are separate save/load problems." }, { - "left": "issue:42898", - "right": "issue:45812", + "left": "issue:43278", + "right": "issue:45137", "accept": false, - "reason": "Both are tokenizer regressions, but the behaviors and affected models differ materially." + "reason": "Embedding dtype drift during evaluate and a ZeRO3 deque pop error are unrelated training/runtime failures." }, { - "left": "issue:43278", - "right": "issue:45137", + "left": "issue:42898", + "right": "issue:45812", "accept": false, - "reason": "Embedding dtype drift during evaluate is unrelated to the ZeRO3 deque failure." + "reason": "Tokenizer cleanup behavior change and Granite token ID regression are both tokenizer issues, but they concern different APIs and distinct regressions." }, { "left": "issue:44734", "right": "issue:45103", "accept": false, - "reason": "Serving KV-cache continuation crash is unrelated to auto-docstring annotation handling." + "reason": "KV-cache continuation crash in transformers serve and auto_docstring annotation handling are unrelated." }, { "left": "issue:30333", "right": "issue:33290", "accept": false, - "reason": "MLFlow failure reporting and DeepSpeed OOM are completely different failure modes." + "reason": "MLflow reporting after failed training and OOM with Adafactor in DeepSpeed are different training/runtime problems." }, { "left": "issue:43723", "right": "issue:44568", "accept": false, - "reason": "Tokenizer loading failure and add_special_tokens BOS/EOS regression are not the same bug." + "reason": "Tokenizer loading in v5 and add_special_tokens BOS/EOS regression are both tokenization-related, but not the same failure." }, { "left": "issue:44060", "right": "issue:45125", "accept": false, - "reason": "Tied-weights warning bug and missing tensor-parallel plan are distinct Qwen3-related issues." + "reason": "Both concern Qwen MoE tensor-related behavior, but one is tied-weights warning logic and the other is missing _tp_plan for tensor parallelism." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Tokenizer cleanup semantics and Kimi-K2.5 codec/regex regression are different tokenizer regressions." } ] }, @@ -11577,6 +11618,7 @@ "issue:44297", "issue:44360", "issue:44373", + "issue:44442", "issue:44448", "issue:44483", "issue:44484", @@ -11591,7 +11633,6 @@ "issue:45092", "issue:45103", "issue:45310", - "issue:45356", "issue:45357", "issue:45405", "issue:45406", @@ -11601,7 +11642,6 @@ ], "soft_pairs": [ "issue:43874|issue:45092", - "issue:42898|issue:45356", "issue:42898|issue:44779", "issue:45405|issue:45712", "issue:42898|issue:44448", @@ -11617,8 +11657,8 @@ "issue:41628|issue:42757", "issue:22355|issue:28282", "issue:30333|issue:36331", - "issue:43232|issue:45834", "issue:30333|issue:34689", + "issue:43232|issue:45834", "issue:44222|issue:44484", "issue:44297|issue:45310", "issue:43723|issue:44977", @@ -11628,252 +11668,253 @@ "issue:44483|issue:45406", "issue:44263|issue:44360", "issue:44373|issue:44485", - "issue:44514|issue:45103", "issue:44295|issue:44623", + "issue:44514|issue:45103", "issue:44261|issue:45092", "issue:45092|issue:45357", "issue:43723|issue:44871", "issue:30990|issue:38175", - "issue:43723|issue:44991" + "issue:43723|issue:44991", + "issue:43723|issue:44442" ], - "prepared_review_unit_hash": "1b369334cc9d1ba1ef4454a6330ae1409d19de7fd9f1c6e0c262ac5c6bd5ae3f", + "prepared_review_unit_hash": "f876f7c2ff604678575108c31162c0dc0bb6a130babbf60e67c7e0a93844f81c", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 38059, - "estimated_input_tokens": 9515, - "estimated_eval_tokens": 19286 + "serialized_chars": 38005, + "estimated_input_tokens": 9502, + "estimated_eval_tokens": 19260 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is heterogeneous: most items are superficially similar v5/tokenizer/import/save_pretrained regressions, but they do not appear to be the same concrete bug. I would not merge any of the soft edges, and there are no PRs here to compare.", - "confidence": 0.84, - "canonical_issue_reason": "No single canonical issue fits the cluster well; the items span unrelated problem classes (imports, training, model loading, tokenizer behavior, processor save/load, and model-specific regressions).", + "summary": "The set is mostly noisy and spans several unrelated regressions. The only noticeable subcluster is around v5 tokenizer loading/import behavior, but none of the soft pairs are close enough to treat as the same underlying bug/change.", + "confidence": 0.81, + "canonical_issue_reason": "Best representative of the main subcluster: a broad v5 tokenizer-loading regression centered on AutoTokenizer.from_pretrained, which is the closest hub for the related tokenizer issues here.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43723 is the closest representative of the recurring tokenizer-loading regression subtheme, but it still only covers one slice of the cluster.", + "best_issue_reason": "Most central and reusable issue for triage; it captures the broad tokenizer-loading failure mode without overfitting to one model or tokenizer implementation.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43874", "right": "issue:45092", "accept": false, - "reason": "Different failures: missing image-patch method vs. old InternVL2 meta-init incompatibility." - }, - { - "left": "issue:42898", - "right": "issue:45356", - "accept": false, - "reason": "Both involve tokenizers, but one is a clean_up_tokenization_spaces change and the other is a codec/warning regression for Kimi-K2.5." + "reason": "Different failures: one is a missing image-processor method, the other is a remote-code/meta-init incompatibility." }, { "left": "issue:42898", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and symptoms; not the same bug." + "reason": "Both are tokenizer-related, but one is about cleanup-space behavior and the other is a model-specific wrong-output regression." }, { "left": "issue:45405", "right": "issue:45712", "accept": false, - "reason": "Unrelated issues: PEFT version bump vs leftover dummy-class cleanup." + "reason": "Unrelated: a version pin problem versus dummy class leakage in the import surface." }, { "left": "issue:42898", "right": "issue:44448", "accept": false, - "reason": "Different concrete regressions: tokenizer cleanup behavior vs Pegasus output drift." + "reason": "Both mention v4/v5 output differences, but the concrete bugs are different and not clearly the same code path." }, { "left": "issue:44295", "right": "issue:45656", "accept": false, - "reason": "Position_ids buffer access error and deepspeed double optimizer-step are unrelated." + "reason": "Different subsystems: buffer access versus DeepSpeed optimizer stepping." }, { "left": "issue:44933", "right": "issue:45712", "accept": false, - "reason": "Both touch imports/repo hygiene, but the actual bugs are different modules and failures." + "reason": "Both touch imports, but one is a bad image_utils import and the other is dummy PT placeholders leaking into dir()." }, { "left": "issue:44222", "right": "issue:44297", "accept": false, - "reason": "Both are save_pretrained-related, but one is FP8 MoE saving and the other is tokenizer metadata mismatch." + "reason": "Both involve save_pretrained, but one is FP8 MoE weights and the other is tokenizer config metadata." }, { "left": "issue:43116", "right": "issue:44190", "accept": false, - "reason": "Different example-script bugs: empty multilabel results vs local dataset loading." + "reason": "Different example scripts and different failures: classification output handling versus local dataset loading." }, { "left": "issue:44297", "right": "issue:45656", "accept": false, - "reason": "Tokenizer save metadata bug vs deepspeed optimizer stepping bug." + "reason": "Tokenizer save metadata bug versus DeepSpeed double optimizer step; no shared concrete bug." }, { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Both are init/state issues, but they describe different mechanisms and code paths." + "reason": "Both are initialization-related, but they affect different components and different failure modes." }, { "left": "issue:44222", "right": "issue:44623", "accept": false, - "reason": "Both mention save_pretrained, but they affect different objects and different missing outputs." + "reason": "Both are save_pretrained-related, but one is model FP8 MoE serialization and the other is missing processor files." }, { "left": "issue:44206", "right": "issue:45103", "accept": false, - "reason": "Feature-extractor arg crash and auto_docstring annotation crash are unrelated." + "reason": "Different code paths: a feature extractor argument regression versus auto-docstring annotation handling." }, { "left": "issue:42831", "right": "issue:44623", "accept": false, - "reason": "FP8 accuracy concern and processor save_pretrained missing files are not the same bug." + "reason": "Accuracy/quantization issue versus missing processor save files; unrelated." }, { "left": "issue:41628", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but for different symbols and different root causes." + "reason": "Both are import errors, but from different packages and different causes." }, { "left": "issue:22355", "right": "issue:28282", "accept": false, - "reason": "Different import/dependency failures: missing transformers.onnx vs missing PyTorch." + "reason": "Both are import failures, but one is a missing submodule and the other is an optional backend dependency check." }, { "left": "issue:30333", "right": "issue:36331", "accept": false, - "reason": "MLFlow reporting on failed training and CustomTrainer compute_loss signature mismatch are unrelated." + "reason": "MLflow reporting and trainer loss signature mismatch are unrelated." }, { - "left": "issue:43232", - "right": "issue:45834", + "left": "issue:30333", + "right": "issue:34689", "accept": false, - "reason": "Generation sync state bug vs Kosmos2.5 OCR index error are unrelated." + "reason": "Different problems: training-job reporting versus model loading regression." }, { - "left": "issue:30333", - "right": "issue:34689", + "left": "issue:43232", + "right": "issue:45834", "accept": false, - "reason": "Training failure reporting and Llama 3.2 model-loading regression are different bugs." + "reason": "Generation kwargs handling is unrelated to OCR index errors." }, { "left": "issue:44222", "right": "issue:44484", "accept": false, - "reason": "FP8 MoE save_pretrained bug is unrelated to the max_shard_size question." + "reason": "Different save_pretrained topics: MoE FP8 serialization versus shard-size default behavior." }, { "left": "issue:44297", "right": "issue:45310", "accept": false, - "reason": "Tokenizer-class metadata bug and Qwen3.5 MoE from_pretrained failure are different issues." + "reason": "Tokenizer config metadata mismatch is not the same bug as Qwen3.5 MoE from_pretrained failure." }, { "left": "issue:43723", "right": "issue:44977", "accept": false, - "reason": "Tokenizer loading regression and flash-attention generation bug are not the same concrete problem." + "reason": "Both are tokenizer/generation regressions in v5, but they are model-specific and do not look like the same bug." }, { "left": "issue:28282", "right": "issue:30333", "accept": false, - "reason": "Missing PyTorch import error and MLFlow failure reporting are unrelated." + "reason": "Completely unrelated import error and MLflow reporting issue." }, { "left": "issue:30333", "right": "issue:35141", "accept": false, - "reason": "Training failure reporting and embedding resize/reinit bug are different code paths." + "reason": "Training-job status reporting versus embedding resize/post_init behavior." }, { "left": "issue:44263", "right": "issue:45092", "accept": false, - "reason": "DSA indexer split behavior and InternVL2 meta-init incompatibility are unrelated." + "reason": "Unrelated GLM indexer bug versus InternVL2 meta-init incompatibility." }, { "left": "issue:44483", "right": "issue:45406", "accept": false, - "reason": "Serve API request rejection and Gemma4Processor missing _tokenizer are different serving bugs." + "reason": "Both involve serving, but one is request handling and the other is a processor attribute crash." }, { "left": "issue:44263", "right": "issue:44360", "accept": false, - "reason": "Same subsystem name, but one is a torch.split return-value issue and the other is a missing ReLU; not the same bug." + "reason": "Same module family, but different concrete bugs in the DSA indexer." }, { "left": "issue:44373", "right": "issue:44485", "accept": false, - "reason": "Docstring mistake vs RoPE implementation bug." + "reason": "A docstring issue is not the same as a RoPE implementation bug." }, { - "left": "issue:44514", - "right": "issue:45103", + "left": "issue:44295", + "right": "issue:44623", "accept": false, - "reason": "Qwen2.5-VL chat-template crash and auto_docstring annotation crash are unrelated." + "reason": "Buffer registration/read bug versus missing processor save files." }, { - "left": "issue:44295", - "right": "issue:44623", + "left": "issue:44514", + "right": "issue:45103", "accept": false, - "reason": "Position_ids buffer access issue and processor.save_pretrained missing files are unrelated." + "reason": "Batched chat-template crash and auto-docstring annotation crash are unrelated." }, { "left": "issue:44261", "right": "issue:45092", "accept": false, - "reason": "Different model/config issues: MLA rms_norm_eps precision vs InternVL2 meta-init incompatibility." + "reason": "MLA layernorm epsilon precision issue versus remote-code/meta-init incompatibility." }, { "left": "issue:45092", "right": "issue:45357", "accept": false, - "reason": "InternVL2 meta-init issue and Qwen3.5 visual key save_pretrained regression are unrelated." + "reason": "Both mention Qwen/vision-related v5 regressions, but the failure modes are different." }, { "left": "issue:43723", "right": "issue:44871", "accept": false, - "reason": "Tokenizer loading regression and Gemma-3 eos_token_id mismatch are different bugs." + "reason": "Tokenizer loading regression versus tokenizer/model eos-token-id mismatch; not the same bug." }, { "left": "issue:30990", "right": "issue:38175", "accept": false, - "reason": "Sentence Transformers loading stall and SigLIP2 zero-probabilities issue are unrelated." + "reason": "Different models and different symptoms: loading stall versus zero probabilities." }, { "left": "issue:43723", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading complaints, but they involve different models and no clear shared code-path bug." + "reason": "Both are tokenizer-loading complaints, but they are specific to different tokenizers and lack a shared concrete failure path." + }, + { + "left": "issue:43723", + "right": "issue:44442", + "accept": false, + "reason": "Tokenizer-loading failure versus AutoTokenizer failing on a specific tokenizer class; related theme, not the same underlying bug." } ] }, @@ -11916,7 +11957,6 @@ "issue:43475", "issue:43479", "issue:43576", - "issue:43723", "issue:43756", "issue:43784", "issue:43824", @@ -11929,8 +11969,9 @@ "issue:44263", "issue:44295", "issue:44297", + "issue:44360", "issue:44373", - "issue:44442", + "issue:44485", "issue:44488", "issue:44514", "issue:44521", @@ -11947,7 +11988,6 @@ "issue:45753" ], "soft_pairs": [ - "issue:43723|issue:44442", "issue:41628|issue:43824", "issue:43479|issue:44913", "issue:42831|issue:45310", @@ -11955,9 +11995,9 @@ "issue:44263|issue:44373", "issue:42757|issue:45278", "issue:42831|issue:44038", - "issue:43232|issue:45593", "issue:43756|issue:44060", "issue:42886|issue:42994", + "issue:43232|issue:45593", "issue:43576|issue:43784", "issue:42831|issue:45753", "issue:42886|issue:43066", @@ -11975,224 +12015,225 @@ "issue:43976|issue:44945", "issue:43784|issue:45600", "issue:39401|issue:44488", - "issue:42907|issue:43937", "issue:43408|issue:43475", - "issue:44261|issue:44373" + "issue:42907|issue:43937", + "issue:44261|issue:44373", + "issue:44360|issue:44485" ], - "prepared_review_unit_hash": "aa9f9638b0ce2d899fd4380af975ca0060f5ce94f315b1a08abf6578dff54370", + "prepared_review_unit_hash": "918337310c9add629896f2dc674f7768575998cd2fce99491e61ac6ead5d6fff", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36512, - "estimated_input_tokens": 9128, - "estimated_eval_tokens": 18512 + "serialized_chars": 36472, + "estimated_input_tokens": 9118, + "estimated_eval_tokens": 18492 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are a heterogeneous set of Transformers issues: tokenizer regressions, import failures, model-specific bugs, config problems, and a few SAM/Qwen/GLM-specific failures. There is no strong duplicate cluster here; the closest thing to a representative umbrella is the broad import-regression report.", - "confidence": 0.91, - "canonical_issue_reason": "issue:45278 is the broadest and most umbrella-like issue in the set: \"Many import errors after update from 4.57.0 to 5.5.0.\" It best represents the general regression theme, even though it is not a duplicate of the others.", + "summary": "This cluster is heterogeneous: it mixes import/API regressions, tokenizer behavior bugs, multimodal/model-specific issues, and unrelated configuration/documentation problems. None of the soft pairs look like true duplicates or the same underlying fix. There are no PRs in the cluster.", + "confidence": 0.89, + "canonical_issue_reason": "issue:45278 is the broadest and most representative issue here: it captures the general upgrade/import-regression theme that overlaps with several other reports in the set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most generally actionable and central report in this set, with a broad regression scope and open status. It is the best single issue to anchor triage, though not a true duplicate target for the rest.", + "best_issue_reason": "issue:45278 is the best standalone issue to anchor the cluster because it is the most general, clearly user-facing, and could plausibly subsume multiple import-related reports without overfitting to a narrow model-specific bug.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43723", - "right": "issue:44442", - "accept": false, - "reason": "Both are tokenizer-loading problems, but one is a generic AutoTokenizer v5 load issue and the other is a specific FastSpeech2ConformerTokenizer failure; different code paths and symptoms." - }, { "left": "issue:41628", "right": "issue:43824", "accept": false, - "reason": "Both are import errors, but they involve different missing symbols and different modules/models; not the same underlying bug." + "reason": "Different missing symbols and different model areas: AutoImageProcessor import vs Qwen2_5_VLForConditionalGeneration import." }, { "left": "issue:43479", "right": "issue:44913", "accept": false, - "reason": "Both concern config handling, but one is default multimodal config initialization and the other is GPTNeoX rotary_pct persistence on reload; distinct failures." + "reason": "Unrelated config bugs: multimodal default config initialization vs GPTNeoX rotary_pct resetting on reload." }, { "left": "issue:42831", "right": "issue:45310", "accept": false, - "reason": "Completely different problems: FineGrainedFP8 accuracy vs a Qwen3.5 MoE from_pretrained error." + "reason": "Different failures: FineGrainedFP8 accuracy degradation vs Qwen3.5 MoE from_pretrained error." }, { "left": "issue:39401", "right": "issue:44568", "accept": false, - "reason": "Both are tokenizer regressions, but one is wrong offset mapping and the other is add_special_tokens not adding BOS/EOS; not the same bug." + "reason": "Tokenizer offset_mapping bug is not the same as add_special_tokens failing to add BOS/EOS tokens." }, { "left": "issue:44263", "right": "issue:44373", "accept": false, - "reason": "Unrelated: a torch.split return-value issue in GLM MoE code vs a wrong docstring for position_ids." + "reason": "DSA indexer torch.split behavior is unrelated to a wrong docstring for position_ids." }, { "left": "issue:42757", "right": "issue:45278", "accept": false, - "reason": "Both are import-related regressions, but one is a specific missing huggingface_hub symbol and the other is a broad import-error report; too different to merge." + "reason": "One is a specific missing is_offline_mode import; the other is a broad collection of import errors after upgrade." }, { "left": "issue:42831", "right": "issue:44038", "accept": false, - "reason": "Unrelated model-specific issues: FineGrainedFP8 accuracy vs Qwen3-VL-Moe behavior in Transformers 5.0." - }, - { - "left": "issue:43232", - "right": "issue:45593", - "accept": false, - "reason": "Generation/sync_gpus bookkeeping is unrelated to D-FINE auxiliary-loss behavior." + "reason": "Different model/tooling problems: FP8 accuracy issue vs Qwen3-VL-Moe bug." }, { "left": "issue:43756", "right": "issue:44060", "accept": false, - "reason": "Different SAM/Qwen model bugs: RoPE layer dropping vs tied-weights warning across layers." + "reason": "Smollm3 RoPE-layer count discrepancy is unrelated to Qwen3-Next tied-weights warning." }, { "left": "issue:42886", "right": "issue:42994", "accept": false, - "reason": "One is tokenizer cache loading offline; the other is quantized model saving failure. Different subsystems and fixes." + "reason": "Tokenizer offline cache loading and quantized model saving are different code paths and symptoms." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation kwargs update after sync_gpus is unrelated to D-FINE auxiliary-loss behavior." }, { "left": "issue:43576", "right": "issue:43784", "accept": false, - "reason": "A broken CLI env command and a sentence-transformers import NameError are unrelated." + "reason": "Broken transformers env command and sentence-transformers import NameError are distinct issues." }, { "left": "issue:42831", "right": "issue:45753", "accept": false, - "reason": "Accuracy regression in FineGrainedFP8 is unrelated to a Qwen3_5 infinite-loop image bug." + "reason": "FineGrainedFP8 accuracy issue is unrelated to Qwen3_5 infinite-loop behavior on a specific image." }, { "left": "issue:42886", "right": "issue:43066", "accept": false, - "reason": "Both involve tokenizers, but one is offline cache loading and the other is an incorrect tokenizer decoder type in v5; different concrete bugs." + "reason": "Offline tokenizer cache loading and wrong tokenizer decoder type are separate tokenizer problems." }, { "left": "issue:22355", "right": "issue:28218", "accept": false, - "reason": "Unrelated: missing transformers.onnx module vs an extra-space tokenizer formatting issue." + "reason": "Missing transformers.onnx module and an extra space after added token are unrelated bugs." }, { "left": "issue:43122", "right": "issue:44521", "accept": false, - "reason": "Both are tokenization-related regressions, but one is a version-to-version tokenization difference and the other is zero assistant masks in multimodal apply_chat_template; not the same bug." + "reason": "Version-to-version tokenization differences do not match all-zero assistant masks for multimodal inputs." }, { "left": "issue:42371", "right": "issue:44263", "accept": false, - "reason": "TF32 settings guidance and a torch.split issue are unrelated." + "reason": "TF32 settings warning is unrelated to the DSA indexer torch.split issue." }, { "left": "issue:44117", "right": "issue:44704", "accept": false, - "reason": "Both touch from_pretrained plumbing, but one is tokenizer mapping fallback logic and the other is AutoProcessor kwargs propagation; not the same fix." + "reason": "TOKENIZER_MAPPING_NAMES returning None and AutoProcessor kwargs propagation are different failure points." }, { "left": "issue:43232", "right": "issue:44743", "accept": false, - "reason": "Both mention generation/cache, but the concrete bug differs: sync_gpus update bookkeeping vs recurrent states resetting in modular_qwen3_5.py." + "reason": "Generation sync_gpus behavior and recurrent-state reset in modular_qwen3_5 are not the same bug." }, { "left": "issue:30990", "right": "issue:33357", "accept": false, - "reason": "Both are loading-related, but one is Sentence Transformers stuck loading and the other is a MacOS bus error on a CLIP model; not clearly the same issue." + "reason": "Sentence Transformers loading hang and MacOS bus error with CLIP are different runtime failures." }, { "left": "issue:42907", "right": "issue:44945", "accept": false, - "reason": "Saving dequantized models and incorrect output under pipeline parallelism are different problems." + "reason": "Dequantized model saving failure and incorrect pipeline-parallel LLM output are unrelated." }, { "left": "issue:44295", "right": "issue:44297", "accept": false, - "reason": "A position_ids buffer read error and tokenizer.save_pretrained class-name mismatch are unrelated." + "reason": "Position_ids buffer read error is unrelated to tokenizer_class mismatch in tokenizer_config.json." }, { "left": "issue:44704", "right": "issue:45290", "accept": false, - "reason": "AutoProcessor.cached_file kwargs propagation is unrelated to apply_chat_template crashing on assistant tool-call messages." + "reason": "AutoProcessor kwargs passing and chat-template crashes with tool calls are different issues." }, { "left": "issue:43122", "right": "issue:44514", "accept": false, - "reason": "Both are tokenizer/template regressions, but one is a general tokenization change across versions and the other is a batched Qwen2_5_VLProcessor padding crash; different code paths." + "reason": "General tokenization differences across versions are not the same as batched apply_chat_template crashes." }, { "left": "issue:41628", "right": "issue:45278", "accept": false, - "reason": "Both are import failures, but one is a specific AutoImageProcessor import error and the other is a broad many-imports regression; not the same underlying bug." + "reason": "Specific AutoImageProcessor import failure is only loosely related to a broad import-error report." }, { "left": "issue:43976", "right": "issue:44945", "accept": false, - "reason": "Python version compatibility failure and pipeline-parallel output corruption are unrelated." + "reason": "Python version compatibility regression and pipeline-parallel output corruption are different problems." }, { "left": "issue:43784", "right": "issue:45600", "accept": false, - "reason": "A NameError during import and a broken auto_mappings reference to removed Sam3LiteText configs are different failures." + "reason": "Import NameError in sentence-transformers is unrelated to stale auto_mappings references to removed configs." }, { "left": "issue:39401", "right": "issue:44488", "accept": false, - "reason": "Wrong offset_mapping for Qwen3 tokenizer and failure to load a specific tokenizer are not the same bug." + "reason": "Tokenizer offset_mapping bug is unrelated to a model loading failure for cjvt/sleng-bert." }, { - "left": "issue:42907", - "right": "issue:43937", + "left": "issue:43408", + "right": "issue:43475", "accept": false, - "reason": "Dequantized model saving failure and invalid GenerationConfig for GLM-5 are unrelated." + "reason": "Wrong model-type warning and missing fpn_position_embeddings attribute are distinct SAM 3 issues." }, { - "left": "issue:43408", - "right": "issue:43475", + "left": "issue:42907", + "right": "issue:43937", "accept": false, - "reason": "Same SAM3 family, but one is a model-type warning and the other is a missing attribute error; they are distinct concrete bugs." + "reason": "Quantized model saving failure and invalid GenerationConfig are different subsystems and symptoms." }, { "left": "issue:44261", "right": "issue:44373", "accept": false, - "reason": "MLA layernorm precision/config issue and a wrong docstring for position_ids are unrelated." + "reason": "Precision error from missing rms_norm_eps is not the same as a wrong docstring." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer lacking ReLU and GLM-5 RoPE implementation are unrelated model-specific bugs." } ] }, @@ -12247,7 +12288,6 @@ "issue:44295", "issue:44297", "issue:44351", - "issue:44360", "issue:44485", "issue:44521", "issue:44625", @@ -12266,7 +12306,6 @@ "issue:45753" ], "soft_pairs": [ - "issue:44360|issue:44485", "issue:44625|issue:44829", "issue:39401|issue:44521", "issue:43116|issue:45200", @@ -12284,9 +12323,9 @@ "issue:43278|issue:44945", "issue:43493|issue:45753", "issue:42907|issue:44222", - "issue:43066|issue:44779", "issue:43122|issue:44977", "issue:43992|issue:44704", + "issue:43066|issue:44779", "issue:43784|issue:45446", "issue:38617|issue:45278", "issue:42673|issue:45072", @@ -12298,220 +12337,214 @@ "issue:42222|issue:43441", "issue:43116|issue:43493" ], - "prepared_review_unit_hash": "6d4998cec299b6607d4bcffbb29e5aab09d8fb1584af767f733f3828a6f47957", + "prepared_review_unit_hash": "cb23cd07c2291dd5f8ec298a4b030edaca47acfaabb2c9dd8af8b0b903656410", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 31, - "serialized_chars": 36601, - "estimated_input_tokens": 9151, - "estimated_eval_tokens": 18558 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 35753, + "estimated_input_tokens": 8939, + "estimated_eval_tokens": 18134 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is not a tight duplicate cluster; it mixes many unrelated model-specific bugs, import errors, tokenizer regressions, and training issues. The closest umbrella issue is the broad import-regression report (#45278), but most others should remain separate.", - "confidence": 0.86, - "canonical_issue_reason": "#45278 is the broadest and most umbrella-like issue here: it explicitly covers many import errors after upgrading to 5.5.0, so it is the best candidate to represent the cluster.", + "summary": "These are mostly unrelated Transformers bug reports across different models, APIs, and failure modes. The only broadly representative issue is the open v4\u2192v5 import-regression report; none of the soft pairs look like true duplicates or mergeable together as one fix.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45278 is the broadest umbrella report here (multiple import errors after upgrading 4.57.0 to 5.5.0), so it best represents the small import-regression subset.", "canonical_pr_reason": null, - "best_issue_reason": "#45278 is the best overall issue candidate because it is open, general, and plausibly subsumes some of the import-related regressions in the set.", + "best_issue_reason": "issue:45278 is the best issue to anchor triage because it is open, cross-cutting, and more general than the model-specific or single-symbol bug reports.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44360", - "right": "issue:44485", - "accept": false, - "reason": "DSA/ReLU-indexer discussion and GLM-5 RoPE implementation are unrelated subsystems and failure modes." - }, { "left": "issue:44625", "right": "issue:44829", "accept": false, - "reason": "One is a Qwen3.5 config propagation bug; the other is a flash-attention training degeneration issue. Different code paths." + "reason": "Different bugs: config propagation for Qwen3.5 vs degenerate training with flash_attention_3." }, { "left": "issue:39401", "right": "issue:44521", "accept": false, - "reason": "Wrong offset mappings and empty assistant masks are both tokenization-related only at a very high level, but they are distinct bugs with different symptoms." + "reason": "Both involve tokenization-adjacent behavior, but one is offset_mapping and the other is assistant_mask generation for multimodal chat templates." }, { "left": "issue:43116", "right": "issue:45200", "accept": false, - "reason": "Empty multi-label classification results and missing mm_token_type_ids defaults are different layers of the stack." + "reason": "Example-script multi-label output bug is unrelated to Gemma 4 mm_token_type_ids defaulting." }, { "left": "issue:29127", "right": "issue:30990", "accept": false, - "reason": "Error-message clarity for LayoutLMv3 boxes and Sentence Transformers loading hang are unrelated." + "reason": "LayoutLMv3 error-message clarity is unrelated to Sentence Transformers loading stalls." }, { "left": "issue:42907", "right": "issue:44964", "accept": false, - "reason": "Saving dequantized Ministral/Devstral models and loading Phi-4 multimodal are separate model-specific regressions." + "reason": "Saving dequantized Ministral/Devstral models is a different code path from loading Phi-4 multimodal instruct." }, { "left": "issue:44038", "right": "issue:44297", "accept": false, - "reason": "Qwen3-VL-Moe loading and tokenizer save_pretrained class mismatch do not look like the same underlying bug." + "reason": "Qwen3-VL-Moe failure is unrelated to tokenizer_config tokenizer_class mismatch during save_pretrained." }, { "left": "issue:43122", "right": "issue:44779", "accept": false, - "reason": "Both mention tokenizer behavior changes in v5, but they affect different models and likely different code paths." + "reason": "Both are tokenization regressions, but they affect different model families and concrete tokenizer behavior." }, { "left": "issue:44038", "right": "issue:44295", "accept": false, - "reason": "Qwen3-VL-Moe loading issues and position_ids buffer access are unrelated." + "reason": "Different failures: Qwen3-VL-Moe vs a position_ids buffer access error." }, { "left": "issue:36296", "right": "issue:43116", "accept": false, - "reason": "Tensor-parallel optimizer/training corruption is unrelated to the multi-label classification example script bug." + "reason": "Tensor-parallel optimizer-state bug is unrelated to the classification example returning empty results." }, { "left": "issue:42907", "right": "issue:43452", "accept": false, - "reason": "Saving dequantized models and gguf_file breaking from_pretrained are different I/O paths and symptoms." + "reason": "Saving dequantized models and gguf_file loading failures are separate issues." }, { "left": "issue:43824", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but they concern different symbols and packages, so they are not the same bug." + "reason": "Both are import errors, but they concern different symbols and different modules." }, { "left": "issue:44038", "right": "issue:44222", "accept": false, - "reason": "Qwen3-VL-Moe loading and FP8 save_pretrained for moe are separate issues." + "reason": "Different MoE/model bugs: Qwen3-VL-Moe vs FP8 save_pretrained for moe models." }, { "left": "issue:45216", "right": "issue:45478", "accept": false, - "reason": "Both involve Qwen3.5 and newer versions, but one is a bad saved checkpoint while the other is a from_pretrained load error; not enough to call them the same bug." + "reason": "Related to Qwen3.5 and version 5.4+, but one is save_pretrained checkpoint correctness and the other is from_pretrained loading; not the same concrete bug." }, { "left": "issue:42757", "right": "issue:44351", "accept": false, - "reason": "Missing is_offline_mode from huggingface_hub and missing HybridCache from transformers are unrelated import failures." + "reason": "Separate missing imports from different packages and symbols." }, { "left": "issue:43278", "right": "issue:44945", "accept": false, - "reason": "Embedding dtype drift and pipeline-parallel output corruption are distinct training/inference bugs." + "reason": "Embedding dtype drift in evaluate is unrelated to incorrect outputs under pipeline parallelism." }, { "left": "issue:43493", "right": "issue:45753", "accept": false, - "reason": "SigLIP2 implementation discrepancy and Qwen3_5 image-specific infinite loop are unrelated model behaviors." + "reason": "SigLIP2 implementation discrepancy and a Qwen3_5 image-specific infinite loop are unrelated." }, { "left": "issue:42907", "right": "issue:44222", "accept": false, - "reason": "Both concern model saving with quantization, but they affect different models and formats, so they are not clearly the same change." - }, - { - "left": "issue:43066", - "right": "issue:44779", - "accept": false, - "reason": "Tokenizer decoder-type mismatch and Deepseek tokenizer regression are related only broadly; the model and concrete failure differ." + "reason": "Both involve saving, but they are different models and different quantization formats." }, { "left": "issue:43122", "right": "issue:44977", "accept": false, - "reason": "Tokenizer regression versus flash-attention generation bug are different problems." + "reason": "Tokenizer regression vs flash-attention generation failure are different bugs." }, { "left": "issue:43992", "right": "issue:44704", "accept": false, - "reason": "Missing embed_tokens.weight in UMT5Encoder and AutoProcessor kwargs forwarding are unrelated load-path bugs." + "reason": "Both use from_pretrained, but one misses a weight load and the other drops kwargs when resolving files." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Tokenizer decoder type bug is not the same as Deepseek tokenization changes in v5." }, { "left": "issue:43784", "right": "issue:45446", "accept": false, - "reason": "Sentence-transformers import NameError and a PyTorch version check bug in flex_attention are different import-time issues." + "reason": "Sentence-transformers import NameError and flex_attention version check are unrelated." }, { "left": "issue:38617", "right": "issue:45278", "accept": false, - "reason": "The latter is a broad umbrella for many import errors, but this specific missing symbol is not enough to prove the same underlying regression." + "reason": "A specific missing import is not the same underlying bug as a broad report of multiple import errors." }, { "left": "issue:42673", "right": "issue:45072", "accept": false, - "reason": "VRAM leak in Qwen3ForCausalLM and bf16 dtype mismatches in CI are unrelated." + "reason": "VRAM leakage in multi-threaded loading is unrelated to bfloat16 dtype mismatches." }, { "left": "issue:43937", "right": "issue:44485", "accept": false, - "reason": "GenerationConfig validation and GLM-5 RoPE implementation are different components." + "reason": "GenerationConfig validation failure and GLM-5 RoPE implementation are distinct issues." }, { "left": "issue:22355", "right": "issue:29127", "accept": false, - "reason": "A missing transformers.onnx module and a LayoutLMv3 error-message complaint are unrelated." + "reason": "onnx module import failure is unrelated to LayoutLMv3 error-message clarity." }, { "left": "issue:42371", "right": "issue:43874", "accept": false, - "reason": "TF32 API guidance and a missing image-patch helper in GLM46V are unrelated." + "reason": "TF32 API guidance and missing get_number_of_image_patches are unrelated." }, { "left": "issue:43976", "right": "issue:45446", "accept": false, - "reason": "Python-version compatibility and flex_attention import version-checking are different compatibility issues." + "reason": "Python version support regression is unrelated to an AuxRequest flex_attention version-check bug." }, { "left": "issue:43064", "right": "issue:43673", "accept": false, - "reason": "FSDP2/PEFT optimizer state corruption and chunked_prefill cache missing are separate training vs generation bugs." + "reason": "Optimizer-state mismatch under FSDP2/PEFT is unrelated to missing GenerationMixin cache during chunked prefill." }, { "left": "issue:42222", "right": "issue:43441", "accept": false, - "reason": "VitPose breakage and Ministral-3 FlashAttention failure are unrelated model-specific regressions." + "reason": "Broken vitpose model code and Ministral-3 flash-attention failure are different model-specific regressions." }, { "left": "issue:43116", "right": "issue:43493", "accept": false, - "reason": "Example-script multi-label output emptiness and SigLIP2 implementation mismatch are not the same bug." + "reason": "Classification example output bug is unrelated to SigLIP2 implementation discrepancy." } ] }, @@ -12589,8 +12622,8 @@ "issue:43116|issue:43825", "issue:45357|issue:45478", "issue:44038|issue:45478", - "issue:43937|issue:45563", "issue:45083|issue:45446", + "issue:43937|issue:45563", "issue:33453|issue:39401", "issue:44117|issue:45588", "issue:43827|issue:45446", @@ -12613,17 +12646,17 @@ "issue:42617|issue:45341", "issue:44933|issue:45278", "issue:42222|issue:45542", - "issue:43937|issue:44496", - "issue:42673|issue:43577" + "issue:42673|issue:43577", + "issue:43937|issue:44496" ], - "prepared_review_unit_hash": "f5a3c0a386beab5110455ba1a1f4c8148f221d93858e7c24901f6faea9930ba5", + "prepared_review_unit_hash": "b9014470ca55687855d80a8ed8a191530da0de530148f63721f6ac9b88e6582c", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, @@ -12638,192 +12671,192 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly false-positive similarity matches across unrelated bugs. The only clear near-duplicate in the set is the repeated Qwen3.5-Moe from_pretrained regression (45310/45478), which is not one of the soft-edge candidates.", - "confidence": 0.88, - "canonical_issue_reason": "Issue 45310 is the best canonical issue: it is a concrete, reproducible regression for Qwen3.5-Moe loading and is an exact duplicate of 45478.", + "summary": "The cluster is highly heterogeneous: most pairs are only superficially similar (often same subsystem, model family, or release, but different concrete failures). I would not merge any of the soft edges as duplicates. If forced to pick a single issue anchor, 45278 is the broadest import-regression umbrella, but it is still not a true match for most of the set.", + "confidence": 0.23, + "canonical_issue_reason": "45278 is the broadest umbrella report in the set ('many import errors after update'), so it is the closest catch-all representative for the import-regression subset.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45310 is the strongest representative for triage because it captures the exact failing code path and has a clear duplicate twin (45478), while most other issues here are only loosely similar or unrelated.", + "best_issue_reason": "45278 is the most general and discussion-rich issue here, making it the best single representative even though the overall cluster is mixed.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44075", "right": "issue:44936", "accept": false, - "reason": "Different bugs: optimizer argument handling vs trainer.evaluate failing after train; same training area but not the same code-path problem." + "reason": "Different bugs: optimizer argument handling vs Trainer.evaluate after train." }, { "left": "issue:43116", "right": "issue:43825", "accept": false, - "reason": "Both involve example/pipeline behavior, but one is multi-label classification output and the other is a v5 translation-task error message; not the same underlying bug." + "reason": "Different topics: example-script multi-label output vs v5 pipeline messaging/docs." }, { "left": "issue:45357", "right": "issue:45478", "accept": false, - "reason": "Same model family, but one is save_pretrained writing wrong visual encoder keys and the other is from_pretrained load failure; different concrete code paths." + "reason": "Different code paths: save_pretrained serialization vs from_pretrained loading failure." }, { "left": "issue:44038", "right": "issue:45478", "accept": false, - "reason": "Both mention Qwen/Moe v5 regressions, but the model variant and failure mode differ; too broad to treat as the same bug." + "reason": "Different model families and failure modes; not the same bug." }, { - "left": "issue:43937", - "right": "issue:45563", + "left": "issue:45083", + "right": "issue:45446", "accept": false, - "reason": "Unrelated generation issues: invalid GenerationConfig vs stale warning about num_return_sequences." + "reason": "Unrelated helper behavior in qwen3_omni_moe vs a flex_attention version-check import issue." }, { - "left": "issue:45083", - "right": "issue:45446", + "left": "issue:43937", + "right": "issue:45563", "accept": false, - "reason": "Different subsystems and symptoms: qwen3_omni_moe feature-length helper vs flex_attention import version check." + "reason": "GenerationConfig validation error vs stale warning text; different problems." }, { "left": "issue:33453", "right": "issue:39401", "accept": false, - "reason": "Both are tokenizer problems, but one is tokenizer loading regression and the other is wrong offset mapping in Qwen3; not the same bug." + "reason": "Both tokenizer-related, but different models and symptoms (loading regression vs offset_mapping bug)." }, { "left": "issue:44117", "right": "issue:45588", "accept": false, - "reason": "Tokenizer mapping None-handling vs flash_attention crash on sink-less models; unrelated failures." + "reason": "Tokenizer mapping fallback bug vs flash_attention crash; unrelated." }, { "left": "issue:43827", "right": "issue:45446", "accept": false, - "reason": "Docs referring to removed pipeline APIs is unrelated to a PyTorch version gate/import check bug." + "reason": "Docs still referencing pipeline() vs a runtime import/version check issue." }, { "left": "issue:39401", "right": "issue:44779", "accept": false, - "reason": "Both are model-specific tokenizer incorrectness after v5, but Qwen3 offset mapping and Deepseek tokenization are different code paths and symptoms." + "reason": "Both are tokenizer regressions, but they affect different models and likely different root causes." }, { "left": "issue:43937", "right": "issue:44977", "accept": false, - "reason": "Different generation failures: GLM-5 invalid config vs Qwen3.5 flash-attention generation issues." + "reason": "GenerationConfig invalid vs flash-attention generation behavior; not the same issue." }, { "left": "issue:42673", "right": "issue:43906", "accept": false, - "reason": "No evidence they share the same bug; one is VRAM leakage across dataloader threads and the other is just an isolated reproduction of another issue." + "reason": "VRAM leak in multi-threaded Qwen3ForCausalLM vs an isolated reproduction of a different issue." }, { "left": "issue:44291", "right": "issue:45071", "accept": false, - "reason": "Both are type/initialization regressions in v5, but one is init_empty_weights with an unexpected argument and the other is PretrainedConfig type checking; not the same concrete fix." + "reason": "Different failures: init_empty_weights TypeError vs PretrainedConfig type-checking regression." }, { "left": "issue:43450", "right": "issue:43986", "accept": false, - "reason": "Different areas: batched video processor output shape vs AutoProcessor crash without torchvision." + "reason": "Batched video processor shape bug vs AutoProcessor crash without torchvision." }, { "left": "issue:45405", "right": "issue:45446", "accept": false, - "reason": "PEFT minimum version bump is unrelated to a flex_attention import/version check bug." + "reason": "Dependency version pinning issue vs PyTorch version gating in flex_attention." }, { "left": "issue:45341", "right": "issue:45561", "accept": false, - "reason": "Both touch testing_utils, but one is a small bug and the other is a pytest-xdist race on captured_info.txt; not the same underlying defect." + "reason": "Both touch testing_utils, but one is a generic bug and the other is a specific xdist race." }, { "left": "issue:38617", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but for different missing symbols and different modules; too broad to merge." + "reason": "Different missing imports/symbols; same broad symptom but not the same underlying export bug." }, { "left": "issue:44117", "right": "issue:45081", "accept": false, - "reason": "Tokenizer mapping fallback issue vs Mistral regex patch crashing on backend_tokenizer; different bugs." + "reason": "Token mapping None vs Mistral regex patch crash; different tokenizer paths." }, { "left": "issue:39401", "right": "issue:43421", "accept": false, - "reason": "Tokenizer offset_mapping bug vs runtime post-processor update behavior in TokenizersBackend; not the same issue." + "reason": "Different tokenizer bugs: offset_mapping correctness vs runtime special-token post-processor sync." }, { "left": "issue:44410", "right": "issue:45310", "accept": false, - "reason": "Different Qwen3.5-family regressions: missing projections in qwen3next vs from_pretrained error for Qwen3.5 Moe." + "reason": "Missing projections in qwen3next vs Qwen3.5 from_pretrained error; different concrete failures." }, { "left": "issue:44484", "right": "issue:45310", "accept": false, - "reason": "save_pretrained shard-size question is unrelated to a Qwen3.5 Moe loading regression." + "reason": "save_pretrained shard-size question vs model loading error; unrelated." }, { "left": "issue:44829", "right": "issue:45216", "accept": false, - "reason": "Different model/task behaviors: flash_attention_3 degenerates training, while 45216 is a Qwen3.5 save_pretrained regression." + "reason": "Training degeneration under flash_attention_3 vs incorrect saved checkpoint contents." }, { "left": "issue:42673", "right": "issue:43582", "accept": false, - "reason": "VRAM leak in multithreaded dataloaders is unrelated to an Apple Silicon TypeError in caching_allocator_warmup." + "reason": "VRAM leak in dataloader threads vs Apple Silicon warmup TypeError; unrelated." }, { "left": "issue:42371", "right": "issue:44261", "accept": false, - "reason": "TF32 settings guidance is unrelated to MLA q_a_layernorm precision/config epsilon behavior." + "reason": "TF32 API guidance vs MLA precision/config epsilon bug." }, { "left": "issue:43408", "right": "issue:43742", "accept": false, - "reason": "Model-type warning for sam3_tracker vs key error loading MobileLLM are unrelated." + "reason": "Model-type warning for sam3 vs key error loading MobileLLM; different loading problems." }, { "left": "issue:42617", "right": "issue:45341", "accept": false, - "reason": "3d_parallel.py execution failure is unrelated to a testing_utils bug." + "reason": "3d_parallel.py execution failure vs a testing_utils bug; no shared underlying defect." }, { "left": "issue:44933", "right": "issue:45278", "accept": false, - "reason": "The latter is a broad umbrella about many import errors, while the former is one specific missing import; not enough to prove the same underlying bug." + "reason": "A specific missing import is not enough to equate it with a broad 'many import errors' report." }, { "left": "issue:42222", "right": "issue:45542", "accept": false, - "reason": "Vitpose breakage and a tensorboard-without-TensorFlow backend error are unrelated." + "reason": "Vitpose model breakage vs TensorFlow backend detection issue; unrelated." }, { - "left": "issue:43937", - "right": "issue:44496", + "left": "issue:42673", + "right": "issue:43577", "accept": false, - "reason": "GenerationConfig validation for GLM-5 is unrelated to an unrecognized model/config loading error." + "reason": "VRAM leak in Qwen3ForCausalLM threads vs dtype propagation in BLIP2; unrelated." }, { - "left": "issue:42673", - "right": "issue:43577", + "left": "issue:43937", + "right": "issue:44496", "accept": false, - "reason": "VRAM leak in dataloader threads vs dtype not propagating in Blip2 loading; different defects." + "reason": "Invalid GenerationConfig vs unrecognized model/config.json missing model_type; different load-time failures." } ] }, @@ -12917,176 +12950,170 @@ "issue:45071|issue:45542", "issue:43931|issue:44410", "issue:43335|issue:45071", - "issue:43931|issue:44038", "issue:45003|issue:45341", - "issue:43986|issue:44484", + "issue:43931|issue:44038", "issue:43986|issue:45341", - "issue:43408|issue:44368", "issue:43097|issue:43452", + "issue:43408|issue:44368", "issue:44117|issue:44987", + "issue:43986|issue:44484", "issue:43874|issue:44263", "issue:43976|issue:45071", "issue:43986|issue:45372", "issue:44829|issue:45245", "issue:43531|issue:45341", "issue:38617|issue:42757", - "issue:43064|issue:44315", - "issue:44261|issue:44485" + "issue:44261|issue:44485", + "issue:43064|issue:44315" ], - "prepared_review_unit_hash": "4d5ef40f0dde068923e7a30b2d6e182d45a53a7d609eea0e4b96f06127cea84f", + "prepared_review_unit_hash": "ce13f384107e6f20636fc52d687e60b09f94239aeb70bcb7fff005b4acf0dfa4", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 38032, - "estimated_input_tokens": 9508, - "estimated_eval_tokens": 19272 + "serialized_chars": 38035, + "estimated_input_tokens": 9509, + "estimated_eval_tokens": 19274 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issues. The closest soft match is issue:44483 \u2194 issue:45464, but the available metadata only supports a shared chat/completions failure area, not a confirmed same bug or model-specific duplicate.", - "confidence": 0.78, - "canonical_issue_reason": "issue:44483 is the best anchor for the chat/completions subgroup because it is earlier and broader, but the provided metadata does not establish it as the definitive master for issue:45464.", + "summary": "These items are mostly unrelated single-bug reports across packaging, tokenizers, model loaders, and optimizer behavior. I found no soft edge that clearly describes the same underlying bug/change with enough specificity to merge the pair.", + "confidence": 0.68, + "canonical_issue_reason": "issue:43122 is the broadest clear regression report in the set, centered on tokenizer behavior changing between 4.57.3 and 5.0, which is a plausible umbrella for nearby tokenizer-related complaints.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44483 is the strongest issue candidate to represent the chat/completions failure subgroup, though the cluster overall is too mixed to justify a strong global duplicate-tracker claim.", + "best_issue_reason": "issue:43122 is the most suitable duplicate target because it describes a user-visible semantic regression across versions, not just a narrow implementation detail.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42175", "right": "issue:42222", "accept": false, - "reason": "Different subsystems: backend/package installation vs ViTPose model breakage." + "reason": "Different problems: backend packaging for TensorFlow vs a broken VitPose model import path." }, { "left": "issue:44075", "right": "issue:45656", "accept": false, - "reason": "Both involve optimization, but one is ignored SGD args and the other is duplicate optimizer.step() calls under deepspeed." + "reason": "Both involve optimizers, but one is about SGD arguments being ignored while the other is about DeepSpeed causing double optimizer steps." }, { "left": "issue:44483", "right": "issue:45464", "accept": false, - "reason": "Shared chat/completions area, but the metadata does not prove the same model, request path, or exact failure mode." + "reason": "Same API area, but one is a general chat/completions failure and the other is a streaming-inference bug on a specific Qwen model." }, { "left": "issue:38617", "right": "issue:44220", "accept": false, - "reason": "Unrelated import error vs audio feature extraction bug." + "reason": "Unrelated code paths: a missing import from configuration_utils vs an issue in _torch_extract_fbank_features()." }, { "left": "issue:42371", "right": "issue:42831", "accept": false, - "reason": "TF32 warning/API guidance and FineGrainedFP8 accuracy regression are different issues." + "reason": "TF32 settings behavior and FineGrainedFP8 accuracy are distinct numerical/configuration issues." }, { "left": "issue:43066", "right": "issue:43122", "accept": false, - "reason": "Tokenizer decoder-type mismatch is not the same as changed tokenization output across versions." + "reason": "Both are tokenizer regressions, but one is about decoder type metadata and the other is a broader tokenization output change; not clearly the same bug." }, { "left": "issue:43012", "right": "issue:43577", "accept": false, - "reason": "Compile-time precision warning vs incorrect dtype loading for model components." + "reason": "Both mention dtype/precision, but they affect different models and different loading behaviors." }, { "left": "issue:43673", "right": "issue:45341", "accept": false, - "reason": "Generation cache handling and a testing_utils bug are unrelated." + "reason": "Generation cache behavior and a testing utility bug are unrelated." }, { "left": "issue:44484", "right": "issue:45478", "accept": false, - "reason": "Save-pretrained shard-size behavior is unrelated to Qwen3.5 MoE loading failure." + "reason": "Saving shard size defaults and a Qwen3.5 MoE from_pretrained error are separate concerns." }, { "left": "issue:43874", "right": "issue:44261", "accept": false, - "reason": "Different model internals: image-patch counting vs MLA norm-epsilon precision issue." + "reason": "Different model components and failures: image processor method missing vs MLA layernorm epsilon handling." }, { "left": "issue:44410", "right": "issue:45478", "accept": false, - "reason": "Qwen3next missing projections is not the same underlying failure as Qwen3.5 MoE loading." + "reason": "A missing projection layer in Qwen3Next is not the same as a Qwen3.5 MoE loading failure." }, { "left": "issue:43097", "right": "issue:44493", "accept": false, - "reason": "Removed config attribute and unexpected position-id keys are separate migration symptoms." + "reason": "One is about a removed config helper; the other is about unexpected position-id keys during loading." }, { "left": "issue:43012", "right": "issue:43408", "accept": false, - "reason": "Dtype/precision warning and model-type mismatch warning are distinct configuration problems." + "reason": "TF32/bfloat16 warnings and a SAM3 model-type mismatch warning are not the same issue." }, { "left": "issue:44263", "right": "issue:44485", "accept": false, - "reason": "torch.split return handling in GLM indexer is unrelated to GLM-5 RoPE implementation." + "reason": "Torch split return handling in one GLM MoE helper is distinct from a RoPE implementation bug." }, { "left": "issue:41628", "right": "issue:45071", "accept": false, - "reason": "Missing AutoImageProcessor import and PretrainedConfig type-checking failure are different API breakages." + "reason": "Importing AutoImageProcessor and PreTrainedConfig type checking are unrelated failures." }, { "left": "issue:42222", "right": "issue:45459", "accept": false, - "reason": "Broken ViTPose models and tokenizer error masking without protobuf are unrelated." + "reason": "A broken VitPose model and tokenizer error masking when protobuf is absent are different bugs." }, { "left": "issue:43441", "right": "issue:44829", "accept": false, - "reason": "FlashAttention-related issues, but different models and outcomes: inference failure vs degenerate training." + "reason": "Both mention FlashAttention, but one is a model-specific inference failure and the other is degenerate training with FlashAttention 3." }, { "left": "issue:45071", "right": "issue:45542", "accept": false, - "reason": "PretrainedConfig type checking and TensorFlow backend detection are separate issues." + "reason": "Type-checking regression in PreTrainedConfig is unrelated to a missing TensorFlow backend when only tensorboard is installed." }, { "left": "issue:43931", "right": "issue:44410", "accept": false, - "reason": "Different Qwen model families and different loading failures." + "reason": "Different model families and failure modes: Qwen3-VL weight mismatch vs Qwen3Next layer projection omission." }, { "left": "issue:43335", "right": "issue:45071", "accept": false, - "reason": "Sparse layer construction bug and type-checking breakage are unrelated." - }, - { - "left": "issue:43931", - "right": "issue:44038", - "accept": false, - "reason": "Different Qwen model families and different failure modes." + "reason": "SwitchTransformers sparse-layer construction and PreTrainedConfig type checking do not share the same code path." }, { "left": "issue:45003", @@ -13095,91 +13122,94 @@ "reason": "Unsafe sys.modules access in modeling_utils is unrelated to a testing_utils bug." }, { - "left": "issue:43986", - "right": "issue:44484", + "left": "issue:43931", + "right": "issue:44038", "accept": false, - "reason": "Missing torchvision crash and max_shard_size behavior are unrelated." + "reason": "Both are model-loading problems, but they involve different Qwen variants and different root causes." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "Video-model dependency handling and testing_utils are unrelated." + "reason": "AutoProcessor/video-model dependency handling is unrelated to testing_utils." }, { - "left": "issue:43408", - "right": "issue:44368", + "left": "issue:43097", + "right": "issue:43452", "accept": false, - "reason": "Different configuration warnings with different triggers." + "reason": "A removed config helper and a gguf_file loading regression are different loader issues." }, { - "left": "issue:43097", - "right": "issue:43452", + "left": "issue:43408", + "right": "issue:44368", "accept": false, - "reason": "Config attribute removal and gguf_file loading breakage are not the same bug." + "reason": "The SAM3 model-type warning is not the same as the tie_word_embeddings warning during Qwen3.5 LoRA fine-tuning." }, { "left": "issue:44117", "right": "issue:44987", "accept": false, - "reason": "Tokenizer mapping None handling and a model-loading failure are too vague to treat as the same code path." + "reason": "Tokenizer mapping None handling and a specific physical-intelligence/fast loading failure are not the same bug." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Video-model auto-processing without torchvision is unrelated to max_shard_size in save_pretrained()." }, { "left": "issue:43874", "right": "issue:44263", "accept": false, - "reason": "Both touch GLM-family multimodal code, but they are distinct bugs." + "reason": "Both touch GLM-related internals, but one is an image-processor method gap and the other is a torch.split indexing issue." }, { "left": "issue:43976", "right": "issue:45071", "accept": false, - "reason": "Python version compatibility and PretrainedConfig type checking are unrelated." + "reason": "Python version/package compatibility and config type checking are different concerns." }, { "left": "issue:43986", "right": "issue:45372", "accept": false, - "reason": "Missing torchvision and missing ReasoningEffort from mistral_common are different optional-dependency import failures." + "reason": "Missing torchvision in AutoProcessor video loading is unrelated to the ReasoningEffort import error for Gemma 4." }, { "left": "issue:44829", "right": "issue:45245", "accept": false, - "reason": "FlashAttention-3 training degeneration is unrelated to the category-count limit error." + "reason": "FlashAttention training degeneration and the 2^24 category limit are unrelated numerical/shape constraints." }, { "left": "issue:43531", "right": "issue:45341", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and a testing_utils bug are unrelated." + "reason": "A sliding_window bug in Qwen3-MoE is unrelated to testing_utils." }, { "left": "issue:38617", "right": "issue:42757", "accept": false, - "reason": "Two import errors, but from different packages and different symbols." + "reason": "Both are import errors, but they are about different missing symbols from different packages." }, { - "left": "issue:43064", - "right": "issue:44315", + "left": "issue:44261", + "right": "issue:44485", "accept": false, - "reason": "FSDP2 optimizer state corruption and Liger Kernel not being applied are different training-stack issues." + "reason": "Different GLM issues: missing rms_norm_eps in config vs a RoPE implementation discussion." }, { - "left": "issue:44261", - "right": "issue:44485", + "left": "issue:43064", + "right": "issue:44315", "accept": false, - "reason": "Missing rms_norm_eps in MLA and GLM-5 RoPE implementation are distinct model-architecture concerns." + "reason": "FSDP2/PEFT optimizer-state corruption and Liger Kernel not applying in model_init are distinct training setup bugs." } ] }, - "evaluator_result": { - "accept": false, - "feedback": "The conservative pair judgments are fine, but the duplicate call on issue:44483 \u2194 issue:45464 is overstated. From the packet we only know both concern chat/completions failures; the exact Qwen3.5-0.8B streaming bug is not established by the provided metadata alone. The summary also goes beyond the evidence by calling this the only clear duplicate pair / canonical tracker. Please soften that conclusion unless the issue bodies explicitly show the same model and failure mode." - }, - "evaluator_used": true, - "retried": true, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -13275,28 +13305,28 @@ "issue:36683|issue:43582", "issue:44483|issue:44514", "issue:44222|issue:45478", - "issue:43618|issue:44380", "issue:43012|issue:44683", "issue:43618|issue:43761", + "issue:43618|issue:44380", "issue:42757|issue:43824", "issue:38617|issue:43502", "issue:44483|issue:45290", "issue:44038|issue:44230" ], - "prepared_review_unit_hash": "7dddff1596eddbf72710670fecc2b858768863a57cca49e03ffdbaad740c55a3", + "prepared_review_unit_hash": "1c6a33e4672d25a3a752225d2fbb1ce7bec779e0c99d9fde8aacd394f3b0a7e3", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 38064, + "serialized_chars": 38062, "estimated_input_tokens": 9516, "estimated_eval_tokens": 19288 }, @@ -13305,162 +13335,162 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of unrelated Transformers issues with only superficial keyword overlap (imports, v5 regressions, save/load, Qwen, CLIP, etc.). I do not see any pair that is clearly the same underlying bug or change, so all soft edges should be rejected.", - "confidence": 0.41, - "canonical_issue_reason": "issue:43097 is the best representative of the largest recurring theme here: Transformers v5 behavior/regression complaints around removed or changed model/config APIs, and it also has the strongest apparent centrality among the candidate set.", + "summary": "This cluster is very broad and mostly contains unrelated Transformers regressions that only share a loose v5/compatibility/theme overlap. I do not see any soft pair that clearly describes the same underlying bug or change, so all soft edges should be rejected. The closest representative issue is the broad v5 API-breakage report around removed configuration behavior.", + "confidence": 0.89, + "canonical_issue_reason": "issue:43097 is the broadest v5 API-compatibility regression in the set and best matches the cluster\u2019s shared theme of breaking changes, even though the rest of the items are largely unrelated.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43097 is the most representative issue for triage because it sits closest to the broad v5 migration/regression theme that appears repeatedly in this cluster, even though the cluster overall is not truly coherent.", + "best_issue_reason": "issue:43097 is the best standalone representative because it captures a core, user-visible v5 breakage rather than a narrow model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:38617", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but for different symbols and modules (`layer_type_validation` vs `HybridCache`) with unrelated failure modes." + "reason": "Both are import errors, but they involve different missing symbols and different failure modes." }, { "left": "issue:44410", "right": "issue:45375", "accept": false, - "reason": "Both mention Qwen3.x vision-related models, but one is missing layer projections and the other is a config field being dropped by strict validation." + "reason": "Both concern Qwen-family configs, but they are different fields, models, and bugs." }, { "left": "issue:44062", "right": "issue:44117", "accept": false, - "reason": "One is an `AddedToken` argument collision, the other is `TOKENIZER_MAPPING_NAMES` returning `None`; these are different tokenizer code paths." + "reason": "Tokenizer-related, but one is an AddedToken construction error and the other is a mapping/None-handling bug." }, { "left": "issue:43097", "right": "issue:45588", "accept": false, - "reason": "`tie_embeddings_and_encoder_decoder` removal and a flash-attention `s_aux=None` crash are unrelated bugs." + "reason": "One is a v5 config/API removal complaint; the other is a flash-attention AttributeError in a different code path." }, { "left": "issue:44038", "right": "issue:44410", "accept": false, - "reason": "Different model families and different failures: Qwen3-VL-Moe compatibility versus missing projections in qwen3next." + "reason": "Both touch Qwen models, but the concrete problem is different: model support vs missing config/projection fields." }, { "left": "issue:44704", "right": "issue:45103", "accept": false, - "reason": "`AutoProcessor.from_pretrained` kwargs forwarding and the auto-docstring annotation crash are separate utilities with different failure points." + "reason": "Both involve kwargs handling, but one is processor loading and the other is auto-docstring introspection." }, { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "Offline/local file handling is unrelated to missing `embed_tokens.weight` during UMT5 encoder loading." + "reason": "Offline/local-file behavior and missing state-dict weights are unrelated bugs." }, { "left": "issue:43986", "right": "issue:44117", "accept": false, - "reason": "A missing `torchvision` dependency crash is not the same as a tokenizer mapping lookup returning `None`." + "reason": "Different subsystems: video AutoProcessor dependency handling vs tokenizer mapping None-handling." }, { "left": "issue:43097", "right": "issue:43873", "accept": false, - "reason": "One is a v5 config/API removal warning; the other is an offloading/quantization bug." + "reason": "A v5 API removal is not the same as quantization offloading misbehavior." }, { "left": "issue:43824", "right": "issue:43986", "accept": false, - "reason": "Importing a specific Qwen VL class and crashing without torchvision are different issues." + "reason": "One is a missing model import; the other is a torchvision dependency crash in AutoProcessor." }, { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "Wrong tokenizer decoder type and an isolated reproduction of another issue are not the same concrete bug." + "reason": "Both mention tokenizer behavior, but there is no evidence they describe the same bug." }, { "left": "issue:43097", "right": "issue:43901", "accept": false, - "reason": "A removed embedding-tie helper and outdated pipeline docs about `return_all_scores` are different topics." + "reason": "API removal and documentation drift for a pipeline argument are different issues." }, { "left": "issue:43825", "right": "issue:44655", "accept": false, - "reason": "Pipeline chat/translation task messaging and saving Pipeline objects involve different code paths and symptoms." + "reason": "Both involve pipeline-related behavior, but one is an error message regression and the other is save_pretrained support." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "A missing weight when loading UMT5 is unrelated to the `max_shard_size` default in `save_pretrained()`." + "reason": "Missing encoder weights and max_shard_size behavior are separate save/load concerns." }, { "left": "issue:43352", "right": "issue:44380", "accept": false, - "reason": "Both involve attention backends, but one is unsupported Flash Attention for a model and the other is GPT-2 scaling being ignored." + "reason": "Both mention attention backends, but they affect different models and different concrete failures." }, { "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "xLSTM training bugs and a Bamba `torch.compile`/SDPA failure are unrelated model-specific problems." + "reason": "Different model families and different training/compile failures." }, { "left": "issue:43012", "right": "issue:43550", "accept": false, - "reason": "A bfloat16 precision warning from compilation is not the same as a model failure under `torch.compile` with SDPA." + "reason": "A precision warning during compile is not the same as a SDPA compile failure." }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "These are different warnings with different triggers: compilation precision versus tied-embeddings config advice." + "reason": "Different symptoms: a PyTorch warning versus a tie_word_embeddings config warning." }, { "left": "issue:42886", "right": "issue:43381", "accept": false, - "reason": "Offline cache loading and gradient checkpointing in eval mode are unrelated behaviors." + "reason": "Offline cache loading and gradient checkpointing eval-mode behavior are unrelated." }, { "left": "issue:42907", "right": "issue:45753", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals and an infinite-loop image generation bug are different model/runtime issues." + "reason": "Saving dequantized models and an infinite-loop generation bug are not the same issue." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Saving a dequantized model and saving a Pipeline object are both save-related but not the same bug." + "reason": "Both are save-related, but one is dequantized-model saving and the other is pipeline serialization." }, { "left": "issue:44380", "right": "issue:45479", "accept": false, - "reason": "GPT-2 attention scaling under backends and degenerate zero-loss classification are completely different code paths." + "reason": "Different problem classes: attention backend scaling vs classification loss degeneracy." }, { "left": "issue:43352", "right": "issue:43976", "accept": false, - "reason": "Flash Attention support for Nemotron and a Python-version compatibility problem are unrelated." + "reason": "A model capability issue and a Python version compatibility issue are unrelated." }, { "left": "issue:43064", "right": "issue:44380", "accept": false, - "reason": "FSDP2/PEFT optimizer-state corruption and GPT-2 attention scaling are unrelated issues." + "reason": "Distributed optimizer-state corruption and attention scaling are unrelated bugs." }, { "left": "issue:43208", "right": "issue:43526", "accept": false, - "reason": "xLSTM training bugs and a Beit image processor label-reduction bug are unrelated." + "reason": "Different models and different preprocessing/training failures." }, { "left": "issue:36683", @@ -13472,55 +13502,37 @@ "left": "issue:44483", "right": "issue:44514", "accept": false, - "reason": "A chat-completions API request rejection and a batched `apply_chat_template` crash are different endpoints and failures." + "reason": "Both involve chat-related APIs, but one is an endpoint acceptance regression and the other is processor batching behavior." }, { "left": "issue:44222", "right": "issue:45478", "accept": false, - "reason": "Both mention Qwen3.5/Moe/FP8, but one is save_pretrained support and the other is a from_pretrained regression; not the same concrete bug." - }, - { - "left": "issue:43618", - "right": "issue:44380", - "accept": false, - "reason": "Missing CLIP attentions and GPT-2 attention scaling are unrelated model behaviors." - }, - { - "left": "issue:43012", - "right": "issue:44683", - "accept": false, - "reason": "A bfloat16 warning is not the same as compiled flex attention failing on torch >= 2.9." - }, - { - "left": "issue:43618", - "right": "issue:43761", - "accept": false, - "reason": "Both are CLIP regressions, but one concerns attentions and the other hidden states; different output fields and fixes." + "reason": "Both concern Qwen3.5 MoE/FP8-ish areas, but one is save_pretrained and the other is from_pretrained; not the same bug." }, { "left": "issue:42757", "right": "issue:43824", "accept": false, - "reason": "These are different import errors from different packages/symbols and do not share the same root cause." + "reason": "External dependency import failure and missing model-class export are different import problems." }, { "left": "issue:38617", "right": "issue:43502", "accept": false, - "reason": "An import failure in configuration utils and unwanted API requests under `local_files_only=True` are unrelated." + "reason": "Different failures: an internal import symbol removal versus local_files_only still making network calls." }, { "left": "issue:44483", "right": "issue:45290", "accept": false, - "reason": "A v1 chat-completions endpoint issue and a chat template crash with tool-call assistant messages are different problems." + "reason": "Both are chat-template/request formatting issues, but the concrete crash conditions differ." }, { "left": "issue:44038", "right": "issue:44230", "accept": false, - "reason": "Both mention Qwen/VL/Moe/FP8, but one is a general v5 compatibility bug and the other is FP8 inference support; not the same code-path problem." + "reason": "Same broad Qwen3/VL/MoE area, but one is a model bug and the other is explicit fp8 inference support." } ] }, @@ -13602,16 +13614,16 @@ "issue:36683|issue:43475", "issue:42994|issue:44655", "issue:43352|issue:43937", - "issue:43867|issue:45216", "issue:44077|issue:44380", + "issue:43867|issue:45216", "issue:36010|issue:42757", "issue:42898|issue:43066", "issue:36683|issue:44077", "issue:43611|issue:44655", "issue:43976|issue:44938", "issue:43352|issue:45593", - "issue:44263|issue:45440", "issue:36683|issue:43525", + "issue:44263|issue:45440", "issue:43618|issue:45216", "issue:44188|issue:44380", "issue:36683|issue:44242", @@ -13628,242 +13640,227 @@ "issue:43867|issue:45357", "issue:43010|issue:45850" ], - "prepared_review_unit_hash": "e82ff03e28a993437c60eecf1de0f4d87863f3ebd09b071dee9a55238a4ecda6", + "prepared_review_unit_hash": "c73576f4bb53809b99edaa0d2bf4d90964d08e6da71ab2bdd85420e5911cc5e2", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37169, - "estimated_input_tokens": 9293, - "estimated_eval_tokens": 18842 + "serialized_chars": 37161, + "estimated_input_tokens": 9291, + "estimated_eval_tokens": 18838 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous Transformers bug reports spanning different models, versions, and subsystems. There is only one clear duplicate-like pair around import-time slowdown; the rest should stay separate.", - "confidence": 0.88, - "canonical_issue_reason": "Issue 41669 is the clearest root-cause report for the import slowdown: it names the concrete cause (`import *` in models) and the observed impact. Its paired report 44246 is the more generic symptom.", + "summary": "This cluster is mostly a loose collection of unrelated Transformers regressions (imports, loading/saving, config mismatches, tokenizer behavior, training, and backend-specific bugs). I do not see any soft-edge pair that is safely the same underlying bug/change, so I would not merge any of the candidate pairs. If one issue must represent the cluster, issue 45278 is the broadest umbrella report.", + "confidence": 0.62, + "canonical_issue_reason": "issue:45278 is the most umbrella-like report: it describes broad import breakage after upgrading, which best matches the cluster\u2019s general upgrade/regression theme.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45278 is the best overall umbrella issue in this set because it is the broadest, open report about widespread import regressions after upgrading and can absorb related symptom reports.", + "best_issue_reason": "issue:45278 is the best single representative because it is broad, open, and central to the recurring \u2018upgrade to v5 breaks things\u2019 pattern seen across this set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43299", "right": "issue:43867", "accept": false, - "reason": "Different failures: Qwen3VL MoE loading vs a state-dict ordering load error. Same general loading area, but not the same bug." + "reason": "Both are model-loading failures, but one is a Qwen3VL MoE loading regression and the other is a state_dict ordering/load error; different code paths and fixes." }, { "left": "issue:44483", "right": "issue:45381", "accept": false, - "reason": "Chat/completions request validation is unrelated to Qwen2.5-VL video position-id handling." + "reason": "One is an API request rejection in /v1/chat/completions, the other is a qwen2.5-vl video position-id bug; unrelated failures." }, { "left": "issue:43756", "right": "issue:44485", "accept": false, - "reason": "Both mention RoPE, but they concern different model implementations and different defects." + "reason": "Both mention RoPE, but they concern different models and different implementation bugs; not the same concrete change." }, { "left": "issue:43299", "right": "issue:45753", "accept": false, - "reason": "One is a model-loading regression; the other is an infinite-loop generation bug on a specific image." + "reason": "A loading regression for Qwen3VL MoE is not the same as an infinite-loop generation bug for a specific image in Qwen3_5." }, { "left": "issue:43901", "right": "issue:44655", "accept": false, - "reason": "Docs mismatch for `return_all_scores` is unrelated to pipeline serialization/save_pretrained failure." + "reason": "Docs mismatch for return_all_scores is unrelated to saving Pipeline objects." }, { "left": "issue:36683", "right": "issue:43475", "accept": false, - "reason": "Gemma3 config attribute loss and Sam3 vision output attribute loss are different model-specific bugs." + "reason": "Different missing-attribute errors on different model configs/outputs; no shared underlying bug is evident." }, { "left": "issue:42994", "right": "issue:44655", "accept": false, - "reason": "Quantized model saving and pipeline object saving are distinct serialization paths and failure modes." + "reason": "Quantized model save failure and Pipeline save_pretrained failure are distinct serialization issues." }, { "left": "issue:43352", "right": "issue:43937", "accept": false, - "reason": "FlashAttention2 support gating is unrelated to invalid GenerationConfig for GLM-5." + "reason": "Flash-Attn support gating for Nemotron is unrelated to GLM-5 GenerationConfig validation." }, { - "left": "issue:43867", - "right": "issue:45216", + "left": "issue:44077", + "right": "issue:44380", "accept": false, - "reason": "State-dict sorting load error and incorrect Qwen3.5 checkpoint saving are different save/load regressions." + "reason": "One is a patchtsmixer post_init constraint, the other is GPT2 attention scaling under SDPA/FlashAttention; different bugs." }, { - "left": "issue:44077", - "right": "issue:44380", + "left": "issue:43867", + "right": "issue:45216", "accept": false, - "reason": "`post_init` policy for patchtsmixer is unrelated to GPT2 attention scaling being ignored under SDPA/FlashAttention." + "reason": "Both involve model persistence/loading, but one is a sorted-state_dict load error and the other is an incorrect Qwen3.5 save_pretrained checkpoint; not the same fix." }, { "left": "issue:36010", "right": "issue:42757", "accept": false, - "reason": "Different missing imports from different packages (`transformers.generation` vs `huggingface_hub`)." + "reason": "Different import errors from different modules/dependencies; same symptom, not same bug." }, { "left": "issue:42898", "right": "issue:43066", "accept": false, - "reason": "Both tokenizer-related, but one is a behavior change in cleanup spaces and the other is a decoder-type regression." + "reason": "Tokenizer cleanup-space behavior and wrong decoder type are separate v5 tokenizer regressions." }, { "left": "issue:36683", "right": "issue:44077", "accept": false, - "reason": "Gemma3Config missing `vocab_size` is unrelated to patchtsmixer `post_init` deprecation." + "reason": "A config attribute error and a patchtsmixer post_init policy issue are unrelated." }, { "left": "issue:43611", "right": "issue:44655", "accept": false, - "reason": "`base_model_prefix` loading breaks and pipeline save_pretrained failures are separate code paths." + "reason": "base_model_prefix loading breakage is not the same as saving Pipeline objects." }, { "left": "issue:43976", "right": "issue:44938", "accept": false, - "reason": "Different Python-version compatibility failures: 5.1.0 vs Python 3.9+ and Python 3.14 import failure." + "reason": "Both are Python-version compatibility complaints, but they point to different releases and likely different root causes." }, { "left": "issue:43352", "right": "issue:45593", "accept": false, - "reason": "Unsupported FlashAttention2 and missing auxiliary losses in D-FINE are unrelated." + "reason": "Flash-Attn support for Nemotron has nothing to do with D-FINE auxiliary-loss handling." }, { - "left": "issue:44263", - "right": "issue:45440", + "left": "issue:36683", + "right": "issue:43525", "accept": false, - "reason": "Both involve MoE-ish internals, but they are different models and different bugs." + "reason": "Different models and different missing config attributes; not a shared code-path bug." }, { - "left": "issue:36683", - "right": "issue:43525", + "left": "issue:44263", + "right": "issue:45440", "accept": false, - "reason": "Gemma3 config missing `vocab_size` and Llama4Config missing `pad_token_id` are separate attribute regressions." + "reason": "A torch.split indexing bug and a DeepseekV3MoE divergence from remote implementation are different issues." }, { "left": "issue:43618", "right": "issue:45216", "accept": false, - "reason": "CLIPOutput attentions regression is unrelated to Qwen3.5 checkpoint save correctness." + "reason": "Missing attentions assignment in CLIPOutput is unrelated to Qwen3.5 save_pretrained checkpoint corruption." }, { "left": "issue:44188", "right": "issue:44380", "accept": false, - "reason": "Torch.compile branching divergence in attention kernels is a different bug from GPT2 attention scaling being ignored." + "reason": "Both touch attention backends, but one is a torch.compile branching divergence and the other is GPT2 attention scaling being ignored; distinct bugs." }, { "left": "issue:36683", "right": "issue:44242", "accept": false, - "reason": "Gemma3Config attribute error is unrelated to router-logits load-balancing loss handling." + "reason": "A missing config field and a missing load-balancing loss are separate model logic issues." }, { "left": "issue:43064", "right": "issue:44492", "accept": false, - "reason": "Distributed optimizer-state corruption is unrelated to a cache-strategy typo." + "reason": "Distributed optimizer-state corruption and a cache-strategy typo are not the same underlying problem." }, { "left": "issue:43352", "right": "issue:43931", "accept": false, - "reason": "FlashAttention2 support error is unrelated to Qwen3-VL weight-shape mismatch." + "reason": "FlashAttention support for Nemotron is unrelated to Qwen3-VL weight-shape mismatch on load." }, { "left": "issue:43010", "right": "issue:45278", "accept": false, - "reason": "Cache update no_grad behavior is unrelated to generic import errors after upgrade." + "reason": "no_grad decoration for cache/layer updates is a code correctness change, while 45278 is a broad import-breakage report; different issues." }, { "left": "issue:43992", "right": "issue:45310", "accept": false, - "reason": "UMT5Encoder missing weights and Qwen3.5 MoE from_pretrained errors are different loading regressions." + "reason": "Both are from_pretrained failures, but they concern different models and different missing/misrouted weights." }, { "left": "issue:43867", "right": "issue:45375", "accept": false, - "reason": "A sorted state_dict load failure is unrelated to missing `deepstack_visual_indexes` in Qwen3_5MoeVisionConfig." + "reason": "State_dict loading error and a missing Qwen3_5MoeVisionConfig field are separate bugs." }, { "left": "issue:41669", "right": "issue:44246", - "accept": true, - "reason": "These match on the same underlying import-time slowdown: 41669 identifies the concrete cause (`import *` bloat), while 44246 reports the user-visible slow import symptom." + "accept": false, + "reason": "Import-time performance slowdown is related in theme, but the issue about removing import * usage is not clearly the same concrete bug as intermittent slow imports." }, { "left": "issue:44355", "right": "issue:44855", "accept": false, - "reason": "Compiled Python file errors and Python 3.13 `@torch.jit.script` parsing failures are different interpreter/decorator issues." - }, - { - "left": "issue:36683", - "right": "issue:43064", - "accept": false, - "reason": "Config attribute regression and distributed training optimizer-state bug are unrelated." - }, - { - "left": "issue:42890", - "right": "issue:45561", - "accept": false, - "reason": "Test flakiness from missing `set_seed()` is unrelated to xdist workers racing on `captured_info.txt`." + "reason": "Compiled Python-file errors and a Python 3.13 torch.jit parsing indentation error are different import/parsing failures." }, { "left": "issue:44938", "right": "issue:45071", "accept": false, - "reason": "Python 3.14 load failure and `PretrainedConfig` type-checking regression are different compatibility problems." + "reason": "Python 3.14 load failure and PretrainedConfig type-checking regression are unrelated." }, { "left": "issue:43867", "right": "issue:45357", "accept": false, - "reason": "State-dict ordering load failure and incorrect visual encoder keys in save_pretrained are different save/load bugs." + "reason": "Both are Qwen3.5 save/load regressions, but one is a load/state_dict ordering problem and the other is incorrect visual-encoder keys on save; not the same exact fix." }, { "left": "issue:43010", "right": "issue:45850", "accept": false, - "reason": "Cache/layer `update()` no_grad behavior is unrelated to the broad complaint about too many models breaking since v5." + "reason": "A cache-update no_grad change does not match a broad \u2018too many models break since 5.0\u2019 complaint." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet: the broad import-regression umbrella around issue 45278 is reasonable, issue 41669 is the clearest root-cause import slowdown report, and the lone accepted soft pair (41669/44246) is supported by the titles. The remaining soft-edge rejections stay conservative and do not overclaim shared bugs." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -13966,248 +13963,251 @@ "issue:30333|issue:36296", "issue:44355|issue:45071" ], - "prepared_review_unit_hash": "b712106ef137fee0c5438014e14a0eb9ac5e93af3403051c425e8b45a14c41e3", + "prepared_review_unit_hash": "43610d24231281fba017f15c58177be24dce2da339c8223ce4ac321dbddee836", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 37518, - "estimated_input_tokens": 9380, - "estimated_eval_tokens": 19016 + "serialized_chars": 37521, + "estimated_input_tokens": 9381, + "estimated_eval_tokens": 19018 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a mix of unrelated issues; only one pair looks plausibly like the same model-loading bug. Most soft matches are broad keyword overlaps rather than true duplicates.", - "confidence": 0.69, - "canonical_issue_reason": "issue:45125 is the most concrete canonical candidate: it names the specific model class and a clear loading/tensor-parallel failure mode that plausibly explains the related from_pretrained report.", + "summary": "The cluster is broad and mostly not duplicate-safe: many items are unrelated load/import/test/model-specific regressions that only share superficial wording. No soft pair looks like the same concrete bug or a PR that should be merged together. The closest common thread is v5-era loading regressions, with one issue serving as the best anchor.", + "confidence": 0.73, + "canonical_issue_reason": "issue:44556 is the clearest representative of the recurring v5 loading-regression theme: it explicitly describes a v4.57 -> v5.2/v5.3 checkpoint reload break with a concrete, user-facing failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45125 is the best representative issue because it is specific, reproducible, and tied to a clear code path; if one duplicate is kept, this is the better anchor than the more generic error report.", + "best_issue_reason": "issue:44556 is the best global anchor because it is explicit about a versioned regression, concrete about the failing code path, and broad enough to represent the main loading-compatibility complaints in this cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43992", "right": "issue:45478", "accept": false, - "reason": "Both are loading errors, but they involve different models and likely different root causes; too broad to treat as duplicates." + "reason": "Same model family and loading symptom, but one is missing embed_tokens.weight while the other is a generic v5.4.0 from_pretrained failure; not the same bug." }, { "left": "issue:43010", "right": "issue:43976", "accept": false, - "reason": "One is about no_grad decoration on cache update, the other about Python version compatibility; unrelated bugs." + "reason": "Cache no_grad behavior vs Python 3.9+ compatibility; unrelated." }, { "left": "issue:44938", "right": "issue:45381", "accept": false, - "reason": "Python 3.14 import/load failure and Qwen2.5-VL video position_ids are different failure modes." + "reason": "Python 3.14 load failure vs qwen2.5-vl video position ids bug; unrelated." }, { "left": "issue:43618", "right": "issue:44077", "accept": false, - "reason": "CLIP attentions assignment and patchtsmixer post_init validation are separate code paths." + "reason": "CLIPOutput attentions regression vs patchtsmixer post_init validation; unrelated." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Both mention tensor parallelism, but one is a generic training bug and the other is a specific OOM on one model; not the same bug." + "reason": "Tensor-parallel training bug vs tp OOM on a specific Qwen model; same area but different failure mode and likely different fix." }, { "left": "issue:44360", "right": "issue:45440", "accept": false, - "reason": "Implementation divergence complaints in different subsystems are not enough to unify these." + "reason": "DSA indexer ReLU discussion vs DeepSeekV3 divergence; unrelated." }, { "left": "issue:28282", "right": "issue:33453", "accept": false, - "reason": "PyTorch missing import and tokenizer loading regression are unrelated." + "reason": "ImportError for AutoModel/PyTorch vs tokenizer loading regression; both loading-related but distinct symbols and code paths." }, { "left": "issue:43867", "right": "issue:44484", "accept": false, - "reason": "Model load failure due to sorted state_dict and save_pretrained shard-size discussion are different topics." + "reason": "State-dict sorting load error vs save_pretrained shard-size question; unrelated." }, { "left": "issue:43618", "right": "issue:44857", "accept": false, - "reason": "CLIPOutput attentions and LwDetrImageLoss AMP crash are unrelated." + "reason": "CLIPOutput attentions regression vs LwDetrImageLoss AMP crash; unrelated." }, { "left": "issue:43618", "right": "issue:44977", "accept": false, - "reason": "These target different models and different runtime issues." + "reason": "CLIPOutput attentions regression vs Qwen3.5 flash-attention generation issue; unrelated." }, { "left": "issue:42890", "right": "issue:44964", "accept": false, - "reason": "Test flakiness from missing seeds is unrelated to a model-loading regression." + "reason": "Flaky integration tests vs model loading regression; unrelated." }, { "left": "issue:42890", "right": "issue:44112", "accept": false, - "reason": "Both are CI/test failures, but the affected tests and causes differ." + "reason": "Both are CI/test issues, but different tests and failure causes." }, { "left": "issue:42890", "right": "issue:45600", "accept": false, - "reason": "Different CI failures in different files; no shared underlying bug." + "reason": "Test seed flakiness vs auto_mappings CI break; unrelated." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "Offline API-request leakage and AddedToken argument duplication are unrelated." + "reason": "local_files_only network leak vs AddedToken duplicate 'special' TypeError; unrelated." }, { "left": "issue:43010", "right": "issue:44075", "accept": false, - "reason": "Generation cache no_grad and SGD arguments not being used are unrelated." + "reason": "Cache no_grad update vs SGD args ignored; unrelated." }, { "left": "issue:43723", "right": "issue:44556", "accept": false, - "reason": "Tokenizer loading in v5 and checkpoint reload after upgrade to v5.2/v5.3 are both version-related, but not the same concrete failure." + "reason": "Both concern loading regressions after v5, but tokenizer loading and checkpoint reload are different concrete bugs." }, { "left": "issue:42890", "right": "issue:45341", "accept": false, - "reason": "Test seed flakiness and a bug in testing_utils are not the same issue." + "reason": "Seed-related test flakiness vs a testing_utils bug; unrelated." }, { "left": "issue:44387", "right": "issue:44556", "accept": false, - "reason": "Int4 quantization memory regression and checkpoint reload incompatibility are different bugs." + "reason": "int4 quantization OOM vs checkpoint reload failure; unrelated." }, { "left": "issue:44222", "right": "issue:45125", "accept": false, - "reason": "Both involve MoE models, but one is FP8 save_pretrained and the other is tensor-parallel loading; different code paths." + "reason": "FP8 save_pretrained issue vs missing _tp_plan for tensor parallelism; different code paths." }, { "left": "issue:43010", "right": "issue:43065", "accept": false, - "reason": "Cache/layer update decoration and Sam3PixelDecoder dummy Conv2d are unrelated." + "reason": "Generation cache update decorator vs dummy Conv2d in Sam3PixelDecoder; unrelated." }, { "left": "issue:42548", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but for different symbols and likely different export issues." + "reason": "Both are import-name errors, but different exported symbols and likely different regressions." }, { "left": "issue:43502", "right": "issue:44485", "accept": false, - "reason": "Offline request leakage and GLM-5 RoPE implementation are unrelated." + "reason": "Offline-loading bug vs GLM-5 RoPE implementation; unrelated." }, { "left": "issue:44355", "right": "issue:44749", "accept": false, - "reason": "Compiled-file import errors and a performance regression after upgrade are not the same bug." + "reason": "Compiled-file errors vs performance slowdown after upgrade; unrelated." }, { "left": "issue:43010", "right": "issue:43232", "accept": false, - "reason": "Both touch generation internals, but the failing behavior and likely fix differ." + "reason": "Both touch generation helpers, but one is cache update/no_grad and the other sync_gpus state handling; not the same bug." }, { "left": "issue:30990", "right": "issue:31795", "accept": false, - "reason": "Sentence Transformers loading hang and forward doc confusion are unrelated." + "reason": "Loading hang vs documentation confusion; unrelated." }, { "left": "issue:45125", "right": "issue:45478", - "accept": true, - "reason": "Both point to the same Qwen3.5 MoE loading path; the generic from_pretrained failure is plausibly the symptom of the missing _tp_plan / tensor-parallel loading bug." + "accept": false, + "reason": "Same Qwen3.5 MoE area, but one is missing _tp_plan and the other is a generic from_pretrained error; not enough to merge." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Tensor-parallel plan issue vs missing vision config field; unrelated." }, { "left": "issue:30333", "right": "issue:33453", "accept": false, - "reason": "MLflow reporting on failed training jobs and tokenizer loading regression are unrelated." + "reason": "MLflow reporting failure vs tokenizer loading regression; unrelated." }, { "left": "issue:43299", "right": "issue:44556", "accept": false, - "reason": "A specific Qwen3VL MoE load failure and a version-upgrade checkpoint reload problem are not the same bug." + "reason": "Version-related model loading failures, but different models and failure modes; not one duplicate." }, { "left": "issue:44315", "right": "issue:44360", "accept": false, - "reason": "Liger Kernel application with model_init and DSA indexer ReLU discussion are unrelated." + "reason": "Liger Kernel application vs DSA ReLU discussion; unrelated." }, { "left": "issue:43502", "right": "issue:44222", "accept": false, - "reason": "Offline request leakage and FP8 MoE save_pretrained issues are different concerns." + "reason": "Offline requests vs FP8 save_pretrained moe; unrelated." }, { "left": "issue:42831", "right": "issue:43874", "accept": false, - "reason": "FineGrainedFP8 accuracy and GLM46V image-patch counting are unrelated." + "reason": "FP8 accuracy issue vs missing image processor method; unrelated." }, { "left": "issue:43502", "right": "issue:44961", "accept": false, - "reason": "API requests despite local_files_only and a stray 'racoon' issue are unrelated." + "reason": "Offline API request bug vs placeholder racoon issue; unrelated." }, { "left": "issue:30333", "right": "issue:36296", "accept": false, - "reason": "Failed MLflow reporting and tensor parallel training bugs are different subsystems and symptoms." + "reason": "MLflow job tracking vs tensor parallel training bug; unrelated." }, { "left": "issue:44355", "right": "issue:45071", "accept": false, - "reason": "Compiled Python file loading errors and PretrainedConfig type-checking breakage are unrelated." + "reason": "Compiled Python load errors vs PretrainedConfig type checking; unrelated." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The cluster-level summary is conservative and grounded in the packet. The single accepted pair (45125, 45478) is a plausible same-bug match based on the shared Qwen3.5 MoE loading path, and the remaining soft matches are appropriately rejected as likely unrelated or too broad." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -14246,6 +14246,7 @@ "issue:44038", "issue:44222", "issue:44230", + "issue:44246", "issue:44261", "issue:44273", "issue:44355", @@ -14273,8 +14274,8 @@ "issue:45850" ], "soft_pairs": [ - "issue:44355|issue:44655", "issue:43827|issue:44961", + "issue:44355|issue:44655", "issue:33453|issue:34689", "issue:42757|issue:43576", "issue:44222|issue:44230", @@ -14298,8 +14299,8 @@ "issue:44273|issue:45542", "issue:44273|issue:45003", "issue:44393|issue:45753", - "issue:43576|issue:44556", "issue:44393|issue:45071", + "issue:43576|issue:44556", "issue:44038|issue:45125", "issue:43525|issue:45850", "issue:41669|issue:43673", @@ -14309,66 +14310,67 @@ "issue:43572|issue:45850", "issue:44704|issue:44908", "issue:42890|issue:44908", - "issue:45278|issue:45850" + "issue:45278|issue:45850", + "issue:44246|issue:44273" ], - "prepared_review_unit_hash": "910f2e2294b9b65fced10702089037b10097808148cbedf143d90d37d8db4b95", + "prepared_review_unit_hash": "8f2030faa0c46c95b94dbaf3fa15d1c308e2e86b2969dd75f74341f7fcc57fde", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 37, - "serialized_chars": 37332, - "estimated_input_tokens": 9333, - "estimated_eval_tokens": 18922 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 38, + "serialized_chars": 38176, + "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 set of Transformers v5 regressions and compatibility issues, especially import/load/save breakages. The only reasonable canonical issue is the broad upgrade-regression report about many import errors after moving to v5.5.0; the rest are too heterogeneous for duplicate merging. No soft edge pair is convincing enough to accept.", - "confidence": 0.33, - "canonical_issue_reason": "issue:45278 is the strongest umbrella candidate: it is open, clearly states a broad post-upgrade regression, and fits the common theme of import/load breakage better than the more vague or highly specific reports.", + "summary": "This cluster is mostly a mix of unrelated Transformers v5 regressions, import/export failures, and model-specific bugs. The best umbrella issue is the broad post-upgrade import-error report.", + "confidence": 0.92, + "canonical_issue_reason": "#45278 is the most representative umbrella: it concretely describes broad import failures after upgrading to 5.5.0, which matches the general regression theme better than the other broad-but-vague reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most useful triage target because it is concrete enough to act on while still broad enough to represent the shared v5 import-regression theme in this cluster.", + "best_issue_reason": "#45278 is the strongest canonical target for duplicate triage because it is broad enough to cover the shared import-regression pattern, while still being specific about the upgrade-triggered failure mode.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44355", - "right": "issue:44655", + "left": "issue:43827", + "right": "issue:44961", "accept": false, - "reason": "Different bugs: compiled Python-file execution errors vs saving Pipeline objects. No shared code-path or concrete fix." + "reason": "Docs about pipeline() references vs an unrelated issue titled \"racoon\"; no shared bug." }, { - "left": "issue:43827", - "right": "issue:44961", + "left": "issue:44355", + "right": "issue:44655", "accept": false, - "reason": "Docs mentioning pipeline() and an unrelated issue titled 'racoon' are not the same underlying bug." + "reason": "Compiled Python file errors and pipeline save_pretrained failures are different code paths." }, { "left": "issue:33453", "right": "issue:34689", "accept": false, - "reason": "Both involve loading, but one is tokenizer loading and the other is model loading for a specific Llama vision checkpoint; too different to merge." + "reason": "Both are loading regressions, but one is tokenizer loading and the other is a specific model-loading breakage." }, { "left": "issue:42757", "right": "issue:43576", "accept": false, - "reason": "One is an import error for is_offline_mode, the other is a broken CLI command; different symptoms and fixes." + "reason": "Missing import from huggingface_hub vs broken env command; different failures." }, { "left": "issue:44222", "right": "issue:44230", "accept": false, - "reason": "Both mention FP8/MoE, but one is about save_pretrained and the other about inference support; not the same bug." + "reason": "Both mention FP8/moe, but one is save_pretrained and the other is inference support." }, { "left": "issue:33453", @@ -14380,61 +14382,61 @@ "left": "issue:43576", "right": "issue:43824", "accept": false, - "reason": "Both are v5 breakages, but one is CLI-related and the other is a specific model import failure; not one concrete code-path problem." + "reason": "Broken env command and missing Qwen2.5-VL export are different issues." }, { "left": "issue:43576", "right": "issue:45071", "accept": false, - "reason": "Broken env command and PretrainedConfig type-checking regression are separate issues." + "reason": "Env command breakage is unrelated to PretrainedConfig type-checking." }, { "left": "issue:31795", "right": "issue:38617", "accept": false, - "reason": "Documentation confusion versus an import error for layer_type_validation are unrelated." + "reason": "Documentation confusion and an import error for layer_type_validation are not the same bug." }, { "left": "issue:44961", "right": "issue:45468", "accept": false, - "reason": "No meaningful overlap beyond being issues in the same repo." + "reason": "Unrelated titles: no evidence of the same underlying defect." }, { "left": "issue:41628", "right": "issue:42548", "accept": false, - "reason": "Both are import errors, but for different missing symbols; not enough evidence of a single shared root cause." + "reason": "Both are import errors, but for different symbols and likely different export regressions." }, { "left": "issue:38617", "right": "issue:43576", "accept": false, - "reason": "Different failure modes: missing import versus broken env command." + "reason": "ImportError for layer_type_validation and a broken env command are unrelated." }, { "left": "issue:43576", "right": "issue:44938", "accept": false, - "reason": "CLI breakage and Python 3.14 load failure are different problems." + "reason": "Env command failure is not the same as Python 3.14 load failure." }, { "left": "issue:43673", "right": "issue:44492", "accept": false, - "reason": "Generation cache missing in chunked_prefill and a typo in cache strategies are not the same underlying bug." + "reason": "Cache missing during chunked prefill vs a typo in cache strategies are different problems." }, { "left": "issue:44908", "right": "issue:44961", "accept": false, - "reason": "Scheduler kwargs bug and an unrelated issue titled 'racoon' do not match." + "reason": "Scheduler kwargs bug and unrelated \"racoon\" issue do not match." }, { "left": "issue:44485", "right": "issue:44961", "accept": false, - "reason": "RoPE implementation discussion and an unrelated issue titled 'racoon' are not duplicates." + "reason": "RoPE implementation discussion vs unrelated \"racoon\" issue." }, { "left": "issue:36296", @@ -14446,121 +14448,127 @@ "left": "issue:41628", "right": "issue:44355", "accept": false, - "reason": "ImportError for AutoImageProcessor and compiled-file execution errors do not share a concrete bug." + "reason": "Different failure classes: import export vs compiled-file runtime errors." }, { "left": "issue:36296", "right": "issue:43526", "accept": false, - "reason": "Tensor parallel training and BeitImageProcessorFast label reduction are unrelated." + "reason": "Tensor parallel training bug and BeitImageProcessorFast label reduction bug are unrelated." }, { "left": "issue:45362", "right": "issue:45850", "accept": false, - "reason": "The latter is a vague umbrella complaint; it is not the same concrete bug as a Qwen3.5 chat crash." + "reason": "A specific Qwen3.5 chat crash is not the same as a broad complaint that v5 breaks too many models." }, { "left": "issue:44261", "right": "issue:44360", "accept": false, - "reason": "Precision error from missing rms_norm_eps and DSA indexer ReLU discussion are unrelated." + "reason": "MLA q_a_layernorm precision issue and DSA indexer ReLU issue are unrelated." }, { "left": "issue:44623", "right": "issue:45230", "accept": false, - "reason": "A specific missing-files save_pretrained bug and a generic bug report are not the same issue." + "reason": "Specific processor.save_pretrained missing files versus a generic bug report." }, { "left": "issue:44273", "right": "issue:45542", "accept": false, - "reason": "Lazy loading problems and undefined tf backend from tensorboard-only installs are different root causes." + "reason": "Lazy loading problem and TensorFlow backend error are different issues." }, { "left": "issue:44273", "right": "issue:45003", "accept": false, - "reason": "Lazy loading not working and unsafe sys.modules access are related in theme but not clearly the same concrete bug." + "reason": "Lazy loading failure and unsafe sys.modules access are related at most broadly, not the same concrete bug." }, { "left": "issue:44393", "right": "issue:45753", "accept": false, - "reason": "Both are Qwen3-VL behavior bugs, but one is 2D box output hallucination/error and the other is an infinite loop on a specific image." + "reason": "Different Qwen3_5VL behaviors: bounding-box output errors vs an infinite loop on one image." }, { - "left": "issue:43576", - "right": "issue:44556", + "left": "issue:44393", + "right": "issue:45071", "accept": false, - "reason": "Both are upgrade regressions, but one is the env command and the other is checkpoint reload compatibility; different fixes." + "reason": "Model output error and PretrainedConfig type checking are unrelated." }, { - "left": "issue:44393", - "right": "issue:45071", + "left": "issue:43576", + "right": "issue:44556", "accept": false, - "reason": "Model output bug and PretrainedConfig typing regression are unrelated." + "reason": "Broken env command and checkpoint reload incompatibility are different code paths." }, { "left": "issue:44038", "right": "issue:45125", "accept": false, - "reason": "Both mention Qwen/MoE and v5, but one is a general bug report and the other is a missing tensor-parallel plan; not the same code-path." + "reason": "Both touch Qwen/moe tensor-parallel space, but one is a model bug and the other is a missing tp plan; not the same fix." }, { "left": "issue:43525", "right": "issue:45850", "accept": false, - "reason": "Specific Llama4Config attribute bug versus a vague complaint that many models break; not a duplicate." + "reason": "Missing pad_token_id in Llama4Config is a specific config bug, not the broad v5 complaint." }, { "left": "issue:41669", "right": "issue:43673", "accept": false, - "reason": "Import-star cleanup/performance issue and chunked_prefill cache issue are not the same bug." + "reason": "Import-star performance cleanup and chunked-prefill cache missing are unrelated." }, { "left": "issue:44556", "right": "issue:45092", "accept": false, - "reason": "Both concern v5 checkpoint compatibility, but one is generic reload failure and the other is remote-code meta-init incompatibility; too different to merge." + "reason": "Both are checkpoint compatibility regressions, but they affect different models and different v5 mechanisms." }, { "left": "issue:44492", "right": "issue:45850", "accept": false, - "reason": "A typo in cache strategies is not the same as a broad v5 breakage complaint." + "reason": "A cache-strategy typo is not the same as a general v5 breakage report." }, { "left": "issue:42371", "right": "issue:44492", "accept": false, - "reason": "TF32 API settings and cache strategy typo are unrelated." + "reason": "TF32 API guidance and cache strategy typo are unrelated." }, { "left": "issue:43572", "right": "issue:45850", "accept": false, - "reason": "Missing pad_token_idx in a specific config is not the same as a generic 'too many models break' report." + "reason": "StableLmConfig pad_token_idx regression is specific and not the same as a vague broad breakage report." }, { "left": "issue:44704", "right": "issue:44908", "accept": false, - "reason": "Both mention kwargs, but they affect different APIs and code paths; not a single fix." + "reason": "AutoProcessor kwargs forwarding and inverse_sqrt scheduler kwargs are different functions and bugs." }, { "left": "issue:42890", "right": "issue:44908", "accept": false, - "reason": "A flaky integration test and scheduler kwargs handling are unrelated." + "reason": "A flaky test seeding issue is unrelated to scheduler kwargs handling." }, { "left": "issue:45278", "right": "issue:45850", "accept": false, - "reason": "Both are broad v5 complaints, but one is specifically about many import errors after upgrade and the other is an unspecific generalization; not the same concrete bug." + "reason": "Both are broad complaints about v5 regressions, but they are not the same concrete bug report." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Slow imports and broken lazy loading may overlap thematically, but there is no evidence they stem from the same bug." } ] }, @@ -14596,6 +14604,7 @@ "issue:43408", "issue:43502", "issue:43519", + "issue:43525", "issue:43576", "issue:43606", "issue:43618", @@ -14604,7 +14613,6 @@ "issue:43874", "issue:43976", "issue:44077", - "issue:44246", "issue:44261", "issue:44263", "issue:44273", @@ -14635,7 +14643,6 @@ ], "soft_pairs": [ "issue:43408|issue:44315", - "issue:44246|issue:44273", "issue:43976|issue:44393", "issue:43352|issue:44393", "issue:43576|issue:44908", @@ -14649,8 +14656,8 @@ "issue:44393|issue:45478", "issue:44077|issue:44861", "issue:43827|issue:44908", - "issue:44315|issue:44829", "issue:31795|issue:33453", + "issue:44315|issue:44829", "issue:41669|issue:44908", "issue:43576|issue:44273", "issue:44655|issue:44861", @@ -14679,312 +14686,313 @@ "issue:44492|issue:44998", "issue:44998|issue:45850", "issue:43618|issue:44998", - "issue:44998|issue:45341" + "issue:44998|issue:45341", + "issue:43525|issue:44861" ], - "prepared_review_unit_hash": "e22cb91f06bebdc16a239008f171efe98eb99791a25c91e92b7591df27a670ee", + "prepared_review_unit_hash": "ba1a4d3ecc1451cc45a577328a41faf5102d85b1c2d3dc3fd1d1849cb6cb6e23", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 46, - "serialized_chars": 40805, - "estimated_input_tokens": 10202, - "estimated_eval_tokens": 20660 + "serialized_chars": 40832, + "estimated_input_tokens": 10208, + "estimated_eval_tokens": 20672 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose bundle of Transformers v5/v5.4 regression reports and unrelated model-specific bugs. The soft matches are generally too broad or too far apart to treat as duplicates, so I would keep them separate.", - "confidence": 0.28, - "canonical_issue_reason": "If forced to pick one issue, issue:45850 is the broadest umbrella complaint about widespread v5 breakage, but the cluster is too heterogeneous for a strong canonical duplicate.", + "summary": "This cluster is mostly a loose collection of closed Transformer v5 regression reports, tokenizer/model-loading issues, and a few model-specific bugs. I don\u2019t see any strong duplicate pairs; the soft matches are mostly broad topical similarity rather than the same underlying defect.", + "confidence": 0.76, + "canonical_issue_reason": "Issue 45850 is the broadest umbrella report in the set: it captures the general theme of Transformers v5 breaking multiple models, which makes it the best top-level canonical bucket for this otherwise fragmented cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45850 is the best cluster anchor only because it is the broadest and most generic report; it is not a clean duplicate target for the other issues.", + "best_issue_reason": "Issue 45850 is the most suitable global representative because it is broad enough to subsume many of the v5 regression-style reports, even though it is less concrete than some of the model-specific issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43408", "right": "issue:44315", "accept": false, - "reason": "Different bugs: model-type warning vs Liger Kernel application." - }, - { - "left": "issue:44246", - "right": "issue:44273", - "accept": false, - "reason": "Import slowdown and lazy-loading failure are not the same issue." + "reason": "Both mention model creation/configuration, but they are different problems: a SAM3 model-type mismatch warning vs. Liger Kernel application with model_init." }, { "left": "issue:43976", "right": "issue:44393", "accept": false, - "reason": "Python version compatibility vs Qwen3-VL output bug." + "reason": "Unrelated failures: Python-version compatibility vs. Qwen3-VL 2D bounding-box output behavior." }, { "left": "issue:43352", "right": "issue:44393", "accept": false, - "reason": "Flash Attention support error vs Qwen3-VL hallucination/output issue." + "reason": "Different model families and different failure modes; Flash Attention 2.0 support is not the same as Qwen3-VL output issues." }, { "left": "issue:43576", "right": "issue:44908", "accept": false, - "reason": "CLI env command breakage is unrelated to scheduler kwargs." + "reason": "CLI/env-command breakage and scheduler kwargs handling are unrelated code paths." }, { "left": "issue:44297", "right": "issue:45230", "accept": false, - "reason": "Tokenizer save_pretrained mismatch vs generic bug report." + "reason": "One is a tokenizer_class metadata mismatch on save_pretrained; the other is only a generic bug report title." }, { "left": "issue:44908", "right": "issue:45071", "accept": false, - "reason": "Scheduler kwargs bug vs PretrainedConfig type checking regression." + "reason": "Scheduler kwargs bug and PretrainedConfig type-checking regression are different issues." }, { "left": "issue:44295", "right": "issue:45230", "accept": false, - "reason": "position_ids buffer error vs generic placeholder bug report." + "reason": "Position_ids buffer access and a generic bug report are not enough to establish a duplicate." }, { "left": "issue:41669", "right": "issue:44351", "accept": false, - "reason": "Import-star cleanup/import cost vs HybridCache import failure." + "reason": "Import * cleanup/performance and missing HybridCache import are different defects." }, { "left": "issue:41669", "right": "issue:45341", "accept": false, - "reason": "Import-star cleanup/import cost vs testing_utils typo." + "reason": "Broad import cleanup issue vs. a small testing_utils bug; not the same change or bug." }, { "left": "issue:44492", "right": "issue:44945", "accept": false, - "reason": "Cache strategy typo vs pipeline-parallelism output bug." + "reason": "Cache-strategy typo and pipeline-parallelism output corruption are unrelated." }, { "left": "issue:43874", "right": "issue:45850", "accept": false, - "reason": "GLM46V patch-count AttributeError vs broad v5 complaints." + "reason": "A specific GLM46V image-processor method is missing; 45850 is just a broad complaint about many models breaking." }, { "left": "issue:44393", "right": "issue:45478", "accept": false, - "reason": "Different Qwen3-VL vs Qwen3.5 MoE from_pretrained failures." + "reason": "Different Qwen model variants and different errors; no shared concrete code path is evident." }, { "left": "issue:44077", "right": "issue:44861", "accept": false, - "reason": "post_init policy change vs tied-weights AttributeError." + "reason": "Optional post_init enforcement and tied-weight-key list handling are separate model-init validations." }, { "left": "issue:43827", "right": "issue:44908", "accept": false, - "reason": "Docs referencing pipeline() vs scheduler kwarg handling." + "reason": "Docs still referencing pipeline() and scheduler kwargs handling are unrelated." }, { - "left": "issue:44315", - "right": "issue:44829", + "left": "issue:31795", + "right": "issue:33453", "accept": false, - "reason": "Liger Kernel model_init issue vs flash_attention_3 training degeneration." + "reason": "Documentation confusion for model.forward is unrelated to tokenizer loading regression." }, { - "left": "issue:31795", - "right": "issue:33453", + "left": "issue:44315", + "right": "issue:44829", "accept": false, - "reason": "Confusing docs vs tokenizer loading regression." + "reason": "Liger Kernel application with model_init is unrelated to flash_attention_3 causing degenerate training." }, { "left": "issue:41669", "right": "issue:44908", "accept": false, - "reason": "Import performance cleanup vs scheduler timescale bug." + "reason": "Import cleanup/performance issue and scheduler kwargs bug do not share the same underlying defect." }, { "left": "issue:43576", "right": "issue:44273", "accept": false, - "reason": "Broken env command vs lazy loading problem." + "reason": "Broken env command and lazy-loading behavior are unrelated." }, { "left": "issue:44655", "right": "issue:44861", "accept": false, - "reason": "Pipeline save_pretrained vs tied weight key crash." + "reason": "Saving Pipeline objects and tied-weight-key handling are different save/load paths." }, { "left": "issue:44360", "right": "issue:45468", "accept": false, - "reason": "DSA indexer ReLU bug vs Gemma-4 audio positional encoding." + "reason": "DSA indexer ReLU discussion and Gemma-4 audio positional encoding are unrelated model internals." }, { "left": "issue:45230", "right": "issue:45310", "accept": false, - "reason": "Generic bug report vs Qwen3.5 MoE from_pretrained error." + "reason": "A generic bug report title is not enough to treat it as the same as the specific Qwen3.5 MoE from_pretrained error." }, { "left": "issue:44273", "right": "issue:44908", "accept": false, - "reason": "Lazy loading bug vs inverse_sqrt scheduler kwargs." + "reason": "Lazy loading and inverse_sqrt scheduler kwargs are unrelated." }, { "left": "issue:43519", "right": "issue:44485", "accept": false, - "reason": "Timestamp calculation bug vs GLM-5 RoPE implementation." + "reason": "Qwen3VL timestamp calculation and GLM-5 RoPE implementation are different multimodal/math bugs." }, { "left": "issue:43519", "right": "issue:44908", "accept": false, - "reason": "Timestamp calculation bug vs scheduler kwargs regression." + "reason": "Timestamp calculation in a processor is unrelated to scheduler kwargs handling." }, { "left": "issue:41669", "right": "issue:43704", "accept": false, - "reason": "Import performance cleanup vs VRAM leak in dataloader threads." + "reason": "Import * usage and VRAM leak under multiple dataloader threads are unrelated." }, { "left": "issue:31795", "right": "issue:34689", "accept": false, - "reason": "Docs confusion vs Llama 3.2 model-loading breakage." + "reason": "Model.forward docs confusion and Llama 3.2 Vision loading regression do not describe the same bug." }, { "left": "issue:31795", "right": "issue:37428", "accept": false, - "reason": "Docs confusion vs flash_attention import error." + "reason": "Documentation confusion is unrelated to the flash_attention_utils import error." }, { "left": "issue:44315", "right": "issue:45092", "accept": false, - "reason": "Liger Kernel model_init issue vs remote-code/meta init incompatibility." + "reason": "Both involve model instantiation, but one is Liger Kernel/model_init and the other is remote-code checkpoint/meta-init compatibility." }, { "left": "issue:44485", "right": "issue:45468", "accept": false, - "reason": "GLM-5 RoPE implementation vs Gemma-4 audio encoding bug." + "reason": "Different model families and different algorithmic areas; no duplicate evidence." }, { "left": "issue:43704", "right": "issue:44485", "accept": false, - "reason": "VRAM leak vs GLM-5 RoPE bug." + "reason": "VRAM leak in dataloader threads is unrelated to GLM-5 RoPE behavior." }, { "left": "issue:42371", "right": "issue:43704", "accept": false, - "reason": "TF32 settings warning vs VRAM leak." + "reason": "TF32 API migration and VRAM leak are unrelated." }, { "left": "issue:43502", "right": "issue:43519", "accept": false, - "reason": "local_files_only network request bug vs timestamp calculation." + "reason": "local_files_only network leakage and timestamp calculation are different bugs." }, { "left": "issue:43519", "right": "issue:44623", "accept": false, - "reason": "Timestamp bug vs processor.save_pretrained missing files." + "reason": "Processor timestamp math and processor.save_pretrained missing files are different save/process flows." }, { "left": "issue:43519", "right": "issue:43704", "accept": false, - "reason": "Timestamp bug vs VRAM leak." + "reason": "Timestamp calculation bug and VRAM leak are unrelated." }, { "left": "issue:44263", "right": "issue:45468", "accept": false, - "reason": "torch.split return values bug vs Gemma-4 audio encoding." + "reason": "torch.split return handling in GlmMoeDsaIndexer is unrelated to Gemma-4 audio positional encoding." }, { "left": "issue:43606", "right": "issue:45468", "accept": false, - "reason": "CPU offload device mismatch vs Gemma-4 audio encoding." + "reason": "CPU offload device mismatch and Gemma-4 audio positional encoding are unrelated." }, { "left": "issue:44261", "right": "issue:45468", "accept": false, - "reason": "MLA q_a_layernorm precision bug vs Gemma-4 audio encoding." + "reason": "MLA q_a_layernorm epsilon precision discussion does not match Gemma-4 audio positional encoding." }, { "left": "issue:43704", "right": "issue:45341", "accept": false, - "reason": "VRAM leak vs testing_utils bug." + "reason": "VRAM leak under multithreaded dataloaders and a testing_utils bug are not the same issue." }, { "left": "issue:42913", "right": "issue:43066", "accept": false, - "reason": "Tokenizer v4/v5 behavior differences vs wrong decoder type." + "reason": "Broad v4-to-v5 tokenizer behavior change and a specific wrong tokenizer decoder type are related by theme but not clearly the same underlying bug." }, { "left": "issue:44908", "right": "issue:44998", "accept": false, - "reason": "Scheduler bug vs unrelated placeholder issue." + "reason": "Inverse_sqrt scheduler kwargs and an irrelevant 'Unemployment' issue are not comparable." }, { "left": "issue:44861", "right": "issue:45290", "accept": false, - "reason": "tied weight crash vs chat template tool-call crash." + "reason": "Tied-weight-key AttributeError and apply_chat_template crashes on tool-call assistant messages are different code paths." }, { "left": "issue:44492", "right": "issue:44998", "accept": false, - "reason": "Cache strategy typo vs unrelated placeholder issue." + "reason": "Cache-strategy typo and unrelated placeholder issue are not duplicates." }, { "left": "issue:44998", "right": "issue:45850", "accept": false, - "reason": "Placeholder issue vs broad v5 breakage complaint." + "reason": "The placeholder 'Unemployment' issue has no substantive relation to broad v5 model breakage." }, { "left": "issue:43618", "right": "issue:44998", "accept": false, - "reason": "CLIPOutput attentions regression vs unrelated placeholder issue." + "reason": "CLIPOutput attentions regression and the placeholder issue are unrelated." }, { "left": "issue:44998", "right": "issue:45341", "accept": false, - "reason": "Unrelated placeholder issue vs testing_utils bug." + "reason": "The placeholder issue is unrelated to the testing_utils bug." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Llama4Config pad_token_id AttributeError and tied-weight-key list handling are separate configuration bugs." } ] }, @@ -15015,7 +15023,6 @@ "issue:36296", "issue:38617", "issue:43298", - "issue:43525", "issue:43572", "issue:44230", "issue:44623", @@ -15023,7 +15030,6 @@ "issue:45310" ], "soft_pairs": [ - "issue:43525|issue:44861", "issue:43572|issue:44861", "issue:31515|issue:38617", "issue:44230|issue:44623", @@ -15034,94 +15040,88 @@ "issue:31515|issue:33453", "issue:43298|issue:44230" ], - "prepared_review_unit_hash": "8c3494c53747e48fd43b20f59b9010e5c96fe086ede2eb6e89a02b62c8fca242", + "prepared_review_unit_hash": "a519aefdec659341cd4b0db535d79bb785e47f909078d7b5647dec77d6c8ed12", "component_budget": { "node_count": 399, "item_count": 399, "soft_pair_count": 1495, - "serialized_chars": 770929, - "estimated_input_tokens": 192733, - "estimated_eval_tokens": 385722 + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 }, "budget": { - "node_count": 13, - "item_count": 13, - "soft_pair_count": 10, - "serialized_chars": 10348, - "estimated_input_tokens": 2587, - "estimated_eval_tokens": 5430 + "node_count": 12, + "item_count": 12, + "soft_pair_count": 9, + "serialized_chars": 9480, + "estimated_input_tokens": 2370, + "estimated_eval_tokens": 4996 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a set of superficially similar but unrelated issues. The only repeatedly linked item is the generic from_pretrained/loading-speed report, but the soft pairs do not look like true duplicates.", - "confidence": 0.81, - "canonical_issue_reason": "Issue 31515 is the broadest and most central item in the similarity graph: it is the only generic loading/from_pretrained report and appears in the most soft-pair matches, even though those matches are not true duplicates.", + "summary": "This cluster is heterogeneous: the issues cover MLflow reporting, checkpoint/tokenizer loading, docs, tensor-parallel training, import errors, audio batching, fp8 support, processor saving, and tied-weight handling. I don\u2019t see a single underlying bug/change that ties them together, so there is no true duplicate anchor here.", + "confidence": 0.96, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "31515 is the best representative issue because it is the most general loading-related report in the set; the other issues are narrowly about different bugs, models, or documentation.", + "best_issue_reason": "issue #45310 is the strongest standalone bug report in the set: it names a concrete version regression (transformers>=5.4.0), a specific model family (Qwen3.5 MoE), and a reproducible from_pretrained failure.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43525", - "right": "issue:44861", - "accept": false, - "reason": "Different failures: Llama4Config missing pad_token_id vs _get_tied_weight_keys crashing on a list-valued _tied_weights_keys. Not the same code path." - }, { "left": "issue:43572", "right": "issue:44861", "accept": false, - "reason": "Different problems: missing pad_token_idx in StableLmConfig vs a tied-weights helper AttributeError. No shared underlying bug." + "reason": "Different bugs: StableLmConfig pad_token_idx regression vs AttributeError in _get_tied_weight_keys. No shared code path or symptom." }, { "left": "issue:31515", "right": "issue:38617", "accept": false, - "reason": "31515 is a slow checkpoint load issue; 38617 is an ImportError for layer_type_validation. Unrelated symptoms and code paths." + "reason": "Checkpoint loading slowness and an ImportError for layer_type_validation are unrelated issues." }, { "left": "issue:44230", "right": "issue:44623", "accept": false, - "reason": "fp8 inference support for Qwen3/VL vs processor.save_pretrained omitting files. Different functionality." + "reason": "fp8 inference support for Qwen models is a feature gap, while processor.save_pretrained missing files is a serialization bug." }, { "left": "issue:30333", "right": "issue:31515", "accept": false, - "reason": "MLFlow job-status reporting on failed training vs checkpoint loading slowness. Different subsystems and bugs." + "reason": "MLflow job-status reporting after failed training is unrelated to slow from_pretrained checkpoint loading." }, { "left": "issue:31515", "right": "issue:36296", "accept": false, - "reason": "from_pretrained load performance vs tensor-parallel training correctness. Not the same bug." + "reason": "These concern different subsystems and failure modes: pretrained checkpoint load performance vs tensor-parallel training correctness." }, { "left": "issue:31515", "right": "issue:31795", "accept": false, - "reason": "Loading performance issue vs documentation confusion about model.forward arguments. Not a duplicate." + "reason": "Loading performance and documentation confusion are not the same underlying bug." }, { "left": "issue:44230", "right": "issue:45310", "accept": false, - "reason": "Both touch Qwen3.5 MoE-related work, but one is fp8 inference support and the other is a from_pretrained error; too different to merge as duplicates." + "reason": "Both mention Qwen3.5/MoE, but one is fp8 support for inference and the other is a from_pretrained regression; the concrete bugs differ." }, { "left": "issue:31515", "right": "issue:33453", "accept": false, - "reason": "Checkpoint loading slowness vs tokenizer loading regression. Similar theme, but different concrete failures." + "reason": "Slow checkpoint loading and tokenizer loading regression are distinct problems." }, { "left": "issue:43298", "right": "issue:44230", "accept": false, - "reason": "Continuous batching audio-model support vs fp8 Qwen3/VL support. Different model capability gaps." + "reason": "Continuous batching not supporting audio models is unrelated to fp8 Qwen model inference support." } ] }, @@ -15153,7 +15153,7 @@ "issue:13244|issue:16618", "issue:15980|issue:16618" ], - "prepared_review_unit_hash": "37eb65c2224b21a7cf81fa2559d493c677fe001ac7f6a03188175c98a339e2ea", + "prepared_review_unit_hash": "2e82dc6d54a8e348c9dc5fb5e46180b2c8ebcf0b48f4de29a93bcbcee8d342e5", "component_budget": { "node_count": 3, "item_count": 3, @@ -15175,24 +15175,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These three issues are all about model/tokenizer loading behavior in a broad sense, but they describe different failures: a TAPAS tokenization mismatch, an auth-related download/error-message problem for private models, and a generic inability to load bert-base-uncased. They do not look like duplicates of the same underlying bug.", - "confidence": 0.89, - "canonical_issue_reason": "Issue #16618 is the most broadly phrased and central to the cluster theme ('Can't load the model for bert-base-uncased'), so it is the best representative issue if one must be chosen.", + "summary": "These three issues do not appear to be duplicates as a cluster: one is about Tapas tokenization, and the other two are separate model-loading/authentication complaints. No soft edge is strong enough to merge.", + "confidence": 0.87, + "canonical_issue_reason": "Issue 16618 is the broadest and most central model-loading report in the set, while 15980 is narrowly about private-model login/error messaging and 13244 is unrelated tokenization.", "canonical_pr_reason": null, - "best_issue_reason": "#16618 is the best fit for the cluster because it is the most general model-loading failure among the three, whereas #13244 is specifically about TAPAS tokenization and #15980 is specifically about private-model download/auth behavior.", + "best_issue_reason": "Issue 16618 is the best representative of the cluster\u2019s model-loading side because it is the most general user-facing failure report among the candidates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:13244", "right": "issue:16618", "accept": false, - "reason": "Different underlying problems: TAPAS tokenization differing from TensorFlow code vs. a generic model-loading failure for bert-base-uncased. Same broad area, but not the same bug." + "reason": "Different problem area: Tapas tokenization mismatch versus failure to load bert-base-uncased." }, { "left": "issue:15980", "right": "issue:16618", "accept": false, - "reason": "#15980 is about an incorrect error message when downloading a private model while logged out, which is an auth/download UX issue, not the same as the model-loading failure described in #16618." + "reason": "Related broadly to model downloads, but one is about private-model login/error messaging and the other is a generic load failure for a public model; not the same concrete bug." } ] }, @@ -15222,7 +15222,7 @@ "soft_pairs": [ "issue:26350|issue:33260" ], - "prepared_review_unit_hash": "3a96138591fd5cb18145c44df4fc9faa0b39ef1eeea9d32e94b8b65cc66795c3", + "prepared_review_unit_hash": "121743d6490668e44bab32e9eb9dbbb8fa72664684a10e99cf4764e18986d96d", "component_budget": { "node_count": 2, "item_count": 2, @@ -15244,18 +15244,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated feature-request issues: one for Flash Attention 2 support on more architectures, the other for GGUF support on more architectures. They share a loose \u201ccommunity contribution\u201d framing and architecture-related wording, but not the same underlying change or bug.", - "confidence": 0.98, + "summary": "These are two separate feature requests with only superficial similarity: one asks for Flash Attention 2 support on more architectures, the other for GGUF support on more architectures. They do not appear to describe the same underlying bug or change.", + "confidence": 0.96, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:26350 is the more established and discussion-heavy request, but it is not a duplicate of issue:33260.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:26350", "right": "issue:33260", "accept": false, - "reason": "Reject: the requests target different features and code paths (Flash Attention 2 vs GGUF support). Similar wording about supporting more architectures is too broad to treat as the same issue." + "reason": "Both are community contribution requests about broad architecture support, but they target different features and code paths (Flash Attention 2 vs GGUF). They are not the same underlying issue." } ] }, @@ -15292,12 +15292,12 @@ "issue:43834|issue:44365", "issue:44365|issue:44370" ], - "prepared_review_unit_hash": "829c0fd5da6cdccf98dd33f13991bb3ba6fe76e02bc5273ec2d5ede5ada3071d", + "prepared_review_unit_hash": "f5ca49f3633697e29c1b1e7a22157f7e05c0802598201785f03dc211be87f4b5", "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 }, @@ -15305,7 +15305,7 @@ "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4455, + "serialized_chars": 4453, "estimated_input_tokens": 1114, "estimated_eval_tokens": 2484 }, @@ -15314,48 +15314,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are language-specific documentation translation requests that look superficially similar because they share a template, but they are not duplicates of the same underlying task. The Turkish issue is concrete; the others appear to be separate per-language instances of the same template.", - "confidence": 0.96, - "canonical_issue_reason": null, + "summary": "These are all translation-doc request issues with very similar template text, but they appear to target different language requests rather than one identical underlying change. The Turkish issue is the most concrete and active representative.", + "confidence": 0.71, + "canonical_issue_reason": "Issue 27088 is the only concrete language-specific request ([i18n-TR] Turkish) and has the most discussion activity, so it is the best representative of this cluster.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 27088 is the only concrete, language-specific request ([i18n-TR]); the others are generic placeholder/template instances and should not be treated as duplicates of it.", + "best_issue_reason": "27088 is the strongest issue candidate because it is specific, oldest, and has the most engagement; the others look like template-like variants rather than a clearer canonical target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27088", "right": "issue:44370", "accept": false, - "reason": "Different translation targets; same template wording, but not the same underlying work item." + "reason": "Same translation-request theme, but the targets are not concretely the same; this is not enough to treat them as one underlying issue." }, { "left": "issue:27088", "right": "issue:44365", "accept": false, - "reason": "Different language translation requests. The similarity is from the shared template, not duplicate content." + "reason": "The Turkish request and the generic placeholder request do not clearly refer to the same language change." }, { "left": "issue:27088", "right": "issue:43834", "accept": false, - "reason": "Turkish-specific request vs. a separate language placeholder instance; these are distinct issues." + "reason": "Similar template wording, but no evidence they are the same exact translation request or language target." }, { "left": "issue:43834", "right": "issue:44370", "accept": false, - "reason": "Both are generic i18n translation template issues, but they likely correspond to different languages and separate tasks." + "reason": "Very similar template text, but still only generic translation-request similarity; no concrete shared target to justify duplication." }, { "left": "issue:43834", "right": "issue:44365", "accept": false, - "reason": "High textual similarity comes from the shared template; no evidence they track the same language translation." + "reason": "These look like near-template issues, but the available evidence does not show the same specific request." }, { "left": "issue:44365", "right": "issue:44370", "accept": false, - "reason": "These appear to be separate per-language translation requests, not a single duplicate bug or change." + "reason": "Close textual match, but not enough evidence that they are the same underlying translation issue." } ] }, @@ -15432,244 +15432,244 @@ "issue:37015|issue:45027", "issue:43837|issue:45177" ], - "prepared_review_unit_hash": "e5c541b9ae2d9dc3a50ec8a5beda37cb9451229e7d271164f70d3fea78b93d0c", + "prepared_review_unit_hash": "38cf906479d70fe7c9213c436fabdb3fbdbb274ae9e7aa9a7031b7a4fce5f0f8", "component_budget": { "node_count": 15, "item_count": 15, "soft_pair_count": 35, - "serialized_chars": 20707, - "estimated_input_tokens": 5177, - "estimated_eval_tokens": 10610 + "serialized_chars": 20711, + "estimated_input_tokens": 5178, + "estimated_eval_tokens": 10612 }, "budget": { "node_count": 15, "item_count": 15, "soft_pair_count": 35, - "serialized_chars": 20707, - "estimated_input_tokens": 5177, - "estimated_eval_tokens": 10610 + "serialized_chars": 20711, + "estimated_input_tokens": 5178, + "estimated_eval_tokens": 10612 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These items are mostly broad model-add/support requests, but they target different model families and often different modalities (vision, text, ASR/TTS, OCR, sparse transformers). The similarity is largely lexical, so there is no true duplicate cluster here.", - "confidence": 0.96, - "canonical_issue_reason": "No single canonical duplicate issue: the tickets describe distinct model requests and code paths, not one underlying bug or change.", + "summary": "This cluster is a loose set of model-support request issues that only share a generic \"add model\" pattern. The titles span unrelated architectures and modalities (text, audio, vision, detection, OCR, sparse transformers), so they should not be treated as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "If a single issue must be chosen, issue #36817 is the broadest and most template-like open request ('Add EuroBert Model To Config'), but it is still only one specific model request and does not truly cover the rest of the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42738 is the closest umbrella-style representative because it is the most general architecture request, but it still does not duplicate the others.", + "best_issue_reason": "#36817 is the best representative only in the sense that it is the most generic/open config-addition request; however, the cluster is too heterogeneous for any issue to serve as a real duplicate umbrella.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36817", "right": "issue:41211", "accept": false, - "reason": "EuroBert and DEIMv2 are different model requests with no shared underlying implementation bug." + "reason": "EuroBert vs DEIMv2 are different model families with different support paths." }, { "left": "issue:36817", "right": "issue:43671", "accept": false, - "reason": "EuroBert model config support and Qwen3-TTS support are unrelated model families/modalities." + "reason": "EuroBert text model request is unrelated to Qwen3-TTS speech generation support." }, { "left": "issue:36817", "right": "issue:36879", "accept": false, - "reason": "EuroBert and RF-DETR are different models with different architecture targets." + "reason": "EuroBert NLP config addition and RF-DETR object detection are different underlying changes." }, { "left": "issue:36879", "right": "issue:41211", "accept": false, - "reason": "RF-DETR and DEIMv2 are separate model additions, not the same change." + "reason": "RF-DETR and DEIMv2 are distinct detection models; not the same concrete fix." }, { "left": "issue:36879", "right": "issue:43909", "accept": false, - "reason": "RF-DETR is a detection model, while LFM2.5 Audio is an audio model request." + "reason": "RF-DETR detection support is unrelated to LFM2.5 Audio model support." }, { "left": "issue:36879", "right": "issue:44121", "accept": false, - "reason": "RF-DETR and the weight-sparse transformer request concern different architectures and code paths." + "reason": "RF-DETR model support and weight-sparse transformer support are unrelated code paths." }, { "left": "issue:36879", "right": "issue:43837", "accept": false, - "reason": "RF-DETR and Qwen3-ASR are different modalities and model integrations." + "reason": "RF-DETR vision detection and Qwen3-ASR speech recognition are different tasks." }, { "left": "issue:36879", "right": "issue:43671", "accept": false, - "reason": "RF-DETR and Qwen3-TTS are unrelated model support requests." + "reason": "RF-DETR and Qwen3-TTS are different modalities and model integrations." }, { "left": "issue:37015", "right": "issue:43837", "accept": false, - "reason": "NeoBERT and Qwen3-ASR are different model families; no shared duplicate bug is evident." + "reason": "NeoBERT language-model support is unrelated to Qwen3-ASR audio transcription support." }, { "left": "issue:37015", "right": "issue:44121", "accept": false, - "reason": "NeoBERT support and weight-sparse transformer support are distinct architecture requests." + "reason": "NeoBERT and the sparse-transformer request target different architectures and fixes." }, { "left": "issue:37015", "right": "issue:43909", "accept": false, - "reason": "NeoBERT and LFM2.5 Audio are unrelated model additions." + "reason": "NeoBERT text model support is unrelated to LFM2.5 Audio support." }, { "left": "issue:36879", "right": "issue:37015", "accept": false, - "reason": "RF-DETR and NeoBERT are different model types and would not be merged as one fix." + "reason": "RF-DETR and NeoBERT are different model families, not one bug/change." }, { "left": "issue:37015", "right": "issue:42503", "accept": false, - "reason": "NeoBERT and ModernVBERT are separate model requests, not the same underlying issue." + "reason": "NeoBERT and ModernVBERT are distinct model additions even if both are BERT-like." }, { "left": "issue:41115", "right": "issue:43837", "accept": false, - "reason": "MiniCPM3 model architecture support and Qwen3-ASR support are different code-path changes." + "reason": "MiniCPM3 architecture support is unrelated to Qwen3-ASR support." }, { "left": "issue:41115", "right": "issue:44121", "accept": false, - "reason": "MiniCPM3 and the weight-sparse transformer request are not the same model integration." + "reason": "MiniCPM3 model architecture work does not match sparse-transformer support." }, { "left": "issue:41115", "right": "issue:42503", "accept": false, - "reason": "MiniCPM3 and ModernVBERT are distinct model requests with different scope." + "reason": "MiniCPM3 and ModernVBERT are different model integrations." }, { "left": "issue:41115", "right": "issue:45027", "accept": false, - "reason": "MiniCPM3 and Voxtral-4B-TTS-2603 are different model modalities and targets." + "reason": "MiniCPM3 and Voxtral TTS support are different tasks/code paths." }, { "left": "issue:41211", "right": "issue:43671", "accept": false, - "reason": "DEIMv2 and Qwen3-TTS are unrelated model additions." + "reason": "DEIMv2 and Qwen3-TTS are unrelated model requests." }, { "left": "issue:42503", "right": "issue:43837", "accept": false, - "reason": "ModernVBERT and Qwen3-ASR are separate model support tickets." + "reason": "ModernVBERT and Qwen3-ASR are different models and modalities." }, { "left": "issue:42503", "right": "issue:43671", "accept": false, - "reason": "ModernVBERT and Qwen3-TTS are different model families/modalities." + "reason": "ModernVBERT text model support is unrelated to Qwen3-TTS audio generation." }, { "left": "issue:42503", "right": "issue:42738", "accept": false, - "reason": "ModernVBERT models and BERT-like models with RoPE overlap only broadly on naming, not on the same concrete change." + "reason": "ModernVBERT and BERT-like models with RoPE are not the same concrete request." }, { "left": "issue:42738", "right": "issue:45027", "accept": false, - "reason": "BERT-like models with RoPE and Voxtral TTS are different feature requests." + "reason": "BERT-like RoPE support and Voxtral TTS support are unrelated changes." }, { "left": "issue:43671", "right": "issue:43837", "accept": false, - "reason": "Qwen3-TTS and Qwen3-ASR are closely named but still distinct support requests for different endpoints." + "reason": "Qwen3-TTS and Qwen3-ASR share a name but fix different speech tasks and code paths." }, { "left": "issue:43671", "right": "issue:45027", "accept": false, - "reason": "Qwen3-TTS and Voxtral TTS are different TTS model integrations." + "reason": "Two different TTS model integrations; not enough to treat as one duplicate." }, { "left": "issue:43671", "right": "issue:45177", "accept": false, - "reason": "Qwen3-TTS and DeepSeek-OCR2 are unrelated modalities and code paths." + "reason": "Qwen3-TTS and DeepSeek-OCR2 are different model domains." }, { "left": "issue:43671", "right": "issue:44121", "accept": false, - "reason": "Qwen3-TTS and the sparse transformer request are different architectures." + "reason": "Speech generation support is unrelated to the sparse-transformer request." }, { "left": "issue:43671", "right": "issue:43909", "accept": false, - "reason": "Qwen3-TTS and LFM2.5 Audio are separate audio-model requests." + "reason": "Qwen3-TTS and LFM2.5 Audio are distinct audio model integrations." }, { "left": "issue:43837", "right": "issue:43909", "accept": false, - "reason": "Qwen3-ASR and LFM2.5 Audio are different audio-model support requests." + "reason": "ASR vs audio-model support are different concrete changes." }, { "left": "issue:43837", "right": "issue:45322", "accept": false, - "reason": "Qwen3-ASR and EUPE are unrelated model requests." + "reason": "Qwen3-ASR and EUPE are different model classes and implementations." }, { "left": "issue:43837", "right": "issue:44121", "accept": false, - "reason": "Qwen3-ASR and the sparse transformer request do not share the same underlying code change." + "reason": "Speech recognition support is unrelated to the sparse-transformer model request." }, { "left": "issue:43909", "right": "issue:44205", "accept": false, - "reason": "LFM2.5 Audio and SAM3-LiteText are different model additions across different modalities." + "reason": "LFM2.5 Audio and SAM3-LiteText are different modalities and code paths." }, { "left": "issue:44121", "right": "issue:45322", "accept": false, - "reason": "Weight-sparse transformer support and EUPE are distinct model requests." + "reason": "Weight-sparse transformer support and EUPE support are unrelated architecture requests." }, { "left": "issue:43837", "right": "issue:45027", "accept": false, - "reason": "Qwen3-ASR and Voxtral TTS are different speech model integrations." + "reason": "Qwen3-ASR and Voxtral-4B-TTS are different speech tasks." }, { "left": "issue:37015", "right": "issue:45027", "accept": false, - "reason": "NeoBERT and Voxtral TTS are unrelated model requests." + "reason": "NeoBERT text-model support is unrelated to Voxtral TTS support." }, { "left": "issue:43837", "right": "issue:45177", "accept": false, - "reason": "Qwen3-ASR and DeepSeek-OCR2 are different modalities and unrelated changes." + "reason": "ASR support and OCR model support are different underlying changes." } ] }, @@ -15710,82 +15710,82 @@ "issue:43668|issue:44031", "issue:43668|issue:43813" ], - "prepared_review_unit_hash": "6791e245b285028e81d4ec776167d14c902d22dd8bb035e6ba4364fca2880c5b", + "prepared_review_unit_hash": "72ac5eb4b655a5d95bf0045e0062b24c1045ca5c59c471d4e1041bd7cb42177c", "component_budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 8, - "serialized_chars": 6291, - "estimated_input_tokens": 1573, - "estimated_eval_tokens": 3402 + "serialized_chars": 6293, + "estimated_input_tokens": 1574, + "estimated_eval_tokens": 3404 }, "budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 8, - "serialized_chars": 6291, - "estimated_input_tokens": 1573, - "estimated_eval_tokens": 3402 + "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": "These are not duplicates: the issues span unrelated components (normalization, Mask2Former loss handling, generic loss kwargs, config typing, a typo, and tokenizer warnings). No soft edge pair looks like the same underlying bug or change.", - "confidence": 0.96, - "canonical_issue_reason": "issue:44031 is the strongest standalone representative: it describes a concrete, user-facing regression with broad impact across tokenizers and a specific failure mode.", + "summary": "These issues are not duplicates of one another; they span unrelated bugs across different models/configs, a tokenizer regression, and a typo. The soft similarity is superficial rather than indicating the same underlying fix.", + "confidence": 0.97, + "canonical_issue_reason": "No true canonical issue: the set is heterogeneous (different models, different failure modes, and one pure typo/type-hint issue), so there isn\u2019t a single underlying bug to centralize.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44031 has the clearest symptom, broadest scope, and most actionable bug report among the set.", + "best_issue_reason": "issue:43240 is the broadest functional bug and has inbound references, so it is the most representative placeholder if one must be chosen, though this cluster should not be treated as a duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "Different models and different problem types: a question about Llama4TextL2Norm semantics vs a ModernBERTConfig type-hint issue." + "reason": "Different subjects: Llama4TextL2Norm/RMS norm question vs. ModernBERTConfig type-hint issue. No shared code-path or fix." }, { "left": "issue:42936", "right": "issue:43240", "accept": false, - "reason": "Both concern loss-related behavior, but one is Mask2Former's ignore_value usage while the other is generic kwargs plumbing; not the same concrete bug." + "reason": "Both are model-related bugs, but one is about Mask2Former ignore_value handling and the other about kwargs not reaching loss computation. Distinct defects." }, { "left": "issue:42936", "right": "issue:44031", "accept": false, - "reason": "Mask2Former loss/ignore_value handling is unrelated to tokenizer regex warnings." + "reason": "Mask2Former ignore_value behavior is unrelated to a tokenizer regex warning regression." }, { "left": "issue:42936", "right": "issue:43813", "accept": false, - "reason": "Different subsystems and failure modes: Mask2Former loss behavior vs a quantization attribute typo." + "reason": "Model ignore_value handling and a typo in a quantization attribute are unrelated." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "Mask2Former model bug vs ModernBERTConfig annotation bug; no shared code path." + "reason": "Mask2Former loss/ignore_value bug vs. a ModernBERT config type annotation problem; not the same bug." }, { "left": "issue:43240", "right": "issue:43668", "accept": false, - "reason": "Generic loss kwargs forwarding is unrelated to a configuration type-hint correction." + "reason": "Passing kwargs to loss calculation is a runtime behavior issue; incorrect norm_eps typing is a separate config metadata issue." }, { "left": "issue:43668", "right": "issue:44031", "accept": false, - "reason": "ModernBERTConfig type-hint issue and tokenizer regex warning regression are different components and defects." + "reason": "ModernBERTConfig type-hint correction and a tokenizer warning regression are different problems in different areas." }, { "left": "issue:43668", "right": "issue:43813", "accept": false, - "reason": "A config type hint typo is not the same as a quantization-operation typo in a different area." + "reason": "A config type-hint bug is not the same as a typo in a quantization_operation attribute." } ] }, @@ -15815,7 +15815,7 @@ "soft_pairs": [ "issue:41084|issue:43037" ], - "prepared_review_unit_hash": "9a3c7b7e44b91d5b2f89d05cb066a7406b671d955cca8a90bd72d096f0f3177c", + "prepared_review_unit_hash": "1a3ee235d7758d2ca8b665178cadcfb1fd068bba138fa1e52b07a45689b66c0f", "component_budget": { "node_count": 2, "item_count": 2, @@ -15837,8 +15837,8 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are not duplicates: one is about block decoding behavior, the other is a feature request for DeepSeek v3.2 support. They describe different user problems and different likely code paths.", - "confidence": 0.97, + "summary": "These two issues are unrelated: one is about block decoding, while the other is a feature request for DeepSeek v3.2 support. There is no duplicate relationship here and no sensible single canonical item for the cluster.", + "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, @@ -15848,7 +15848,7 @@ "left": "issue:41084", "right": "issue:43037", "accept": false, - "reason": "Different topics and remedies: decoding logic versus model/version support. No evidence they refer to the same underlying bug or change." + "reason": "Titles and likely scopes do not match: 'Set Block Decoding' suggests a decoding/configuration bug or feature, while 'DeepSeek v3.2 support' is an LLM compatibility request. No shared target or concrete code-path overlap is evident." } ] }, @@ -15878,7 +15878,7 @@ "soft_pairs": [ "issue:43598|issue:43600" ], - "prepared_review_unit_hash": "d85e916ddb872ae6a3e035106e7114948db1046f0ec1a16ed2264fd9cf5d0dec", + "prepared_review_unit_hash": "60adfd38b83ae2ddf1b843b2a3648e16e34f2aa3557860055d88cdddf23b647f", "component_budget": { "node_count": 2, "item_count": 2, @@ -15900,18 +15900,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two closed issues are related by theme (Trainer customizability), but they describe different underlying problems: one is about `remove_unused_column` behavior, the other is about FLOPS estimation for custom trainers. They should not be clustered as duplicates.", - "confidence": 0.84, - "canonical_issue_reason": null, + "summary": "Both items are about Trainer customizability, but they describe different problems: one is about `remove_unused_column`, the other about FLOPS estimation for custom trainers. Similar subsystem, not the same underlying bug.", + "confidence": 0.96, + "canonical_issue_reason": "issue #43598 is the more concrete and actionable problem statement, centered on a specific Trainer behavior (`remove_unused_column`) affecting custom trainers.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43598 is the better representative of the cluster because it is broader and centers on the main Trainer customizability concern, whereas 43600 is a narrower, separate accounting problem.", + "best_issue_reason": "issue #43598 is the stronger anchor for the cluster because it is narrower and more clearly tied to a single code path.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43598", "right": "issue:43600", "accept": false, - "reason": "Different code-path problems: 43598 concerns `remove_unused_column` customizability in Trainer, while 43600 concerns FLOPS estimation logic for custom trainers. Related topic, but not the same bug." + "reason": "These are related only at a high level through custom Trainer behavior. One concerns column pruning customizability; the other concerns FLOPS estimation. They would not be merged as the same issue." } ] }, @@ -15941,7 +15941,7 @@ "soft_pairs": [ "issue:43684|issue:44405" ], - "prepared_review_unit_hash": "6d18c4fab9309c0b201b9174c290127fedcbc7f40ae2e6aa3f46a8249b14fd04", + "prepared_review_unit_hash": "7c42814c8935a3309a29d91343eb018190d36bbc0edcf471fbc72c6d13833df9", "component_budget": { "node_count": 2, "item_count": 2, @@ -15963,18 +15963,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are both about adding model-class support in Transformers auto-mapping, but they target different model families and different auto classes/tasks, so they are not duplicates.", - "confidence": 0.95, - "canonical_issue_reason": "Issue #43684 is the broader and clearer canonical item: it requests explicit registration for Qwen3-Omni in AutoModel and AutoModelForConditionalGeneration, which is a concrete auto-mapping gap.", + "summary": "These two issues are not duplicates: one requests broad AutoModel/AutoModelForConditionalGeneration registration for Qwen3-Omni, while the other requests AutoModelForSequenceClassification support for Qwen3.5. They concern different model families and different task heads.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue #43684 is the better representative because it is more directly phrased as a model-registration task and covers the core auto-class support problem.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43684", "right": "issue:44405", "accept": false, - "reason": "Different targets: Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration vs Qwen3.5 support for AutoModelForSequenceClassification. They are related only at a broad subsystem level, not the same underlying bug or change." + "reason": "Different underlying requests: general model registration for Qwen3-Omni versus sequence-classification support for Qwen3.5/Qwen3_5Config. They are not the same bug/change and would not plausibly be merged into one issue." } ] }, @@ -16011,77 +16011,80 @@ "issue:43845|issue:43859", "issue:43846|issue:43859" ], - "prepared_review_unit_hash": "079eaf0e3fa1ce898bd7dc2778ab6850c8dd176e848f36d6f2533b8243e72a87", + "prepared_review_unit_hash": "104a0e5a3cf03a205c228facfa7e74dc1fe6491630323f8ea30909a2428551dd", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4264, - "estimated_input_tokens": 1066, - "estimated_eval_tokens": 2388 + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4264, - "estimated_input_tokens": 1066, - "estimated_eval_tokens": 2388 + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Four very low-information closed issues, three with the identical title \"huggingface \" and one unrelated title \"Fett\". There is not enough concrete evidence to confirm any true duplicate relationship, so this looks more like noisy/spammy issue traffic than a well-defined duplicate cluster.", - "confidence": 0.9, - "canonical_issue_reason": "If a representative must be chosen, issue #43859 is the least-bad canonical issue because it has the only discussion activity and matches the repeated \"huggingface\" title. That said, the cluster lacks actionable problem details, so this is only a placeholder canonical choice.", + "summary": "Three issues share the same generic title \"huggingface\" and appear to be duplicates of each other; issue 43835 ('Fett') looks unrelated. I would keep 43845 as the canonical issue among the huggingface duplicates.", + "confidence": 0.68, + "canonical_issue_reason": "43845 is the earliest of the near-identical \"huggingface\" issues and is the best representative of that duplicate set.", "canonical_pr_reason": null, - "best_issue_reason": "#43859 is the best single issue to keep as the representative of the repeated \"huggingface\" submissions because it has slightly more activity than the others. It is not a strong canonical issue in substance, just the most representative of the set.", + "best_issue_reason": "43845 is the strongest issue candidate because it matches the duplicate title pattern and predates the other matching issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43835", "right": "issue:43845", "accept": false, - "reason": "Different titles and no body/detail evidence tying them to the same underlying problem; only weak textual similarity." + "reason": "Different titles and no evidence they describe the same underlying problem." }, { "left": "issue:43835", "right": "issue:43846", "accept": false, - "reason": "Same as above: no concrete bug description, only generic soft similarity." + "reason": "Different titles and no concrete shared bug description." }, { "left": "issue:43835", "right": "issue:43859", "accept": false, - "reason": "No shared issue target or actionable context; \"Fett\" and \"huggingface\" do not establish the same underlying issue." + "reason": "Different titles and no concrete shared bug description." }, { "left": "issue:43845", "right": "issue:43846", - "accept": false, - "reason": "Identical titles, but the issues contain no substantive description to prove they are the same bug rather than repeated/noisy submissions; insufficient evidence to merge as duplicates." + "accept": true, + "reason": "Same title, very close creation times, and no distinguishing details; they look like duplicate reports of the same issue." }, { "left": "issue:43845", "right": "issue:43859", - "accept": false, - "reason": "Same broad title text, but no concrete problem statement or target to confirm they refer to the same underlying issue." + "accept": true, + "reason": "Same title and similar sparse content strongly suggest the same underlying report." }, { "left": "issue:43846", "right": "issue:43859", - "accept": false, - "reason": "Same broad title text, but no concrete problem statement or target to confirm they refer to the same underlying issue." + "accept": true, + "reason": "Same title and similarly minimal content indicate a duplicate issue." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet: 43835 has a distinct title, while 43845/43846/43859 share the same near-identical title and are plausibly a duplicate set. The canonical choice of 43845 is conservative because it is the earliest of that trio. The soft-edge verdicts also stay within the evidence available." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -16115,88 +16118,88 @@ "issue:44572|issue:44937", "issue:44573|issue:44937" ], - "prepared_review_unit_hash": "758ee8ceb3bbbf9c3b33703eeb19793946b4073d10122e3c3c9dd2e03c727b75", + "prepared_review_unit_hash": "e22f90c55f83e5e3c40c265d1d9b89169994dd1607f7092fc9e6e6fd158be0f2", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5791, - "estimated_input_tokens": 1448, - "estimated_eval_tokens": 3152 + "serialized_chars": 5785, + "estimated_input_tokens": 1447, + "estimated_eval_tokens": 3150 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5791, - "estimated_input_tokens": 1448, - "estimated_eval_tokens": 3152 + "serialized_chars": 5785, + "estimated_input_tokens": 1447, + "estimated_eval_tokens": 3150 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All items look like low-signal spam or generic placeholder issues; there is no substantive bug report to cluster, so no duplicate relationship is justifiable.", - "confidence": 0.94, - "canonical_issue_reason": "If a representative must be chosen, issue 44371 is the earliest of the spam/placeholder issues, but none of these contain enough real content to establish a true canonical bug.", + "summary": "No convincing duplicate cluster here: the issues are mostly spam/generic placeholders, and the one specific title does not match the others on a concrete bug or change. I rejected all soft edges.", + "confidence": 0.89, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 44371 is the least bad representative only because it is earliest; it is not a meaningful canonical issue because the cluster lacks actual problem descriptions.", + "best_issue_reason": "issue:44937 is the most specific and actionable title in the set, so it is the least-bad representative if one must be chosen.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44371", "right": "issue:44572", "accept": false, - "reason": "Both are generic placeholder spam titles with no concrete bug details; identical text alone is not enough to infer the same underlying issue." + "reason": "Both are generic spam placeholders; no evidence of the same underlying issue." }, { "left": "issue:44371", "right": "issue:44573", "accept": false, - "reason": "Both appear to be spam/placeholder issues, but there is no describable problem or code-path evidence tying them together." + "reason": "Both are generic spam placeholders; no shared concrete bug or change." }, { "left": "issue:44371", "right": "issue:44418", "accept": false, - "reason": "One is spam and the other is a vague documentation suggestion; they do not describe the same issue." + "reason": "Placeholder spam vs. a vague docs suggestion; too little evidence to be duplicates." }, { "left": "issue:44371", "right": "issue:44937", "accept": false, - "reason": "The titles are unrelated: placeholder spam versus a specific quoted phrase. No common underlying bug is evident." + "reason": "Spam placeholder and a specific 'Google LLC' topic are not the same issue." }, { "left": "issue:44418", "right": "issue:44572", "accept": false, - "reason": "Generic documentation suggestion and placeholder spam are not the same underlying report." + "reason": "Vague documentation suggestion vs. spam placeholder; no concrete overlap." }, { "left": "issue:44418", "right": "issue:44573", "accept": false, - "reason": "Generic documentation suggestion and placeholder spam are not the same underlying report." + "reason": "Vague documentation suggestion vs. spam placeholder; no concrete overlap." }, { "left": "issue:44572", "right": "issue:44573", "accept": false, - "reason": "They share the same placeholder spam title, but there is no real issue description to confirm they refer to the same bug." + "reason": "Both are spam placeholders, but there is no evidence they describe the same underlying bug." }, { "left": "issue:44572", "right": "issue:44937", "accept": false, - "reason": "Placeholder spam and a specific named phrase are unrelated; no common bug or change is visible." + "reason": "Spam placeholder and a specific title about Google LLC do not match on a concrete problem." }, { "left": "issue:44573", "right": "issue:44937", "accept": false, - "reason": "Placeholder spam and a specific named phrase are unrelated; no common bug or change is visible." + "reason": "Spam placeholder and a specific title about Google LLC do not match on a concrete problem." } ] }, @@ -16226,7 +16229,7 @@ "soft_pairs": [ "issue:45145|issue:45151" ], - "prepared_review_unit_hash": "4cb91d182802b3e0562bb8c38960f2be39c4e3048b0fc0d0398653a9bf6e3cd7", + "prepared_review_unit_hash": "312c1d76aa102015b6de5527a4e60ad1d7c93ab55ca8d00e8b4982a7430ba7bc", "component_budget": { "node_count": 2, "item_count": 2, @@ -16248,27 +16251,27 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are near-identical duplicates with the same title and very similar body content; issue #45145 is the better canonical choice because it was created earlier and already has inbound references.", - "confidence": 0.98, - "canonical_issue_reason": "Issue #45145 was opened earlier and has the same title/body as #45151, so it is the best duplicate cluster anchor.", + "summary": "The two issues look like likely duplicates: they share an identical title, similar body sizes, and a soft-similarity signal. #45145 is the better canonical anchor because it has one inbound reference, but the evidence stops short of proving the bodies are effectively identical.", + "confidence": 0.83, + "canonical_issue_reason": "#45145 is the better anchor because it has the same title as #45151 and also has an inbound reference, which makes it the more established copy for deduplication.", "canonical_pr_reason": null, - "best_issue_reason": "#45145 is preferable as the representative issue because it predates #45151 and already has an inbound reference.", + "best_issue_reason": "#45145 is the strongest issue representative since it has the same title as #45151 and slightly stronger linkage via inbound references.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45145", "right": "issue:45151", "accept": true, - "reason": "Same underlying issue: identical title and highly similar content, with no distinguishing targets or filenames." + "reason": "Same title and similar body size, plus a soft-similarity signal, make these plausibly the same underlying report, though the match is not definitive." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The assessment is grounded in the packet: both issues share the same title, have similar body sizes, and the pair is marked soft-similar. Choosing #45145 as canonical because it was created earlier and has an inbound reference is conservative and supported. No unsupported claims or overreach are apparent." + "accept": false, + "feedback": "The pair looks plausibly duplicate, but the analyst is a bit too certain for the evidence provided. The packet only shows identical titles, similar body lengths, one inbound reference, and a soft-similarity signal; it does not directly establish that the bodies are \u201chighly similar,\u201d \u201ceffectively identical,\u201d or that both are definitively the same report. The canonical choice based on inbound reference is reasonable, but the confidence and duplicate wording are somewhat overstated." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -16346,7 +16349,7 @@ "soft_pairs": [ "issue:45672|issue:45674" ], - "prepared_review_unit_hash": "455c8c94d193241c4ea7bb5b237d40ed3a869babcff8fe4166293774b2a46cb5", + "prepared_review_unit_hash": "18c5850ebed6226e67fa8d1c49fe56e7e6326ace503057a887728e8ac41e3cc1", "component_budget": { "node_count": 2, "item_count": 2, @@ -16368,18 +16371,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are about different 4-bit inference failures with different root causes: one is a Gemma4/audio-module torch.finfo() TypeError on uint8 weights, the other is a BitsAndBytesConfig skip-module regression that drops the default lm_head exclusion and triggers an AssertionError. They should not be clustered as duplicates.", - "confidence": 0.95, - "canonical_issue_reason": "No single canonical issue for the pair: 45672 is a Gemma4/audio-module-specific dtype bug, while 45674 is a separate BitsAndBytesConfig regression affecting module exclusion behavior in 4-bit inference.", + "summary": "These are two separate 4-bit inference bugs. One is a Gemma4/audio-module dtype issue with `torch.finfo()` on `uint8` weights; the other is a BitsAndBytesConfig regression where `llm_int8_skip_modules` drops the default `lm_head` exclusion and triggers an assertion. They share the broad area of quantized inference but not the same underlying defect.", + "confidence": 0.97, + "canonical_issue_reason": "Issue #45674 is the better cluster representative because it describes a concrete, configuration-driven regression with a clear failure mode and likely wider applicability across models using BitsAndBytesConfig.", "canonical_pr_reason": null, - "best_issue_reason": "45674 is the better representative only in the sense that it is the clearer, more general bug report with a specific config regression and an explicit failure mode; 45672 is a narrower model/module-specific symptom and not the same underlying issue.", + "best_issue_reason": "Issue #45674 is more suitable as the primary issue for this cluster since it is a crisp, reproducible bug report about a specific regression path. Issue #45672 is a model-specific dtype failure and not the same problem.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45672", "right": "issue:45674", "accept": false, - "reason": "Different underlying bugs: torch.finfo() on uint8 weights in Gemma4 audio modules versus BitsAndBytesConfig llm_int8_skip_modules clearing the default lm_head exclusion and causing an AssertionError. Same broad 4-bit area, but not the same fix." + "reason": "Different code paths and failure mechanisms: Gemma4/audio-module `torch.finfo()` on `uint8` vs. BitsAndBytesConfig `llm_int8_skip_modules` removing default `lm_head` exclusion. Shared 4-bit inference context is too broad to treat as a duplicate." } ] }, @@ -16409,30 +16412,30 @@ "soft_pairs": [ "issue:45762|issue:45763" ], - "prepared_review_unit_hash": "4140fd384c3350dce98c8b69d3703a6ec331a6dcb67c8bbb5f3e99aca5431108", + "prepared_review_unit_hash": "a68e0224ec8852e86540007d82f8bf05fa84444146295384145b97c018669537", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1476, - "estimated_input_tokens": 369, - "estimated_eval_tokens": 994 + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1476, - "estimated_input_tokens": 369, - "estimated_eval_tokens": 994 + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two closed issues are unrelated and only share superficial name/date formatting. They do not describe the same bug, request, or code-path problem, so the soft duplicate edge should be rejected.", - "confidence": 0.98, + "summary": "The two issues are unrelated closed records with different personal-name titles and no substantive overlap; they should not be clustered as duplicates.", + "confidence": 0.99, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, @@ -16442,7 +16445,7 @@ "left": "issue:45762", "right": "issue:45763", "accept": false, - "reason": "Different personal names and dates; no evidence of a shared defect, feature request, or technical context." + "reason": "Different people and dates in the titles, with no body text, links, targets, or shared context indicating the same underlying bug or request." } ] }, @@ -16472,7 +16475,7 @@ "soft_pairs": [ "issue:45764|issue:45765" ], - "prepared_review_unit_hash": "8932e429868beaa13199451da59cddfdea6e76481efa9913fb0e2c5bf5f9d639", + "prepared_review_unit_hash": "3b7ce52105dcfbe9367fcef09c1789a9c13831341046ddf82ee6338636f760d7", "component_budget": { "node_count": 2, "item_count": 2, @@ -16494,18 +16497,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues do not look like duplicates or the same bug report; they appear to be unrelated low-information entries with different personal-name titles and no body detail.", - "confidence": 0.93, - "canonical_issue_reason": null, + "summary": "Two closed issues with only loosely similar names and dates in the titles, but no body, no shared targets, and no evidence of the same underlying bug or request. They do not look like duplicates beyond superficial name similarity.", + "confidence": 0.97, + "canonical_issue_reason": "Neither issue has substantive content; if one must be chosen as the cluster representative, issue #45764 is marginally earlier and just as content-free as #45765.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Issue #45764 is the best representative only by timestamp; there is no technical signal to distinguish it from #45765.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45764", "right": "issue:45765", "accept": false, - "reason": "Titles reference different people and dates; there is no shared problem description, code path, or evidence of the same underlying issue." + "reason": "The titles are different personal names with dates, and there is no body, code target, or discussion evidence tying them to the same bug or change." } ] }, @@ -16545,88 +16548,88 @@ "issue:45871|issue:45873", "issue:45872|issue:45873" ], - "prepared_review_unit_hash": "14d8b67f84abbc2e684b403d91333be59c909c3d8f1726f0b353444c613c95b2", + "prepared_review_unit_hash": "cc1ebd6e5a56fb6a8fdf8bb2139092a6529f99e85fdb299e9a7041af46685f4a", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5726, - "estimated_input_tokens": 1432, - "estimated_eval_tokens": 3120 + "serialized_chars": 5724, + "estimated_input_tokens": 1431, + "estimated_eval_tokens": 3118 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5726, - "estimated_input_tokens": 1432, - "estimated_eval_tokens": 3120 + "serialized_chars": 5724, + "estimated_input_tokens": 1431, + "estimated_eval_tokens": 3118 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The cluster contains two separate exact-duplicate issue pairs: 45870/45871 and 45872/45873. Issue 45869 is a different dataset-sales post and should not be merged with either pair.", - "confidence": 0.97, - "canonical_issue_reason": "45870 is the earliest of the exact duplicate pair 45870/45871, with identical title/body content and no sign it differs in substance.", + "summary": "The cluster is not one single duplicate set: it contains two exact duplicate issue pairs (45870/45871 and 45872/45873) plus an unrelated dataset-promo issue (45869). The cross-pair matches are only broad topical similarity around financial data/DataForge, not the same underlying request.", + "confidence": 0.96, + "canonical_issue_reason": "There is no single canonical issue for the whole cluster because it splits into two separate duplicate pairs with different asks. If one representative must be chosen, 45872 is the most detailed and specific of the DataForge pair, but it should not absorb the other unrelated pair.", "canonical_pr_reason": null, - "best_issue_reason": "45870 is the best issue representative because it is an exact duplicate twin with the earliest timestamp; the rest of the cluster splits into a second duplicate pair plus an unrelated dataset item.", + "best_issue_reason": "45872 is the strongest representative within the cluster because it is a concrete, specific request and has an exact duplicate in 45873. However, it only represents one of the two duplicate subclusters, not the entire cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45869", "right": "issue:45870", "accept": false, - "reason": "Different topics: a dataset bundle/free-sample listing versus a financial time-series example in DataForge." + "reason": "Different artifacts: one is a financial dataset promo, the other is a DataForge example request. Similar keywords only, not the same issue." }, { "left": "issue:45869", "right": "issue:45871", "accept": false, - "reason": "Different topics: a dataset bundle/free-sample listing versus a financial time-series example in DataForge." + "reason": "Different artifacts: one is a financial dataset promo, the other is a DataForge example request. Similar keywords only, not the same issue." }, { "left": "issue:45870", "right": "issue:45871", "accept": true, - "reason": "Exact duplicate pair: same title and matching body/content." + "reason": "Exact duplicate: same title and apparent same request, so they are the same underlying issue." }, { "left": "issue:45870", "right": "issue:45872", "accept": false, - "reason": "Both are finance/DataForge themed, but one is an example post and the other is a scraping-service request; not the same underlying issue." + "reason": "Both mention financial time series/DataForge, but one is an example request and the other adds a scraping service component. Related topic, different underlying issue." }, { "left": "issue:45870", "right": "issue:45873", "accept": false, - "reason": "Both are finance/DataForge themed, but one is an example post and the other is a scraping-service request; not the same underlying issue." + "reason": "Both mention financial time series/DataForge, but one is an example request and the other adds a scraping service component. Related topic, different underlying issue." }, { "left": "issue:45871", "right": "issue:45872", "accept": false, - "reason": "Different concrete request: example prediction writeup versus scraping-service enhancement." + "reason": "Both mention financial time series/DataForge, but one is an example request and the other adds a scraping service component. Related topic, different underlying issue." }, { "left": "issue:45871", "right": "issue:45873", "accept": false, - "reason": "Different concrete request: example prediction writeup versus scraping-service enhancement." + "reason": "Both mention financial time series/DataForge, but one is an example request and the other adds a scraping service component. Related topic, different underlying issue." }, { "left": "issue:45872", "right": "issue:45873", "accept": true, - "reason": "Exact duplicate pair: same title and matching body/content." + "reason": "Exact duplicate: same title and apparent same request, so they are the same underlying issue." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and the soft-edge verdicts are conservative. The titles, body lengths, and Jaccard 1.0 scores support treating 45870/45871 and 45872/45873 as duplicate pairs, while 45869 is clearly a different dataset-sales item. No mergeability issue is present here." + "feedback": "The summary is well grounded in the packet: 45870/45871 and 45872/45873 are exact duplicate pairs, while 45869 is only softly related by topic. The cross-pair rejections are appropriately conservative, and the canonical-issue note correctly avoids collapsing two different asks into one cluster-wide representative." }, "evaluator_used": true, "retried": false, @@ -16637,7 +16640,7 @@ { "label": "pull_request", "component_index": 1, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -16678,8 +16681,8 @@ "pull_request:43823|pull_request:45655", "pull_request:43823|pull_request:45922", "pull_request:43823|pull_request:43995", - "pull_request:43995|pull_request:44827", "pull_request:43995|pull_request:45213", + "pull_request:43995|pull_request:44827", "pull_request:43995|pull_request:45899", "pull_request:44827|pull_request:45899", "pull_request:44827|pull_request:45170", @@ -16695,214 +16698,214 @@ "pull_request:44827|pull_request:45922", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "7ac2ccd4b942bc1e4b28e81e55ecec6066c37e1e896d0993d47146f2e7d73aaa", + "prepared_review_unit_hash": "8b6de8f398270997940292a09e62c61e2afdedf856a8210ece612a16d77998f8", "component_budget": { "node_count": 16, "item_count": 16, "soft_pair_count": 29, - "serialized_chars": 46650, - "estimated_input_tokens": 11663, - "estimated_eval_tokens": 23582 + "serialized_chars": 46661, + "estimated_input_tokens": 11666, + "estimated_eval_tokens": 23588 }, "budget": { "node_count": 16, "item_count": 16, "soft_pair_count": 29, - "serialized_chars": 46650, - "estimated_input_tokens": 11663, - "estimated_eval_tokens": 23582 + "serialized_chars": 46661, + "estimated_input_tokens": 11666, + "estimated_eval_tokens": 23588 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly heterogeneous PRs with one true duplicate pair: two nearly identical fixes for `fixed_cross_entropy` kwargs. The rest are weak title-level similarities across unrelated model additions, docs, or refactors and should stay separate.", - "confidence": 0.98, + "summary": "This cluster is mostly a set of unrelated PRs that only overlap at a broad topic level (new-model additions, docs, tokenizer utilities, etc.). The only true duplicate-looking pair is the two `fixed_cross_entropy` PRs, which both change the same function to forward supported `cross_entropy` kwargs for the same issue.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43254 is the cleanest representative of the only true duplicate pair: it directly adds supported kwargs to `fixed_cross_entropy`, has the explicit issue target, and matches the same code path as #43251.", + "canonical_pr_reason": "PR #43251 is the best canonical PR for the duplicate pair because it targets the shared issue `#43240`, patches the same `fixed_cross_entropy` code path, and appears to be the more issue-aligned version of the same fix.", "best_issue_reason": null, - "best_pr_reason": "PR #43254 is the best global representative because it is the most focused, issue-linked implementation of the actual duplicated fix; the other PRs in this set are unrelated despite superficial similarity.", + "best_pr_reason": "PR #43251 is the best representative PR in this cluster because it is the only clearly duplicated change and the one explicitly linked to the underlying bug report; the rest of the cluster is not duplicate-suitable.", "soft_edge_verdicts": [ { "left": "pull_request:39895", "right": "pull_request:43424", "accept": false, - "reason": "Unrelated PRs: VideoPrism model addition vs. Executorch dynamic-shape test." + "reason": "Both are unrelated model/runtime PRs; one adds VideoPrism, the other adds an Executorch dynamic-shape test." }, { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Unrelated PRs: Mistral tokenizer methods vs. adding MobileLLM." + "reason": "Tokenizer method additions for MistralCommon are unrelated to adding MobileLLM." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Tokenizer API fix and Executorch export test are different changes." + "reason": "Mistral tokenizer cleanup and Executorch export testing fix different code paths." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor robustness changes are unrelated to a new model addition." + "reason": "Processor auto-loading robustness is not the same change as adding a new model implementation." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Loss helper kwargs fix is unrelated to MobileLLM model scaffolding." + "reason": "Loss-function kwargs forwarding is unrelated to adding MobileLLM." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "Same broad repo area is not enough; loss utility fix and Mistral4 test fixes are different bugs." + "reason": "A cross-entropy helper fix and Mistral4 test/model cleanup are different bugs." }, { "left": "pull_request:43424", "right": "pull_request:45919", "accept": false, - "reason": "Executorch dynamic-shape test and Sapiens2 model addition are unrelated." + "reason": "Executorch dynamic-shape testing is unrelated to adding Sapiens2." }, { "left": "pull_request:43424", "right": "pull_request:45899", "accept": false, - "reason": "Executorch export testing and continuous batching docs are unrelated." + "reason": "Executorch export tests and continuous-batching docs cover different concerns." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "Different targets: exportability test vs. Mistral4 test/model integration work." + "reason": "Dynamic-shape export tests do not match the Mistral4 model/test fix." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "No shared bug/change; export test vs. MobileLLM addition." + "reason": "Executorch export testing is unrelated to MobileLLM model addition." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "New model addition and `cls.config` resolution fix are unrelated." + "reason": "Adding MobileLLM is unrelated to fixing `cls.config` resolution order." }, { "left": "pull_request:43823", "right": "pull_request:45922", "accept": false, - "reason": "MobileLLM addition and vision-model cache leak fix are unrelated." + "reason": "MobileLLM model scaffolding and vision-model lru-cache leak fixes are unrelated." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "MobileLLM addition and Falcon refactor are different model changes." + "reason": "Adding a new model is not the same as refactoring Falcon output collection behavior." }, { "left": "pull_request:43995", - "right": "pull_request:44827", + "right": "pull_request:45213", "accept": false, - "reason": "Falcon output-collection refactor and Mistral4 test fix do not address the same concrete bug." + "reason": "Falcon refactoring and a model-creation skill are unrelated changes." }, { "left": "pull_request:43995", - "right": "pull_request:45213", + "right": "pull_request:44827", "accept": false, - "reason": "Falcon refactor and model-creation skill PR are unrelated." + "reason": "Falcon output-interface refactoring does not match Mistral4 test fixes." }, { "left": "pull_request:43995", "right": "pull_request:45899", "accept": false, - "reason": "Falcon model refactor and docs for continuous batching decode fast path are unrelated." + "reason": "Falcon model internals and continuous-batching documentation are not the same underlying change." }, { "left": "pull_request:44827", "right": "pull_request:45899", "accept": false, - "reason": "Mistral4 test fixes and decode-fast-path docs are unrelated; shared similarity is only broad infra/model context." + "reason": "Mistral4 test/model fixes do not overlap with a docs-only continuous-batching update." }, { "left": "pull_request:44827", "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 test fixes and `layrnorm` renaming are different bugs and code paths." + "reason": "Mistral4 fixes are unrelated to the CLIP-like `layrnorm` spelling correction." }, { "left": "pull_request:44827", "right": "pull_request:45795", "accept": false, - "reason": "Mistral4 tests and docs for audio/video processors are unrelated." + "reason": "Mistral4 tests/model adjustments are unrelated to audio/video processor docs." }, { "left": "pull_request:44827", "right": "pull_request:45655", "accept": false, - "reason": "Mistral4 test fixes and config-resolution ordering are unrelated." + "reason": "Mistral4 changes are unrelated to fixing config resolution order." }, { "left": "pull_request:45170", "right": "pull_request:45899", "accept": false, - "reason": "Weight-renaming cleanup and decode-fast-path docs are unrelated." + "reason": "A weight-renaming/model spelling fix is unrelated to continuous-batching documentation." }, { "left": "pull_request:44827", "right": "pull_request:45213", "accept": false, - "reason": "Mistral4 test fix and model-creation skill PR are not the same underlying change." + "reason": "Mistral4 model/test fixes are unrelated to the model-creation skill PR." }, { "left": "pull_request:43823", "right": "pull_request:45213", "accept": false, - "reason": "MobileLLM addition and model-creation skill are unrelated." + "reason": "Adding MobileLLM is not the same as introducing a model creation skill." }, { "left": "pull_request:45213", "right": "pull_request:45899", "accept": false, - "reason": "Model-creation skill and decode-fast-path docs are unrelated." + "reason": "A model-creation skill and decode-fast-path docs are unrelated." }, { "left": "pull_request:45795", "right": "pull_request:45899", "accept": false, - "reason": "Docs for audio/video processors and decode-fast-path docs are different documentation changes." + "reason": "Docs for audio/video processors do not duplicate decode-fast-path documentation." }, { "left": "pull_request:43823", "right": "pull_request:45795", "accept": false, - "reason": "MobileLLM addition and processor docs update are unrelated." + "reason": "MobileLLM model code and processor docs are different changes." }, { "left": "pull_request:43823", "right": "pull_request:45899", "accept": false, - "reason": "MobileLLM addition and decode-fast-path docs are unrelated." + "reason": "Adding a model and documenting a decode optimization are unrelated." }, { "left": "pull_request:44827", "right": "pull_request:45922", "accept": false, - "reason": "Mistral4 tests and vision lru-cache leak fix are unrelated." + "reason": "Mistral4 fixes are unrelated to the BEiT/vision-model memory leak cache cleanup." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "True duplicate: both modify `src/transformers/loss/loss_utils.py` to pass supported kwargs through `fixed_cross_entropy` to `nn.functional.cross_entropy`." + "reason": "Both PRs modify `fixed_cross_entropy` in the same way to forward supported kwargs to `nn.functional.cross_entropy`, so they are the same concrete fix and could be merged into one PR." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The accepted 43251/43254 pair is the only clearly mergeable same-bugfix duplicate, and the other soft-edge rejections are appropriately treated as unrelated or only superficially similar." + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted duplicate-like pair is the shared `fixed_cross_entropy` change (`43251`/`43254`), which is supported by the shared file and issue target `#43240`. The other pairwise rejections are consistent with the listed titles/diffs and do not overstate duplication." }, "evaluator_used": true, "retried": false, @@ -16913,7 +16916,7 @@ { "label": "pull_request", "component_index": 2, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -16929,7 +16932,7 @@ "soft_pairs": [ "pull_request:40546|pull_request:42781" ], - "prepared_review_unit_hash": "f7f0c169e233bf4bf6fda03535c7e51c604cc77899d171200e33cfcd978ac56c", + "prepared_review_unit_hash": "c8eb94239f6eb8f3cfd36781ac9b2debfceedb6fe09310db64376120a7c0691f", "component_budget": { "node_count": 2, "item_count": 2, @@ -16951,18 +16954,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are VibeVoice-related, but they are not the same change: #40546 implements the base VibeVoice model, while #42781 extends VibeVoice with realtime/acoustic-tokenizer work and additional generation/model plumbing. The overlap is mostly shared infrastructure, not a single duplicate fix or feature.", - "confidence": 0.88, + "summary": "These are related VibeVoice pull requests, but they are not duplicates: PR 40546 is the main initial VibeVoice implementation, while PR 42781 adds a separate VibeVoice Realtime path plus an acoustic tokenizer and generation changes. Shared files are mostly the same model-family scaffolding, not the same concrete fix/change.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "#40546 is the better canonical PR because it is the earlier, broader base VibeVoice implementation and is closer to the core model bring-up that the later PR builds on.", + "canonical_pr_reason": "PR 40546 is the better canonical PR because it is the broader, more complete initial VibeVoice implementation and has much higher review activity; PR 42781 looks like a later extension for realtime support rather than the primary change.", "best_issue_reason": null, - "best_pr_reason": "#40546 is the best representative PR for the cluster: it is the main VibeVoice implementation, has substantial review activity, and is not a draft.", + "best_pr_reason": "PR 40546 is the strongest representative of the cluster: it introduces the core VibeVoice integration across docs, auto-mapping, model code, processing, and pipeline hooks. PR 42781 is narrower and draft-status, focused on realtime/acoustic-tokenizer follow-up work.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Reject: these are related VibeVoice features, but not the same concrete code-path change. #40546 brings up VibeVoice itself; #42781 adds realtime-specific support plus an acoustic tokenizer and extra generation changes, so they should not be merged as duplicates." + "reason": "Do not merge: the PRs share VibeVoice files and auto-registration, but 42781 adds a separate realtime/acoustic-tokenizer feature set and generation changes rather than fixing or implementing the same concrete code path as 40546." } ] }, @@ -16976,7 +16979,7 @@ { "label": "pull_request", "component_index": 3, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -16992,12 +16995,12 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "e1b06ed09717aca7a00352a8053fb13e278d77f18ff485c907cf3281f616a914", + "prepared_review_unit_hash": "77c2685f7696b2e48898c1fade88c87477d3dbde2820f4cb494c453f9109c025", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4674, + "serialized_chars": 4675, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -17005,7 +17008,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4674, + "serialized_chars": 4675, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -17014,8 +17017,8 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two pull requests are unrelated feature additions in different model families, so this is not a duplicate cluster.", - "confidence": 0.99, + "summary": "The two pull requests are unrelated feature additions: one adds TimesFM covariate forecasting support, the other adds DINOv3 image classification support. They do not appear to fix the same underlying bug or implement the same change.", + "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, @@ -17025,7 +17028,7 @@ "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different underlying changes: TimesFM covariate forecasting vs. DINOv3 image classification support. They touch unrelated code paths and would not be merged into one PR." + "reason": "Different models and different capabilities; no shared code-path problem or mergeable duplicate change." } ] }, @@ -17039,7 +17042,7 @@ { "label": "pull_request", "component_index": 4, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17102,7 +17105,7 @@ { "label": "pull_request", "component_index": 5, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17120,46 +17123,46 @@ "pull_request:41797|pull_request:45186", "pull_request:45186|pull_request:45630" ], - "prepared_review_unit_hash": "11ae9d3e39e46e3a1eec16590aa810135a766cf3168076c78c2b955fd7e3523b", + "prepared_review_unit_hash": "6feb716c29ddcc88592b8f12d8208ba4891a135a91bc0f2a0d15cb689fd14647", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10408, - "estimated_input_tokens": 2602, - "estimated_eval_tokens": 5460 + "serialized_chars": 10409, + "estimated_input_tokens": 2603, + "estimated_eval_tokens": 5462 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10408, - "estimated_input_tokens": 2602, - "estimated_eval_tokens": 5460 + "serialized_chars": 10409, + "estimated_input_tokens": 2603, + "estimated_eval_tokens": 5462 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate pull requests adding different models (DeepSeekOCR, Isaac, and Kimi2-6). They share only the usual Transformers boilerplate around docs and auto-mapping files, so they are not duplicates or the same underlying change.", - "confidence": 0.97, + "summary": "These are three separate open PRs adding different models (DeepSeekOCR, Isaac, and Kimi2-6). They share the same broad scaffolding pattern for new model integration, but they are not the same underlying change.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "No canonical PR: each PR adds a different model and touches its own model-specific code paths.", "best_issue_reason": null, - "best_pr_reason": "PR #45630 is the strongest representative of the cluster only in the sense that it has the most review activity and broadest model-integration surface, but it is still a distinct Kimi2-6 addition rather than a duplicate of the others.", + "best_pr_reason": "No single PR is a good representative duplicate target because the changes are distinct model additions rather than one shared fix.", "soft_edge_verdicts": [ { "left": "pull_request:41797", "right": "pull_request:45186", "accept": false, - "reason": "Different models and different code paths: DeepSeekOCR vs Isaac. Shared auto-registration/docs files are boilerplate for new-model additions, not evidence of the same fix/change." + "reason": "Different models and different implementations; shared new-model boilerplate is not enough to treat them as the same change." }, { "left": "pull_request:45186", "right": "pull_request:45630", "accept": false, - "reason": "These add unrelated models (Isaac vs Kimi2-6). Overlap in model registry and docs is expected for separate integrations and is too broad to treat as one underlying change." + "reason": "Isaac and Kimi2-6 are separate model integrations with different files, configs, and code paths." } ] }, @@ -17173,7 +17176,7 @@ { "label": "pull_request", "component_index": 6, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17189,7 +17192,7 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "1a0dab776a6b016ed600fd77c2bb052e519a9a8c4bcf0797a5ed0ce756640d9c", + "prepared_review_unit_hash": "595a8e3e100876655e759346220735e4e2aa96afc32318d50673a0a50ac53524", "component_budget": { "node_count": 2, "item_count": 2, @@ -17211,18 +17214,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests are dependency-version bump sweeps in GitHub Actions workflows, but they target different action families and different maintenance goals: one is specifically for Node 24 compatibility of core actions, the other upgrades docker/build-push-action. They are related at the workflow-maintenance level, but not the same underlying change.", - "confidence": 0.95, + "summary": "These two PRs both touch GitHub Actions workflows, but they are different upgrade efforts: one focuses on checkout/cache/artifact actions for Node 24 compatibility, while the other bumps docker/build-push-action and a separate set of workflow actions. Shared files alone are not enough to treat them as duplicates.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #42975 is the better canonical representative because its title and diff clearly describe a concrete compatibility-driven upgrade across many workflows, making it the more specific and central change in this cluster.", + "canonical_pr_reason": "No canonical PR for a duplicate cluster: the PRs address different action-version upgrade tracks and do not fix the same underlying change.", "best_issue_reason": null, - "best_pr_reason": "PR #42975 is the best overall representative: it has the clearest problem statement, broader workflow coverage, and a more concrete maintenance target than the more generic latest-version bump in #42976.", + "best_pr_reason": "If one PR must represent the cluster, 42975 is the broader GitHub Actions compatibility update, but it still should not absorb 42976 as a duplicate.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "Reject: both are GitHub Actions version upgrades, but they change different actions and different code paths. #42975 updates checkout/cache/upload-artifact for Node 24 compatibility; #42976 updates docker/build-push-action. They are not the same concrete fix and would not plausibly be merged as one PR." + "reason": "They upgrade different GitHub Actions families and target different maintenance goals; overlapping workflow files do not make them the same change." } ] }, @@ -17236,7 +17239,7 @@ { "label": "pull_request", "component_index": 7, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17265,12 +17268,12 @@ "pull_request:44601|pull_request:45838", "pull_request:45490|pull_request:45838" ], - "prepared_review_unit_hash": "454a7e364cefe1519c2cf573123e9c03f2a25fe4369fc01532b2ee9e18462b7d", + "prepared_review_unit_hash": "eef849d5186c2f253d507ddd1180de42ad4402dde2758aae2dbf4e621aede737", "component_budget": { "node_count": 7, "item_count": 7, "soft_pair_count": 9, - "serialized_chars": 23543, + "serialized_chars": 23541, "estimated_input_tokens": 5886, "estimated_eval_tokens": 12028 }, @@ -17278,7 +17281,7 @@ "node_count": 7, "item_count": 7, "soft_pair_count": 9, - "serialized_chars": 23543, + "serialized_chars": 23541, "estimated_input_tokens": 5886, "estimated_eval_tokens": 12028 }, @@ -17287,66 +17290,66 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are unrelated additions across different models/subsystems; the similarity is mostly from shared Transformers boilerplate and docs/auto-registration files, not the same underlying change.", - "confidence": 0.98, + "summary": "This cluster is mostly superficial similarity from model-addition PR boilerplate (docs, auto mappings, tests), plus one unrelated distributed-loading PR. None of the soft edges look like true duplicates or mergeable into a single change.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "If one PR must represent the set, 43451 is the strongest standalone candidate because it is a large, cohesive model-addition PR with the most review activity and the broadest implementation surface.", + "canonical_pr_reason": "No canonical PR: these are separate feature PRs for different model families or pipeline-parallel infrastructure, and the overlap is mostly docs/registry boilerplate.", "best_issue_reason": null, - "best_pr_reason": "43451 is the best standalone PR here: it appears comparatively complete and well-reviewed, but it is not a duplicate of the other PRs.", + "best_pr_reason": "No PR is a true duplicate anchor here; if forced, 43451 is the most substantial standalone model-addition PR, but it does not subsume the others.", "soft_edge_verdicts": [ { "left": "pull_request:42978", "right": "pull_request:43448", "accept": false, - "reason": "Different model additions: ViT NEPA vs Molmo. Shared docs/auto files are generic scaffolding, not the same fix/change." + "reason": "Different model families (ViT NEPA vs Molmo) with distinct codepaths; shared docs/auto-registration files are boilerplate, not the same fix." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "Different model additions: ViT NEPA vs Molmo2. Same broad Transformers integration patterns, but not one underlying bug or mergeable change." + "reason": "ViT NEPA and Molmo2 are separate model integrations with different implementations and files; not the same underlying change." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "Different model additions: ViT NEPA vs CTSM. Overlap is only in generic model-registration/docs files." + "reason": "Different models (ViT NEPA vs CTSM) and different model code; overlap is only generic transformers registration/docs." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are related names but distinct model families with separate codepaths and files; not the same concrete change." + "reason": "Molmo and Molmo2 are related by name but are distinct model releases with different configs/modeling/processing code; not a duplicate PR." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Molmo vs CTSM are unrelated model integrations; shared auto/docs edits are too broad to indicate duplication." + "reason": "Molmo and CTSM are unrelated model additions; shared auto/model registry edits are broad subsystem boilerplate." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 vs CTSM are distinct model additions with different implementation details and targets." + "reason": "Molmo2 and CTSM are separate model integrations with different model paths; no concrete bug or change overlap beyond registration/docs." }, { "left": "pull_request:43973", "right": "pull_request:45838", "accept": false, - "reason": "Lfm2Audio adds an audio model, while PP-OCRv6 adds text recognition models; different domains and codepaths." + "reason": "Audio model addition vs OCR text-recognition model support; different domains, code paths, and artifacts." }, { "left": "pull_request:44601", "right": "pull_request:45838", "accept": false, - "reason": "Pipeline-parallel runtime support is a distributed loading change, not the same as adding OCR model support." + "reason": "Pipeline-parallel loading infrastructure is unrelated to PP-OCRv6 model support; not the same bug or change." }, { "left": "pull_request:45490", "right": "pull_request:45838", "accept": false, - "reason": "CTSM and PP-OCRv6 are different model families and separate integration efforts, despite shared boilerplate files." + "reason": "CTSM model addition and PP-OCRv6 text recognition support are distinct features with different model implementations and purposes." } ] }, @@ -17360,7 +17363,7 @@ { "label": "pull_request", "component_index": 8, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17379,7 +17382,7 @@ "pull_request:43291|pull_request:44815", "pull_request:43665|pull_request:44815" ], - "prepared_review_unit_hash": "e0ab4169e723ce3fa65213534d29e34ed3c40858fd0aba3ba6b52347b32351ab", + "prepared_review_unit_hash": "356185cd66553e5d8185cfd6093f4b59457cfbf2293784af67f3df8b395e0efc", "component_budget": { "node_count": 3, "item_count": 3, @@ -17401,30 +17404,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs do not look like duplicates: one is a Whisper tokenizer/test fix, one is a test-only patch with placeholder assertions, and one is a separate FP8 dequantization fix. They touch different subsystems and address different code paths.", + "summary": "All three items are pull requests with unrelated fixes: Whisper tokenizer/test changes, CLIP/ViT test changes, and a dequantization/fine-grained FP8 fix. None are the same underlying change.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "pull_request:44815 is the strongest standalone candidate: it has a coherent implementation fix in loading/FP8 quantization code, whereas the others are narrower test-oriented changes with weak or noisy diffs.", + "canonical_pr_reason": "PR 44815 is the strongest canonical PR because it contains a substantive code-path fix across model loading and FP8 quantization, whereas 43291 and 43665 are narrow test-oriented changes with no clear shared bug.", "best_issue_reason": null, - "best_pr_reason": "pull_request:44815 is the most substantive and clearly scoped fix in the set; it is the best representative if one PR must be chosen.", + "best_pr_reason": "PR 44815 is the best representative artifact: it addresses a concrete implementation bug, touches production code, and has the most complete multi-file fix. The other PRs appear to be test-only or localized cleanup changes.", "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "Both are test-focused, but they target different models and different failures; the diffs do not indicate the same underlying bug or mergeable change." + "reason": "Both are test-related, but they target different models and different failures (Whisper tokenizer/tests vs CLIP/ViT test edits). They do not look like the same bug or one mergeable PR." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper tokenizer/test changes are unrelated to FP8 dequant/loading changes; no shared code path or concrete duplicate bug is visible." + "reason": "Whisper tokenizer/test fixes and FP8 dequantization/loading fixes are unrelated code paths and subsystems. Shared 'fix' wording is too broad to treat as a duplicate." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "The CLIP/ViT test edits are unrelated to the FP8 quantization/dequantization implementation fix, so these are not duplicates." + "reason": "PR 43665 only changes CLIP/ViT tests, while PR 44815 changes model loading and FP8 quantization internals. These are not the same change and should not be merged as duplicates." } ] }, @@ -17438,7 +17441,7 @@ { "label": "pull_request", "component_index": 9, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17454,12 +17457,12 @@ "soft_pairs": [ "pull_request:43488|pull_request:43532" ], - "prepared_review_unit_hash": "e602f29816038f192f3238eee35a60498d0d0658c59dadc66c08c932173cbaed", + "prepared_review_unit_hash": "d3ab2ef564602e14e357ccfde7525e4489bd7dc77d1335f263188828cb8d7be1", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4562, + "serialized_chars": 4561, "estimated_input_tokens": 1141, "estimated_eval_tokens": 2538 }, @@ -17467,7 +17470,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4562, + "serialized_chars": 4561, "estimated_input_tokens": 1141, "estimated_eval_tokens": 2538 }, @@ -17476,18 +17479,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated open test/demo PRs: one touches packaging/versioning and a Vit test, while the other changes CI/workflow and utility files. They only overlap on a shared test filename, not on the same underlying bug or change.", - "confidence": 0.97, + "summary": "The two PRs are both marked as non-merging experiments, but they change different areas: one is a packaging/versioning and test-dependency tweak, while the other rewires CI/workflow and testing utilities. They do not look like duplicates.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "No clear canonical PR: the two PRs are not duplicates and do not fix the same concrete problem; they only share a broad test file and are both '[don't merge]' style validation/demo changes.", + "canonical_pr_reason": "No clear canonical PR: the patches target different code paths and are not plausibly mergeable into one fix despite sharing a test filename.", "best_issue_reason": null, - "best_pr_reason": "PR 43532 is the better representative only because it is the larger, more system-level change, but it still should not be treated as a duplicate of 43488.", + "best_pr_reason": "No single PR stands out as the representative duplicate target because the changes are unrelated beyond superficial overlap.", "soft_edge_verdicts": [ { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "Reject: these PRs address different concerns. 43488 changes setup/version/dependency metadata and a Vit test, while 43532 rewires CI/workflow and notification/testing utilities. Shared test-file overlap is not enough to indicate the same underlying fix." + "reason": "Reject: shared vit test filename is incidental; one PR changes setup/versioning and pytest deps, the other changes workflow automation and testing utilities, so they are not the same underlying change." } ] }, @@ -17501,7 +17504,7 @@ { "label": "pull_request", "component_index": 10, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17517,7 +17520,7 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "5de78cd7792db62060fbba0a998d11415b2badc4c76fa9e13aba9af9a2083831", + "prepared_review_unit_hash": "5a4c7cf47869a9849de41c35f61ba0f8ae1249c3de1978a6cecaede877f08404", "component_budget": { "node_count": 2, "item_count": 2, @@ -17539,18 +17542,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two similarly structured PRs add different segmentation pipelines: one for promptable concept segmentation and one for promptable visual segmentation. They share docs/scaffolding, but they are not the same underlying change.", + "summary": "These are related pipeline/docs additions, but they implement different features: promptable concept segmentation vs promptable visual segmentation. Shared infrastructure files are not enough to treat them as duplicates, so the soft edge should be rejected.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "Neither PR is a canonical duplicate of the other; they implement different pipeline features for different segmentation tasks.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "No single PR is a better dedupe representative here, since the PRs target different code paths and should remain separate.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Reject: these are distinct features (concept vs visual segmentation). The overlap is mostly shared docs/config/pipeline scaffolding, not the same concrete bug or change." + "reason": "Different underlying feature work and code paths: one adds PromptableConceptSegmentationPipeline, the other adds PromptableVisualSegmentationPipeline with different model/test coverage. They are related but not mergeable as one PR." } ] }, @@ -17564,7 +17567,7 @@ { "label": "pull_request", "component_index": 11, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17580,7 +17583,7 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "a5005858f7500487883c76673bd732651fe985d349917c5dd4a2ed769ccb7deb", + "prepared_review_unit_hash": "6b5a55abae0eab118852981dc3fc5e232dbc8d0033a4df319b2692a4a6449631", "component_budget": { "node_count": 2, "item_count": 2, @@ -17602,18 +17605,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two PRs touch the same file but address different trainer extension points: one changes gradient-accumulation loss scaling, the other adds a reusable signature-column helper for column filtering. They are not duplicates and should not be merged into one change.", - "confidence": 0.97, + "summary": "Two separate trainer extensibility PRs: one adds a loss-scaling hook for gradient accumulation, the other adds an overridable signature-column helper for column filtering. Same file, but different code paths and behaviors; not duplicates.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43663 is the better representative of the cluster because it cleanly extracts a reusable trainer helper with an explicit override hook; PR 43651 is a separate, narrower loss-scaling change.", + "canonical_pr_reason": "PR 43663 is the more general trainer API refactor: it extracts signature-column computation into an overridable helper and is the cleaner representative of the cluster.", "best_issue_reason": null, - "best_pr_reason": "PR 43663 is the stronger standalone PR for the cluster: it has a clear refactor boundary, a specific customization point, and a more self-contained behavior change than the gradient-scaling tweak in 43651.", + "best_pr_reason": "PR 43663 is the strongest standalone change here because it introduces a reusable trainer hook with broad extensibility value and a self-contained code-path adjustment.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Same file and same trainer class, but different code paths and goals: one fixes GA loss normalization, the other refactors signature-column inspection for column filtering. They are not the same underlying bug/change." + "reason": "Both touch trainer.py, but one changes gradient-accumulation loss scaling and the other changes input-column filtering via signature inspection; they fix different problems and would not plausibly merge as one PR." } ] }, @@ -17627,7 +17630,7 @@ { "label": "pull_request", "component_index": 12, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17645,55 +17648,55 @@ "pull_request:43656|pull_request:43842", "pull_request:43836|pull_request:43842" ], - "prepared_review_unit_hash": "f6626ba4f1125e30c3bf41a31d11523b210bbf0b0bec0ec3878904644b8ce820", + "prepared_review_unit_hash": "eede7eb815cf7f8b49b29ca3f8ee65f7b579dec35362cfc21762f445a0b8fc1d", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 5993, - "estimated_input_tokens": 1499, - "estimated_eval_tokens": 3254 + "serialized_chars": 5992, + "estimated_input_tokens": 1498, + "estimated_eval_tokens": 3252 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 5993, - "estimated_input_tokens": 1499, - "estimated_eval_tokens": 3254 + "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": "PR 43842 is the strongest representative for the concrete TypeAdapter/pydantic NameError fix in `src/transformers/cli/serve.py`. PR 43836 looks closely related and likely duplicates the same runtime-import issue, while PR 43656 is broader and not supported well enough to treat as the same exact fix.", - "confidence": 0.71, + "summary": "All three PRs modify `src/transformers/cli/serve.py` and are aimed at the same import-time `TypeAdapter` annotation NameError in the CLI. PR 43842 is the most focused fix; 43656 and 43836 look like broader/earlier variants of the same underlying problem.", + "confidence": 0.79, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43842 is the most targeted fix: it explicitly moves `TypeAdapter` behind `TYPE_CHECKING` and stringifies the annotation at the call site, directly addressing the reported NameError with minimal scope.", + "canonical_pr_reason": "43842 is the cleanest direct fix: it makes `TypeAdapter` a type-checking-only import and stringizes the annotation at the failing call site, with minimal unrelated churn.", "best_issue_reason": null, - "best_pr_reason": "PR 43842 is the best cluster anchor because it cleanly matches the concrete failure mode and has the smallest, most specific change set among the candidates.", + "best_pr_reason": "43842 is the best representative PR because it targets the exact runtime NameError in the most explicit and narrowly scoped way.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", - "accept": false, - "reason": "Too little concrete evidence that this broader annotation-focused change fixes the same exact TypeAdapter/pydantic NameError; it also targets a different issue and appears more general than the specific runtime import fix." + "accept": true, + "reason": "Both address the same `TypeAdapter`-related NameError in `serve.py`; 43656 is a broader postponed-annotations variant of the same fix, while 43842 is the narrower follow-up." }, { "left": "pull_request:43836", "right": "pull_request:43842", "accept": true, - "reason": "Both changes are in `cli/serve.py` and directly address the same TypeAdapter NameError by deferring runtime evaluation of the annotation/import path, so they look like duplicates of the same bug fix." + "reason": "Same file, same import-time `TypeAdapter` bug, and both make the annotation safe at runtime; 43836 is just a broader refactor around the same concrete problem." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The cluster summary overstates the evidence for PR 43656. The packet only shows shared-file/soft-similarity evidence for 43656 vs 43842, and 43656 is linked to a different issue target (43576) with a broader annotation-related diff preview; there is not enough concrete support to assert it fixes the same TypeAdapter/pydantic NameError. The 43836\u219443842 pairing is more grounded, but the 43656 pairing is not conservative enough, so I can\u2019t accept the cluster as written." + "accept": true, + "feedback": "The summary and pair verdicts are grounded in the packet. All three PRs touch `src/transformers/cli/serve.py` and are clearly in the same TypeAdapter/annotation bug family. The soft-edge accepts are conservative enough: 43836 and 43842 share the same issue target, and 43656 is plausibly the same CLI NameError class of fix even though it is broader and only linked by soft similarity. No overstatement that would warrant rejection." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -17701,7 +17704,7 @@ { "label": "pull_request", "component_index": 13, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17767,7 +17770,7 @@ { "label": "pull_request", "component_index": 14, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 2, "status": "reviewed", @@ -17845,10 +17848,10 @@ "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:44018|pull_request:44059", - "pull_request:44073|pull_request:44161", "pull_request:44056|pull_request:44101", - "pull_request:44059|pull_request:44068", + "pull_request:44073|pull_request:44161", "pull_request:44101|pull_request:44161", "pull_request:44017|pull_request:44018", "pull_request:43997|pull_request:44056", @@ -17879,20 +17882,20 @@ "pull_request:44044|pull_request:44129", "pull_request:44044|pull_request:44154" ], - "prepared_review_unit_hash": "89dfac3f6eaa5e43ddfeb3c3df6e3ffb0d8b1133ea18bff8f90db8e929aafdec", + "prepared_review_unit_hash": "6281b41118f98e3a6c462158e3b164825d2cab748e960cf0764db645ced7ce7f", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84898, - "estimated_input_tokens": 21225, - "estimated_eval_tokens": 42706 + "serialized_chars": 84896, + "estimated_input_tokens": 21224, + "estimated_eval_tokens": 42704 }, "budget": { "node_count": 27, "item_count": 27, "soft_pair_count": 72, - "serialized_chars": 75094, + "serialized_chars": 75093, "estimated_input_tokens": 18774, "estimated_eval_tokens": 37804 }, @@ -17901,405 +17904,351 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of separate model-specific output-tracing refactors, not one big duplicate set. There are two clear duplicate subclusters: GPT-J/CodeGen (44066, 44085, 44722) and GPT-Neo (44018, 44068). The rest are different model code paths and should stay separate.", - "confidence": 0.95, + "summary": "This cluster is mostly a broad wave of standardized output-tracing refactors across many unrelated models. The only clear duplicate-like overlaps are within the GPT-J/CodeGen subgroup and the GPT-Neo subgroup; the rest are just thematically similar, not the same underlying code path.", + "confidence": 0.84, "canonical_issue_reason": null, - "canonical_pr_reason": "44722 is the best canonical PR for the main GPT-J/CodeGen subgroup: it is the broadest representative of that concrete output-tracing refactor, with overlapping GPT-J/CodeGen files and an explicit issue link. The GPT-Neo pair forms a separate duplicate subcluster, so 44722 is canonical only for the GPT-J/CodeGen side of the cluster.", + "canonical_pr_reason": "PR 44066 is the strongest representative of the shared GPT-J/CodeGen output-tracing change: it is issue-linked, touches the core files in the duplicate set, and cleanly expresses the standardized decorator migration.", "best_issue_reason": null, - "best_pr_reason": "44722 is the strongest single PR representative overall, but only for the GPT-J/CodeGen duplicate subgroup. It best captures the shared code-path change there, while the GPT-Neo duplicates should be handled as a separate subgroup.", + "best_pr_reason": "PR 44066 is the best single PR to keep as the representative change because it is the most complete, issue-linked GPT-J/CodeGen refactor and has the clearest overlap with the accepted duplicate candidates.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different model families and code paths; same output-tracing umbrella is not enough to merge them as duplicates." + "reason": "Same general refactor theme, but different models and code paths (CVT/FNet vs GPT-J)." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "CVT/FNet vs DeBERTa V2; unrelated concrete refactors." + "reason": "Both are output-tracing refactors, but they affect unrelated model implementations and are not the same change." }, { "left": "pull_request:44066", "right": "pull_request:44085", "accept": true, - "reason": "Same GPT-J output-tracing code path; 44066 is a broader variant that plausibly subsumes 44085." + "reason": "Both target GPT-J output tracing in the same file and are the same underlying decorator migration." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "Different models and different implementations; not the same underlying change." + "reason": "Different vision models and separate implementations; only the refactor theme is shared." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "EfficientNet vs GPT-J/CodeGen; only share the generic refactor theme." + "reason": "Different model families and no concrete code-path overlap." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen work vs EfficientNet work; different code paths." + "reason": "GPT-J/CodeGen versus EfficientNet; same modernization pattern, not the same bug or change." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 and DeBERTa V2 are unrelated model-specific changes." + "reason": "MobileNetV2 and DeBERTa v2 are unrelated code paths." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J/CodeGen vs MGP-STR; similar pattern, different concrete fix." + "reason": "Different models and behavior; no shared concrete change." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "GPT-J/CodeGen vs MPT; different model implementation and change set." + "reason": "GPT-J/CodeGen refactor versus MPT refactor; unrelated implementation details." }, { "left": "pull_request:44018", "right": "pull_request:44068", "accept": true, - "reason": "Both target the same GPT-Neo output-tracing refactor in the same file; these look like duplicate variants." + "reason": "Both are GPT-Neo output-tracing refactors on the same model/code path, so they are plausibly duplicate iterations of the same change." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPT-J/CodeGen and GPT-Neo are different model-specific refactors." + "reason": "GPT-J versus GPT-Neo; same refactor family but not the same code path." }, { "left": "pull_request:44018", "right": "pull_request:44066", "accept": false, - "reason": "GPT-Neo vs GPT-J/CodeGen; different code paths." + "reason": "Different model families and different concrete changes." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet refactor and GPT-J/CodeGen refactor are unrelated despite similar decorator wording." + "reason": "ResNet refactor and GPT-J/CodeGen refactor are unrelated." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "MGP-STR vs GPT-J/CodeGen; different model code paths." + "reason": "MGP-STR and GPT-J/CodeGen are different model implementations." }, { "left": "pull_request:44019", "right": "pull_request:44085", "accept": false, - "reason": "ResNet and GPT-J are unrelated model implementations." + "reason": "ResNet versus GPT-J; only superficial similarity in refactor style." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "GPT-Neo vs ResNet; same initiative, different concrete change." + "reason": "GPT-Neo and ResNet are unrelated code paths." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet and MPT are different model code paths." + "reason": "ResNet and MPT are unrelated model changes." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo vs Flaubert; different model families." + "reason": "GPT-Neo refactor versus Flaubert/XLM refactor; not the same underlying change." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo refactor is separate from the GPT-J/CodeGen refactor." - }, - { - "left": "pull_request:44059", - "right": "pull_request:44086", - "accept": false, - "reason": "GPT-2 output tracing vs MGP-STR output tracing; not mergeable as one fix." - }, - { - "left": "pull_request:44085", - "right": "pull_request:44086", - "accept": false, - "reason": "GPT-J vs MGP-STR; only broad refactor similarity." + "reason": "GPT-Neo and GPT-J/CodeGen are different model families and changes." }, { - "left": "pull_request:44098", - "right": "pull_request:44101", + "left": "pull_request:44018", + "right": "pull_request:44116", "accept": false, - "reason": "ViLT and Flaubert/XLM are different model code paths." + "reason": "GPT-Neo and Flaubert/XLM are unrelated." }, { - "left": "pull_request:44017", - "right": "pull_request:44161", + "left": "pull_request:44071", + "right": "pull_request:44722", "accept": false, - "reason": "Segformer vs LongT5; unrelated implementations." + "reason": "MPT and GPT-J/CodeGen are unrelated implementation changes." }, { - "left": "pull_request:44017", - "right": "pull_request:44073", + "left": "pull_request:44018", + "right": "pull_request:44086", "accept": false, - "reason": "Segformer and VisualBert are different models with different output handling." + "reason": "GPT-Neo and MGP-STR do not share the same concrete code path." }, { - "left": "pull_request:43997", - "right": "pull_request:44073", + "left": "pull_request:44018", + "right": "pull_request:44085", "accept": false, - "reason": "RegNet vs VisualBert; same umbrella only." + "reason": "GPT-Neo and GPT-J are separate model refactors." }, { - "left": "pull_request:44073", + "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "VisualBert vs Flaubert/XLM; not the same concrete change." + "reason": "ViLT and Flaubert/XLM are unrelated model implementations." }, { - "left": "pull_request:44074", - "right": "pull_request:44161", + "left": "pull_request:44059", + "right": "pull_request:44722", "accept": false, - "reason": "TextNet and LongT5 are unrelated model refactors." + "reason": "GPT-2 and GPT-J/CodeGen are different model-specific changes." }, { - "left": "pull_request:44056", - "right": "pull_request:44101", + "left": "pull_request:44018", + "right": "pull_request:44071", "accept": false, - "reason": "MPNet vs Flaubert/XLM; different code paths." + "reason": "GPT-Neo and MPT are different code paths." }, { - "left": "pull_request:44056", + "left": "pull_request:44017", "right": "pull_request:44161", "accept": false, - "reason": "MPNet and LongT5 are distinct implementations." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44018", - "accept": false, - "reason": "RegNet and GPT-Neo are different models and different fixes." + "reason": "Segformer and LongT5 are unrelated models." }, { - "left": "pull_request:43997", - "right": "pull_request:44161", + "left": "pull_request:44017", + "right": "pull_request:44722", "accept": false, - "reason": "RegNet and LongT5 are unrelated changes." + "reason": "Segformer output tracing and GPT-J/CodeGen output tracing are not the same change." }, { - "left": "pull_request:43997", - "right": "pull_request:44071", + "left": "pull_request:44017", + "right": "pull_request:44073", "accept": false, - "reason": "RegNet and MPT are different model code paths." + "reason": "Segformer and VisualBert are separate implementations." }, { - "left": "pull_request:44010", + "left": "pull_request:44068", "right": "pull_request:44071", "accept": false, - "reason": "SqueezeBert and MPT are unrelated implementations." - }, - { - "left": "pull_request:44018", - "right": "pull_request:44129", - "accept": false, - "reason": "GPT-Neo and SpeechT5 are different model families." - }, - { - "left": "pull_request:44010", - "right": "pull_request:44073", - "accept": false, - "reason": "SqueezeBert vs VisualBert; same pattern, different concrete fix." + "reason": "GPT-Neo and MPT are unrelated model changes." }, { "left": "pull_request:43997", - "right": "pull_request:44074", + "right": "pull_request:44073", "accept": false, - "reason": "RegNet and TextNet are separate model-specific changes." + "reason": "RegNet and VisualBert are different code paths." }, { "left": "pull_request:44056", "right": "pull_request:44161", "accept": false, - "reason": "MPNet vs LongT5; not the same underlying bug." + "reason": "MPNet and LongT5 are unrelated implementations." }, { "left": "pull_request:43997", - "right": "pull_request:44018", + "right": "pull_request:44056", "accept": false, - "reason": "RegNet and GPT-Neo are different code paths." + "reason": "RegNet and MPNet are different model-specific refactors." }, { "left": "pull_request:43997", - "right": "pull_request:44071", + "right": "pull_request:44161", "accept": false, - "reason": "RegNet and MPT are unrelated model refactors." + "reason": "RegNet and LongT5 are unrelated." }, { - "left": "pull_request:44010", - "right": "pull_request:44044", + "left": "pull_request:44074", + "right": "pull_request:44101", "accept": false, - "reason": "SqueezeBert vs DeBERTa V2; different code paths." + "reason": "TextNet and Flaubert/XLM are different code paths." }, { - "left": "pull_request:44044", - "right": "pull_request:44074", + "left": "pull_request:44056", + "right": "pull_request:44161", "accept": false, - "reason": "DeBERTa V2 and TextNet are unrelated model implementations." + "reason": "MPNet and LongT5 are unrelated model changes." }, { - "left": "pull_request:44071", - "right": "pull_request:44722", + "left": "pull_request:43997", + "right": "pull_request:44018", "accept": false, - "reason": "MPT and GPT-J/CodeGen are different model-specific changes." + "reason": "RegNet and GPT-Neo are unrelated." }, { - "left": "pull_request:44018", - "right": "pull_request:44086", + "left": "pull_request:44010", + "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo vs MGP-STR; different models and different code paths." + "reason": "SqueezeBERT and MPT are different implementations." }, { "left": "pull_request:44018", - "right": "pull_request:44085", - "accept": false, - "reason": "GPT-Neo vs GPT-J; related theme but not the same concrete change." - }, - { - "left": "pull_request:44071", - "right": "pull_request:44085", - "accept": false, - "reason": "MPT and GPT-J are unrelated model refactors." - }, - { - "left": "pull_request:44068", - "right": "pull_request:44086", - "accept": false, - "reason": "GPT-Neo vs MGP-STR; different implementations." - }, - { - "left": "pull_request:44059", - "right": "pull_request:44086", + "right": "pull_request:44129", "accept": false, - "reason": "GPT-2 and MGP-STR are different model code paths." + "reason": "GPT-Neo and SpeechT5 are unrelated model changes." }, { - "left": "pull_request:44085", - "right": "pull_request:44086", + "left": "pull_request:44010", + "right": "pull_request:44073", "accept": false, - "reason": "GPT-J vs MGP-STR; not the same underlying bug or change." + "reason": "SqueezeBERT and VisualBert are unrelated." }, { - "left": "pull_request:44059", - "right": "pull_request:44071", + "left": "pull_request:43997", + "right": "pull_request:44074", "accept": false, - "reason": "GPT-2 and MPT are unrelated output-tracing refactors." + "reason": "RegNet and TextNet are different code paths." }, { - "left": "pull_request:44018", - "right": "pull_request:44059", + "left": "pull_request:43997", + "right": "pull_request:44129", "accept": false, - "reason": "GPT-Neo vs GPT-2; different model families." + "reason": "RegNet and SpeechT5 are unrelated model refactors." }, { - "left": "pull_request:44073", - "right": "pull_request:44161", + "left": "pull_request:44071", + "right": "pull_request:44129", "accept": false, - "reason": "VisualBert and LongT5 are different concrete refactors." + "reason": "MPT and SpeechT5 are unrelated." }, { - "left": "pull_request:44059", - "right": "pull_request:44068", + "left": "pull_request:44010", + "right": "pull_request:44018", "accept": false, - "reason": "GPT-2 vs GPT-Neo; same theme, different code paths." + "reason": "SqueezeBERT and GPT-Neo are different model implementations." }, { - "left": "pull_request:44101", + "left": "pull_request:43996", "right": "pull_request:44161", "accept": false, - "reason": "Flaubert/XLM and LongT5 are unrelated implementations." - }, - { - "left": "pull_request:44017", - "right": "pull_request:44018", - "accept": false, - "reason": "Segformer and GPT-Neo are unrelated model refactors." + "reason": "CVT/FNet and LongT5 are unrelated." }, { - "left": "pull_request:43997", - "right": "pull_request:44056", + "left": "pull_request:43996", + "right": "pull_request:44017", "accept": false, - "reason": "RegNet and MPNet are different model code paths." + "reason": "Different models and different code paths." }, { "left": "pull_request:44010", "right": "pull_request:44044", "accept": false, - "reason": "SqueezeBert and DeBERTa V2 are unrelated." + "reason": "SqueezeBERT and DeBERTa v2 are unrelated." }, { "left": "pull_request:44044", "right": "pull_request:44074", "accept": false, - "reason": "DeBERTa V2 and TextNet are different model-specific changes." + "reason": "DeBERTa v2 and TextNet are different implementations." }, { "left": "pull_request:44066", "right": "pull_request:44722", "accept": true, - "reason": "Same GPT-J/CodeGen output-tracing refactor with overlapping concrete files; 44722 is the strongest representative." + "reason": "Same GPT-J/CodeGen output-tracing refactor with substantial file overlap." }, { "left": "pull_request:44085", "right": "pull_request:44722", "accept": true, - "reason": "Both touch the same GPT-J output-tracing code path; 44722 subsumes the narrower GPT-J variant." + "reason": "Both are GPT-J output-tracing changes to the same model code path, so they are duplicate-like." }, { "left": "pull_request:44085", "right": "pull_request:44129", "accept": false, - "reason": "GPT-J and SpeechT5 are unrelated model changes." + "reason": "GPT-J and SpeechT5 are unrelated." }, { "left": "pull_request:43996", "right": "pull_request:44007", "accept": false, - "reason": "Different model-specific output-tracing refactors; not the same fix." + "reason": "Different vision model families and separate implementations." }, { "left": "pull_request:43996", "right": "pull_request:44129", "accept": false, - "reason": "CVT/FNet vs SpeechT5; unrelated code paths." + "reason": "CVT/FNet and SpeechT5 are unrelated code paths." }, { "left": "pull_request:44044", "right": "pull_request:44129", "accept": false, - "reason": "DeBERTa V2 and SpeechT5 are different implementations." + "reason": "DeBERTa v2 and SpeechT5 are different model-specific changes." }, { "left": "pull_request:44044", "right": "pull_request:44154", "accept": false, - "reason": "DeBERTa V2 vs ViTS; same umbrella pattern but different code paths." + "reason": "DeBERTa v2 and VITS are unrelated implementations." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The summary is not fully grounded: the packet also supports a separate clear duplicate subcluster for GPT-Neo (pull_request:44018 and pull_request:44068), so it is overstated to say the only duplicate subcluster is GPT-J/CodeGen. The GPT-J/CodeGen representative choice of 44722 looks reasonable, but the overall cluster description should acknowledge the GPT-Neo duplicates as well." + "accept": true, + "feedback": "Grounded overall. The summary stays conservative about the non-overlapping model families, and the accepted pairs are limited to the clearly mergeable GPT-J/CodeGen and GPT-Neo iterations with shared file/code-path overlap." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -18307,7 +18256,7 @@ { "label": "pull_request", "component_index": 14, - "component_count": 28, + "component_count": 27, "review_unit_index": 2, "review_unit_count": 2, "status": "reviewed", @@ -18334,70 +18283,70 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "940e277c31e9a8e2ab6928ec432fe0273b43c7bbdcaa6dbed1721a6d69e2050c", + "prepared_review_unit_hash": "8b0dd9ed8cc612b756a51bd4a909c5d635a5f71b03d1adf05ac785f0888bafe5", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84898, - "estimated_input_tokens": 21225, - "estimated_eval_tokens": 42706 + "serialized_chars": 84896, + "estimated_input_tokens": 21224, + "estimated_eval_tokens": 42704 }, "budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 6, - "serialized_chars": 17125, - "estimated_input_tokens": 4282, - "estimated_eval_tokens": 8820 + "serialized_chars": 17124, + "estimated_input_tokens": 4281, + "estimated_eval_tokens": 8818 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These PRs all participate in a broad repository-wide migration to standardized output tracing/capture, but they are model-specific refactors touching different architectures and code paths. They should not be treated as duplicates of one another.", - "confidence": 0.95, + "summary": "All items are separate model-specific pull requests that apply the same general output-tracing refactor pattern, but they target different architectures and code paths. They should not be deduplicated as the same change.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44129 is the clearest representative of the pattern because it has an explicit issue link and a focused output-tracing refactor, but it is still only one model-specific instance of a broader migration.", + "canonical_pr_reason": "44129 is the best representative PR because it has an explicit issue target and the most clearly scoped end-to-end refactor within a single model, but it is still not a duplicate of the others.", "best_issue_reason": null, - "best_pr_reason": "PR 44129 is the best single representative of the cluster theme, but there is no true duplicate-worthy canonical PR because each change is scoped to a different model implementation.", + "best_pr_reason": "44129 is the strongest standalone PR to keep: it is explicit about the target issue, has a clear model-specific change, and is more concretely described than the others.", "soft_edge_verdicts": [ { "left": "pull_request:44024", "right": "pull_request:44076", "accept": false, - "reason": "Both are output-capture migrations, but FocalNet and ImageGPT are different model implementations with distinct forward paths; not the same concrete change." + "reason": "Both are standardized-output refactors, but they touch different models (FocalNet vs ImageGPT) and different forward paths, so they are not the same underlying change." }, { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "Same refactor pattern, but VisualBert and TextNet are different models and the diffs target different code paths; too broad to merge as one PR." + "reason": "These are separate refactors for VisualBert and TextNet. The shared theme is output capture modernization, not one concrete bug or mergeable PR." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "Both modernize output tracing, but MPNet and VisualBert are separate architectures with different tracing hooks and output structures." + "reason": "MPNet and VisualBert use different model code paths and output-recording hooks; this is only broad subsystem similarity." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "Different model codepaths and different refactor details; this is a shared migration theme, not a duplicate fix/change." + "reason": "MPNet vs TextNet are distinct model implementations with different tracing changes, so they should not be merged as duplicates." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are unrelated model implementations; the PRs only share the generic output-tracing modernization pattern." + "reason": "RegNet and SqueezeBert are unrelated model-specific output refactors; same refactor style, not the same concrete fix." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "Both refactor output handling, but they apply to different models and do not appear to fix the same concrete bug or change." + "reason": "SqueezeBert attention tracing and SpeechT5 output capture are different code paths in different models, so they are not duplicative." } ] }, @@ -18411,7 +18360,7 @@ { "label": "pull_request", "component_index": 15, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18477,274 +18426,274 @@ "pull_request:44028|pull_request:44030", "pull_request:44029|pull_request:44030" ], - "prepared_review_unit_hash": "b075c75c5968dfd0d3142d250e86d4dc6cee4aa01e4bf113918346ff9a2f23af", + "prepared_review_unit_hash": "18e007608d35ef73251ef95ed87e6fade473a9f2480d66db9bf2ced88c754da2", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40168, - "estimated_input_tokens": 10042, - "estimated_eval_tokens": 20340 + "serialized_chars": 40178, + "estimated_input_tokens": 10045, + "estimated_eval_tokens": 20346 }, "budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40168, - "estimated_input_tokens": 10042, - "estimated_eval_tokens": 20340 + "serialized_chars": 40178, + "estimated_input_tokens": 10045, + "estimated_eval_tokens": 20346 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "A cluster of independent PRs that all refactor output tracing in different model implementations. They share the same broad theme, but they are not duplicates because they target different code paths and model families.", - "confidence": 0.93, + "summary": "This cluster is not a true duplicate set: all items are separate, model-specific output-tracing refactors that share a tooling pattern (`can_return_tuple` / output capturing) but not the same concrete bug or code path. I would keep them separate and, if forced to pick a representative, use the broadest mamba-related PR.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44003 is the strongest representative: it is the broadest refactor in the set, touches multiple related model files, and introduces the shared output-capturing machinery used by this tracing work.", + "canonical_pr_reason": "PR 44003 is the broadest representative of the shared refactor pattern, touching two related model files and introducing the new tracing helpers, but it is still model-specific rather than a duplicate of the others.", "best_issue_reason": null, - "best_pr_reason": "PR 44003 is the best overall exemplar of the cluster because it most clearly shows the common output-tracing refactor pattern and has the widest scope among the candidates.", + "best_pr_reason": "PR 44003 is the strongest single representative of the cluster\u2019s theme because it is the most substantial output-tracing change among the set and shows the new helper usage most clearly.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Same refactor theme, but different model implementations and call paths; not the same underlying change." + "reason": "Different model wrappers (`timm_backbone` vs `upernet`) with separate forward paths; same refactor style, not the same change." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "Both are output-tracing refactors, but they affect different models and code paths." + "reason": "Different models (`timm_backbone` vs `vision_text_dual_encoder`) and different APIs; only the tracing pattern overlaps." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "Same broad subsystem, but separate model-specific refactors, so not duplicates." + "reason": "Unrelated model implementations; both adopt output-tracing helpers but do not fix the same code path." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "Related tracing cleanup, but these are different model files and different concrete fixes." + "reason": "Separate model families (`timm_backbone` vs `mamba/falcon_mamba`) and distinct forward implementations." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "Both touch output handling, but in different model architectures and implementations." + "reason": "Different architectures (`timm_backbone` vs `rwkv`); shared utility usage is not enough for duplication." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "Same refactor pattern, but separate model-specific changes; not one mergeable PR." + "reason": "`mobilenet_v1` and `univnet` are unrelated models; these are parallel refactors, not the same bug." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "They share the tracing refactor theme, but target different models and code paths." + "reason": "Different subsystems (`mobilenet_v1` vs `upernet`) with separate output-shape/tracing code." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "Similar output-tracing cleanup, but on different model families and not the same bug/change." + "reason": "`mobilenet_v1` and `depth_anything` are distinct models; they only share a common tracing migration." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "Both are tracing refactors, but they are distinct architecture-specific implementations." + "reason": "Different encoder-decoder wrapper families and unrelated forward paths." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "Same broad area, but different encoder-decoder model code; not a duplicate." + "reason": "`mobilenet_v1` and `speech_encoder_decoder` are different code paths; not mergeable as one bug fix." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "These are separate model refactors with different forward paths, so they are not the same change." + "reason": "Different models (`vision_text_dual_encoder` vs `univnet`); same refactor theme only." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Same high-level refactor type, but different models and concrete edits." + "reason": "`vision_text_dual_encoder` and `upernet` are separate model wrappers with different outputs and call signatures." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "Related output-tracing work, but the affected code paths are different enough that these would not be one PR." + "reason": "Different model stacks (`vision_text_dual_encoder` vs `mamba/falcon_mamba`) and different output-capture plumbing." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "Both are output tracing refactors, but they target unrelated model implementations." + "reason": "Both are output-tracing refactors, but they target different models and cannot plausibly be one PR." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "Similar theme, but separate model-specific tracing updates; not duplicates." + "reason": "Unrelated models (`univnet` vs `upernet`); no shared concrete defect." }, { "left": "pull_request:44001", "right": "pull_request:44003", "accept": false, - "reason": "Different model families and different concrete code changes, despite the shared refactor theme." + "reason": "Different model families and forward code paths; only broad helper adoption overlaps." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "Both touch output handling, but in different model code paths and with different refactors." + "reason": "`univnet` and `depth_anything` are separate model-specific changes, not one bug." }, { "left": "pull_request:44002", "right": "pull_request:44029", "accept": false, - "reason": "Same general tracing cleanup, but these are distinct model implementations." + "reason": "`upernet` and `rwkv` are unrelated implementations; shared output-tracing refactor is too broad." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "They are not the same underlying bug/change; only the refactor style matches." + "reason": "Different models and separate forward semantics; not the same underlying issue." }, { "left": "pull_request:44003", "right": "pull_request:44028", "accept": false, - "reason": "Both are tracing refactors, but they target different models and cannot plausibly be merged into one PR." + "reason": "`mamba/falcon_mamba` vs `superpoint` are entirely different code paths; same refactor motif only." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "Different code paths and different refactor scopes; not a duplicate fix." + "reason": "`mamba`/`falcon_mamba` and `codegen` are unrelated model implementations with distinct outputs." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "Shared output-tracing theme only; the actual model code and behavior changes differ." + "reason": "Different model families and APIs; the overlap is only the new output-tracing helper pattern." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "These touch different models and different output semantics, so they are not the same change." + "reason": "`univnet` and `codegen` touch different forward paths and output contracts." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "Same subsystem theme, but the code paths are unrelated and not mergeable as one PR." + "reason": "Separate models (`upernet` vs `codegen`); no shared concrete bug or change." }, { "left": "pull_request:44000", "right": "pull_request:44004", "accept": false, - "reason": "Both mention tracing/output handling, but they are separate model-specific refactors." + "reason": "`vision_text_dual_encoder` and `codegen` are unrelated; similar refactor language is not duplication." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "Both are output-tracing refactors, but they are for different encoder-decoder wrappers and different files." + "reason": "`depth_anything` and `vision_encoder_decoder` are different model wrappers and different code paths." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "Similar refactor pattern, but different model families and forward paths." + "reason": "`depth_anything` and `speech_encoder_decoder` are unrelated models; not mergeable as one fix." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "Not the same underlying change; only the output-tracing cleanup theme overlaps." + "reason": "`depth_anything` and `superpoint` are separate vision models with distinct forward behavior." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "These are distinct model refactors, not one concrete bug fix." + "reason": "`depth_anything` and `rwkv` are unrelated model families; only the tracing migration matches." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "Same broad tracing refactor idea, but separate model-specific code paths." + "reason": "`depth_anything` and `dpr` are different architectures and output contracts." }, { "left": "pull_request:44026", "right": "pull_request:44029", "accept": false, - "reason": "Different model implementations; the shared pattern is too broad to be duplicate evidence." + "reason": "`vision_encoder_decoder` and `rwkv` are unrelated; same helper adoption does not imply duplicate fixes." }, { "left": "pull_request:44026", "right": "pull_request:44027", "accept": false, - "reason": "Both are encoder-decoder refactors, but they are still separate model-specific changes." + "reason": "Vision vs speech encoder-decoder wrappers are different code paths, not the same bug." }, { "left": "pull_request:44026", "right": "pull_request:44028", "accept": false, - "reason": "Related output handling changes, but not the same concrete fix." + "reason": "`vision_encoder_decoder` and `superpoint` are distinct model implementations." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "Different models and different forward signatures; not a duplicate." + "reason": "`vision_encoder_decoder` and `dpr` do not share a concrete defect; only the refactor pattern overlaps." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "Same general refactor style, but distinct models and code paths." + "reason": "Speech encoder-decoder and SuperPoint are unrelated models with different output structures." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "These are separate model-specific output-tracing changes, not one issue." + "reason": "Different model families and different forward signatures; not a duplicate." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "Different model code and behavior; the similarity is only thematic." + "reason": "`speech_encoder_decoder` and `dpr` are separate wrappers with no shared concrete bug." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "Both refactor model outputs, but in unrelated architectures and implementations." + "reason": "`superpoint` and `rwkv` are unrelated code paths; the commonality is only output tracing." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "Not the same concrete code-path problem; just a similar refactor pattern." + "reason": "Different models (`superpoint` vs `dpr`) and different output contracts." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "Separate model-specific tracing changes; too broad to treat as duplicates." + "reason": "`rwkv` and `dpr` are unrelated implementations; no evidence they fix the same underlying issue." } ] }, @@ -18758,7 +18707,7 @@ { "label": "pull_request", "component_index": 16, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18777,52 +18726,52 @@ "pull_request:44662|pull_request:45075", "pull_request:45075|pull_request:45401" ], - "prepared_review_unit_hash": "b5f86e0d5b074d6f5a851a5bcd515f4cc305ac91425e4d8ad37023615f31c8a1", + "prepared_review_unit_hash": "82528b93c5452b71a0585d099a0ef667a2130bfc4e37324d0d77b0243ecd5010", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 10912, - "estimated_input_tokens": 2728, - "estimated_eval_tokens": 5712 + "serialized_chars": 10913, + "estimated_input_tokens": 2729, + "estimated_eval_tokens": 5714 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 10912, - "estimated_input_tokens": 2728, - "estimated_eval_tokens": 5712 + "serialized_chars": 10913, + "estimated_input_tokens": 2729, + "estimated_eval_tokens": 5714 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate model-addition PRs (PenguinVL, DeepSeek-OCR-2, and Voxtral TTS). They share Transformers boilerplate and auto-registration files, but they target different models and codepaths, so they are not duplicates.", - "confidence": 0.99, - "canonical_issue_reason": "No canonical issue: there are no issue artifacts in the cluster, and the PRs are about different model integrations.", - "canonical_pr_reason": "No canonical PR for deduping: each PR adds a different model family with its own configuration/modeling/processing code.", - "best_issue_reason": "No issue candidate present.", - "best_pr_reason": "PR 45075 is the strongest standalone PR: it is more complete, has an explicit issue target, and includes a conversion script plus broader model/test coverage.", + "summary": "These are three separate model-addition PRs (PenguinVL, DeepSeek-OCR-2, Voxtral TTS) that share only common registry/doc boilerplate. They are not duplicate fixes or mergeable into one PR.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "pull_request:45401 is the strongest standalone representative because it has an explicit issue target and a focused, self-contained implementation for one model family.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different model additions (PenguinVL vs Voxtral TTS) with distinct files and behaviors; shared auto/doc boilerplate is not enough to treat them as the same change." + "reason": "Different models and different code paths; overlap is limited to shared Transformers registration/docs scaffolding, which is not enough to treat them as the same change." }, { "left": "pull_request:44662", "right": "pull_request:45075", "accept": false, - "reason": "Different model integrations (PenguinVL vs DeepSeek-OCR-2); they touch similar registration plumbing but implement unrelated models and codepaths." + "reason": "Both add distinct model implementations. Shared auto/registry files are generic model-onboarding changes, not evidence of the same underlying bug or feature." }, { "left": "pull_request:45075", "right": "pull_request:45401", "accept": false, - "reason": "DeepSeek-OCR-2 and Voxtral TTS are separate model families with different model logic, processing, and tests, so they are not the same underlying fix/change." + "reason": "Separate model introductions for unrelated architectures; they do not fix the same concrete problem and would not plausibly be merged as one PR." } ] }, @@ -18836,7 +18785,7 @@ { "label": "pull_request", "component_index": 17, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18852,12 +18801,12 @@ "soft_pairs": [ "pull_request:44979|pull_request:45363" ], - "prepared_review_unit_hash": "28353579dff7465bd08e240e8b1f5040e488d5ec36d8fb973363a074fa239622", + "prepared_review_unit_hash": "52ccd303c16513f897ff3b645d1fb989edb836e7a54953a42f557edb5664e91b", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4406, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -18865,7 +18814,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4406, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -18874,18 +18823,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests are about model/module fusion, but they target different layers of the stack: PR 44979 introduces a new Module Fusion API, while PR 45363 implements n-to-1 kernel fusion through KernelConfig and integration changes. They are related in theme but not the same underlying change.", - "confidence": 0.93, + "summary": "These two PRs are related around module/kernel fusion, but they target different layers: #44979 introduces a general Module Fusion API, while #45363 wires that API into KernelConfig for n-to-1 kernel fusion. They do not appear to be the same concrete fix/change, so they should not be merged as duplicates.", + "confidence": 0.86, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45363 is the more concrete and complete fusion change: it wires fusion through modeling, integration, and KernelConfig paths, whereas PR 44979 is a new API surface and helper module. If one PR should represent the cluster, 45363 is the better anchor.", + "canonical_pr_reason": "PR #44979 is the broader foundational change: it adds the new module_fusion abstraction and tests, which can underpin later integrations.", "best_issue_reason": null, - "best_pr_reason": "PR 45363 is the strongest representative of the cluster because it changes the actual runtime code path for fusion and includes a usage example, making it more substantial and central than the API-only scaffolding in 44979.", + "best_pr_reason": "PR #44979 is the best representative of the cluster because it captures the underlying API being introduced, rather than one specific consumer integration.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "Related feature area, but not the same bug/change: 44979 adds a Module Fusion API and test scaffold, while 45363 implements n-to-1 kernel fusion via KernelConfig and modifies different runtime integration points. They would not plausibly be merged as one PR." + "reason": "#44979 adds a generic Module Fusion API; #45363 adds KernelConfig-based n-to-1 kernel fusion wiring and an example. They are adjacent but not the same concrete bug/change, and would not plausibly be merged into one PR without conflating API foundation with a specific integration." } ] }, @@ -18899,7 +18848,7 @@ { "label": "pull_request", "component_index": 18, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18915,7 +18864,7 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "1e79a20410b04dd3909704f328d6cec58cd677fda0ef30f63a30af18a39aa4a5", + "prepared_review_unit_hash": "40e27eed167f0db272861bf2643d7355bfda449adca9e5a03af1fce8973cfaac", "component_budget": { "node_count": 2, "item_count": 2, @@ -18937,18 +18886,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs fix the same Mistral pre-tokenizer AttributeError by switching from backend_tokenizer.pre_tokenizer to tokenizer.pre_tokenizer in the same code path; the second PR appears to be a fuller version with tests, so they are duplicate-fix candidates.", + "summary": "Both pull requests fix the same Mistral regex patching AttributeError in `src/transformers/tokenization_utils_tokenizers.py`; 45317 is the more complete candidate because it includes a test update in addition to the same code fix.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45317 is the better canonical PR because it addresses the same bug as #45086 but is more complete, with an added regression test and more discussion/activity.", + "canonical_pr_reason": "PR 45317 is the better canonical PR because it implements the same bug fix as 45086 and also adds a regression test, making it the more complete representative of the change.", "best_issue_reason": null, - "best_pr_reason": "PR #45317 is the best representative of this cluster: same concrete fix, same target issue, and it includes test coverage beyond the smaller earlier patch.", + "best_pr_reason": "PR 45317 is the best overall PR candidate: same underlying fix, same touched code path, and it includes tests, whereas 45086 is just the minimal code change.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Both PRs make the same code change in src/transformers/tokenization_utils_tokenizers.py to fix the same AttributeError in _patch_mistral_regex; the later PR simply adds tests, so they could plausibly be merged as one fix." + "reason": "Same concrete bug in `_patch_mistral_regex`, same file, same fix lines, and the second PR is a superset with tests; these are effectively the same change." } ] }, @@ -18962,7 +18911,7 @@ { "label": "pull_request", "component_index": 19, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18978,12 +18927,12 @@ "soft_pairs": [ "pull_request:45115|pull_request:45569" ], - "prepared_review_unit_hash": "d6e06a85f81c3681d98740785adbe0b9db9bd6b6f8e8b5f9b927e7de9ebbb53a", + "prepared_review_unit_hash": "9ea6e766baff0f1a7a71109bc1e794b286bf314bcbecb56e39ef747f464a31e6", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6066, + "serialized_chars": 6065, "estimated_input_tokens": 1517, "estimated_eval_tokens": 3290 }, @@ -18991,7 +18940,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6066, + "serialized_chars": 6065, "estimated_input_tokens": 1517, "estimated_eval_tokens": 3290 }, @@ -19000,18 +18949,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related NemotronH architecture refactors, but they are not the same PR-level change: one is a conversion/inheritance refactor, the other is a broader model split into NemotronH Dense/Sparse with docs and tests. I would not deduplicate them.", - "confidence": 0.79, + "summary": "These are related NemotronH refactors, but they are not the same change: PR #45115 is a narrow conversion-mapping/inheritance refactor, while PR #45569 is a broader architecture split introducing separate dense/sparse model variants, docs, and tests.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45569 is the more complete and user-facing implementation: it introduces the Dense/Sparse model variants, updates auto-mapping, docs, and tests, so it better represents the end-state change.", + "canonical_pr_reason": "PR #45569 is the better canonical representative because it is the broader, more complete model-structure change and includes the user-facing docs/tests alongside code organization updates.", "best_issue_reason": null, - "best_pr_reason": "PR 45569 is the best representative of the cluster because it covers the full NemotronH restructuring rather than just the intermediate inheritance/conversion refactor in 45115.", + "best_pr_reason": "PR #45569 is the strongest standalone PR in the cluster due to its wider scope and clearer end-state; PR #45115 is a smaller supporting refactor and not a substitute for the full change.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "Related NemotronH work, but not the same concrete change. 45115 focuses on inheritance/conversion mapping and internal renames, while 45569 splits NemotronH into Dense/Sparse variants with additional docs/tests; they could coexist as separate PRs." + "reason": "They share the NemotronH subsystem, but they address different concrete changes: one focuses on conversion/name-mapping and inheritance cleanup, the other splits the architecture into dense/sparse variants with documentation and test updates. They do not look mergeable as one PR fixing the same underlying bug or implementing the same change." } ] }, @@ -19025,7 +18974,7 @@ { "label": "pull_request", "component_index": 20, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19041,12 +18990,12 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "6b0537aaf535d1f5dba284e5720fe3b94b6fe456651d475fa3cb40e5f57162e5", + "prepared_review_unit_hash": "1720418058ade51b413fce50d329a043c5017e8b8b103b89da3ff77f6f0503d1", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5397, + "serialized_chars": 5398, "estimated_input_tokens": 1350, "estimated_eval_tokens": 2956 }, @@ -19054,7 +19003,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5397, + "serialized_chars": 5398, "estimated_input_tokens": 1350, "estimated_eval_tokens": 2956 }, @@ -19063,18 +19012,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs in the same heterogeneous-config feature area, but they are not duplicates: one adds per-layer config support, the other adds broader per-layer modeling/cache integration.", - "confidence": 0.86, + "summary": "Two related but distinct PRs: one adds heterogeneous per-layer config support, the other adds heterogeneous model/runtime support. They share helper files, but they are not the same concrete change.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45332 is the better cluster representative because it is the broader, more feature-complete change and sits on top of the config groundwork, touching the core modeling/caching paths.", + "canonical_pr_reason": "PR #45332 is the broader umbrella change and appears to subsume the model-side heterogeneous support along with the config plumbing.", "best_issue_reason": null, - "best_pr_reason": "PR #45332 is the strongest single representative of the cluster because it covers the full heterogeneous-model support surface rather than only config plumbing.", + "best_pr_reason": "#45332 is the best representative PR for the cluster because it covers the larger end-to-end heterogeneous support work rather than only the config layer.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "Related feature work, but not the same concrete bug/change: #45333 adds heterogeneous per-layer config support, while #45332 adds broader heterogeneous model and cache behavior. They could be sequenced, but they are not the same PR-worthy fix." + "reason": "Related infrastructure, but not the same underlying change: #45333 is focused on per-layer config support, while #45332 adds broader model/runtime heterogeneous support. They would not naturally collapse into one single PR without mixing separate scopes." } ] }, @@ -19088,7 +19037,7 @@ { "label": "pull_request", "component_index": 21, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19104,12 +19053,12 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "f7a626ae9499246ae46d58f14ae927910cc19e0e66c0812a9031706c666f67f5", + "prepared_review_unit_hash": "3b37c09dc8cf81299d8aec42538ec2a802e8a8c6f306a4c34b5e9873abb8ee0c", "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 }, @@ -19117,7 +19066,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5513, + "serialized_chars": 5514, "estimated_input_tokens": 1379, "estimated_eval_tokens": 3014 }, @@ -19126,18 +19075,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both artifacts are typing-related, but they are not the same change: one is a broad type-checking sweep across many `src/transformers/*` files with a pyproject ignore tweak, while the other is a more focused typing-protocol update around `_typing.py`, `modeling_utils`, and related helpers. They should not be deduplicated.", - "confidence": 0.88, + "summary": "Both PRs are typing-related, but they address different layers: #45415 is a broad repo-wide type-checking enablement pass, while #45425 adds specific typing aliases/protocols to satisfy a few code paths. They overlap in a few touched files, but not in the same concrete fix, so they should not be merged as duplicates.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45425 is the more focused, concrete typing change: it adds shared typing protocols/aliases and updates the specific code paths that need them, rather than a broad mechanical type-checking sweep.", + "canonical_pr_reason": "#45425 is the more focused, concrete typing-change PR with a clear implementation surface and narrower scope.", "best_issue_reason": null, - "best_pr_reason": "PR 45425 is the better representative PR for this cluster because it captures the core typing-shim work in a self-contained way, whereas PR 45415 is a larger repo-wide typing pass with mixed-file mechanical edits.", + "best_pr_reason": "#45425 is the best single PR in the cluster because it targets a specific typing compatibility problem, whereas #45415 is a broad type-checking cleanup across many files.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Shared typing theme and a few overlapping files are not enough; 45415 is a broad type-checking cleanup across many modules, while 45425 adds specific typing protocols/aliases and targeted integrations. They are different scopes and not the same concrete fix." + "reason": "Shared typing theme and a few filenames, but #45415 is a broad type-checking pass and #45425 is a targeted typing-compatibility change; they do not look like the same underlying bug or one mergeable fix." } ] }, @@ -19151,7 +19100,7 @@ { "label": "pull_request", "component_index": 22, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19159,56 +19108,87 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-45487-3", + "cluster_id": "cluster-45487-6", "nodes": [ "pull_request:45487", "pull_request:45817", - "pull_request:45843" + "pull_request:45843", + "pull_request:45858", + "pull_request:45890", + "pull_request:45926" ], "soft_pairs": [ "pull_request:45487|pull_request:45843", - "pull_request:45817|pull_request:45843" + "pull_request:45487|pull_request:45926", + "pull_request:45817|pull_request:45843", + "pull_request:45858|pull_request:45890", + "pull_request:45858|pull_request:45926", + "pull_request:45890|pull_request:45926" ], - "prepared_review_unit_hash": "f70f311331d3ecc0aab9af19627b6f627ce98266875ee93d1508edebd0330e18", + "prepared_review_unit_hash": "c163ddb41c91c999a53df217577aaa1cb38c4082d9475705f53408369f6dca1b", "component_budget": { - "node_count": 3, - "item_count": 3, - "soft_pair_count": 2, - "serialized_chars": 7563, - "estimated_input_tokens": 1891, - "estimated_eval_tokens": 4038 + "node_count": 6, + "item_count": 6, + "soft_pair_count": 6, + "serialized_chars": 14704, + "estimated_input_tokens": 3676, + "estimated_eval_tokens": 7608 }, "budget": { - "node_count": 3, - "item_count": 3, - "soft_pair_count": 2, - "serialized_chars": 7563, - "estimated_input_tokens": 1891, - "estimated_eval_tokens": 4038 + "node_count": 6, + "item_count": 6, + "soft_pair_count": 6, + "serialized_chars": 14704, + "estimated_input_tokens": 3676, + "estimated_eval_tokens": 7608 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate pull requests that all address model-parallel / device-placement bugs, but in different model families and code paths. They are thematically similar, not duplicates.", - "confidence": 0.95, + "summary": "This cluster is mostly unrelated PRs that happen to mention parallelism or XPU. The only real overlap is the broad theme of model-parallel/device-placement fixes, but the concrete bugs and code paths are different enough that they should stay separate.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "PR 45487 is the strongest representative: it fixes an actual model-parallel device-placement bug in model code, spans multiple affected implementations, and includes tests. It is more substantive than the test-only expectation updates and broader than the single-model parallel fixes.", "best_issue_reason": null, - "best_pr_reason": "PR 45817 is the strongest standalone fix: it targets a concrete Gemma4 model-parallel bug, updates the relevant no-split/device handling, and includes a focused test file. The other PRs are similarly valid but for different models, so none is a better duplicate canonical than the others.", + "best_pr_reason": "PR 45487 is the best single PR for the cluster because it addresses a concrete runtime bug rather than only updating test expectations, and it has the clearest cross-model impact among the candidates.", "soft_edge_verdicts": [ { "left": "pull_request:45487", "right": "pull_request:45843", "accept": false, - "reason": "Both are model-parallel fixes, but they affect different architectures and code paths: AltCLIP/ChineseCLIP versus Deimv2. The shared theme is too broad to treat as the same underlying bug or merge into one PR." + "reason": "Both are model-parallel fixes, but for different underlying problems: altclip/chinese_clip device placement versus deimv2 backbone/_no_split_modules propagation. Not the same concrete bug." + }, + { + "left": "pull_request:45487", + "right": "pull_request:45926", + "accept": false, + "reason": "45487 is a runtime model-parallel bug fix; 45926 only adds XPU test expectations for a different model. Different artifact type and different underlying issue." }, { "left": "pull_request:45817", "right": "pull_request:45843", "accept": false, - "reason": "Gemma4 and Deimv2 have different parallelization issues in different model implementations. Similar symptom, but not the same concrete bug or fix." + "reason": "Gemma4 fixes device handling and no-split modules in one model; deimv2 fixes wrapper/backbone split-module propagation in another. Same theme, not the same change or bug." + }, + { + "left": "pull_request:45858", + "right": "pull_request:45890", + "accept": false, + "reason": "Both are test expectation updates, but they cover different models and different outputs. They do not appear to fix the same underlying defect." + }, + { + "left": "pull_request:45858", + "right": "pull_request:45926", + "accept": false, + "reason": "These are separate XPU expectation updates for different model test files with different logits. Shared XPU context is too broad to merge as duplicates." + }, + { + "left": "pull_request:45890", + "right": "pull_request:45926", + "accept": false, + "reason": "Same rationale as the other test-only pairs: distinct model-specific expectation updates, not one shared bug or change." } ] }, @@ -19222,7 +19202,7 @@ { "label": "pull_request", "component_index": 23, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19238,7 +19218,7 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "c399be9cd1192047a8e4a99ec895fea3780210e08df158581f44bf0b762d2548", + "prepared_review_unit_hash": "d442628b16118ce14ecb8c5551fae8c80edfe64713ee5a8a7a509f0088abe73c", "component_budget": { "node_count": 2, "item_count": 2, @@ -19260,18 +19240,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are related to expert-parallel loading, but they fix different bugs: 45548 addresses EP + DeepSpeed ZeRO-3 loading/config serialization, while 45662 fixes experts being overwritten during EP + FSDP2/rank-0 broadcast. They should remain separate.", - "confidence": 0.95, + "summary": "These are two related but distinct EP-loading fixes: one targets serialization/loading under accelerate + DeepSpeed ZeRO-3, the other targets expert parameters being overwritten during FSDP2/rank-0 broadcast. They overlap in modeling_utils but address different failure modes and code paths, so they should not be merged as duplicates.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "pull_request:45662 is the better representative PR because it fixes a specific concrete failure in the loading path and has the broader, more targeted EP/FSDP2 scope.", "best_issue_reason": null, - "best_pr_reason": "45662 is the stronger representative of the cluster topic because it targets a specific, concrete overwrite bug in the EP/FSDP2 code path with a focused fix.", + "best_pr_reason": "pull_request:45662 is the strongest standalone PR in the cluster: the bug description is concrete, the diff is tightly focused on the affected loading path, and the fix matches a clear runtime failure mode.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "Different failure modes and code paths: 45548 is about ZeRO-3/accelerate loading and config handling, while 45662 fixes expert parameters being overwritten under FSDP2/rank-0 broadcast. Shared EP/modeling_utils context is not enough to merge them." + "reason": "Same EP subsystem, but different underlying bugs: 45548 is about config serialization/loading with accelerate + ZeRO-3, while 45662 prevents expert tensors being overwritten during FSDP2 rank-0 broadcast. They are not the same code-path problem and should not be merged." } ] }, @@ -19285,7 +19265,7 @@ { "label": "pull_request", "component_index": 24, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19301,7 +19281,7 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "79bd36fef7e9ef36eb4eae2de2e52e34400db4a19f7aa0ec7706cd28683e23be", + "prepared_review_unit_hash": "a376a926c25f8ede7d60b3e3561461811140cdec4f978c345ceb1f0669f28392", "component_budget": { "node_count": 2, "item_count": 2, @@ -19323,18 +19303,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs target the same xdist collision problem around captured_info artifacts in testing_utils; 45645 is the broader, more complete fix and can subsume 45639.", - "confidence": 0.89, + "summary": "Both PRs address the same underlying xdist/captured_info artifact collision problem in transformers testing utilities; PR 45645 is the broader, more complete variant because it also updates CI log aggregation for the new multi-file output scheme.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45645 is the better canonical PR because it addresses the same captured_info xdist collision bug but also updates CI log collection, making it the more complete end-to-end fix.", + "canonical_pr_reason": "PR 45645 is the better canonical PR because it contains the core xdist-safe testing_utils change plus the related CI workflow updates needed to consume multiple captured_info files, making it the more complete fix for the shared issue.", "best_issue_reason": null, - "best_pr_reason": "PR #45645 is the best standalone PR: it fixes the core artifact collision in code and also preserves CI debug log aggregation for the new per-worker files.", + "best_pr_reason": "PR 45645 is the best representative because it covers the full end-to-end change, not just the library-side file naming fix.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", "accept": true, - "reason": "Both PRs fix the same underlying xdist collision in captured_info handling; 45645 appears to extend the same fix rather than address a different bug." + "reason": "Same concrete bug: xdist collisions in captured_info output. They share the same issue target and files, and 45645 looks like the superset that also updates CI handling for the new per-worker artifacts." } ] }, @@ -19348,7 +19328,7 @@ { "label": "pull_request", "component_index": 25, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19411,7 +19391,7 @@ { "label": "pull_request", "component_index": 26, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19477,70 +19457,7 @@ { "label": "pull_request", "component_index": 27, - "component_count": 28, - "review_unit_index": 1, - "review_unit_count": 1, - "status": "reviewed", - "reason": null, - "source": "llm", - "cache_hit": false, - "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-45858-2", - "nodes": [ - "pull_request:45858", - "pull_request:45890" - ], - "soft_pairs": [ - "pull_request:45858|pull_request:45890" - ], - "prepared_review_unit_hash": "42cc5fc1abc9ba885b89bd2dcc4ad00f52d3b8dce325d41877cf62101addee7a", - "component_budget": { - "node_count": 2, - "item_count": 2, - "soft_pair_count": 1, - "serialized_chars": 4139, - "estimated_input_tokens": 1035, - "estimated_eval_tokens": 2326 - }, - "budget": { - "node_count": 2, - "item_count": 2, - "soft_pair_count": 1, - "serialized_chars": 4139, - "estimated_input_tokens": 1035, - "estimated_eval_tokens": 2326 - }, - "overflow_policy": "truncate_then_skip", - "trimmed": false, - "aggressively_trimmed": false, - "split": false, - "analyst_result": { - "summary": "Two independent test-only XPU expectation PRs for different model families. They look similar at a high level, but they address separate model-specific golden outputs and are not the same underlying fix.", - "confidence": 0.93, - "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45890 is the better representative because it includes the XPU determinism guard plus updated expectations, making it the more complete example of the pattern in this cluster.", - "best_issue_reason": null, - "best_pr_reason": "PR 45890 is the strongest standalone candidate since it addresses a concrete XPU test adjustment in a self-contained way; PR 45858 is similarly scoped but narrower and unrelated in model scope.", - "soft_edge_verdicts": [ - { - "left": "pull_request:45858", - "right": "pull_request:45890", - "accept": false, - "reason": "Both are XPU expectation updates, but for different model test files and different logits; they do not fix the same code-path problem and would not plausibly be merged into one PR." - } - ] - }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, - "accepted_nontrivial_soft_edge": false, - "error_kind": null, - "error_message": null - }, - { - "label": "pull_request", - "component_index": 28, - "component_count": 28, + "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19556,40 +19473,40 @@ "soft_pairs": [ "pull_request:45861|pull_request:45903" ], - "prepared_review_unit_hash": "eec454b24656c4fbbb53a12e5422cb93c7688bfe6c4988c868967296029fb31d", + "prepared_review_unit_hash": "798d0a6f2ef2b4ded29c5f297dd5c1b2da29fecc1d9b695a7a2963ad882829b9", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5529, - "estimated_input_tokens": 1383, - "estimated_eval_tokens": 3022 + "serialized_chars": 5546, + "estimated_input_tokens": 1387, + "estimated_eval_tokens": 3030 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5529, - "estimated_input_tokens": 1383, - "estimated_eval_tokens": 3022 + "serialized_chars": 5546, + "estimated_input_tokens": 1387, + "estimated_eval_tokens": 3030 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs target the Qwen3-VL family M-RoPE path, and 45903 subsumes the device-mismatch fix from 45861 while adding the meta/to_empty re-init handling. They are closely related enough to treat as the same underlying change cluster, with 45903 as the broader canonical PR.", - "confidence": 0.93, + "summary": "Both PRs target the same Qwen3-VL M-RoPE `inv_freq` device-handling path, with 45903 extending the narrower device-mismatch fix in 45861 to also cover meta/to_empty re-initialization and buffer lifecycle correctness.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45903 is the more complete and newer fix: it includes the same inv_freq device placement change as 45861 and additionally addresses meta -> to_empty re-initialization in the same code path.", + "canonical_pr_reason": "45903 is the more complete and robust change: it covers the same Qwen3-VL M-RoPE device mismatch plus the meta/to_empty re-init edge case, with explicit buffer state tracking and a forward-time reinitialization guard.", "best_issue_reason": null, - "best_pr_reason": "PR 45903 is the best representative because it covers the shared bug plus the extra lifecycle fix, so it is the most complete single PR for the cluster.", + "best_pr_reason": "45903 is the best representative because it addresses the underlying `inv_freq` placement problem more comprehensively and includes the additional initialization fix that 45861 does not.", "soft_edge_verdicts": [ { "left": "pull_request:45861", "right": "pull_request:45903", "accept": true, - "reason": "Accept: 45903 contains the exact device-mismatch fix from 45861 and extends it with closely related M-RoPE initialization handling in the same Qwen3-VL family code paths, so they could plausibly be merged into one PR." + "reason": "Same concrete M-RoPE `inv_freq` device-path bug in Qwen3-VL family; 45903 is a broader superset of the narrower CPU-offload device fix in 45861 and could plausibly be merged as one change." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index 9ad92ce02d594b56e6c2ae56f9286cb5ee512595..a9d03b9993f3e9ef281ccfbdfea970569d881109 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260513t000101z", + "analysis_id": "hybrid-model-20260513t060025z", "archived_artifacts": { - "hybrid": "snapshots/20260513T000101Z/analysis-runs/hybrid-model-20260513t000101z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260513T000101Z/analysis-runs/hybrid-model-20260513t000101z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-05-13T00:09:44Z", + "published_at": "2026-05-13T06:15:26Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260513T000101Z", + "snapshot_id": "20260513T060025Z", "variant": "hybrid" } diff --git a/snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.json b/snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..5150ad2b1d6288832298807e05130020f605cd48 --- /dev/null +++ b/snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.json @@ -0,0 +1,2498 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260513T060025Z", + "generated_at": "2026-05-13T06:15:24Z", + "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-45812-4", + "summary": "Cluster of 2 issues and 2 PRs centered on issue #45812.", + "status": "open", + "confidence": 0.6, + "canonical_issue_number": 45812, + "canonical_pr_number": 45813, + "issue_numbers": [ + 45812, + 45920 + ], + "pr_numbers": [ + 45813, + 45924 + ], + "evidence_types": [ + "closing_reference", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45813, + "right_pr_number": 45924, + "code_similarity": 0.492, + "size_similarity": 0.833, + "file_overlap": 1.0, + "area_overlap": 0.0, + "patch_similarity": 0.167, + "shared_filenames": [ + "src/transformers/models/auto/tokenization_auto.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/auto/tokenization_auto.py", + "left_ranges": [ + [ + 146, + 155 + ] + ], + "right_ranges": [ + [ + 383, + 393 + ] + ] + } + ] + } + ] + }, + { + "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-45861-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 45861, + "issue_numbers": [], + "pr_numbers": [ + 45861, + 45903 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45861, + "right_pr_number": 45903, + "code_similarity": 0.506, + "size_similarity": 0.32, + "file_overlap": 1.0, + "area_overlap": 0.243, + "patch_similarity": 0.379, + "shared_filenames": [ + "src/transformers/models/qwen3_5/modeling_qwen3_5.py", + "src/transformers/models/qwen3_5_moe/modeling_qwen3_5_moe.py", + "src/transformers/models/qwen3_omni_moe/modeling_qwen3_omni_moe.py", + "src/transformers/models/qwen3_vl/modeling_qwen3_vl.py", + "src/transformers/models/qwen3_vl/modular_qwen3_vl.py", + "src/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py", + "tests/models/qwen3_vl/test_modeling_qwen3_vl.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/qwen3_5/modeling_qwen3_5.py", + "left_ranges": [ + [ + 146, + 154 + ] + ], + "right_ranges": [ + [ + 106, + 114 + ], + [ + 145, + 158 + ], + [ + 165, + 208 + ] + ] + }, + { + "filename": "src/transformers/models/qwen3_5_moe/modeling_qwen3_5_moe.py", + "left_ranges": [ + [ + 142, + 150 + ] + ], + "right_ranges": [ + [ + 102, + 110 + ], + [ + 141, + 154 + ], + [ + 161, + 204 + ] + ] + }, + { + "filename": "src/transformers/models/qwen3_omni_moe/modeling_qwen3_omni_moe.py", + "left_ranges": [ + [ + 1315, + 1323 + ] + ], + "right_ranges": [ + [ + 1277, + 1285 + ], + [ + 1314, + 1327 + ], + [ + 1334, + 1377 + ] + ] + }, + { + "filename": "src/transformers/models/qwen3_vl/modeling_qwen3_vl.py", + "left_ranges": [ + [ + 354, + 362 + ] + ], + "right_ranges": [ + [ + 316, + 324 + ], + [ + 353, + 366 + ], + [ + 373, + 416 + ] + ] + }, + { + "filename": "src/transformers/models/qwen3_vl/modular_qwen3_vl.py", + "left_ranges": [ + [ + 294, + 302 + ] + ], + "right_ranges": [ + [ + 32, + 38 + ], + [ + 269, + 315 + ], + [ + 331, + 344 + ] + ] + }, + { + "filename": "src/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py", + "left_ranges": [ + [ + 865, + 873 + ] + ], + "right_ranges": [ + [ + 827, + 835 + ], + [ + 864, + 877 + ], + [ + 884, + 927 + ] + ] + }, + { + "filename": "tests/models/qwen3_vl/test_modeling_qwen3_vl.py", + "left_ranges": [ + [ + 17, + 26 + ], + [ + 29, + 35 + ], + [ + 45, + 78 + ], + [ + 522, + 547 + ] + ], + "right_ranges": [ + [ + 17, + 26 + ], + [ + 29, + 35 + ], + [ + 45, + 78 + ], + [ + 522, + 607 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-44018-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44068, + "issue_numbers": [], + "pr_numbers": [ + 44018, + 44068 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44018, + "right_pr_number": 44068, + "code_similarity": 0.766, + "size_similarity": 0.939, + "file_overlap": 1.0, + "area_overlap": 0.425, + "patch_similarity": 0.866, + "shared_filenames": [ + "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "tests/models/gpt_neo/test_modeling_gpt_neo.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "left_ranges": [ + [ + 26, + 46 + ], + [ + 138, + 143 + ], + [ + 182, + 187 + ], + [ + 283, + 295 + ], + [ + 327, + 341 + ], + [ + 348, + 354 + ], + [ + 360, + 369 + ], + [ + 399, + 406 + ], + [ + 411, + 419 + ], + [ + 428, + 433 + ], + [ + 472, + 492 + ], + [ + 507, + 513 + ], + [ + 519, + 528 + ], + [ + 541, + 559 + ], + [ + 562, + 567 + ], + [ + 595, + 601 + ], + [ + 607, + 614 + ], + [ + 627, + 643 + ], + [ + 685, + 690 + ], + [ + 708, + 714 + ], + [ + 720, + 727 + ], + [ + 740, + 757 + ], + [ + 761, + 766 + ], + [ + 780, + 786 + ], + [ + 791, + 798 + ], + [ + 807, + 822 + ], + [ + 840, + 845 + ] + ], + "right_ranges": [ + [ + 26, + 31 + ], + [ + 34, + 44 + ], + [ + 136, + 141 + ], + [ + 180, + 185 + ], + [ + 281, + 293 + ], + [ + 325, + 339 + ], + [ + 346, + 352 + ], + [ + 358, + 367 + ], + [ + 397, + 404 + ], + [ + 409, + 417 + ], + [ + 426, + 434 + ], + [ + 463, + 483 + ], + [ + 498, + 504 + ], + [ + 510, + 519 + ], + [ + 532, + 550 + ], + [ + 553, + 564 + ], + [ + 586, + 592 + ], + [ + 598, + 605 + ], + [ + 618, + 634 + ], + [ + 676, + 688 + ], + [ + 699, + 705 + ], + [ + 711, + 718 + ], + [ + 731, + 748 + ], + [ + 752, + 762 + ], + [ + 771, + 777 + ], + [ + 782, + 789 + ], + [ + 798, + 813 + ], + [ + 831, + 836 + ] + ] + }, + { + "filename": "tests/models/gpt_neo/test_modeling_gpt_neo.py", + "left_ranges": [ + [ + 458, + 464 + ] + ], + "right_ranges": [ + [ + 458, + 464 + ] + ] + } + ] + } + ] + } + ], + "duplicate_issues": [ + { + "cluster_id": "cluster-45812-4", + "canonical_issue_number": 45812, + "duplicate_issue_numbers": [ + 45920 + ], + "reason": "Issues in cluster-45812-4 are treated as duplicates because they share closing_reference, soft_similarity evidence." + } + ], + "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-45081-3", + "canonical_pr_number": 45317, + "duplicate_pr_numbers": [ + 45086 + ], + "target_issue_number": 45081, + "reason": "PRs in cluster-45081-3 are treated as duplicates because they converge on issue #45081 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45561-3", + "canonical_pr_number": 45645, + "duplicate_pr_numbers": [ + 45639 + ], + "target_issue_number": 45561, + "reason": "PRs in cluster-45561-3 are treated as duplicates because they converge on issue #45561 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45647-3", + "canonical_pr_number": 45738, + "duplicate_pr_numbers": [ + 45737 + ], + "target_issue_number": 45647, + "reason": "PRs in cluster-45647-3 are treated as duplicates because they converge on issue #45647 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-45812-4", + "canonical_pr_number": 45813, + "duplicate_pr_numbers": [ + 45924 + ], + "target_issue_number": 45812, + "reason": "PRs in cluster-45812-4 are treated as duplicates because they converge on issue #45812 with closing_reference, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45861-2", + "canonical_pr_number": 45861, + "duplicate_pr_numbers": [ + 45903 + ], + "target_issue_number": null, + "reason": "PRs in cluster-45861-2 are treated as duplicates because they share soft_similarity evidence." + } + ], + "best_issue": { + "cluster_id": "cluster-45812-4", + "issue_number": 45812, + "reason": "Issue #45812 is the strongest global issue candidate because it is open, belongs to a cluster with 4 artifacts, and carries 2 discussion comments plus 5 inbound references.", + "score": 149.0 + }, + "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/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..98a260e3021a6081f67aebc3b6977b6c038e2e24 --- /dev/null +++ b/snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,19521 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260513T060025Z", + "generated_at": "2026-05-13T06:15:24Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:24643", + "issue:30277", + "issue:34634", + "issue:34928", + "issue:35545", + "issue:36010", + "issue:36331", + "issue:39290", + "issue:41093", + "issue:41720", + "issue:41762", + "issue:42915", + "issue:43039", + "issue:43425", + "issue:43541", + "issue:43596", + "issue:43638", + "issue:43716", + "issue: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:44805", + "issue:44841", + "issue:44863", + "issue:44918", + "issue:45081", + "issue:45161", + "issue:45237", + "issue:45464", + "issue:45507", + "issue:45685", + "issue:45812", + "issue:45823", + "issue:45859", + "issue:45874", + "issue:45901", + "issue:45902", + "issue:45920" + ], + "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:44805|issue:45823", + "issue:43638|issue:44805", + "issue:43596|issue:44805", + "issue:44805|issue:45161", + "issue:44805|issue:44918", + "issue:34634|issue:35545", + "issue:30277|issue:43638", + "issue:43596|issue:43638", + "issue:43039|issue:45161", + "issue:43975|issue:44457", + "issue:44292|issue:44457", + "issue:34928|issue:44805", + "issue:24643|issue:30277", + "issue:45859|issue:45902", + "issue:43638|issue:45161", + "issue:34928|issue:36331", + "issue:43596|issue:45161", + "issue:43716|issue:45507", + "issue:39290|issue:45081", + "issue:45823|issue:45874", + "issue:43828|issue:45161", + "issue:43425|issue:44292", + "issue:34928|issue:41720", + "issue:44918|issue:45161", + "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:44918|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:45237|issue:45901", + "issue:44841|issue:45823", + "issue:43039|issue:44560", + "issue:45812|issue:45920", + "issue:43866|issue:44863", + "issue:43854|issue:43866", + "issue:41093|issue:43638", + "issue:43792|issue:45901" + ], + "prepared_review_unit_hash": "a2f3104273f2a6b93ba3d9fc44eeaf994fecaf3c740760c999a2eb16e1520e75", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 62, + "serialized_chars": 46367, + "estimated_input_tokens": 11592, + "estimated_eval_tokens": 23440 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster contains a few real duplicate families amid lots of unrelated regressions: the broad v5 tokenizer wrong-ID reports and the ZeRO-3 zero-size index-out-of-bounds reports. The mask/tensor shape mismatch issue 41093 should not be grouped with the ZeRO-3 load failures; it is a different failure mode.", + "confidence": 0.71, + "canonical_issue_reason": "Issue 45920 is the best canonical issue because it is the broadest report in the clearest duplicate family: a v5 AutoTokenizer regression affecting multiple model families, with 45812 as a narrower Granite-specific instance.", + "canonical_pr_reason": null, + "best_issue_reason": "45920 is the best global issue candidate because it is the most general, user-visible regression in the cluster and spans the widest set of affected models.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Different Qwen3.5 failures and code paths: missing config attribute vs streaming chat/completions inference error." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Different backends and bugs: AMD GPU failure vs MPS histogram op not implemented for int tensors." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Different model families and symptoms: image-preprocessor dtype mismatch vs GPT-OSS on AMD GPUs." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Both are dtype/device issues, but on different models and different failure points; not the same bug." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Different model-specific issues: AMD GPU execution failure vs GraniteMoEHybrid calling an invalid method." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Different hardware/model problems with no shared concrete code path." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "DeepSpeed collective mismatch is unrelated to ModernBERT ONNX export." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "ONNX export failure and Qwen3Moe FP8 failure are different subsystems and symptoms." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "ModernBERT export and Qwen3 auto-device-map cuda assert are unrelated issues." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs regressions in v5, but they target different removed features and likely different pages/edits." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "Different problems: export-time failure vs import-time GenerationMixin breakage." + }, + { + "left": "issue:44805", + "right": "issue:45823", + "accept": false, + "reason": "Mask/tensor shape IndexError is unrelated to Gemma4 device mismatch during forward." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Both are IndexErrors, but one is ZeRO-3 zero-size indexing and the other is a mask shape mismatch; not the same bug." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "Different failure modes: ZeRO-3 zero-size indexing vs mask shape mismatch." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Shape-mismatch IndexError is unrelated to GPT-OSS tensor-parallel behavior." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Different code paths: mask indexing error vs TRL SFT/input-embedding unpacking failure." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "BarkProcessor voice preset handling is unrelated to ONNX export." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Both involve DeepSpeed, but one is a collective mismatch and the other is a ZeRO-3 zero-size indexing bug." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "Same ZeRO-3 load-time index 0 is out of bounds for dimension 0 with size 0 failure; model names differ, but the bug appears identical." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger Kernel cross-entropy dispatch and GPT-OSS tensor-parallel failure are different issues." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "Incorrect detokenization and LoRA merge/save/reload inconsistency are separate bugs." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Qwen NVFP4 runtime error is unrelated to LoRA merge/save behavior." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "FSDP activation-checkpointing size mismatch is a different bug from the mask shape IndexError." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "Different DeepSpeed failures: weight must be 2-D vs collective mismatch on ranks." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": false, + "reason": "Same component family, but different concrete bugs: CPU offload incompatibility vs uninitialized-memory read after to_empty(cuda)." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "BERT ZeRO-3 zero-size indexing is unrelated to GPT-OSS tensor-parallel support." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "FSDP recomputed tensor size mismatch and compute_loss() signature breakage are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "ZeRO-3 BERT loading failure is unrelated to GPT-OSS tensor-parallel behavior." + }, + { + "left": "issue:43716", + "right": "issue:45507", + "accept": false, + "reason": "Different model-specific failures with no shared code path." + }, + { + "left": "issue:39290", + "right": "issue:45081", + "accept": false, + "reason": "Gemma3 config missing attribute in vLLM and Mistral tokenizer regex crash are unrelated." + }, + { + "left": "issue:43828", + "right": "issue:44292", + "accept": false, + "reason": "Different model/runtime combinations and different dtype-related failures." + }, + { + "left": "issue:43039", + "right": "issue:44918", + "accept": false, + "reason": "Cross-entropy dispatch under Liger Kernel and TRL SFT embedding unpacking are different bugs." + }, + { + "left": "issue:41762", + "right": "issue:43638", + "accept": true, + "reason": "Same ZeRO-3 load-time index 0 is out of bounds for dimension 0 with size 0 error; only the model wrapper differs." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Both are v5 docs cleanups, but they remove references to different features and are not clearly the same change." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Image-preprocessor dtype mismatch and Qwen3-vl video StopIteration are different issues." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Autocast dtype mismatch is unrelated to the MPS histogram kernel error." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "TRL SFT embedding unpacking and MPS int-histogram failure are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Different models and different failure points, despite both involving loss/dtype behavior." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch and trainer compute_loss signature mismatch are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP checkpointing mismatch and GenerationMixin import failure are different subsystems." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Liger Kernel cross-entropy issue is unrelated to the mask shape mismatch IndexError." + }, + { + "left": "issue:43901", + "right": "issue:44512", + "accept": false, + "reason": "Docs mentioning return_all_scores and docs mentioning the removed run command are separate documentation issues." + }, + { + "left": "issue:41762", + "right": "issue:43596", + "accept": true, + "reason": "Same ZeRO-3 zero-size index-out-of-bounds load failure; only the model wrapper differs." + }, + { + "left": "issue:39290", + "right": "issue:41720", + "accept": false, + "reason": "Gemma3 config missing attribute in vLLM and Qwen3 auto-device-map cuda assert are different bugs." + }, + { + "left": "issue:45237", + "right": "issue:45901", + "accept": false, + "reason": "GPT-OSS AMD GPU failure is unrelated to the table-question-answering crash." + }, + { + "left": "issue:44841", + "right": "issue:45823", + "accept": false, + "reason": "Voxtral processor failure and Gemma4 device mismatch are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44560", + "accept": false, + "reason": "Liger Kernel cross-entropy issue and Qwen3-vl video StopIteration are unrelated." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same tokenizer regression family: AutoTokenizer returns wrong token IDs across multiple model families; 45920 is the broader report and 45812 is a narrower Granite-specific instance." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Checkpoint corruption and checkpoint loading implementation failure are not the same bug." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "Unit-test model loading failure and corrupted checkpoint are different root causes." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "The mask/tensor shape mismatch in 41093 is a different failure mode from the ZeRO-3 zero-size index-out-of-bounds load error in 43638." + }, + { + "left": "issue:43792", + "right": "issue:45901", + "accept": false, + "reason": "Whisper model load/run failure is unrelated to the table-question-answering crash." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Rejecting because one soft-edge verdict is too aggressive: issue:41093 is not grounded as a duplicate of the DeepSpeed ZeRO-3 zero-size indexing reports. Its title is a mask/tensor shape mismatch, while issue:43638 is a load-time `index 0 is out of bounds for dimension 0 with size 0` failure. Those look like different bugs, so the accepted duplicate link is not conservative enough. The tokenizer cluster acceptance (45812 vs 45920) and the ZeRO-3 pairings are plausible, but the overreach on 41093 means the overall output should not be accepted." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 2, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:36010", + "issue:39186", + "issue:39290", + "issue:40990", + "issue:41093", + "issue:41762", + "issue:43054", + "issue:43296", + "issue:43366", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43645", + "issue:43716", + "issue:43792", + "issue:43828", + "issue:43866", + "issue:43975", + "issue:43994", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44530", + "issue:44589", + "issue:44841", + "issue:44918", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45084", + "issue:45161", + "issue:45362", + "issue:45399", + "issue:45464", + "issue:45542", + "issue:45632", + "issue:45685", + "issue:45698", + "issue:45759", + "issue:45800", + "issue:45823", + "issue:45901", + "issue:45902" + ], + "soft_pairs": [ + "issue:39290|issue:43531", + "issue:43425|issue:45070", + "issue:43596|issue:45800", + "issue:43975|issue:45901", + "issue:45362|issue:45464", + "issue:43792|issue:45070", + "issue:44292|issue:45901", + "issue:43404|issue:43866", + "issue:39290|issue:45823", + "issue:45464|issue:45685", + "issue:43541|issue:45685", + "issue:43541|issue:43828", + "issue:36010|issue:39290", + "issue:44841|issue:45084", + "issue:41093|issue:41762", + "issue:45800|issue:45823", + "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:43366|issue:45464", + "issue:45020|issue:45800", + "issue:43531|issue:45070", + "issue:43541|issue:44589", + "issue:41093|issue:43596", + "issue:43054|issue:43994", + "issue:45759|issue:45902", + "issue:45542|issue:45800", + "issue:40990|issue:44960", + "issue:43716|issue:44918", + "issue:39186|issue:44918", + "issue:45632|issue:45698", + "issue:43404|issue:45005", + "issue:43645|issue:45399", + "issue:44530|issue:45005" + ], + "prepared_review_unit_hash": "fba4059aae310db279bfa674131424140ddba385b0d45a5017683d34a96867bd", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 37, + "serialized_chars": 37920, + "estimated_input_tokens": 9480, + "estimated_eval_tokens": 19216 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous transformer bug reports with no strong duplicate core. The only plausible thematic pair, 45632/45698, is still too weak to merge because the evidence is only soft similarity and the two reports are not shown to share the same concrete failure path.", + "confidence": 0.28, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45020 is the broadest open umbrella for recent `remote_code` regressions, but this cluster remains mixed and does not support a stronger canonical duplicate anchor.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Both mention `sliding_window`, but they involve different model families and error paths." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Torch-version incompatibility and a `PretrainedConfig`/pydantic breakage are different failures." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "Both are runtime/version problems, but the concrete bugs are unrelated." + }, + { + "left": "issue:43975", + "right": "issue:45901", + "accept": false, + "reason": "Detokenization regression and table-question-answering crash do not share a code path." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Both touch Qwen3.5 chat, but one is a chat crash and the other is a streaming API failure." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "Whisper load failure is unrelated to the pydantic/config regression." + }, + { + "left": "issue:44292", + "right": "issue:45901", + "accept": false, + "reason": "Qwen NVFP4 runtime error and table QA crash are different model/task failures." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "`lm_head` tying and a checkpoint corruption report are not the same bug." + }, + { + "left": "issue:39290", + "right": "issue:45823", + "accept": false, + "reason": "Gemma3 missing-attribute and Gemma4 device-mismatch failures are different issues." + }, + { + "left": "issue:45464", + "right": "issue:45685", + "accept": false, + "reason": "Chat streaming failure and an MPS histogram/int error are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:45685", + "accept": false, + "reason": "Both are MoE-related, but the traced `grouped_mm` error and the MPS int histogram error are different." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Both are dtype-sensitive runtime errors, but they affect different models and execution paths." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "`GenerationMixin` import failure is unrelated to the Gemma3 `sliding_window` attribute issue." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Processor failure and non-template-node compilation are different subsystems." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both are index errors, but the mask-shape mismatch and zero-sized tensor index-out-of-bounds are distinct bugs." + }, + { + "left": "issue:45800", + "right": "issue:45823", + "accept": false, + "reason": "Torch/transformers incompatibility and a device placement mismatch are not the same defect." + }, + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "PaddleOCR-VL loading and GGUF gpt-oss support are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "Missing `pad_token_idx` in `StableLmConfig` is separate from the pydantic `PretrainedConfig` regression." + }, + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Int4 memory/OOM behavior and tied-weights translation issues are different bugs." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Grouped-`mm` tracing failure and GPT-OSS tensor-parallel-only failure are different MoE issues." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "Autocast dtype mismatch and chat streaming failure are different code paths." + }, + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "`init_empty_weights` TypeError is not the same as the `PretrainedConfig` pydantic break." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "GGUF model support and chat/completions streaming are unrelated." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "`remote_code` regressions and torch/transformers incompatibility are different problems." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Qwen3-MoE `sliding_window` behavior does not match the pydantic config regression." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Grouped-`mm` tracing failure and missing Float8 storage are different runtime bugs." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Both are zero-3-related index errors, but with different models, shapes, and failure mechanisms." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": false, + "reason": "Both concern Siglip2 quality, but the symptoms differ and the data do not establish the same root cause." + }, + { + "left": "issue:45759", + "right": "issue:45902", + "accept": false, + "reason": "Composite config unwrapping and uninitialized rotary-embedding memory are different Qwen issues." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "TensorBoard backend error is unrelated to the torch compatibility issue." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "High perplexity on gpt-oss and a GLM5 report are not the same bug." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Gemma dtype mismatch and Qwen3.5 embedding unpacking are different model behaviors." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "FSDP weight-dimension error and Qwen3.5 embedding unpacking are unrelated." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": false, + "reason": "The reports are thematically similar, but the packet only supports soft similarity; the shared custom-code loading theme is not enough to confirm the same concrete bug." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Tied-weights issues are related at a high level, but these are not the same concrete defect." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Jupyter custom-model initialization breakage and flash-attn fallback gating are different issues." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "`linear_attention` crash and tied-weights translation issues are unrelated." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Reject: the response overstates evidence on the only accepted soft edge. In the packet, issue:45632 vs issue:45698 is marked deterministic_accept=false and supported only by soft_similarity, so treating it as a confirmed same-bug pair is too strong. The canonical_issue_reason also goes beyond the packet by claiming an explicit cache/path-collision root cause and a duplicate-anchor relationship that are not established by the provided data. The rest of the soft-edge rejections are conservative, but this one accepted edge is not." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 3, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:36331", + "issue:38175", + "issue:39692", + "issue:41553", + "issue:41628", + "issue:41762", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43493", + "issue:43643", + "issue:43644", + "issue:43782", + "issue:43828", + "issue:43856", + "issue:43866", + "issue:43872", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:43950", + "issue:43975", + "issue:44292", + "issue:44360", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44492", + "issue:44512", + "issue:44534", + "issue:44589", + "issue:44863", + "issue:44910", + "issue:44918", + "issue:44928", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45237", + "issue:45399", + "issue:45507", + "issue:45520", + "issue:45663", + "issue:45701", + "issue:45800", + "issue:45859", + "issue:45902" + ], + "soft_pairs": [ + "issue:44360|issue:44512", + "issue:42915|issue:45399", + "issue:43873|issue:45005", + "issue:44863|issue:45005", + "issue:44492|issue:44512", + "issue:44918|issue:45507", + "issue:43975|issue:44292", + "issue:43950|issue:44534", + "issue:44292|issue:45507", + "issue:43644|issue:43950", + "issue:39692|issue:43054", + "issue:43828|issue:44928", + "issue:36331|issue:41762", + "issue:43866|issue:44403", + "issue:41628|issue:45800", + "issue:43950|issue:45902", + "issue:43054|issue:45701", + "issue:43872|issue:44589", + "issue:41553|issue:43883", + "issue:38175|issue:43054", + "issue:44910|issue:45859", + "issue:45663|issue:45902", + "issue:41553|issue:43881", + "issue:42947|issue:44387", + "issue:43054|issue:43493", + "issue:44403|issue:44451", + "issue:43643|issue:45020", + "issue:43975|issue:45520", + "issue:43975|issue:45237", + "issue:44403|issue:44488", + "issue:42947|issue:43856", + "issue:43782|issue:44960", + "issue:43881|issue:43883", + "issue:41762|issue:42915" + ], + "prepared_review_unit_hash": "fccb4f311e0d441938ca68053e799f9afd09637c3f918aba80561cf77144e63a", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37475, + "estimated_input_tokens": 9369, + "estimated_eval_tokens": 18994 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is still mostly heterogeneous. Issue 43950 remains the best canonical issue for the non-persistent-buffer corruption regression, and 44534 is a clear duplicate. Issue 45902 is related in area but not conservative enough to accept as a duplicate from the packet alone, so it should stay separate/uncertain.", + "confidence": 0.79, + "canonical_issue_reason": "43950 is the clearest umbrella report for the Transformers v5 non-persistent-buffer corruption regression (`register_buffer(persistent=False)` / junk on load).", + "canonical_pr_reason": null, + "best_issue_reason": "43950 best represents the shared bug family because it states the regression broadly and precisely, while 44534 is a direct duplicate and 45902 is only a possible related manifestation.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "Unrelated topics: DSA indexer discussion versus docs/CLI removal." + }, + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "Different failure modes: Qwen3Moe FP8 config versus flash-attn fallback gating." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization/offloading behavior is not the same bug as tied-weights handling in translation models." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint loading is a different defect from tied-weights issues in v5." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "A typo in cache strategies is unrelated to the removed CLI/docs reference." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Different code paths and symptoms: TRL SFT embedding unpacking versus GraniteMoEHybrid method dispatch." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "Different runtime failures: detokenization regression versus NVFP4 run failure." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "Both report the same v5 regression where non-persistent buffers are corrupted/fill with junk on load." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": false, + "reason": "Remote-code config field loss is a different bug from non-persistent-buffer corruption." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "Both mention SigLIP2, but one is a docs/example problem and the other is an embedding-quality complaint." + }, + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Autocast dtype mismatch is not the same bug as Qwen3.5 position-id/SDPA instability." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "Trainer API kwarg mismatch is unrelated to Gemma3 ZeRO-3 loading failures." + }, + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "Checkpoint corruption is not the same as generic loading noise." + }, + { + "left": "issue:41628", + "right": "issue:45800", + "accept": false, + "reason": "Import/export symbol issues and torch-version incompatibility are different problems." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": false, + "reason": "Related area, but 45902 describes uninitialized `inv_freq` after `meta -> to_empty(cuda)` materialization and is not clearly the same regression as the non-persistent-buffer junk-on-load bug." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "Tokenization/model-behavior complaints, but the symptoms and likely root causes differ." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "Different quantization backend errors: Int8Params kwarg mismatch versus Float8 storage lookup failure." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Bad AutoTokenizer error reporting and missing tied-weights metadata are unrelated." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "Both are SigLIP2-related, but zero probabilities and lower embedding quality are not clearly the same bug." + }, + { + "left": "issue:44910", + "right": "issue:45859", + "accept": false, + "reason": "Flash-attn illegal memory access from 3D position_ids is not the same as CPU-offload incompatibility in rotary embedding." + }, + { + "left": "issue:45663", + "right": "issue:45902", + "accept": false, + "reason": "Gemma4 FSDP2 shared_kv_states KeyError is unrelated to buffer materialization corruption." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Bad AutoTokenizer error reporting and GLM4V loading failure are different issues." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "Gradient checkpointing ineffectiveness and int4 memory bloat/OOM are distinct bugs." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": false, + "reason": "SigLIP2 embedding quality complaint and HF-vs-JAX discrepancy are related in theme but not clearly the same bug." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic loading noise and a specific model loading failure are not the same underlying defect." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "Missing fields from trust_remote_code config loading is a specific bug; 45020 is a broader umbrella report." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "Detokenization regression and Python 3.13 flash_attn import KeyError are unrelated." + }, + { + "left": "issue:43975", + "right": "issue:45237", + "accept": false, + "reason": "Detokenization regression and AMD GPU support/loading are different failures." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "Different model loading reports with no clear shared code-path bug." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "Both mention memory, but one is LoRA checkpointing behavior and the other is Qwen3 MoE training memory usage." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Qwen3VL weight_only loading error is unrelated to the GLM5 report." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "Loading failure for glm-4v-9b and missing all_tied_weights_keys are different defects." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Gemma3 ZeRO-3 loading error is unrelated to Qwen3Moe FP8 configuration failure." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster summary is mostly grounded, but one soft-edge verdict is not conservative enough: treating issue 45902 as a duplicate of 43950 goes beyond what the packet supports from titles alone. 45902 describes uninitialized `inv_freq` memory after `meta -> to_empty(cuda)` materialization, which may be related, but it is not clearly the same regression as the non-persistent-buffer junk-on-load bug in 43950/44534. The 44534 link is well supported; the 45902 link should be downgraded to uncertain rather than accepted." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 4, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36331", + "issue:41093", + "issue:41553", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43366", + "issue:43404", + "issue:43638", + "issue:43742", + "issue:43792", + "issue:43824", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43872", + "issue:43940", + "issue:43950", + "issue:44220", + "issue:44291", + "issue:44368", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44530", + "issue:44560", + "issue:44661", + "issue:44805", + "issue:44863", + "issue:44928", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45305", + "issue:45313", + "issue:45325", + "issue:45356", + "issue:45399", + "issue:45538", + "issue:45588", + "issue:45685", + "issue:45823", + "issue:45902", + "issue:45923" + ], + "soft_pairs": [ + "issue:42491|issue:44368", + "issue:45685|issue:45823", + "issue:43824|issue:43872", + "issue:45020|issue:45823", + "issue:44661|issue:45020", + "issue:44991|issue:45399", + "issue:43828|issue:45823", + "issue:43827|issue:44509", + "issue:45081|issue:45399", + "issue:45020|issue:45356", + "issue:43404|issue:43950", + "issue:43742|issue:45399", + "issue:43257|issue:43866", + "issue:42915|issue:45020", + "issue:43638|issue:45823", + "issue:43792|issue:44220", + "issue:43872|issue:44291", + "issue:43940|issue:45313", + "issue:42915|issue:43872", + "issue:43854|issue:44863", + "issue:41553|issue:44991", + "issue:42947|issue:45923", + "issue:43257|issue:45305", + "issue:36331|issue:41093", + "issue:44928|issue:45588", + "issue:41093|issue:44805", + "issue:43872|issue:45538", + "issue:41553|issue:44488", + "issue:44530|issue:44863", + "issue:44479|issue:44560", + "issue:45325|issue:45902", + "issue:24643|issue:30064", + "issue:43366|issue:43828" + ], + "prepared_review_unit_hash": "c236b628e099e50cf069dfc2fb32a015813810e1678d11bd06adc896faa53e66", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 38062, + "estimated_input_tokens": 9516, + "estimated_eval_tokens": 19288 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a heterogeneous set of recent Transformers bug reports with superficial title similarity; only a few pairs look like true duplicates (docs cleanup, mask-shape IndexError, and the `_is_hf_initialized` loading regression).", + "confidence": 0.69, + "canonical_issue_reason": "issue:45020 is the broadest umbrella-style report in the set and the closest thing to a cluster anchor, covering recent-version model-loading breakage for `remote_code` models.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the most generally representative issue here: broad, recent, and centered on model-loading regressions rather than a narrow model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "Both involve Qwen/LoRA wording, but one is a version-compatibility failure and the other is a warning about `tie_word_embeddings`; not the same bug." + }, + { + "left": "issue:45685", + "right": "issue:45823", + "accept": false, + "reason": "Different failure modes on different paths: MPS `histogram_mps` int issue vs Gemma4 device placement mismatch." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "Import error for a missing Qwen2.5-VL class vs bitsandbytes `Int8Params` constructor incompatibility; unrelated code paths." + }, + { + "left": "issue:45020", + "right": "issue:45823", + "accept": false, + "reason": "`remote_code` loading regressions are too broad to equate with a specific Gemma4 forward device mismatch." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "Tokenizer-mapping failure in `add-new-model-like` is a distinct tooling bug, not the same as remote-code loading breakage." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer loading failure for a specific model vs flash-attn fallback gating; no shared concrete failure path." + }, + { + "left": "issue:43828", + "right": "issue:45823", + "accept": false, + "reason": "Autocast dtype mismatch in Phi MoE and Gemma4 device mismatch are different runtime errors." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": true, + "reason": "Both report stale docs still referencing removed pipeline tasks after v5; same documentation cleanup issue." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer regex crash vs flash-attn backend fallback logic; different subsystems and symptoms." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "Generic `remote_code` regression vs a specific Kimi-K2.5 tokenizer/codec regression; not the same bug." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "One is a tied-weights bug in Mistral3, the other a `from_pretrained()` non-persistent-buffer regression; unrelated behavior." + }, + { + "left": "issue:43742", + "right": "issue:45399", + "accept": false, + "reason": "MobileLLM loading failure and flash-attn fallback blocking are not the same code-path." + }, + { + "left": "issue:43257", + "right": "issue:43866", + "accept": false, + "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed vs a corrupted Ovis2 checkpoint; different underlying problems." + }, + { + "left": "issue:42915", + "right": "issue:45020", + "accept": false, + "reason": "A specific LoRA incompatibility on hf4.x vs a broad `remote_code` loading regression; not the same concrete bug." + }, + { + "left": "issue:43638", + "right": "issue:45823", + "accept": false, + "reason": "DeepSpeed ZeRO-3 index error on a Bert model vs Gemma4 PLE device mismatch; unrelated." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": false, + "reason": "Whisper not running is too vague to merge with a specific `_torch_extract_fbank_features()` report without stronger evidence." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "Both center on the unexpected `_is_hf_initialized` kwarg during model init/loading, indicating the same regression." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Both are ZeRO-3 weight-loading failures, but they target different model families and are not clearly the same concrete defect." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "LoRA compatibility on version upgrades vs bitsandbytes `_is_hf_initialized` constructor breakage; different failures." + }, + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "GLM-4.7 test loading issue and NemotronH checkpoint loading failure are separate model-specific bugs." + }, + { + "left": "issue:41553", + "right": "issue:44991", + "accept": false, + "reason": "Tokenizer loading error for Voxtral vs tokenizer loading failure for EMBEDDIA/est-roberta; no shared concrete mechanism shown." + }, + { + "left": "issue:42947", + "right": "issue:45923", + "accept": false, + "reason": "Gradient checkpointing not effective with LoRA is different from a missing `supports_gradient_checkpointing` flag on a remote-code model." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Weight conversion failure under accelerate+deepspeed vs gradient averaging under GAS with DeepSpeed; unrelated training bugs." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "Unexpected `compute_loss` kwarg vs mask/tensor shape mismatch; different error classes and code paths." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "RLHF NaN explosion from Qwen3.5 position ids is not the same as flash_attention `s_aux=None` crashing." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": true, + "reason": "Same IndexError pattern about mask shape not matching indexed tensor shape; 44805 is a later generic reproduction of the same bug." + }, + { + "left": "issue:43872", + "right": "issue:45538", + "accept": false, + "reason": "`_is_hf_initialized` loading regression and CLIPTokenizer max length constant issue are unrelated." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Different tokenizer/model-loading failures for different models; not the same root cause." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash and NemotronH checkpoint loading failure are distinct model/runtime issues." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": false, + "reason": "Both involve video/VL models, but one is a broad v5.3 regression and the other a specific `StopIteration` in a different embedding flow." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "RoPE/position-id scaling bug for Qwen2.5-VL vs uninitialized `inv_freq` after meta-to-empty materialization in Qwen3.5 MoE; different defects." + }, + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "DeepSpeed training `weight must be 2-D` and void segmentation map processing are unrelated issues." + }, + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "GGUF architecture support for gpt-oss is unrelated to Phi MoE autocast dtype mismatch." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The selected anchor issue 45020 is a reasonable umbrella for broad remote-code/model-loading regressions, and the accepted soft pairs are supported by the issue titles and described failure modes. No soft-edge verdict appears overstated." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 5, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:36010", + "issue:38175", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42617", + "issue:42915", + "issue:43317", + "issue:43366", + "issue:43531", + "issue:43550", + "issue:43596", + "issue:43632", + "issue:43638", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43720", + "issue:43824", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44075", + "issue:44155", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44756", + "issue:44805", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:44960", + "issue:44977", + "issue:45030", + "issue:45276", + "issue:45313", + "issue:45357", + "issue:45584", + "issue:45663", + "issue:45685", + "issue:45823", + "issue:45834" + ], + "soft_pairs": [ + "issue:45663|issue:45823", + "issue:44155|issue:45584", + "issue:43366|issue:43531", + "issue:43950|issue:45834", + "issue:43550|issue:44910", + "issue:44560|issue:45685", + "issue:43824|issue:43957", + "issue:43645|issue:43646", + "issue:43366|issue:44589", + "issue:42915|issue:43957", + "issue:42617|issue:43366", + "issue:44560|issue:44918", + "issue:36010|issue:42915", + "issue:44661|issue:45823", + "issue:38175|issue:43994", + "issue:43632|issue:45834", + "issue:44560|issue:44805", + "issue:44877|issue:45030", + "issue:43638|issue:44661", + "issue:43596|issue:44661", + "issue:44661|issue:44805", + "issue:43317|issue:44756", + "issue:43644|issue:44534", + "issue:38175|issue:39692", + "issue:44960|issue:44977", + "issue:42491|issue:43720", + "issue:45313|issue:45357", + "issue:44451|issue:44488", + "issue:41720|issue:42915", + "issue:42491|issue:45276", + "issue:43708|issue:44075" + ], + "prepared_review_unit_hash": "9abd19eeba6a0164ec1d1bc13a5dc6ec8056bd1f54171f646b4713a33ead26d6", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36322, + "estimated_input_tokens": 9081, + "estimated_eval_tokens": 18418 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated issue set with a few clear duplicate pairs around the same concrete regression; the rest are broad theme matches but not the same bug.", + "confidence": 0.86, + "canonical_issue_reason": "issue:43644 is the clearest representative of the core v5 regression and has an exact duplicate (44534).", + "canonical_pr_reason": null, + "best_issue_reason": "It is a broad, concrete regression report with the cleanest reproduction and the strongest evidence of duplication.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Same model family, but different failures: FSDP2 KeyError in attention vs device_map='auto' forward-device mismatch." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "Different subsystems and symptoms: AudioFlamingo3 track leakage vs Whisper empty-transcription generation failure." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "Both involve Qwen-related models, but one is GGUF/gpt-oss support and the other is a sliding_window bug in Qwen3-MoE." + }, + { + "left": "issue:43950", + "right": "issue:45834", + "accept": false, + "reason": "Different bugs: non-persistent buffer corruption vs an OCR long-input index error." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different code paths: torch.compile with SDPA on Bamba vs Flash Attention misreading 3D position_ids on Qwen3.5." + }, + { + "left": "issue:44560", + "right": "issue:45685", + "accept": false, + "reason": "Unrelated failures: Qwen3-VL video StopIteration vs an MPS histogram kernel not implemented for Int." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "ImportError for a missing class is not the same as meta-device loading breaking some models." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "These describe the same custom-model initialization regression; one is notebook-specific, the other is the general bug." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "GGUF/gpt-oss support request and a Float8 storage lookup error are different issues." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "Qwen3-MoE FineGrainedFP8Config failure is unrelated to meta-device loading regressions." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "3d_parallel.py execution failure and GGUF gpt-oss support are different problems." + }, + { + "left": "issue:44560", + "right": "issue:44918", + "accept": false, + "reason": "Both touch Qwen3.5, but one is a video-generation StopIteration and the other is TRL SFT embedding unpacking." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "A missing GenerationMixin import is unrelated to a Qwen3-MoE FP8 runtime failure." + }, + { + "left": "issue:44661", + "right": "issue:45823", + "accept": false, + "reason": "Tokenizer-mapping-name edge case and Gemma4 PLE device mismatch are different code-path problems." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": true, + "reason": "Same SigLIP2 model regression: unexpected zero probabilities / nonsensical outputs from AutoModel and pipeline." + }, + { + "left": "issue:43632", + "right": "issue:45834", + "accept": false, + "reason": "The _is_hf_initialized flag regression is unrelated to the long OCR index error." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Different Qwen3-VL issues: video StopIteration vs a generic mask/tensor shape mismatch." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": false, + "reason": "Both are config-loading failures, but they affect different model configs and likely different validation checks." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "DeepSpeed ZeRO-3 Bert init failure is not the same as add-new-model-like failing inside TOKENIZER_MAPPING_NAMES." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Same as above: a ZeRO-3 BertModel index error versus a tokenizer-mapping bug." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "Different storage/offload problems: dequantized model loading on gpu+cpu offload vs mmap OOM on Strix Halo." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Exact same regression about non-persistent buffers being filled with junk in Transformers v5." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "Both are SigLIP2-related, but 39692 is a docs/example mismatch plus quantization failure, not the same concrete bug as zero probabilities." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "GLM5 and Qwen3.5 Flash Attention generation failures are different model-specific issues." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "LoRA/Qwen3-MoE incompatibility across versions and BitNet packed-weight loading are unrelated." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "One is ZeRO-3 weight loading for language_model; the other is save_pretrained writing incorrect visual encoder keys." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": false, + "reason": "Both are model-loading failures, but they reference different checkpoints and there is no clear evidence of the same root cause." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Qwen3 auto device mapping cudaErrorAssert and Qwen3-MoE FP8 failures are different runtime bugs." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "Qwen3-MoE LoRA/version incompatibility is unrelated to Gemma4 resize_token_embeddings not updating tied outputs." + }, + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Trainer checkpoint-resume max_steps math and optimizer SGD args not being used are separate trainer/optimizer issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative: the clear duplicate/near-duplicate pairs are identified, and the non-matches are rejected with appropriately cautious reasoning based on the issue titles. I did not see any overstatement that would require rejection." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 6, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "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:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44155", + "issue:44164", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44402", + "issue:44488", + "issue:44743", + "issue:44756", + "issue:44910", + "issue:44912", + "issue:44960", + "issue:44991", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45406", + "issue:45488", + "issue:45588", + "issue:45663", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:45859" + ], + "soft_pairs": [ + "issue:43854|issue:43931", + "issue:43257|issue:43854", + "issue:45488|issue:45812", + "issue:42915|issue:44756", + "issue:42491|issue:43931", + "issue:39692|issue:43994", + "issue:43856|issue:43873", + "issue:44488|issue:44991", + "issue:43644|issue:43645", + "issue:45362|issue:45406", + "issue:43856|issue:44387", + "issue:44910|issue:45588", + "issue:42491|issue:45292", + "issue:36010|issue:39404", + "issue:45276|issue:45335", + "issue:43404|issue:43454", + "issue:44164|issue:44291", + "issue:41720|issue:44155", + "issue:44402|issue:44488", + "issue:39692|issue:44960", + "issue:43425|issue:43957", + "issue:42947|issue:45704", + "issue:43873|issue:44387", + "issue:43335|issue:45663", + "issue:45663|issue:45759", + "issue:45663|issue:45859", + "issue:44291|issue:44756", + "issue:44155|issue:45823", + "issue:44387|issue:44756", + "issue:43854|issue:45216", + "issue:44292|issue:44912", + "issue:44743|issue:44960", + "issue:43957|issue:44756" + ], + "prepared_review_unit_hash": "7919e3b5a55e53e37a2139c8a0a16e560ddce8c086b4ea7f97f6b88543ee379a", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37140, + "estimated_input_tokens": 9285, + "estimated_eval_tokens": 18826 + }, + "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 issues (tokenization, model loading, quantization, device mapping, and model-specific architecture bugs). The soft pairs share broad themes, but none look like the same concrete underlying bug or change.", + "confidence": 0.83, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:44991 is the cleanest broad regression report in the set: it describes a general tokenizer-loading failure on transformers>=5.0.0, with a clear, stable title and scope.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Both are model-loading failures, but they affect different models and different failure modes (unit-test load issue vs weight-shape mismatch)." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Different code paths: Qwen3 MoE checkpoint conversion under accelerate/deepspeed vs GLM-4.7-Flash loading in tests." + }, + { + "left": "issue:45488", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but one is a LlamaTokenizer pre-tokenizer override and the other is wrong token IDs for Granite models; not the same bug." + }, + { + "left": "issue:42915", + "right": "issue:44756", + "accept": false, + "reason": "Unrelated problems: Qwen3Moe FP8 failure vs a Strix Halo mmap OOM workaround." + }, + { + "left": "issue:42491", + "right": "issue:43931", + "accept": false, + "reason": "LoRA compatibility on hf5.x is a save/load format issue, not the same as Qwen3-VL shape-mismatch loading." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but one is a broken doc example and the other is a runtime inference regression; different fixes." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "Both mention memory, but one is Qwen3 MoE training memory use and the other is quantization/offloading behavior." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading failures, but they involve different models and likely different tokenizer files/call paths." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "One is junk-filled non-persistent buffers, the other is a Jupyter custom-model initialization issue; not the same bug." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Different components and models: transformers chat crash on Qwen3.5 vs serve crashing on Gemma4Processor missing _tokenizer." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both are memory-related, but one is training-memory inefficiency and the other is reserved-memory growth under int4 quantization." + }, + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Both hit flash_attention, but one is a Qwen3.5 position_ids crash and the other is an s_aux=None AttributeError on sink-less models." + }, + { + "left": "issue:42491", + "right": "issue:45292", + "accept": false, + "reason": "LoRA checkpoint incompatibility is unrelated to resize_token_embeddings behavior." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "Importing GenerationMixin and Whisper pipeline return_language are unrelated failures." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": false, + "reason": "Both involve resize_token_embeddings, but they hit different architectures and different submodules; too model-specific to treat as one bug." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both are lm_head tying bugs, but they are on different model classes (Mistral3 vs AyaVision) and likely separate fixes." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "Different serialization/loading paths: extra_state handling vs init_empty_weights/_is_hf_initialized argument mismatch." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "Qwen3 auto device-map cuda assert and AudioFlamingo3 batched inference leakage are unrelated bugs." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "Both concern specific model/tokenizer loading, but one is vocab-size token placement and the other is a separate tokenizer load failure." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "Doc-example errors for SigLIP2 are unrelated to the generic GLM5 issue report." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Torch version incompatibility is not the same as meta-device loading failures on specific models." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing with PEFT LoRA and apex RMSNorm memory leak are unrelated subsystems." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Both involve memory pressure, but one is offloading not working and the other is increased reserved memory under int4 quantization." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug is unrelated to Gemma4 FSDP2 shared_kv_states behavior." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "Different model families and failures: Gemma4 FSDP2 KeyError vs Qwen multimodal config unwrapping." + }, + { + "left": "issue:45663", + "right": "issue:45859", + "accept": false, + "reason": "Gemma4 FSDP2 shared_kv_states bug and Qwen3_5MoeTextRotaryEmbedding CPU-offload issue are separate code paths." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "init_empty_weights argument handling is unrelated to disabling mmap for Strix Halo OOM." + }, + { + "left": "issue:44155", + "right": "issue:45823", + "accept": false, + "reason": "Audio track leakage in batched inference is unrelated to Gemma4 device mismatch during forward." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "One is quantization-related CUDA memory growth; the other is a platform-specific mmap OOM workaround." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "Different regression surfaces: GLM-4.7-Flash loading in tests vs Qwen3.5 save_pretrained checkpoint correctness." + }, + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Both are quantized-model loading issues, but they concern different models, formats, and failure outcomes." + }, + { + "left": "issue:44743", + "right": "issue:44960", + "accept": false, + "reason": "Qwen3.5 recurrent-state cache bug is unrelated to the generic GLM5 issue." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "Meta-device loading failures and Strix Halo mmap OOM are different problem classes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 7, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:39404", + "issue:42947", + "issue:43257", + "issue:43299", + "issue:43381", + "issue:43425", + "issue:43611", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43716", + "issue:43828", + "issue:43901", + "issue:43906", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44164", + "issue:44265", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44509", + "issue:44530", + "issue:44589", + "issue:44671", + "issue:44683", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44877", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44928", + "issue:45084", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45588", + "issue:45663", + "issue:45684", + "issue:45698", + "issue:45701", + "issue:45759", + "issue:45823" + ], + "soft_pairs": [ + "issue:45684|issue:45698", + "issue:45276|issue:45292", + "issue:42947|issue:43381", + "issue:43381|issue:44928", + "issue:44164|issue:44756", + "issue:44479|issue:44877", + "issue:43646|issue:43708", + "issue:45588|issue:45663", + "issue:45292|issue:45335", + "issue:43257|issue:44863", + "issue:44402|issue:44451", + "issue:43645|issue:44164", + "issue:43828|issue:43957", + "issue:43425|issue:43828", + "issue:39404|issue:45701", + "issue:44589|issue:44756", + "issue:44683|issue:45823", + "issue:43931|issue:44863", + "issue:43906|issue:44530", + "issue:43425|issue:45823", + "issue:43611|issue:43645", + "issue:44530|issue:45084", + "issue:43425|issue:43716", + "issue:43901|issue:44509", + "issue:39404|issue:44912", + "issue:44265|issue:44898", + "issue:43299|issue:43931", + "issue:44479|issue:44805", + "issue:43994|issue:44079", + "issue:39404|issue:44671", + "issue:44589|issue:45084", + "issue:43425|issue:44918", + "issue:44164|issue:45759" + ], + "prepared_review_unit_hash": "579d0df0b5fa4accc9c8297106d96f7d1bb5b01b2a82880bb22b48e2280ca080", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37052, + "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 a grab bag of unrelated Transformers regressions and compatibility issues. I found one plausible duplicate pair around `resize_token_embeddings` not updating output embeddings; the rest look like distinct bugs or model-specific failures.", + "confidence": 0.64, + "canonical_issue_reason": "issue:43611 is the clearest broad regression-style issue in the set: it has a specific broken code path, a version boundary (Transformers 5.0.0), and a load-time compatibility failure that makes it a reasonable anchor for the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43611 is the most suitable representative issue because it is explicit, actionable, and broadly about Transformers 5 model-loading breakage rather than an incidental symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "Both involve custom-model save/load, but one is about permissions on source files and the other about loading the wrong custom module; different bugs." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same underlying `resize_token_embeddings` bug: output embeddings are not updated; 45276 is just a Gemma4-specific manifestation with extra layer details." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Both mention gradient checkpointing, but one is PEFT/LoRA ineffectiveness and the other is eval-mode restriction; not the same failure." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Unrelated issues: eval-mode gradient checkpointing vs RLHF NaNs from Qwen3.5 3D position IDs and SDPA fallback." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "`extra_state` serialization failure and Strix Halo mmap OOM are different subsystems and different bugs." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Video-input regression for Qwen VL models vs strict-config loading failure for granite_speech; not the same code path." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model initialization in notebooks is unrelated to `resume_from_checkpoint` max-step calculation." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "Flash-attention `s_aux=None` crash and Gemma4 FSDP2 shared-key/value rebuild error are different model-specific forward-path bugs." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": false, + "reason": "Both are `resize_token_embeddings`-related, but one targets generic output embeddings while the other is about `decoder.embed_tokens`; too model-specific to treat as the same bug." + }, + { + "left": "issue:43257", + "right": "issue:44863", + "accept": false, + "reason": "Qwen3 MOE weight conversion under accelerate/deepspeed and NemotronH checkpoint loading are distinct loading issues." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Tokenizer vocab-size mismatch and inability to load ScandiBERT are not the same underlying defect." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Jupyter custom model initialization and `extra_state` save/from_pretrained handling are separate problems." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Autocast dtype mismatch for Phi-tiny-MoE and meta-device loading failures are different runtime environments and failure modes." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Torch 2.10 compatibility is a version/platform issue; the other is a specific autocast dtype mismatch in a model forward path." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper `return_language` pipeline breakage and tokenizer version-dependent behavior are unrelated." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Float8 storage lookup failure and mmap-related OOM are unrelated runtime/storage issues." + }, + { + "left": "issue:44683", + "right": "issue:45823", + "accept": false, + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 PLE device mismatch are different subsystems." + }, + { + "left": "issue:43931", + "right": "issue:44863", + "accept": false, + "reason": "Qwen3-VL shape mismatch and NemotronH checkpoint loading are different model-loading bugs." + }, + { + "left": "issue:43906", + "right": "issue:44530", + "accept": false, + "reason": "A reproduction of an earlier issue and a PagedAttentionCache `linear_attention` crash are not the same bug." + }, + { + "left": "issue:43425", + "right": "issue:45823", + "accept": false, + "reason": "General Torch 2.10 incompatibility is not the same as Gemma4 device-map forward mismatch." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "Both are Transformers 5 loading regressions, but one is about `base_model_prefix` and the other about notebook-defined custom models; different code paths." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache group-type crash and non-template node compilation error are unrelated." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "Torch version incompatibility and Mistral-3 image-preprocessor dtype mismatch are different bugs." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "These are both docs issues, but they cover different removed/changed pipeline behaviors." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper pipeline behavior and MXFP4 quantization fallback are unrelated." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "Torch export failure via `torch_compilable_check` is unrelated to Perceiver non-default-resolution failures." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Both are Qwen3-VL loading regressions, but one is a generic v5 breakage and the other is a specific checkpoint shape mismatch; not clearly the same bug." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "Qwen video-input regression and a mask/tensor shape IndexError are different failure modes." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "Siglip2 wrong outputs and `ModelOutput` key assignment behavior are unrelated." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline language return regression and CamemBERT masked-LM prediction regression are unrelated model bugs." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage lookup and non-template node compilation failures are unrelated." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch version incompatibility and Qwen3.5 embedding unpacking with TRL are different issues." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "`extra_state` serialization and `from_config` text-config unwrapping for Qwen multimodal models are different code paths." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The representative issue choice is defensible, and the only accepted soft-edge pair (issue:45276 / issue:45292) is supported by the shared `resize_token_embeddings` / `output_embeddings` wording in the titles. The rejected pairs are generally framed with caution and do not overstate duplicate evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 8, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:32090", + "issue:34567", + "issue:36032", + "issue:39404", + "issue:40444", + "issue:40990", + "issue:42915", + "issue:43262", + "issue:43317", + "issue:43377", + "issue:43404", + "issue:43421", + "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:43856", + "issue:43883", + "issue:44164", + "issue:44206", + "issue:44220", + "issue:44451", + "issue:44466", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44661", + "issue:44671", + "issue:44683", + "issue:44737", + "issue:44756", + "issue:44811", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45584", + "issue:45859" + ], + "soft_pairs": [ + "issue:36032|issue:43452", + "issue:44671|issue:44991", + "issue:43596|issue:44683", + "issue:43856|issue:44756", + "issue:32090|issue:39404", + "issue:44991|issue:45081", + "issue:44737|issue:45859", + "issue:44164|issue:44479", + "issue:43582|issue:43632", + "issue:44488|issue:45081", + "issue:44661|issue:44683", + "issue:40990|issue:43421", + "issue:43611|issue:43644", + "issue:44488|issue:44987", + "issue:44206|issue:44220", + "issue:43716|issue:45084", + "issue:43643|issue:44530", + "issue:43883|issue:44488", + "issue:34567|issue:40444", + "issue:44811|issue:45584", + "issue:36032|issue:42915", + "issue:43644|issue:43646", + "issue:43452|issue:43482", + "issue:43541|issue:45084", + "issue:44451|issue:44987", + "issue:42915|issue:44530", + "issue:44849|issue:45325", + "issue:43317|issue:43482", + "issue:36032|issue:43645", + "issue:44987|issue:44991", + "issue:43404|issue:44466", + "issue:43883|issue:44991", + "issue:43262|issue:43377" + ], + "prepared_review_unit_hash": "7d2e2c866b202148be9a01a1cb43344f1843651457ab01603a1d08b08e138524", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37690, + "estimated_input_tokens": 9423, + "estimated_eval_tokens": 19102 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly conservative clustering. I keep the audio feature-extraction duplicate, but I reject the 43644/43646 soft edge: both are v5 regressions, yet the packet does not show they are the same bug or even the same failure mode.", + "confidence": 0.83, + "canonical_issue_reason": "issue:43646 is the best umbrella issue for the custom-model-initialization regression described in this packet; it is broader and more representative than the more specific notebook-related report nearby.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the strongest general issue candidate because its title is broad, user-facing, and likely to absorb nearby custom-model-init reports without overclaiming cross-duplication.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "Different failure modes: T5 tokenizer special-token conflict vs GGUF loader breakage." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "Both are v5 regressions, but one is CamemBERT MLM behavior and the other is tokenizer loading for a different model." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "Deepspeed zero3/BertModel init error is unrelated to compiled flex_attention on torch >= 2.9." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "Qwen3 MoE memory usage and Strix Halo mmap OOM are unrelated resource issues." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer GPU broadcast NoneType error is unrelated to Whisper return_language pipeline behavior." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Both involve loading/tokenizer paths, but they target different models and different bugs." + }, + { + "left": "issue:44737", + "right": "issue:45859", + "accept": false, + "reason": "XLNet CPU arange inefficiency is unrelated to Qwen3.5 MoE CPU offload compatibility." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "Serialization of extra_state is unrelated to the Qwen video-input regression." + }, + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError is unrelated to the _is_hf_initialized flag regression." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Both are model-loading reports, but the affected models and tokenizer code paths differ." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES handling and flex_attention compilation are unrelated subsystems." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "Perplexity regression on gpt-oss is unrelated to runtime special-token post-processing updates." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "Base-model-prefix loading failure and non-persistent-buffer junk are different v5 regressions." + }, + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "Both are loading failures, but there is no evidence they share the same underlying bug." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": true, + "reason": "Both point to the same audio feature-extraction bug around unsupported center handling / _torch_extract_fbank_features()." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Mistral-3 image-preprocessor dtype mismatch is unrelated to template compilation errors." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "Missing fields from AutoConfig with trust_remote_code is unrelated to Qwen3.5 PagedAttention cache crashes." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Missing all_tied_weights_keys and model-loading failure for a different checkpoint are not the same bug." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "Trainer token-count tracking and multi-image IterableDataset finetuning failure are unrelated." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Whisper decode skipping special tokens and empty-transcription generation failure are different behaviors." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "T5 tokenizer load conflict is unrelated to Qwen3 MoE FineGrainedFP8 failure." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 regressions, but the packet does not establish the same underlying bug or code path." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": false, + "reason": "gguf_file loading breakage and Qwen2.5-GGUF loading failure are separate loader issues." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Torch dynamo grouped_mm tracing error is unrelated to template compilation failure." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Loading vesteinn/ScandiBERT and loading physical-intelligence/fast are different model-specific regressions." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "Qwen3 MoE FP8 training and Qwen3.5 PagedAttention linear_attention crash are different code paths." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Qwen3.5 hidden-states output bug and Qwen2.5-VL rope-index scaling bug are unrelated." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "Dequantized model offload failure and Qwen2.5-GGUF loading failure are not the same bug." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "Tokenizer special-token conflict is unrelated to custom model initialization in notebooks." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Both fail on model loading in v5.x, but they are different checkpoints and different failure modes." + }, + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Mistral3 lm_head tying and inconsistent lm_head serialization are related to weights, but not the same concrete bug from the titles." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Missing all_tied_weights_keys and tokenizer-loading failure are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Audio processor chat-template sampling-rate default and MIMI padding-mask batching are different audio bugs." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The overall clustering is mostly conservative, but one accepted soft-edge is too strong for the packet. `issue:43644` (\u201cfills non-persistent buffers with junk\u201d) and `issue:43646` (\u201cbreaks custom model initialization\u201d) are both v5 regressions, but the packet does not ground them as the same bug or even clearly the same failure mode; the rationale overstates evidence by calling them the same underlying change. Because the acceptability of a soft-edge verdict must stay conservative, I can\u2019t accept this summary as-is." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 9, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:36032", + "issue:40444", + "issue:40990", + "issue:41553", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43296", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43540", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43645", + "issue:43646", + "issue:43742", + "issue:43818", + "issue:43819", + "issue:43994", + "issue:44062", + "issue:44079", + "issue:44220", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44479", + "issue:44534", + "issue:44589", + "issue:44610", + "issue:44683", + "issue:44756", + "issue:44811", + "issue:44877", + "issue:44987", + "issue:45081", + "issue:45084", + "issue:45127", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45507", + "issue:45542", + "issue:45663" + ], + "soft_pairs": [ + "issue:45084|issue:45507", + "issue:43994|issue:44610", + "issue:43646|issue:44534", + "issue:43322|issue:43540", + "issue:43645|issue:44534", + "issue:44403|issue:45081", + "issue:43377|issue:43819", + "issue:42915|issue:45084", + "issue:44265|issue:44610", + "issue:45216|issue:45357", + "issue:44451|issue:45081", + "issue:44220|issue:44683", + "issue:44987|issue:45081", + "issue:43404|issue:45127", + "issue:43632|issue:44756", + "issue:42915|issue:44291", + "issue:44062|issue:44987", + "issue:44479|issue:45325", + "issue:36032|issue:43742", + "issue:43296|issue:43531", + "issue:40444|issue:45325", + "issue:42915|issue:44589", + "issue:43611|issue:44534", + "issue:43531|issue:44877", + "issue:36032|issue:41720", + "issue:40990|issue:44811", + "issue:43425|issue:44403", + "issue:44079|issue:45663", + "issue:41553|issue:43582", + "issue:43632|issue:44534", + "issue:42175|issue:45542", + "issue:43818|issue:45357" + ], + "prepared_review_unit_hash": "22ec130f16e76344a898e5938bf0593e1140e0345c3f214267bf21608f471de5", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37215, + "estimated_input_tokens": 9304, + "estimated_eval_tokens": 18864 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose mix of unrelated transformer/model bugs, with several broad Transformers v5 regression reports but no clear duplicate set. I found no soft-edge pair that looks like the same underlying bug.", + "confidence": 0.89, + "canonical_issue_reason": "issue:43646 is the broadest representative of the Transformers v5 initialization regressions in this set, and it is easier to treat as the umbrella issue than the more narrowly scoped siblings.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the strongest single issue candidate because it is a clear, general regression report about custom model initialization, which is a recurring theme across the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "Different failures: template-node compilation vs an invalid method call in GraniteMoEHybrid." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Different models and symptoms: nonsensical outputs in SigLIP2 vs a processor/model input-size mismatch in OmDet-Turbo." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Both are Transformers v5 regressions, but one is custom model init while the other is non-persistent buffer corruption; different bugs." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Different multimodal code paths: Llava Next loading segfault vs Qwen3OmniMoe video input validation failure." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Jupyter/custom-model initialization breakage is not the same bug as junk-filled non-persistent buffers." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Unnecessary loading noise is unrelated to the Mistral tokenizer regex patch crash." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "Different audio/model implementations and failures: batching/padding-mask mismatch in MIMI vs DAC STE mismatch." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "FineGrainedFP8Config failure in Qwen3Moe is unrelated to the non-template-node compilation error." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "torch.export with torch_compilable_check is a different export-time issue from a processor size mismatch." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": false, + "reason": "Both concern Qwen3.5 save_pretrained, but one is a general checkpoint regression and the other is specifically incorrect visual encoder keys; not clearly the same bug." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "Different model-loading failures: ScandiBERT cannot load, while Mistral regex patching hits a tokenizer backend attribute error." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Unrelated subsystems: audio feature extraction vs compiled flex_attention on newer torch." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Both are loading-related, but one is a model-specific failure on physical-intelligence/fast and the other is a Mistral tokenizer patch crash." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both touch tied embeddings/lm_head behavior, but one is a missing weight tie in Mistral3 while the other is LoRA merge collapse with extended vocab." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "Different regressions: _is_hf_initialized handling vs mmap/OOM behavior on Strix Halo." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "Qwen3Moe FP8 config failure is not the same as init_empty_weights rejecting an unexpected _is_hf_initialized argument." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer AddedToken argument duplication is unrelated to a model-loading failure for physical-intelligence/fast." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Both are Qwen multimodal regressions, but one is a video-input regression and the other is still-image temporal position-id scaling; different code paths." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "T5 tokenizer method conflict and MobileLLM key error are different loading bugs." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "PaddleOCR-VL loading in vLLM is unrelated to Qwen3-MoE sliding_window behavior." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Multiple-image iterable training failure is different from Qwen2.5-VL still-image rope-index scaling." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "Different failure modes: FP8 config handling vs missing Float8 storage class." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "base_model_prefix loading breakage is unrelated to junk-filled non-persistent buffers." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Sliding-window handling in Qwen3-MoE is different from strict config rejection for granite_speech." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Tokenizer load conflict and Qwen3 device-mapping CUDA assert are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "High perplexity on WikiText and Whisper batch_decode skip_special_tokens are unrelated bugs." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Torch 2.10 compatibility and transformer loading noise are not the same problem." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key assignment and Gemma4 FSDP2 shared_kv_states rebuilding are different issues." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Bad AutoTokenizer error messaging for Voxtral is unrelated to the Apple Silicon caching_allocator_warmup TypeError." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "_is_hf_initialized regression and junk non-persistent buffers are different Transformers v5 bugs." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": false, + "reason": "Both involve TensorFlow/backend packaging, but one is missing TF in the torch extra and the other is a runtime undefined-backend error when only tensorboard is installed." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "Video-LLaVA tower/temporal-attention bug is unrelated to Qwen3.5 save_pretrained visual-key serialization." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 10, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:33666", + "issue:35707", + "issue:40444", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43381", + "issue:43421", + "issue:43454", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43749", + "issue:43792", + "issue:43856", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44462", + "issue:44466", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44683", + "issue:44829", + "issue:44898", + "issue:44991", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45381", + "issue:45479", + "issue:45491", + "issue:45538", + "issue:45812", + "issue:45834", + "issue:45923" + ], + "soft_pairs": [ + "issue:43317|issue:43856", + "issue:43421|issue:44568", + "issue:44462|issue:45812", + "issue:33666|issue:40444", + "issue:43792|issue:44589", + "issue:44683|issue:45020", + "issue:44829|issue:45479", + "issue:44403|issue:44991", + "issue:43381|issue:45923", + "issue:43482|issue:45538", + "issue:35707|issue:40444", + "issue:44079|issue:44534", + "issue:44291|issue:44589", + "issue:43482|issue:43525", + "issue:43377|issue:45491", + "issue:43749|issue:44079", + "issue:40444|issue:45381", + "issue:43644|issue:44079", + "issue:43531|issue:44291", + "issue:41720|issue:45081", + "issue:44265|issue:45072", + "issue:43632|issue:45072", + "issue:45325|issue:45381", + "issue:43994|issue:44898", + "issue:43611|issue:43646", + "issue:42175|issue:43296", + "issue:43454|issue:44466", + "issue:45325|issue:45834", + "issue:43792|issue:43957", + "issue:42915|issue:43482" + ], + "prepared_review_unit_hash": "3ba9e645c1e1ab0f5115603cefa6b13e76de31999cd0a56d6135056f84d3e84f", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36830, + "estimated_input_tokens": 9208, + "estimated_eval_tokens": 18672 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is dominated by Transformers v5/model-loading regression reports, but most pairs are model-specific or subsystem-specific rather than true duplicates. Only the Qwen2.5-VL position-id pair looks like the same underlying bug.", + "confidence": 0.78, + "canonical_issue_reason": "issue:43646 is the broadest representative of the cluster\u2019s main theme: Transformers v5 breaking model initialization/loading paths. Most other items are narrower variants or unrelated bugs.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the best single issue to represent the cluster because it is broad, central, and closest to the recurring v5 initialization/loading regression pattern across the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Different failures: device offload/dequantized loading vs Qwen3 MoE training memory usage." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Both involve tokenizer special tokens, but one is runtime post-processor updates and the other is add_special_tokens/BOS-EOS insertion for a specific tokenizer." + }, + { + "left": "issue:44462", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer-loading regressions, but they affect different behaviors and models; not the same concrete bug." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Different Qwen-VL training problems: multi-GPU training vs iterable dataset with multiple images per prompt." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Different root causes: Whisper model runtime failure vs missing Float8 storage type." + }, + { + "left": "issue:44683", + "right": "issue:45020", + "accept": false, + "reason": "Unrelated: compiled flex_attention on torch>=2.9 vs remote_code model breakage." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Both mention degenerate training, but one is flash_attention_3 behavior and the other is a classification-loss setup bug." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Different issues: noisy loading logs vs tokenizer loading failure for a specific model." + }, + { + "left": "issue:43381", + "right": "issue:45923", + "accept": false, + "reason": "One is gradient checkpointing in eval mode; the other is a missing supports_gradient_checkpointing flag on a remote-code model." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "Different tokenizer/model-loading regressions with no clear shared code-path." + }, + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Different generation/training contexts and different multimodal models." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "ModelOutput key assignment vs non-persistent buffer initialization are different bugs." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "Different loading failures: unexpected _is_hf_initialized arg vs Float8 storage lookup." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "Different v5 loading failures; one is GGUF loading, the other is missing pad_token_id on Llama4Config." + }, + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "Both are input-shape/masking issues, but one is MIMI encoder padding masks and the other is Gemma3 sliding-window all-padding NaNs." + }, + { + "left": "issue:43749", + "right": "issue:44079", + "accept": false, + "reason": "FSDP CPU-efficient loading is unrelated to ModelOutput key handling." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Different Qwen2.5-VL problems: finetuning with multiple images vs wrong video position ids." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "Non-persistent buffer junk vs ModelOutput key assignment are unrelated bugs." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Qwen3-MoE sliding_window bug and init_empty_weights/_is_hf_initialized bug are different failure modes." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "Qwen3 device-mapping cuda assert is unrelated to Mistral tokenizer regex patching." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "torch.export with torch_compilable_check is unrelated to dtype mismatches in inference." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "_is_hf_initialized regression is a loading/init issue, not the same as dtype mismatch in bfloat16 inference." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": true, + "reason": "Both describe wrong Qwen2.5-VL position-id generation in transformers 5.3.0; they likely share the same get_rope_index/rope-index code path." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Different vision model issues: nonsensical SigLIP2 outputs vs Perceiver non-default resolution failure." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 loading regressions, but the specific problems differ: base_model_prefix handling vs custom model initialization." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "Packaging/backend dependency issue vs a model load failure in vLLM are unrelated." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "One is tied lm_head weights in AyaVision; the other is device-dependent lm_head serialization in v5." + }, + { + "left": "issue:45325", + "right": "issue:45834", + "accept": false, + "reason": "Wrong Qwen2.5-VL position ids is unrelated to Kosmos2.5 OCR long-input index errors." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Both involve model loading on transformers 5+, but the concrete failures and affected code paths differ." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Different model families and different version-specific failures: FineGrainedFP8Config vs GGUF loading." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: many items are indeed v5/model-loading regressions, and the only accepted duplicate pair (45325/45381) is a conservative match. The rejected soft-edge verdicts generally stay within the evidence from titles and avoid overclaiming sameness." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 11, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42947", + "issue:43257", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43454", + "issue:43526", + "issue:43550", + "issue:43572", + "issue:43606", + "issue:43701", + "issue:43720", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44186", + "issue:44265", + "issue:44291", + "issue:44387", + "issue:44466", + "issue:44479", + "issue:44488", + "issue:44534", + "issue:44568", + "issue:44610", + "issue:44617", + "issue:44877", + "issue:44898", + "issue:45072", + "issue:45325", + "issue:45357", + "issue:45584", + "issue:45701", + "issue:45759", + "issue:45800", + "issue:45823", + "issue:45920", + "issue:45923" + ], + "soft_pairs": [ + "issue:44387|issue:44617", + "issue:43296|issue:45800", + "issue:43701|issue:45325", + "issue:44479|issue:45823", + "issue:43881|issue:44488", + "issue:43262|issue:43746", + "issue:44291|issue:44877", + "issue:43606|issue:44265", + "issue:44568|issue:45920", + "issue:40990|issue:45584", + "issue:34634|issue:35707", + "issue:43873|issue:45701", + "issue:44186|issue:44265", + "issue:43317|issue:43873", + "issue:35707|issue:41720", + "issue:43606|issue:43819", + "issue:43454|issue:43550", + "issue:41720|issue:45823", + "issue:43377|issue:44610", + "issue:39692|issue:40444", + "issue:43454|issue:45072", + "issue:43377|issue:44898", + "issue:43257|issue:45923", + "issue:43572|issue:45759", + "issue:40990|issue:42947", + "issue:43526|issue:43883", + "issue:44466|issue:45357", + "issue:43749|issue:44534", + "issue:43454|issue:43720" + ], + "prepared_review_unit_hash": "69b2aa48d5f5c75d7f719fccec54f8edaf87002f65c1a7e82c629fdaddaf43ed", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36584, + "estimated_input_tokens": 9146, + "estimated_eval_tokens": 18548 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The soft pairs are mostly superficially similar but not true duplicates: they span unrelated model-specific loading, tokenizer, offloading, multimodal, and serialization bugs. I would not merge any of the proposed soft pairs.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 43873 is the broadest representative of the offloading/quantization subtheme, but the overall set is heterogeneous and does not form a single duplicate cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43873 is the closest thing to a cluster representative because it is the most generic, broadly worded bug in the set, centered on offloading/quantization behavior.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both mention OOM, but one is int4 quantization memory growth and the other is a Sam3Video inference OOM; different failures and code paths." + }, + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "One is a PaddleOCR-VL model load failure in vLLM, the other is a torch/transformers version incompatibility; not the same bug." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint resume key mismatch and Qwen2.5-VL rope/position-id scaling are unrelated issues." + }, + { + "left": "issue:44479", + "right": "issue:45823", + "accept": false, + "reason": "Both are multimodal regressions involving device/multimodal handling, but they affect different models and different failure modes." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "Both are model loading problems, but they concern different models and distinct root causes." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and GraniteSpeech PEFT checkpoint loading are unrelated." + }, + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "Both involve loading/config behavior in v5, but one is an init_empty_weights argument mismatch and the other is strict config validation; not the same bug." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "CPU offload device mismatch and torch.export failing on torch_compilable_check are different code paths." + }, + { + "left": "issue:44568", + "right": "issue:45920", + "accept": false, + "reason": "One is a specific add_special_tokens regression for a tokenizer, the other is broad wrong-token-id behavior across several models." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "High perplexity on gpt-oss and Whisper empty-transcription generation failure are unrelated." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "Bark voice_preset and progressive generation with inputs_embeds/past_key_values are different bugs." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Offloading/quantization issues are not the same as tokenization changes across versions." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Tokenizer crashes on NER/padding versus export-time torch_compilable_check failure; no shared underlying bug." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "Both involve offload/quantization language, but one is dequantized model loading with gpu+cpu offload and the other is a generic offloading complaint; too broad to merge." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Progressive generation with cached embeddings and a Qwen3 cudaErrorAssert under auto device mapping are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "Bark CPU offload device mismatch and DAC from_latents vs forward mismatch are different model behaviors." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "lm_head tying/garbage generation and torch.compile SDPA failure are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:45823", + "accept": false, + "reason": "Both mention device mapping, but they affect different models and likely different backend faults." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "Missing padding-mask support in MIMI and OmDet-Turbo processor/model size mismatch are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "Doc-example model/processor mismatch plus quantization failure is not the same as Qwen2.5-VL multi-image finetuning failure." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "Weight-tying serialization and bfloat16 dtype mismatch are different problems." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Both are shape/batching-related vision issues, but one is padding-mask support and the other is resolution/interpolation behavior." + }, + { + "left": "issue:43257", + "right": "issue:45923", + "accept": false, + "reason": "MOE weight conversion during accelerate/deepspeed loading and a missing supports_gradient_checkpointing flag are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "StableLm pad_token_idx config regression and composite Qwen config unwrapping are different config-loading bugs." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "Perplexity regression and gradient checkpointing ineffectiveness are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and missing all_tied_weights_keys are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Both touch serialization, but one is lm_head weight behavior and the other is incorrect visual encoder keys; not the same issue." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "FSDP CPU RAM-efficient loading and non-persistent buffer initialization are different v5 regressions." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "Weight tying in AyaVision and packed-weight unpacking in BitNet loading are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 12, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:36331", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:43295", + "issue:43296", + "issue:43316", + "issue:43317", + "issue:43322", + "issue:43404", + "issue:43450", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43708", + "issue:43720", + "issue:43873", + "issue:43881", + "issue:44008", + "issue:44186", + "issue:44387", + "issue:44423", + "issue:44479", + "issue:44610", + "issue:44671", + "issue:44743", + "issue:44756", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44918", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45161", + "issue:45357", + "issue:45406", + "issue:45632", + "issue:45684", + "issue:45701", + "issue:45704", + "issue:45812", + "issue:45823" + ], + "soft_pairs": [ + "issue:43881|issue:44991", + "issue:41720|issue:43322", + "issue:43296|issue:44877", + "issue:41720|issue:44977", + "issue:44008|issue:45823", + "issue:45632|issue:45684", + "issue:43404|issue:45357", + "issue:40990|issue:45701", + "issue:42617|issue:42915", + "issue:44186|issue:45072", + "issue:43873|issue:44756", + "issue:43577|issue:43708", + "issue:44387|issue:45704", + "issue:44423|issue:45406", + "issue:36331|issue:42617", + "issue:44877|issue:44991", + "issue:43550|issue:43720", + "issue:36331|issue:42915", + "issue:43450|issue:45701", + "issue:44610|issue:44898", + "issue:44186|issue:44610", + "issue:44479|issue:44918", + "issue:43550|issue:43606", + "issue:41720|issue:42617", + "issue:43316|issue:44913", + "issue:42915|issue:43317", + "issue:43295|issue:45701", + "issue:43526|issue:43881", + "issue:44671|issue:44743", + "issue:45081|issue:45812", + "issue:41720|issue:45161", + "issue:43531|issue:43572" + ], + "prepared_review_unit_hash": "1b1c244f137251e30d2079577daaaaf0da1ea5a4e61fc8eb570daa568bf9d06a", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37427, + "estimated_input_tokens": 9357, + "estimated_eval_tokens": 18970 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is loosely centered on Transformers regressions around loading, offloading, quantization, tokenizer/config handling, and multimodal processors, but the soft pairs mostly describe different concrete failures rather than true duplicates.", + "confidence": 0.58, + "canonical_issue_reason": "issue:43873 is the broadest and most central report in the set: it is still open, has the most engagement/inbound references, and sits closest to the recurring offloading/quantization theme shared by several nearby issues.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best representative issue for this cluster because it is the most general, active, and thematically central; the other issues are more model- or API-specific edge cases.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "Different models and symptoms: a model loading failure vs tokenizer loading regression." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Both are loading failures, but they affect different models and different crash modes." + }, + { + "left": "issue:43296", + "right": "issue:44877", + "accept": false, + "reason": "Both involve loading/config handling, but the models and underlying failure paths are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Same family name, but one is a device-map assert during loading and the other is a flash-attention generation issue." + }, + { + "left": "issue:44008", + "right": "issue:45823", + "accept": false, + "reason": "Different Gemma generations and different bug classes: attribute collision vs device mismatch." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Both are path/file-handling related, but one is cache collision and the other is permission propagation on save." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Different models and different regressions: tied weights vs save_pretrained key mapping." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "Perplexity regression and tokenization change are not the same concrete bug." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "Both are model execution failures, but they concern different scripts, models, and code paths." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer/padding crash is unrelated to dtype mismatch in inference." + }, + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Both touch memory/offload behavior, but one is quantization offloading and the other is mmap/OOM on Strix Halo." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "Different problem domains: dtype propagation in BLIP-2 vs Trainer checkpoint-resume math." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "Both mention memory pressure, but one is int4 CUDA reserved memory and the other is an apex RMSNorm leak." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Both crash in serving, but the processor errors are different and likely separate code paths." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "Trainer API incompatibility is unrelated to a 3d_parallel execution failure." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Config strictness and tokenizer loading regression are different failure sources." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "Both involve advanced execution settings, but one is torch.compile+SDPA and the other is accelerate loading/unpacking." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "A Trainer signature mismatch is unrelated to a FineGrainedFP8 model load failure." + }, + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Video processor shape bug does not match a tokenizer behavior regression." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Both concern image sizes, but the models and expected resolutions differ." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Tokenizer batching crash and image resolution mismatch are different subsystems and bugs." + }, + { + "left": "issue:44479", + "right": "issue:44918", + "accept": false, + "reason": "Video input regression and embedding unpacking failure are not the same underlying issue." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Different failure modes: compile/SDPA bug vs CPU offload device mismatch." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Auto device-map CUDA assert and 3d_parallel runtime failure are not the same bug." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Both are config-related, but one is API inconsistency and the other is config value not persisting." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "Both are load/offload problems, but they target different quantization/dequantization scenarios." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "Processor/tokenizer API regression is not the same as a general tokenization change report." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "BEiT label reduction bug is unrelated to a model loading failure." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "CamemBERT masked LM output bug and Qwen recurrent-state bug affect different model families and code paths." + }, + { + "left": "issue:45081", + "right": "issue:45812", + "accept": false, + "reason": "Tokenizer patch crash and Granite token-ID regression are different tokenizer bugs." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Both mention GPT/Qwen-style model work, but they are different models and different execution modes." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to a StableLm config missing pad_token_idx." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 13, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:30064", + "issue:33357", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:36010", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:43322", + "issue:43388", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43873", + "issue:43927", + "issue:44077", + "issue:44186", + "issue:44361", + "issue:44368", + "issue:44451", + "issue:44492", + "issue:44509", + "issue:44514", + "issue:44545", + "issue:44625", + "issue:44898", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45059", + "issue:45072", + "issue:45357", + "issue:45381", + "issue:45406", + "issue:45464", + "issue:45479", + "issue:45701", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45859" + ], + "soft_pairs": [ + "issue:43540|issue:43782", + "issue:36010|issue:42617", + "issue:43421|issue:45812", + "issue:42915|issue:43322", + "issue:44186|issue:44361", + "issue:43388|issue:45059", + "issue:45406|issue:45464", + "issue:33357|issue:35707", + "issue:44492|issue:44509", + "issue:43421|issue:45701", + "issue:41720|issue:43746", + "issue:43606|issue:43720", + "issue:30064|issue:34634", + "issue:43550|issue:43653", + "issue:43873|issue:45704", + "issue:44514|issue:44545", + "issue:44625|issue:45479", + "issue:43606|issue:43653", + "issue:43531|issue:45759", + "issue:34634|issue:41720", + "issue:34567|issue:35707", + "issue:43421|issue:43927", + "issue:44368|issue:45357", + "issue:43782|issue:45381", + "issue:44898|issue:45072", + "issue:44451|issue:44991", + "issue:44077|issue:44625", + "issue:45759|issue:45859", + "issue:44987|issue:45042" + ], + "prepared_review_unit_hash": "2a21a3b4327dee18bff08e3cd8ce45de3c3f5e85db4d1a84a31d299d2c1370d5", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36464, + "estimated_input_tokens": 9116, + "estimated_eval_tokens": 18488 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are unrelated despite superficial subsystem overlap. The only clear duplicate is the two identical Qwen2_5_VLProcessor batched-input crash reports; the rest should stay separate issues.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 44545 is the best canonical representative for the Qwen2_5_VLProcessor padding=False crash: it matches 44514 exactly on title/symptom and has slightly more discussion/inbound references.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44545 is the strongest issue to keep as the main reference for this duplicate bug because it is the clearest, most central report of the exact same crash.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Different models and failure modes: video input ValueError vs from_pretrained weight_only load error." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "Unrelated problems: GenerationMixin import error vs 3d_parallel.py execution failure." + }, + { + "left": "issue:43421", + "right": "issue:45812", + "accept": false, + "reason": "Both concern tokenization internals, but one is runtime post-processor updates and the other is incorrect Granite token IDs; not the same bug." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Different code paths and symptoms: Qwen3Moe FP8 config failure vs Llava Next segmentation fault." + }, + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Both tokenizer-related, but different models and errors (NER/batching crash vs AttributeError on MLuke tasks)." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "Different features and failures: gather_for_metrics label handling vs SAM3 text/bbox behavior." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Unrelated API crashes in different processors/models." + }, + { + "left": "issue:33357", + "right": "issue:35707", + "accept": false, + "reason": "Completely different bugs: MacOS bus error with CLIP vs generation with inputs_embeds/past_key_values." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache strategy docs is unrelated to removed pipeline-task docs." + }, + { + "left": "issue:43421", + "right": "issue:45701", + "accept": false, + "reason": "Generic tokenization regression is too broad; this is not the same specific runtime special-token bug." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Different model-specific issues: device-mapping cudaErrorAssert vs PEFT local checkpoint loading." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "Both involve loading/offload, but one is CPU offload device mismatch and the other is packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Unrelated processors: void segmentation map handling vs Bark voice_preset." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "Different components and symptoms: torch.compile/SDPA failure vs tokenizer special-token registration." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization offloading issue and apex RMSNorm memory leak are unrelated." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Exact same Qwen2_5_VLProcessor.apply_chat_template batched-input crash with padding=False; these are duplicates of the same bug." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "Different classification bugs: config propagation in Qwen3.5 vs degenerate zero-loss behavior across sequence-classification models." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "Device mismatch on CPU offload is unrelated to BigBirdTokenizer special-token decode behavior." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "Both mention Qwen3-family configs, but one is sliding_window behavior and the other is text_config unwrapping in from_config." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "No shared underlying bug: BarkProcessor voice_preset vs Qwen3 auto device mapping CUDA assert." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState token counting and progressive generation are unrelated." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "Both involve token IDs/special tokens, but one is runtime post-processing and the other is config save/load corruption." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "A warning about tie_word_embeddings is unrelated to incorrect visual encoder keys on save_pretrained." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Different Qwen VL bugs: weight_only load failure vs wrong video vision_position_ids." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Different model classes and failures: Perceiver interpolation bug vs bfloat16 dtype mismatch." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Both are loading failures, but they affect different models and likely different tokenizer code paths; not enough evidence of the same bug." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "Optional post_init policy and num_labels propagation are unrelated." + }, + { + "left": "issue:45759", + "right": "issue:45859", + "accept": false, + "reason": "Different Qwen3.5 multimodal issues: config unwrapping vs CPU-offload rotary embedding compatibility." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "Unrelated regressions: model loading failure vs PIL image processors requiring torchvision." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: 44514 and 44545 are the same Qwen2_5_VLProcessor.apply_chat_template batched-input/padding=False crash, and 44545 does have slightly more discussion/inbound references. The other soft-edge rejections are cautious and consistent with the issue titles; no pair is overstated as a duplicate without clear evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 14, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:39692", + "issue:40444", + "issue:42886", + "issue:43295", + "issue:43335", + "issue:43381", + "issue:43421", + "issue:43525", + "issue:43531", + "issue:43540", + "issue:43596", + "issue:43618", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43746", + "issue:43756", + "issue:43819", + "issue:43844", + "issue:43873", + "issue:43994", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44220", + "issue:44451", + "issue:44589", + "issue:44610", + "issue:44661", + "issue:44671", + "issue:44843", + "issue:44918", + "issue:44936", + "issue:45081", + "issue:45183", + "issue:45356", + "issue:45381", + "issue:45479", + "issue:45588", + "issue:45663", + "issue:45704", + "issue:45759", + "issue:45834", + "issue:45859", + "issue:45907", + "issue:45923" + ], + "soft_pairs": [ + "issue:45704|issue:45923", + "issue:43531|issue:44589", + "issue:44918|issue:44936", + "issue:44062|issue:44451", + "issue:45663|issue:45923", + "issue:39692|issue:44610", + "issue:43421|issue:43653", + "issue:45588|issue:45759", + "issue:42886|issue:44843", + "issue:43381|issue:43844", + "issue:43421|issue:44186", + "issue:43994|issue:44671", + "issue:45704|issue:45859", + "issue:43525|issue:43531", + "issue:45081|issue:45356", + "issue:39692|issue:43873", + "issue:45356|issue:45907", + "issue:43746|issue:43873", + "issue:40444|issue:43746", + "issue:43596|issue:44220", + "issue:43746|issue:45381", + "issue:43540|issue:44008", + "issue:43720|issue:43819", + "issue:43618|issue:43697", + "issue:45183|issue:45479", + "issue:44220|issue:44661", + "issue:43653|issue:43720", + "issue:30064|issue:36010", + "issue:43335|issue:43756", + "issue:43295|issue:45834" + ], + "prepared_review_unit_hash": "f44e7b8298df2b14fc752c6b63af296954196354218b8a0461b8f0bb2556289c", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 37150, + "estimated_input_tokens": 9288, + "estimated_eval_tokens": 18832 + }, + "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 Hugging Face/Transformers issues. I found one plausible duplicate pair around offline tokenizer loading; the rest are similar only at the subsystem level and should stay separate.", + "confidence": 0.78, + "canonical_issue_reason": "Issue 44843 is the most concrete root-cause report in the only plausible duplicate pair: it identifies the exact offline failure path in `_patch_mistral_regex` / `model_info()`, whereas 42886 is the broader symptom report.", + "canonical_pr_reason": null, + "best_issue_reason": "44843 is the best representative issue because it pins the bug to a specific code path and failure mode, making it the most useful canonical anchor for triage.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45704", + "right": "issue:45923", + "accept": false, + "reason": "Different areas: T5 RMSNorm/apex memory leak vs missing gradient_checkpointing flag on a Nemotron remote-code variant." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Unrelated failures: Qwen3-MoE sliding-window logic vs a Float8 storage deserialization error." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "Both are trainer-related, but the concrete bugs differ: SFT embedding unpacking vs evaluate() failing after train()." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "Tokenizer construction bug vs model loading failure for a specific checkpoint; no shared underlying code path is evident." + }, + { + "left": "issue:45663", + "right": "issue:45923", + "accept": false, + "reason": "Different model families and mechanisms: Gemma-4 FSDP2 shared_kv_states bug vs Nemotron gradient_checkpointing metadata." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "Documentation/example mismatch vs processor/model image-size mismatch; same broad vision space, but not the same bug." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Both involve tokenizer special tokens, but one is runtime post-processor updates and the other is BigBird mask-token registration/decoding." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention sinkless-model crash and composite config unwrapping are unrelated code paths." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": true, + "reason": "Both describe the same offline tokenizer-loading failure; 44843 names the concrete root cause in `_patch_mistral_regex`/`model_info()`." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Gradient checkpointing in eval mode and ZeRO-3 gradient growth are different training-time issues." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Both are tokenizer-related, but the bugs are different: runtime special-token updates vs LayoutLMv2 NER/padding crashes." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "Different model behaviors: SigLIP2 AutoModel/pipeline mismatch vs CamemBERT masked-LM prediction regression." + }, + { + "left": "issue:45704", + "right": "issue:45859", + "accept": false, + "reason": "T5 apex RMSNorm memory leak and Qwen3_Moe CPU-offload incompatibility are unrelated." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Llama4 pad_token_id attribute error is unrelated to Qwen3-MoE sliding_window handling." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Both touch tokenizer regex handling, but one is a hard crash on Mistral tokenizer load and the other is a Kimi-K2.5 codec/regression warning issue." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "SigLIP2 docs/example problems are not the same as the quantization/offloading bug in the other issue." + }, + { + "left": "issue:45356", + "right": "issue:45907", + "accept": false, + "reason": "Kimi tokenizer regression and list_repo_files revision forwarding are unrelated." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "PEFT/local-checkpoint loading for GraniteSpeech does not match the quantization offloading problem." + }, + { + "left": "issue:40444", + "right": "issue:43746", + "accept": false, + "reason": "Qwen2.5-VL IterableDataset multi-image finetuning failure is unrelated to GraniteSpeech PEFT local-load issues." + }, + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "Deepspeed ZeRO-3/BertModel index error and `_torch_extract_fbank_features()` are different failure modes." + }, + { + "left": "issue:43746", + "right": "issue:45381", + "accept": false, + "reason": "PEFT checkpoint loading and Qwen2.5-VL video vision_position_ids are separate multimodal bugs." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Qwen3OmniMoe video-processing ValueError and Gemma 3n audio_mel_mask name collision are unrelated." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "BitNet packed-weight loading and DAC.from_latents STE mismatch are different models and different code paths." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "CLIPOutput attentions regression is unrelated to RTDetrV2 output drift in Transformers v5." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "Input validation messaging across models is not the same bug as the single-label classification zero-loss issue." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Audio feature extraction and add-new-model-like/TOKENIZER_MAPPING_NAMES are unrelated tokenizer/model registry issues." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "BigBird special-token decode bug is unrelated to BitNet accelerate loading of packed weights." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Void segmentation-map processing and a GenerationMixin import error are completely unrelated." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "SwitchTransformers sparse-layer config and Smollm3 RoPE-layer mismatch are different architecture/configuration bugs." + }, + { + "left": "issue:43295", + "right": "issue:45834", + "accept": false, + "reason": "Custom processor.tokenizer regression in v4.57.5 is not the same as Kosmos2.5 long OCR index errors." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is conservative and grounded in the packet. The only accepted pair, 42886/44843, is plausibly the same offline tokenizer-loading bug, while the other pairwise rejections are adequately supported by differing titles and failure modes. The summary avoids overclaiming and stays within the evidence available." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 15, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:30064", + "issue:36331", + "issue:39692", + "issue:41720", + "issue:43322", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43577", + "issue:43582", + "issue:43643", + "issue:43644", + "issue:43653", + "issue:43716", + "issue:43749", + "issue:43782", + "issue:43818", + "issue:43819", + "issue:43825", + "issue:43827", + "issue:43994", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44442", + "issue:44448", + "issue:44479", + "issue:44485", + "issue:44488", + "issue:44509", + "issue:44561", + "issue:44671", + "issue:44811", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45381", + "issue:45584", + "issue:45685", + "issue:45834" + ], + "soft_pairs": [ + "issue:43716|issue:44008", + "issue:43550|issue:45072", + "issue:43550|issue:43819", + "issue:43322|issue:45685", + "issue:44062|issue:44488", + "issue:44186|issue:44898", + "issue:44843|issue:45081", + "issue:43577|issue:43818", + "issue:43531|issue:43643", + "issue:43653|issue:43819", + "issue:43643|issue:44877", + "issue:43421|issue:44811", + "issue:43540|issue:45381", + "issue:43782|issue:44977", + "issue:44448|issue:44671", + "issue:45584|issue:45834", + "issue:43825|issue:43994", + "issue:44561|issue:45020", + "issue:39692|issue:45834", + "issue:44485|issue:44509", + "issue:43644|issue:43749", + "issue:44821|issue:45084", + "issue:30064|issue:36331", + "issue:41720|issue:44991", + "issue:43825|issue:43827", + "issue:44186|issue:44442", + "issue:44479|issue:45834", + "issue:43577|issue:43582" + ], + "prepared_review_unit_hash": "22fa7d1edcd72870d1a8f0afa0c32ed56a5d443358ed7ad664d341dfbcc40fd4", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36172, + "estimated_input_tokens": 9043, + "estimated_eval_tokens": 18342 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most candidate pairs are clearly unrelated. The only close theme is the v5 pipeline-removal/docs cluster, but the paired issues there still describe different artifacts (runtime error text vs documentation cleanup), so I rejected all soft edges. No PRs were provided.", + "confidence": 0.67, + "canonical_issue_reason": "issue:43825 is the most concrete, user-facing item in the nearest thematic cluster (v5 pipeline-removal behavior), but it is still not a clean duplicate anchor for the rest of the set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43825 is the best single issue to anchor triage among the near-match set because it describes a specific runtime regression rather than a broad or documentation-only report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Different models and different failure modes: image preprocessor dtype mismatch vs Gemma 3n variable-name collision." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Both involve model execution regressions, but one is Bamba + torch.compile/SDPA and the other is SwitchTransformers/TimmWrapper dtype mismatch; not the same bug." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Unrelated code paths: torch.compile/SDPA failure versus DAC latent/forward mismatch." + }, + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Different subsystems and symptoms: Llava Next loading segfault vs MPS histogram Int error." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "One is an AddedToken constructor argument error; the other is a tokenizer/model load failure for a specific repo." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "Different tokenization/image-classification failures with no shared concrete code path." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Both touch _patch_mistral_regex, but they fail for different reasons: offline model_info call vs missing backend_tokenizer attribute." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "Different models and defects: BLIP2 dtype propagation vs Video-LLaVA missing temporal attention/shared weights." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior is unrelated to trust_remote_code config field loss." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "Tokenizer special-token registration bug is unrelated to DAC from_latents/forward mismatch." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "Both concern config loading, but one is missing fields with trust_remote_code and the other is strict-config rejection of granite_speech." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Both mention tokenizer behavior, but one is runtime post-processor updates and the other is Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Related model family, but different regressions and different affected video-processing paths." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Different Qwen variants and different failure modes: weight_only load error vs flash-attention generation bug." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Both are output regressions, but they affect different models and distinct inference paths." + }, + { + "left": "issue:45584", + "right": "issue:45834", + "accept": false, + "reason": "Whisper empty-transcription generation and Kosmos2.5 long-OCR index error are unrelated." + }, + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "Both are about docs/model usage, but one is pipeline error-message text and the other is AutoModel/pipeline output quality for SigLIP2." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "The former is a specific API-removal breakage; the latter is a broad umbrella report about remote_code breakage." + }, + { + "left": "issue:39692", + "right": "issue:45834", + "accept": false, + "reason": "SigLIP2 docs/model mismatch and Kosmos2.5 OCR index error are unrelated." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "GLM-5 RoPE discussion is unrelated to v5 pipeline-task documentation cleanup." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Non-persistent buffer initialization bug is unrelated to FSDP CPU RAM-efficient loading." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "AutoImageProcessor URL loading and template-node compilation errors are different failures." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Image segmentation-map processing and CustomTrainer compute_loss signature mismatch are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:44991", + "accept": false, + "reason": "Qwen3 auto device-mapping CUDA assert and est-roberta tokenizer loading are different issues." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "Both relate to v5 pipeline removals, but one reports a runtime error-message regression and the other is a documentation cleanup issue; not the same concrete bug." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "LayoutLMv2 tokenizer NER/padding crash and FastSpeech2Conformer AutoTokenizer loading failure are unrelated." + }, + { + "left": "issue:44479", + "right": "issue:45834", + "accept": false, + "reason": "Qwen video-input regression and Kosmos2.5 long OCR index error are different model paths and symptoms." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "BLIP2 dtype propagation and Apple Silicon caching_allocator_warmup TypeError are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 16, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:34634", + "issue:36010", + "issue:36331", + "issue:42175", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43452", + "issue:43526", + "issue:43582", + "issue:43606", + "issue:43638", + "issue:43650", + "issue:43653", + "issue:43696", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43746", + "issue:43749", + "issue:43782", + "issue:43819", + "issue:43906", + "issue:44060", + "issue:44291", + "issue:44368", + "issue:44554", + "issue:44589", + "issue:44610", + "issue:44671", + "issue:44805", + "issue:44811", + "issue:44857", + "issue:44912", + "issue:44936", + "issue:44991", + "issue:45020", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45406", + "issue:45459", + "issue:45538", + "issue:45600", + "issue:45701", + "issue:45803" + ], + "soft_pairs": [ + "issue:44805|issue:44936", + "issue:43606|issue:45072", + "issue:44671|issue:45701", + "issue:34634|issue:36010", + "issue:44610|issue:45072", + "issue:43782|issue:45325", + "issue:43720|issue:45072", + "issue:42175|issue:45081", + "issue:45081|issue:45406", + "issue:43749|issue:44291", + "issue:44060|issue:44368", + "issue:43650|issue:43906", + "issue:43582|issue:44857", + "issue:43262|issue:43819", + "issue:43526|issue:43697", + "issue:43526|issue:44554", + "issue:45459|issue:45803", + "issue:34634|issue:36331", + "issue:45084|issue:45538", + "issue:43329|issue:45600", + "issue:43653|issue:44912", + "issue:43742|issue:45020", + "issue:43638|issue:44936", + "issue:43746|issue:44811", + "issue:43257|issue:43329", + "issue:43452|issue:44991", + "issue:43257|issue:43696", + "issue:44589|issue:45042" + ], + "prepared_review_unit_hash": "b5589a7e7dc6d6a9bb85b23a8522e4c73b7c581b7eb2c5ee818bbb7b4b6c0e02", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36442, + "estimated_input_tokens": 9111, + "estimated_eval_tokens": 18478 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: a mix of unrelated model-loading, processor, tokenizer, trainer, and backend-specific regressions. I did not find any soft pair that looks like the same underlying bug or change.", + "confidence": 0.82, + "canonical_issue_reason": "issue:45020 is the broadest and most representative report in the set: a general regression where newer Transformers versions break remote-code models, which matches the cluster\u2019s overall loading/regression theme better than the narrower model-specific issues.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the best standalone issue here because it is broad, clearly described, and likely to serve as the most useful umbrella reference for related version-regression reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "Different failure modes: a mask/index shape mismatch versus an evaluation-after-training regression. No evidence they share the same code path." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "Different models and different bug classes: CPU offload device mismatch vs bfloat16 dtype mismatch." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "CamemBERT masked-LM prediction regression is not the same as a broad tokenization-change report." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "BarkProcessor voice_preset failure and GenerationMixin import error are unrelated." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Processor output size mismatch and dtype mismatch are distinct problems." + }, + { + "left": "issue:43782", + "right": "issue:45325", + "accept": false, + "reason": "Different Qwen variants and different bugs: weight_only loading error vs rope-index scaling." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Packed-weight unpacking during accelerate loading is unrelated to generic dtype mismatch behavior." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Package/backend dependency issue vs tokenizer regex crash; no shared underlying bug." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Both are tokenizer/processor loading problems, but they affect different components and failure causes." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "Both involve model loading, but one is FSDP RAM-efficient loading and the other is init_empty_weights argument handling; not the same bug." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "These are different tied-weights warnings with different triggers and configurations." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "One is non-technical placeholder content; the other is a reproduction of a specific upstream issue." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "AppleSilicon allocator warmup TypeError and CUDA AMP loss crash are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and DAC latent conversion mismatch are separate issues." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug does not match an RTDetrV2 output-difference regression." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "Label reduction bug and MPS attention correctness issue are unrelated." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Both concern broad exception handling, but they are different code paths and different observable bugs." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "BarkProcessor preset behavior and CustomTrainer compute_loss signature mismatch are unrelated." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "Template-compilation TypeError and tokenizer model_max_length are different tokenizer-layer issues." + }, + { + "left": "issue:43329", + "right": "issue:45600", + "accept": false, + "reason": "Undefined multimodal-token video branch symbols and removed auto-mapping configs are unrelated CI/codegen problems." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "BigBird special-token decoding bug and MXFP4 quantization fallback bug are distinct." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "A key error for a specific model is not the same as the broader remote_code regression." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Zero3 mask indexing failure and train-then-evaluate failure are different runtime issues." + }, + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "PEFT local checkpoint loading and Whisper batch_decode skip_special_tokens behavior do not share a concrete code path." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "MOE weight conversion during accelerate/deepspeed loading is unrelated to multimodal token counting." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "Both involve from_pretrained/tokenizer loading, but they target different APIs and failure mechanisms." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "Loading/conversion bug and CUDA OOM are clearly unrelated." + }, + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Float8 storage lookup failure and PIL/torchvision dependency regression are different classes of issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 17, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:34567", + "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:43531", + "issue:43550", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43643", + "issue:43645", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43819", + "issue:43854", + "issue:43927", + "issue:43931", + "issue:44291", + "issue:44479", + "issue:44568", + "issue:44610", + "issue:44756", + "issue:44811", + "issue:44849", + "issue:44912", + "issue:44977", + "issue:44987", + "issue:45042", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45488", + "issue:45538", + "issue:45701", + "issue:45759", + "issue:45859" + ], + "soft_pairs": [ + "issue:43854|issue:44977", + "issue:38175|issue:44610", + "issue:39692|issue:44912", + "issue:43819|issue:43927", + "issue:36010|issue:43645", + "issue:43720|issue:44610", + "issue:43454|issue:43577", + "issue:43550|issue:44912", + "issue:43262|issue:44811", + "issue:43329|issue:44479", + "issue:43531|issue:45042", + "issue:36010|issue:36331", + "issue:43749|issue:44756", + "issue:44568|issue:45356", + "issue:43643|issue:45759", + "issue:44568|issue:45488", + "issue:44849|issue:45701", + "issue:34567|issue:40990", + "issue:43329|issue:43450", + "issue:43257|issue:43931", + "issue:43606|issue:45859", + "issue:43257|issue:43756", + "issue:43697|issue:45357", + "issue:43452|issue:43582", + "issue:44291|issue:45042", + "issue:44987|issue:45538", + "issue:44479|issue:45381", + "issue:42915|issue:43742" + ], + "prepared_review_unit_hash": "cc46032d771ace898aa070f2da3becd97461984e766e308f7e4fd16805687f4c", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36345, + "estimated_input_tokens": 9087, + "estimated_eval_tokens": 18430 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a heterogeneous set of transformer bug reports, not one duplicate family. I only see one strong duplicate-style overlap: the Qwen2.5-VL/Qwen3.5 video-input regression pair.", + "confidence": 0.44, + "canonical_issue_reason": "issue:44479 is the best canonical anchor for the one clear duplicate family here because it captures the broader Qwen video-input regression that also covers the specific qwen2.5-vl symptom in issue:45381.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45701 is the broadest open issue, but the set is too mixed to make it a strong global duplicate canonical; issue:44479 is the best local match to the only accepted overlap.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "Different models and failure modes: GLM-4.7-Flash loading/test issue vs Qwen3.5 flash-attention generation behavior." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Both are vision-related, but one is a SigLIP2 probability bug and the other is an OmDet-Turbo processor size mismatch." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "Both mention quantization, but they affect different models and different code paths (docs example/processor vs MXFP4 fallback)." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "Unrelated bugs: DAC latent conversion mismatch vs DiaConfig losing token IDs on save/load." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "Both are import/config compatibility regressions, but they are distinct failures with different triggers and code paths." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "Packed-weight loading bug in BitNet is unrelated to an OmDet-Turbo image processor shape issue." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "Different models and symptoms: AyaVision lm_head tying vs BLIP2 dtype propagation." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "Different problem classes: SDPA/torch.compile failure in Bamba vs quantization fallback in git-oss-20b." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Both audio-related, but one is chat-template sampling rate defaulting and the other is batch_decode skip_special_tokens handling." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "Both involve video multimodal processing, but one is an undefined-helper bug in token counting and the other is a broader video-input regression." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to PIL image processors requiring torchvision." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "ImportError for GenerationMixin and compute_loss signature mismatch are separate API breakages." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "Different loading/memory issues: FSDP CPU RAM-efficient loading vs mmap OOM on Strix Halo." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Both tokenizer regressions, but the BOS/EOS special-token bug is unrelated to Kimi-K2.5 codec handling." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "Both are config-loading issues, but one is trust_remote_code field loss and the other is composite Qwen multimodal text_config unwrapping." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Tokenizer regressions in different tokenizers and different mechanisms; not the same underlying bug." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Qwen3.5 hidden-states output bug is unrelated to the broad tokenization-change report." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "TrainerState token-count tracking and GPT-OSS perplexity are unrelated." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Both are video-related, but one is token counting and the other is batched output shape." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Different Qwen3 loading failures: accelerate/deepspeed weight conversion vs VL weight-shape mismatch." + }, + { + "left": "issue:43606", + "right": "issue:45859", + "accept": false, + "reason": "Both mention CPU offload, but they affect different models and different failure mechanisms." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Qwen3 MoE loading/accelerate issue and Smollm3 RoPE-layer dropout are unrelated." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "RTDetrV2 output regression and Qwen3.5 save_pretrained key mapping are distinct bugs." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "gguf_file loading breakage and AppleSilicon warmup TypeError are unrelated." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "init_empty_weights/_is_hf_initialized incompatibility is unrelated to the PIL image processor torchvision regression." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Different tokenizer issues: loading physical-intelligence/fast vs CLIPTokenizer max-length sentinel." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "Same underlying video-input regression in the Qwen2.5-VL/Qwen3.5 family; one report is broad and the other is a specific symptom of the same code path." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Different model-loading failures: FineGrainedFP8Config on Qwen3Moe vs key error loading MobileLLM-125M." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded and conservative overall. The only accepted overlap is the Qwen video-input regression family, and the summary does not overclaim broader duplication across the heterogeneous set." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 18, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:28282", + "issue:29127", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:36331", + "issue:38175", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43317", + "issue:43493", + "issue:43526", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43782", + "issue:43824", + "issue:43906", + "issue:43927", + "issue:43994", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44560", + "issue:44737", + "issue:44821", + "issue:44898", + "issue:44912", + "issue:44938", + "issue:44987", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45137", + "issue:45325", + "issue:45381", + "issue:45405", + "issue:45479", + "issue:45538", + "issue:45588", + "issue:45823", + "issue:45907" + ], + "soft_pairs": [ + "issue:33357|issue:34567", + "issue:41720|issue:43540", + "issue:44987|issue:45907", + "issue:42915|issue:44821", + "issue:43927|issue:44912", + "issue:42915|issue:45538", + "issue:43317|issue:45538", + "issue:42175|issue:43824", + "issue:43540|issue:44479", + "issue:43577|issue:45072", + "issue:43540|issue:44560", + "issue:42175|issue:44938", + "issue:44062|issue:44991", + "issue:38175|issue:44898", + "issue:45137|issue:45405", + "issue:42175|issue:44488", + "issue:43906|issue:45381", + "issue:28282|issue:34689", + "issue:44737|issue:45588", + "issue:42915|issue:43749", + "issue:43742|issue:44291", + "issue:43582|issue:43782", + "issue:43526|issue:45479", + "issue:28282|issue:29127", + "issue:43540|issue:45823", + "issue:43756|issue:45325", + "issue:43493|issue:43994", + "issue:34567|issue:36331", + "issue:44291|issue:44821", + "issue:44062|issue:45081" + ], + "prepared_review_unit_hash": "c81dbd0d05f05e52d103ec8225a26e221a6bb63cc9cfac2d8a96863e2becc8ab", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36326, + "estimated_input_tokens": 9082, + "estimated_eval_tokens": 18420 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a grab bag of unrelated issues, so only a few soft pairs look like true duplicates. The clearest overlap is around Qwen video/position-id regressions and a separate SigLIP2 discrepancy pair.", + "confidence": 0.62, + "canonical_issue_reason": "Issue 44479 is the broadest concrete regression report in the set, covering multiple Qwen VL/video-input variants and serving as the best umbrella for the accepted Qwen-related duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "44479 is the most representative issue overall: it is specific enough to be actionable, but broad enough to subsume the closest matching reports in the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "Different subsystems and failures: MacOS bus error in CLIP vs TrainerState token counter update." + }, + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "Both involve Qwen, but one is device-mapping CUDA assert and the other is a video-input ValueError for a different model path." + }, + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "Loading a model regression is unrelated to a repo-files API forwarding bug in CI." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "FineGrainedFP8/Qwen3Moe loading failure is unrelated to AutoImageProcessor URL loading." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "Custom config save/load corruption is not the same as an MXFP4 quantization fallback issue." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "Qwen3Moe FP8 loading and CLIPTokenizer max_length are different bugs." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "Device-map/offload loading failure is unrelated to tokenizer max_length." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "Packaging/backend installation issue vs model import error for Qwen2.5-VL." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Related domain, but different model families and symptoms; not clearly the same concrete bug." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Both mention dtype, but they affect different architectures and code paths." + }, + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "Same broad video-input area, but the titles do not establish the same underlying bug with enough confidence." + }, + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "TensorFlow backend packaging vs Python 3.14 import failure are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "Tokenizer construction TypeError and a model-specific tokenizer load failure are too indirect to treat as duplicates." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Different models and different failure modes; both are vision issues only at a very broad level." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque error is unrelated to PEFT version pinning." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "Installation/backend packaging issue vs model load failure for a specific checkpoint." + }, + { + "left": "issue:43906", + "right": "issue:45381", + "accept": false, + "reason": "The reproduction issue has no clear evidence it matches the Qwen2.5-VL position-id bug." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "Missing PyTorch dependency is unrelated to a model-loading regression for Llama 3.2 Vision." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "XLNet CPU positional encoding overhead and flash_attention sink handling are different code paths." + }, + { + "left": "issue:42915", + "right": "issue:43749", + "accept": false, + "reason": "Qwen3Moe FP8 loading and FSDP CPU RAM efficient loading are separate training/loading mechanisms." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "KeyError on a specific model load is not the same as the init_empty_weights API break." + }, + { + "left": "issue:43582", + "right": "issue:43782", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError and Qwen3VL weight_only load error are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "Image processor label reduction bug is unrelated to sequence-classification loss degeneracy." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "Missing PyTorch import error and LayoutLMv3 box validation message are distinct issues." + }, + { + "left": "issue:43540", + "right": "issue:45823", + "accept": false, + "reason": "Different models and different failure symptoms; not the same device-mapping bug." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "RoPE-layer mismatch in SmolLM3 is not the same as Qwen2.5-VL temporal position-id scaling." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": true, + "reason": "Both describe the same SigLIP2 implementation discrepancy, one as HF-vs-JAX mismatch and the other as nonsensical outputs in AutoModel/pipeline." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "TrainerState token counting and CustomTrainer compute_loss API mismatch are different trainer bugs." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "Init-empty-weights loading break is unrelated to AutoImageProcessor URL loading." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "Both are tokenizer-related, but the specific failure mechanisms and affected code paths are different." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": true, + "reason": "Same Qwen video-input regression family in transformers 5.3.0; the specific StopIteration report fits the broader video-input breakage." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": true, + "reason": "Both point to the same Qwen2.5-VL position-id/rope-index regression affecting still-image and video inputs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet and stays conservative. The accepted overlaps are the ones with clear title-level alignment (SigLIP2 pair, Qwen video/position-id regressions), and the rejected pairs avoid over-claiming duplicates when the evidence is only broad topical similarity." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 19, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:24643", + "issue:29127", + "issue:38175", + "issue:41762", + "issue:43295", + "issue:43452", + "issue:43531", + "issue:43577", + "issue:43643", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43819", + "issue:44464", + "issue:44466", + "issue:44484", + "issue:44488", + "issue:44514", + "issue:44561", + "issue:44568", + "issue:44589", + "issue:44625", + "issue:44704", + "issue:44743", + "issue:44792", + "issue:44811", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44936", + "issue:44964", + "issue:45005", + "issue:45042", + "issue:45081", + "issue:45216", + "issue:45290", + "issue:45356", + "issue:45357", + "issue:45375", + "issue:45381", + "issue:45405", + "issue:45600", + "issue:45823", + "issue:45920" + ], + "soft_pairs": [ + "issue:45357|issue:45920", + "issue:43643|issue:45375", + "issue:44792|issue:44964", + "issue:43452|issue:44488", + "issue:45042|issue:45081", + "issue:43742|issue:44589", + "issue:43531|issue:44821", + "issue:45405|issue:45600", + "issue:43697|issue:45216", + "issue:43688|issue:43749", + "issue:44514|issue:45381", + "issue:43720|issue:44898", + "issue:44484|issue:44704", + "issue:41762|issue:44936", + "issue:24643|issue:29127", + "issue:43452|issue:43577", + "issue:44625|issue:44877", + "issue:44514|issue:45290", + "issue:43452|issue:44843", + "issue:44466|issue:45005", + "issue:43819|issue:44811", + "issue:43295|issue:45356", + "issue:43643|issue:44561", + "issue:44464|issue:44743", + "issue:44936|issue:45823", + "issue:43653|issue:44568", + "issue:38175|issue:43295" + ], + "prepared_review_unit_hash": "d9e9f681de5af9f56d610906072b6176b1422c544750a068fb370df35c27e2a7", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 36144, + "estimated_input_tokens": 9036, + "estimated_eval_tokens": 18328 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a noisy mix of unrelated regression reports around tokenizers, configs, loading, and generation. The best representative issue is the broad AutoTokenizer token-ID regression, but none of the soft pairs look like the same concrete bug/change.", + "confidence": 0.44, + "canonical_issue_reason": "issue:45920 is the broadest and most central report: it describes a cross-model AutoTokenizer regression affecting multiple model families, which best matches the cluster\u2019s dominant tokenizer/config regression theme.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45920 is the strongest cluster representative because it is open, broad in scope, and explicitly aggregates failures across several model families rather than a single narrow edge case.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45357", + "right": "issue:45920", + "accept": false, + "reason": "Both concern Qwen-related regressions, but one is about incorrect save_pretrained visual encoder keys while the other is about wrong token IDs from AutoTokenizer; different failure modes and code paths." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "Both involve config-field handling, but one is trust_remote_code returning missing fields and the other is strict dropping a specific Qwen vision field; not the same bug." + }, + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Different models and symptoms: a janus test failure versus loading Phi-4 multimodal; no clear shared underlying code-path bug." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "Both are model loading issues, but one is gguf_file handling for AutoTokenizer/AutoModel and the other is a specific model load failure; too broad and not the same concrete defect." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "Unrelated subsystems: PIL image processor torchvision dependency versus Mistral regex patch crash." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "Different failure classes: key error during model loading versus missing Float8 storage object." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to AutoImageProcessor URL loading." + }, + { + "left": "issue:45405", + "right": "issue:45600", + "accept": false, + "reason": "One is a version pin bump for PEFT; the other is a stale auto_mappings reference causing CI breakage." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "Different regressions in different areas: RTDetrV2 output changes versus Qwen3.5 save_pretrained serialization." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Auxiliary-loss normalization and FSDP CPU RAM efficient loading are unrelated issues." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen/VL-related, but one is chat template batching with padding=False and the other is wrong vision_position_ids for video input; different bugs." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "BitNet packed-weight loading and Perceiver interpolation behavior are unrelated code paths." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "A save_pretrained shard-size question is not the same as AutoProcessor failing to forward kwargs to cached_file." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "DeepSpeed ZeRO-3 model loading error versus trainer.evaluate failure after train; different layers of the stack and different symptoms." + }, + { + "left": "issue:24643", + "right": "issue:29127", + "accept": false, + "reason": "DeepSpeed weight-dimension runtime error and LayoutLMv3 box-info error-message clarity are unrelated." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "gguf_file loading breakage and Blip2 dtype propagation are distinct model-loading regressions." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Both touch config handling, but one is Qwen3.5 num_labels propagation and the other is granite_speech strict config loading; not the same bug." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Both are chat-template crashes, but the triggering conditions differ materially: batched input with padding=False versus assistant tool-call messages with no content." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "AutoTokenizer gguf_file loading and mistral-regex offline-mode behavior are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": false, + "reason": "Both mention tied weights, but one is device-dependent serialization of lm_head.weight while the other is a broader translation-model issue; insufficient evidence they share one concrete bug." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "DAC latent decoding mismatch and Whisper batch_decode skip_special_tokens handling are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Both are regressions around processors/tokenizers, but one is processor.tokenizer access and image passing in custom code, while the other is Kimi-K2.5 codec handling and regex warnings." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Both involve trust_remote_code, but one is missing config fields and the other is removal of is_torch_fx_available breaking remote code; different root causes." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Both mention generation/cache behavior, but one is compiled-forward chunked generation inconsistency and the other is recurrent states resetting in modular_qwen3_5; not the same bug." + }, + { + "left": "issue:44936", + "right": "issue:45823", + "accept": false, + "reason": "Trainer lifecycle failure and Gemma4 PLE device-map mismatch are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Both concern tokenizer special tokens, but one is BigBird mask-token registration and the other is mdeberta BOS/EOS addition; different tokenizers and defects." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Siglip2 zero probabilities and processor.tokenizer regression are unrelated." + }, + { + "left": "issue:45375", + "right": "issue:45920", + "accept": false, + "reason": "Both are Qwen3.5-related, but one is a missing vision-config field dropped by strict and the other is wrong token IDs from AutoTokenizer across many models; different layers and symptoms." + }, + { + "left": "issue:45381", + "right": "issue:44792", + "accept": false, + "reason": "Qwen2.5-VL video position-id bug and janus test failure are not the same issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 20, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:28282", + "issue:30064", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:34567", + "issue:35141", + "issue:36010", + "issue:41720", + "issue:43122", + "issue:43278", + "issue:43329", + "issue:43452", + "issue:43493", + "issue:43504", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43606", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43825", + "issue:43906", + "issue:43927", + "issue:44361", + "issue:44442", + "issue:44451", + "issue:44464", + "issue:44479", + "issue:44554", + "issue:44610", + "issue:44625", + "issue:44821", + "issue:44843", + "issue:44849", + "issue:44977", + "issue:44991", + "issue:45325", + "issue:45357", + "issue:45440", + "issue:45588", + "issue:45701", + "issue:45704" + ], + "soft_pairs": [ + "issue:43688|issue:45440", + "issue:43653|issue:44361", + "issue:44361|issue:44442", + "issue:41720|issue:43906", + "issue:43688|issue:45704", + "issue:30064|issue:33666", + "issue:28282|issue:32090", + "issue:43577|issue:43697", + "issue:28282|issue:33357", + "issue:28282|issue:35141", + "issue:43688|issue:45701", + "issue:43646|issue:43749", + "issue:44843|issue:44991", + "issue:43653|issue:43927", + "issue:43531|issue:43742", + "issue:43550|issue:43927", + "issue:44464|issue:44977", + "issue:43329|issue:45325", + "issue:43577|issue:45588", + "issue:43577|issue:44554", + "issue:44479|issue:44625", + "issue:43550|issue:44361", + "issue:43122|issue:43906", + "issue:32090|issue:34567", + "issue:41720|issue:44464", + "issue:43504|issue:43550", + "issue:43504|issue:43720", + "issue:43504|issue:44610", + "issue:32090|issue:33666", + "issue:44849|issue:45357", + "issue:43452|issue:44451", + "issue:43278|issue:43688", + "issue:43742|issue:44821", + "issue:33666|issue:36010", + "issue:43493|issue:43825", + "issue:43606|issue:43927", + "issue:43720|issue:43927" + ], + "prepared_review_unit_hash": "db154134b66dd838e029b544a663f40c12cc90116d2c9a79079e04331324a05e", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 37, + "serialized_chars": 38809, + "estimated_input_tokens": 9703, + "estimated_eval_tokens": 19662 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous: it mixes unrelated tokenizer, vision, training, generation, loading, and dtype regressions. The soft-similarity links are mostly superficial, so none of the candidate pairs look like true duplicates or mergeable PR-equivalents.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the broadest regression-style report and the closest thing to an anchor, but the set is too mixed to treat it as a real duplicate hub.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Different model families and different bugs: auxiliary-loss normalization vs DeepSeekV3 implementation divergence." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "Both involve tokenizers, but they are different tokenizer classes with different failure modes and fixes." + }, + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Both are tokenizer-loading issues, but the affected tokenizer and error path are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:43906", + "accept": false, + "reason": "No evidence they describe the same CUDA/device-mapping bug; 43906 is just an isolated repro reference." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "Unrelated subsystems: expert-loss normalization vs T5 RMSNorm memory leak." + }, + { + "left": "issue:30064", + "right": "issue:33666", + "accept": false, + "reason": "Image segmentation map processing and multi-GPU training are different problems." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "Import-time PyTorch detection error vs trainer broadcast NoneType error; not the same bug." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "Different model families and different issues: dtype propagation vs incorrect outputs in RTDetrV2." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "PyTorch missing import vs MacOS bus error on CLIP are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:35141", + "accept": false, + "reason": "ImportError and embedding reinitialization are distinct code paths and bugs." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "Aux-loss normalization is unrelated to a tokenization-version regression." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Custom model initialization breakage and FSDP CPU RAM efficient loading are different failures." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-related, but one is an offline hub lookup bug and the other is a specific tokenizer load failure." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "Special-token registration bug vs saved custom token IDs being lost; different artifacts and causes." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to a key error loading MobileLLM-125M." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "torch.compile/SDPA failure and config save/load token-ID loss are unrelated." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Compiled-forward chunked generation and flash-attention generation failures are different code paths." + }, + { + "left": "issue:43329", + "right": "issue:45325", + "accept": false, + "reason": "Video-token counting bug and still-image rope-index scaling bug are distinct multimodal issues." + }, + { + "left": "issue:43577", + "right": "issue:45588", + "accept": false, + "reason": "BLIP2 dtype propagation and flash_attention s_aux=None crash are unrelated." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "Dtype loading for BLIP2 does not match an MPS attention correctness issue." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Video-input regression in Qwen VL models is unrelated to num_labels propagation in Qwen3.5." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Different bugs: hidden-states output mismatch vs save_pretrained writing wrong visual keys." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "General gguf tokenizer/model loading breakage and a specific model tokenizer load failure are not the same underlying defect." + }, + { + "left": "issue:43278", + "right": "issue:43688", + "accept": false, + "reason": "Embedding dtype mismatch during eval is unrelated to auxiliary-loss normalization." + }, + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "Loading a text model with a key error is unrelated to loading an image processor from a URL." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Multi-GPU training and a GenerationMixin import error are unrelated." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and pipeline translation-task messaging are different issues." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "CPU offload device mismatch has nothing to do with DiaConfig token-ID persistence." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "Packed-weight unpacking during accelerate loading is unrelated to DiaConfig save/load token-ID loss." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 21, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:28218", + "issue:33357", + "issue:33666", + "issue:36331", + "issue:39692", + "issue:43257", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43299", + "issue:43334", + "issue:43493", + "issue:43504", + "issue:43525", + "issue:43540", + "issue:43572", + "issue:43606", + "issue:43646", + "issue:43653", + "issue:43720", + "issue:43756", + "issue:43824", + "issue:43844", + "issue:43927", + "issue:44291", + "issue:44361", + "issue:44423", + "issue:44464", + "issue:44466", + "issue:44488", + "issue:44568", + "issue:44589", + "issue:44625", + "issue:44734", + "issue:44843", + "issue:44913", + "issue:44964", + "issue:44977", + "issue:45083", + "issue:45200", + "issue:45229", + "issue:45405", + "issue:45520", + "issue:45584", + "issue:45759", + "issue:45812", + "issue:45823" + ], + "soft_pairs": [ + "issue:43262|issue:43295", + "issue:44568|issue:45812", + "issue:43756|issue:44913", + "issue:43525|issue:43572", + "issue:28218|issue:33357", + "issue:43720|issue:44361", + "issue:33666|issue:36331", + "issue:39692|issue:43493", + "issue:43493|issue:43844", + "issue:43540|issue:44977", + "issue:43278|issue:43844", + "issue:44964|issue:45405", + "issue:44625|issue:45584", + "issue:43334|issue:43572", + "issue:43257|issue:43299", + "issue:43756|issue:45083", + "issue:45200|issue:45229", + "issue:44488|issue:44843", + "issue:43334|issue:45759", + "issue:43646|issue:44466", + "issue:39692|issue:43295", + "issue:45229|issue:45823", + "issue:43824|issue:44589", + "issue:33357|issue:33666", + "issue:44589|issue:45520", + "issue:43504|issue:43606", + "issue:43572|issue:44291", + "issue:44423|issue:44734", + "issue:43927|issue:44361", + "issue:43504|issue:43653", + "issue:44464|issue:45584" + ], + "prepared_review_unit_hash": "d4f3abd4771b6d1eef72afec65cb8f4792ae5aa9977fd0c44655e99c95a3678f", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36769, + "estimated_input_tokens": 9193, + "estimated_eval_tokens": 18642 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: most pairs share only broad subsystem or model-family similarity, not the same underlying bug. I would not merge any of the soft-edge pairs.", + "confidence": 0.92, + "canonical_issue_reason": "No single canonical issue stands out for this mixed set; if one must be chosen, issue 45812 is the clearest broad-impact regression with a concrete tokenizer failure.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45812 is the strongest standalone candidate because it describes a silent, model-wide regression affecting Granite tokenization and is clearly reproducible.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Both involve processor/tokenizer behavior, but one is an audio sampling-rate default and the other is a multimodal regression around processor.tokenizer/images; different code paths." + }, + { + "left": "issue:44568", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and fail differently: BOS/EOS handling for mdeberta-v3 vs wrong token IDs for Granite." + }, + { + "left": "issue:43756", + "right": "issue:44913", + "accept": false, + "reason": "Both mention RoPE-related config behavior, but one is about Smollm3 layer dropping and the other is a GPTNeoX config value not persisting on reload." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both are missing token-id fields in configs, but they concern different model families and different attributes (pad_token_id vs pad_token_idx)." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer spacing bug vs MacOS bus error in a CLIP model; no shared underlying defect." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "BitNet accelerate-loading unpacking bug vs tokenizer AttributeError on tasks; unrelated." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Multi-GPU training request for Qwen2-VL vs a CustomTrainer signature break; different problems." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "Both touch SigLIP2, but one is a documentation/example mismatch and the other is an implementation discrepancy with JAX; not the same bug." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "SigLIP2 implementation mismatch vs a gradient blow-up with ZeRO-3; unrelated." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3 OmniMoe video-input validation vs Qwen3.5 flash-attention generation failure; different code paths." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Embedding dtype drift in eval vs abnormal gradients under DeepSpeed; no direct match." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Phi-4 multimodal loading failure vs an unreleased PEFT version bump; unrelated." + }, + { + "left": "issue:44625", + "right": "issue:45584", + "accept": false, + "reason": "Qwen3.5 config propagation bug vs Whisper empty-transcription generation failure; different models and failures." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Both are missing pad-token metadata in configs, but they affect different model classes and arise in different loading paths." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Both involve Qwen3 MOE loading, but one is accelerate/deepspeed weight conversion and the other is a dev-version multimodal config break; not the same issue." + }, + { + "left": "issue:43756", + "right": "issue:45083", + "accept": false, + "reason": "Both are Qwen3 OmniMoe-related, but one is RoPE-layer selection and the other is a feature-length helper behaving unexpectedly." + }, + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Text-only token-type-id defaulting vs multi-GPU inference OOM; unrelated." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "Model-loading failure for a specific checkpoint vs offline-mode break in AutoTokenizer patching; different causes." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "Both relate to composite Qwen multimodal configs, but one is a missing pad_token_id attribute and the other is from_config not unwrapping text_config." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Both are v5 regressions, but one is custom model initialization and the other is tied-weight serialization; unrelated code paths." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "Documentation/example issues vs a processor/tokenizer regression; not the same underlying bug." + }, + { + "left": "issue:45229", + "right": "issue:45823", + "accept": false, + "reason": "Both involve Gemma4 and distributed/device behavior, but one is CUDA OOM during multi-GPU inference and the other is a PLE device mismatch on forward." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "Import error for a Qwen2.5-VL class vs a missing Float8 storage type during load; unrelated." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "MacOS bus error in CLIP vs Qwen2-VL multi-GPU training; no shared bug." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Different import/load failures: missing Float8 storage type vs flash_attn KeyError on Python 3.13." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Legacy-field model loading bug vs CPU offload device mismatch; unrelated." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig vs init_empty_weights rejecting _is_hf_initialized; different compatibility issues." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Both are serving crashes, but one is multimodal continuous-batching and the other is KV-cache continuation indexing; not the same root cause." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "DiaConfig save/load token-ID loss vs MLukeTokenizer AttributeError; unrelated." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "Beit legacy-field loading issue vs BigBird tokenizer special-token registration; different components and failures." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency vs Whisper empty-transcription handling; unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 22, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:15354", + "issue:28218", + "issue:29942", + "issue:30333", + "issue:34567", + "issue:37428", + "issue:38175", + "issue:43232", + "issue:43278", + "issue:43295", + "issue:43493", + "issue:43540", + "issue:43575", + "issue:43582", + "issue:43606", + "issue:43701", + "issue:43844", + "issue:43873", + "issue:43957", + "issue:43994", + "issue:44112", + "issue:44315", + "issue:44464", + "issue:44466", + "issue:44493", + "issue:44521", + "issue:44554", + "issue:44610", + "issue:44792", + "issue:44849", + "issue:44898", + "issue:44964", + "issue:44977", + "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:45563", + "issue:45823" + ], + "soft_pairs": [ + "issue:43994|issue:44493", + "issue:43606|issue:44112", + "issue:43575|issue:45405", + "issue:43278|issue:43701", + "issue:44792|issue:45405", + "issue:29942|issue:30333", + "issue:44521|issue:45290", + "issue:43844|issue:44964", + "issue:28218|issue:34567", + "issue:43582|issue:44977", + "issue:44554|issue:45446", + "issue:45084|issue:45520", + "issue:43232|issue:44849", + "issue:15354|issue:38175", + "issue:43540|issue:45200", + "issue:45072|issue:45198", + "issue:45200|issue:45823", + "issue:37428|issue:43957", + "issue:44315|issue:44849", + "issue:43493|issue:43701", + "issue:44554|issue:45356", + "issue:43295|issue:44610", + "issue:43873|issue:44464", + "issue:44466|issue:45127", + "issue:44898|issue:45412", + "issue:44792|issue:45563", + "issue:45200|issue:45325" + ], + "prepared_review_unit_hash": "5b0a397b57349e1108c6ee4006be66b6636fc4fe9159e4b6db54ebf75ca5485b", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35280, + "estimated_input_tokens": 8820, + "estimated_eval_tokens": 17896 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The items are mostly heterogeneous, and the soft-similarity pairs do not look like true duplicates. They share broad themes or model families, but not the same concrete bug or fix path.", + "confidence": 0.83, + "canonical_issue_reason": "No single canonical issue stands out for the whole set; the cluster is too mixed to represent one underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "If one issue must represent the set, issue 44521 is the most broadly impactful open regression with active discussion, but it is not a duplicate target for the other items.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "Both mention position-id/output regressions, but one is a SigLIP2 inference quality bug and the other is a generic unexpected-key issue; different failures." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "Both are CI/device-mismatch failures, but they affect different models and code paths." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "OOM during tensor-parallel loading is unrelated to a PEFT version pin regression." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "A dtype shift in eval is not the same bug as a resume-from-checkpoint key mismatch." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "Janus test failure and MIN_PEFT_VERSION bump are unrelated." + }, + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention 2 test failures do not match the MLflow reporting issue." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Both involve chat templating, but one is all-zero assistant masks for multimodal inputs and the other is a crash on tool-call assistant messages." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "ZeRO-3 gradient behavior and latest-transformers model loading are different bugs." + }, + { + "left": "issue:28218", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer spacing after added tokens is unrelated to TrainerState token-count updates." + }, + { + "left": "issue:43582", + "right": "issue:44977", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError and Qwen3.5 flash-attention generation bug are unrelated." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "An MPS attention correctness issue is not the same as a PyTorch version-check import bug." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Compilation failure for non-template nodes and a Python 3.13 flash_attn import KeyError are separate issues." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation state update after sync_gpus is unrelated to the hidden-states output bug." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "TorchScript GeneratorExp support and SigLIP2 zero-probabilities are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Video processing failure in Qwen3OmniMoe is not the same as Gemma 4 token-type defaults." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "A dtype mismatch in CI and a Wav2Vec2 save/tokenization failure are different concrete bugs." + }, + { + "left": "issue:45200", + "right": "issue:45823", + "accept": false, + "reason": "Same model family, but one is missing mm_token_type_ids defaults and the other is a PLE device mismatch; not one fix." + }, + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "flash_supports_window_size import failure and meta-device model loading regressions are unrelated." + }, + { + "left": "issue:44315", + "right": "issue:44849", + "accept": false, + "reason": "Liger Kernel application with model_init is unrelated to the Qwen3.5 hidden-states bug." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 JAX discrepancy and checkpoint key mismatch do not share a code path." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "MPS attention dim mismatch and tokenizer codec/warning regression are different subsystems." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Processor/tokenizer regression and OmDet-Turbo image-size mismatch are separate bugs." + }, + { + "left": "issue:43873", + "right": "issue:44464", + "accept": false, + "reason": "Quantization offloading behavior and compiled-forward chunked generation are not the same issue." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "Both touch weights/serialization, but one is inconsistent lm_head serialization and the other is LoRA merge collapse with extended vocab." + }, + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Perceiver interpolation failure and RT-DETR memory release are unrelated." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "Janus image-generation test failure and a stale paged-generate warning are different problems." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Gemma 4 token-type defaults and Qwen2.5-VL rope-index scaling are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 23, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:28218", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:42915", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43334", + "issue:43335", + "issue:43388", + "issue:43450", + "issue:43493", + "issue:43525", + "issue:43526", + "issue:43550", + "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:44112", + "issue:44206", + "issue:44360", + "issue:44442", + "issue:44514", + "issue:44521", + "issue:44561", + "issue:44610", + "issue:44857", + "issue:44936", + "issue:44987", + "issue:45083", + "issue:45127", + "issue:45198", + "issue:45245", + "issue:45412", + "issue:45584" + ], + "soft_pairs": [ + "issue:43572|issue:44987", + "issue:37428|issue:43824", + "issue:43844|issue:44936", + "issue:43572|issue:43957", + "issue:43295|issue:43450", + "issue:43334|issue:43525", + "issue:43824|issue:43975", + "issue:43295|issue:43873", + "issue:43388|issue:43526", + "issue:38175|issue:45412", + "issue:42915|issue:43824", + "issue:38175|issue:43493", + "issue:28218|issue:35141", + "issue:44514|issue:44521", + "issue:44442|issue:45198", + "issue:43278|issue:43856", + "issue:43653|issue:44112", + "issue:43827|issue:43901", + "issue:43450|issue:43575", + "issue:43450|issue:43493", + "issue:44206|issue:45584", + "issue:45127|issue:45245", + "issue:43819|issue:45083", + "issue:43262|issue:43450", + "issue:43819|issue:44112", + "issue:44610|issue:44857", + "issue:43278|issue:43825", + "issue:43606|issue:45083", + "issue:44360|issue:44561", + "issue:43335|issue:43550" + ], + "prepared_review_unit_hash": "a7e47e411776c1f8a67130787f0f5338df0a3c08412da6b34a8966f5b51eb669", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36271, + "estimated_input_tokens": 9068, + "estimated_eval_tokens": 18392 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is heterogeneous: the soft-similarity pairs mostly share only broad themes (tokenizers, configs, processors, loading, training) but not the same concrete bug or fix path. I would not merge any of the proposed pairs as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44521 is the strongest standalone issue here: it is active, has multiple inbound references, and describes a concrete, reproducible multimodal apply_chat_template bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "Both involve model loading regressions, but one is a specific missing config field and the other is a broad load failure on a different model; not the same underlying bug." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "These are unrelated ImportErrors for different missing symbols in different code paths." + }, + { + "left": "issue:43844", + "right": "issue:44936", + "accept": false, + "reason": "One is a gradient explosion under ZeRO-3, the other is an evaluate-after-train failure; different symptoms and fixes." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Both are loading issues, but they point to different failure modes and code paths: missing pad_token_idx vs meta-device loading breakage." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Custom model processor/tokenizer regression and batched video-shape bug are different processor problems." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": false, + "reason": "Both mention missing pad_token_id, but they are model-specific config bugs for different config classes and likely separate fixes." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "An import/export failure is not the same as a detokenization regression." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor API regression and offloading-with-quantization behavior are unrelated." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Both concern labels, but one is metrics gathering in Trainer and the other is BeitImageProcessorFast label reduction." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 probability issues and RT-DETR memory release problems are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "Qwen3Moe FP8 configuration failure is unrelated to a Qwen2.5-VL import error." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": false, + "reason": "Both mention SigLIP2, but one is zero probabilities and the other is an implementation discrepancy versus JAX; not clearly the same bug." + }, + { + "left": "issue:28218", + "right": "issue:35141", + "accept": false, + "reason": "Tokenizer spacing after added tokens and embedding reinitialization after resize are different tokenizer/model initialization bugs." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Both involve apply_chat_template on multimodal inputs, but one is a batched-padding crash and the other is incorrect assistant masks; similar area, not the same concrete bug." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Different tokenizer/model failures on different architectures; no shared code-path evidence." + }, + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "dtype drift in evaluation and MoE training memory inefficiency are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "Tokenizer special-token registration and a stale CI device-override test are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs around pipelines, but they address different API/documentation mismatches." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "A video processor shape bug and a large-model OOM loading issue are unrelated." + }, + { + "left": "issue:43450", + "right": "issue:43493", + "accept": false, + "reason": "Video batching shape and SigLIP2 implementation discrepancy are different problems." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Feature-extractor arg validation and Whisper empty-transcription handling are distinct code paths." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA/vocab collapse and category-limit runtime error are unrelated." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "DAC latent conversion mismatch and qwen3_omni_moe feature-length helper behavior are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sample-rate defaulting and batched video output shape are different processor bugs." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "DAC forward mismatch and a stale CI device override test are unrelated." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Wrong processor output size and AMP/CUDA loss crash are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Embedding dtype regression and pipeline error-message wording are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "CPU offload device mismatch and a qwen3_omni_moe helper-function issue are unrelated." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "A model-architecture discussion and a trust_remote_code breakage from removed helper API are not the same bug." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "Sparse-layer creation and torch.compile/SDPA failure are different model/config regressions." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 24, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:29127", + "issue:29942", + "issue:32090", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:40444", + "issue:43335", + "issue:43425", + "issue:43504", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43653", + "issue:43720", + "issue:43723", + "issue:43819", + "issue:43931", + "issue:44112", + "issue:44206", + "issue:44292", + "issue:44351", + "issue:44448", + "issue:44462", + "issue:44493", + "issue:44554", + "issue:44561", + "issue:44743", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45083", + "issue:45200", + "issue:45229", + "issue:45290", + "issue:45310", + "issue:45381", + "issue:45478", + "issue:45520", + "issue:45588", + "issue:45593", + "issue:45823", + "issue:45859" + ], + "soft_pairs": [ + "issue:45290|issue:45381", + "issue:37428|issue:43425", + "issue:44448|issue:44743", + "issue:44857|issue:45593", + "issue:45200|issue:45588", + "issue:29942|issue:32090", + "issue:43611|issue:43931", + "issue:37428|issue:44351", + "issue:39692|issue:43526", + "issue:44206|issue:45823", + "issue:43335|issue:43653", + "issue:43819|issue:44493", + "issue:43550|issue:44112", + "issue:43335|issue:43504", + "issue:29127|issue:32090", + "issue:44855|issue:45083", + "issue:45310|issue:45478", + "issue:43335|issue:45083", + "issue:45588|issue:45859", + "issue:39692|issue:44493", + "issue:44462|issue:44561", + "issue:44561|issue:44779", + "issue:44292|issue:45520", + "issue:38175|issue:44493", + "issue:43335|issue:45072", + "issue:44448|issue:44779", + "issue:43335|issue:45593", + "issue:43720|issue:44112", + "issue:44977|issue:45229", + "issue:40444|issue:43526", + "issue:34689|issue:35141", + "issue:43723|issue:44554", + "issue:44448|issue:44977", + "issue:43606|issue:44857", + "issue:44857|issue:45072" + ], + "prepared_review_unit_hash": "e34c6b721a9979253b9557110240d67301454c973c1d4ac721f89d28b3581c3c", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38238, + "estimated_input_tokens": 9560, + "estimated_eval_tokens": 19376 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: nearly all soft pairs are false duplicates across different models/subsystems. The only clear duplicate is the Qwen3.5 MoE from_pretrained regression pair (issues 45310 and 45478).", + "confidence": 0.94, + "canonical_issue_reason": "Issue 45310 is the earliest and most complete report of the only true duplicate theme in this set, and 45478 repeats the same Qwen3.5 MoE from_pretrained failure on transformers>=5.4.0.", + "canonical_pr_reason": null, + "best_issue_reason": "45310 is the best representative issue because it has the clearest description of the shared regression and anchors the one real duplicate pair in the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Different bugs: chat template tool-call handling vs Qwen2.5-VL video vision_position_ids." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Flash-attention import error vs general Torch 2.10 incompatibility; not the same issue." + }, + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Pegasus v4/v5 output drift and Qwen recurrent-state reset are unrelated model behaviors." + }, + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "LwDetr AMP/CUDA crash and D-FINE auxiliary-loss behavior are different code paths." + }, + { + "left": "issue:45200", + "right": "issue:45588", + "accept": false, + "reason": "Gemma4 text-only mm_token_type_ids problem vs flash_attention sink-less model crash." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention 2 test failures vs Trainer _gpu_broadcast_one NoneType error are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:43931", + "accept": false, + "reason": "base_model_prefix loading regression vs Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Different missing symbols: _flash_supports_window_size vs HybridCache." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "SigLIP2 docs/quantization issues vs BeitImageProcessorFast reduce_labels bug." + }, + { + "left": "issue:44206", + "right": "issue:45823", + "accept": false, + "reason": "Feature-extractor center arg regression vs Gemma4 PLE device mismatch." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug vs BigBirdTokenizer special-token registration bug." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "DAC.from_latents STE mismatch vs unexpected position-id key issues; different failures." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "Bamba torch.compile/SDPA issue vs GraniteSpeech stale device override CI failure." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "SwitchTransformers config bug and Beit pretrained legacy-field loading bug are unrelated." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "LayoutLMv3 error-message clarity vs Trainer broadcast TypeError." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "DebertaV2 Python 3.13 IndentationError vs qwen3_omni_moe helper-length behavior." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Same reported regression: transformers>=5.4.0 breaks Qwen3.5 MoE from_pretrained loading." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "SwitchTransformers sparse-layer creation and qwen3_omni_moe feature-length logic are unrelated." + }, + { + "left": "issue:45588", + "right": "issue:45859", + "accept": false, + "reason": "Both involve Qwen-related attention/code, but one is sink-less flash_attention aux handling and the other is CPU-offload rotary embedding incompatibility." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 doc example errors vs generic position-id key warnings are not the same bug." + }, + { + "left": "issue:44462", + "right": "issue:44561", + "accept": false, + "reason": "AutoTokenizer tokenizer.json precedence issue vs removal of is_torch_fx_available breaking trust_remote_code models." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer/loading regressions affecting different mechanisms and models." + }, + { + "left": "issue:44292", + "right": "issue:45520", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime failure vs flash_attn import KeyError on Python 3.13." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "SIGLIP2 zero-probabilities issue vs unexpected key with position id; not the same defect." + }, + { + "left": "issue:43335", + "right": "issue:45072", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug vs dtype mismatch in bfloat16 inference." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Pegasus output regression and Deepseek tokenizer regression are unrelated." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "SwitchTransformers config bug vs D-FINE auxiliary-loss omission are different bugs." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "BitNet packed-weight loading issue vs GraniteSpeech device override CI failure." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug vs Gemma4 multi-GPU OOM are unrelated." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "Qwen2.5-VL iterable dataset/multi-image fine-tuning failure vs BeitImageProcessorFast label reduction bug." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Llama 3.2 Vision model loading regression vs post_init embedding reinitialization bug; different code paths." + }, + { + "left": "issue:43723", + "right": "issue:44554", + "accept": false, + "reason": "AutoTokenizer.from_pretrained v5 loading issue vs MPS attention correctness bug." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Pegasus output drift and Qwen3.5 flash-attention generation failure are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU offload device mismatch vs float16 AMP image-loss crash; different subsystems." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "Image-loss AMP crash and dtype mismatch in SwitchTransformers/TimmWrapperModel are not duplicates." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is well grounded in the packet. The only accepted duplicate pair is supported by identical titles and matching regression wording for issues 45310 and 45478, while the other soft-edge rejections stay conservative and model/subsystem-specific." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 25, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:10000", + "issue:15354", + "issue:28218", + "issue:29127", + "issue:29942", + "issue:32090", + "issue:33357", + "issue:35141", + "issue:37428", + "issue:41720", + "issue:43065", + "issue:43066", + "issue:43278", + "issue:43329", + "issue:43381", + "issue:43412", + "issue:43550", + "issue:43653", + "issue:43701", + "issue:43873", + "issue:43927", + "issue:43976", + "issue:43994", + "issue:44448", + "issue:44462", + "issue:44466", + "issue:44493", + "issue:44568", + "issue:44743", + "issue:44857", + "issue:44977", + "issue:44987", + "issue:45083", + "issue:45198", + "issue:45229", + "issue:45245", + "issue:45372", + "issue:45406", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45563", + "issue:45588", + "issue:45698", + "issue:45712", + "issue:45803", + "issue:45823" + ], + "soft_pairs": [ + "issue:44568|issue:45491", + "issue:28218|issue:32090", + "issue:44977|issue:45563", + "issue:43927|issue:45698", + "issue:43701|issue:45083", + "issue:45198|issue:45803", + "issue:43550|issue:45083", + "issue:29942|issue:35141", + "issue:45372|issue:45588", + "issue:43873|issue:44493", + "issue:45459|issue:45712", + "issue:32090|issue:33357", + "issue:37428|issue:44987", + "issue:43873|issue:45229", + "issue:43412|issue:45588", + "issue:45245|issue:45563", + "issue:43278|issue:43329", + "issue:41720|issue:45229", + "issue:10000|issue:15354", + "issue:43976|issue:45440", + "issue:45372|issue:45406", + "issue:28218|issue:29127", + "issue:43065|issue:43994", + "issue:45372|issue:45459", + "issue:44857|issue:45588", + "issue:43065|issue:45412", + "issue:44448|issue:44466", + "issue:43278|issue:43381", + "issue:43653|issue:44743", + "issue:45588|issue:45823", + "issue:43066|issue:44462" + ], + "prepared_review_unit_hash": "4031cbe16c5e8621a4aa40589ea46299347d103b2d678760a247dd5a43d00042", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37193, + "estimated_input_tokens": 9299, + "estimated_eval_tokens": 18854 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are a heterogeneous grab-bag of unrelated Transformers bugs. The soft pairs mostly share superficial subsystem keywords (tokenizers, Gemma, flash attention, save/load) but do not describe the same underlying defect or mergeable fix.", + "confidence": 0.92, + "canonical_issue_reason": "No strong canonical duplicate anchor exists because the cluster is not coherent. If one issue must be chosen as a representative, issue:45372 is the most discussion-rich and currently active, but it is not a true umbrella for the rest.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue:45372 is the best single representative only in the sense of being recent, user-facing, and relatively high-signal; however, it should not be treated as a duplicate target for the other issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Both concern tokenizer/model output behavior, but one is about special tokens not being added and the other is a Gemma3 NaN embedding bug from mixed-length batching. Different code paths and fixes." + }, + { + "left": "issue:28218", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer spacing after added tokens is unrelated to Trainer distributed broadcast NoneType errors." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Qwen3.5 flash-attention generation failure and a stale generate() warning are different bugs with different symptoms and fixes." + }, + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "Both involve save/load behavior, but one is config token IDs being lost and the other is loading the wrong custom module. Not the same defect." + }, + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Resume-from-checkpoint key mismatch is unrelated to a Qwen3 Omni MoE helper-function bug." + }, + { + "left": "issue:45198", + "right": "issue:45803", + "accept": false, + "reason": "Wav2Vec2 save_pretrained/tokenization failure and a bare-except swallowing errors in Fuyu conversion are distinct issues." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "A torch.compile/SDPA failure in Bamba is unrelated to a feature-extraction length helper bug in qwen3_omni_moe." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Failing Flash Attention 2 tests and embedding reinitialization after resizing token embeddings are different regressions." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "Gemma4 processor import failure due to ReasoningEffort is a separate problem from flash_attention.py crashing on s_aux=None." + }, + { + "left": "issue:43873", + "right": "issue:44493", + "accept": false, + "reason": "Quantization/offloading behavior and unexpected position-id keys are unrelated runtime/configuration bugs." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "A broad tokenizer exception-handling bug and leftover dummy classes leaking into imports are not the same underlying change." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "Trainer GPU broadcast NoneType error and MacOS bus error with a CLIP model have no shared concrete root cause." + }, + { + "left": "issue:37428", + "right": "issue:44987", + "accept": false, + "reason": "Both mention loading failures, but one is a missing flash-attention utility import and the other is a model-specific transformers>=5.1.0 load failure; too little evidence they are the same bug." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Both are Gemma4-related, but one is processor loading import failure and the other is serve crashing because Gemma4Processor lacks _tokenizer. Different failure points." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Executorch export failures for Gemma3n and flash-attention crashes on sink-less models are unrelated." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Category-count runtime error and stale num_return_sequences warning are not the same defect." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Embedding dtype drift in eval and undefined video-token helper names are distinct issues." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3 auto device-map cudaErrorAssert and Gemma4 multi-GPU OOM are different runtime problems." + }, + { + "left": "issue:10000", + "right": "issue:15354", + "accept": false, + "reason": "German DistilBertModel issue and torch.jit GeneratorExp export failure are unrelated model/export bugs." + }, + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Python-version compatibility and DeepSeekV3MoE divergence are unrelated." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Repeated pair: same rationale as above; import-time processor loading bug vs serve-time missing attribute bug are not duplicates." + }, + { + "left": "issue:28218", + "right": "issue:29127", + "accept": false, + "reason": "Tokenizer added-space behavior and LayoutLMv3 error-message clarity are unrelated." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "Dummy nn.Conv2d in Sam3PixelDecoder and SigLIP2 nonsensical outputs are different model bugs." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "ReasoningEffort import failure in Gemma4 processor loading and protobuf error handling in tokenizer decode are unrelated." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "AMP crash in LwDetrImageLoss and flash_attention s_aux=None are different code paths." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Dummy layer definition in Sam3PixelDecoder and RT-DETR memory release issues are not the same underlying bug." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "Different model output in v4 vs v5 and inconsistent lm_head serialization are separate versioning/serialization problems." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Eval/training state problems, but one is dtype regression and the other is gradient checkpointing restriction in eval; not the same fix." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and Qwen3.5 recurrent-state reset are unrelated." + }, + { + "left": "issue:45588", + "right": "issue:45823", + "accept": false, + "reason": "Both are Gemma4-adjacent, but flash_attention.py s_aux handling and PLE device mismatch under device_map='auto' are different concrete bugs." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Tokenizer decoder-type selection and AutoTokenizer ignoring tokenizer.json are distinct tokenizer-loading issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 26, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:25251", + "issue:29127", + "issue:29942", + "issue:30064", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34567", + "issue:35141", + "issue:36010", + "issue:39692", + "issue:42915", + "issue:42994", + "issue:43257", + "issue:43278", + "issue:43334", + "issue:43381", + "issue:43421", + "issue:43526", + "issue:43531", + "issue:43653", + "issue:43701", + "issue:43756", + "issue:43761", + "issue:43824", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44279", + "issue:44361", + "issue:44442", + "issue:44451", + "issue:44479", + "issue:44610", + "issue:44871", + "issue:44977", + "issue:45103", + "issue:45198", + "issue:45245", + "issue:45372", + "issue:45412", + "issue:45440", + "issue:45491", + "issue:45520", + "issue:45593", + "issue:45834", + "issue:45859" + ], + "soft_pairs": [ + "issue:25251|issue:32090", + "issue:43994|issue:45412", + "issue:44186|issue:45198", + "issue:44206|issue:44479", + "issue:44451|issue:45520", + "issue:44361|issue:45198", + "issue:42915|issue:42994", + "issue:44871|issue:44977", + "issue:44265|issue:45198", + "issue:43756|issue:45440", + "issue:43421|issue:45491", + "issue:29942|issue:30064", + "issue:43257|issue:43278", + "issue:33357|issue:39692", + "issue:29127|issue:34567", + "issue:32090|issue:33290", + "issue:44610|issue:45412", + "issue:45103|issue:45372", + "issue:45440|issue:45593", + "issue:43824|issue:44279", + "issue:43653|issue:44442", + "issue:43381|issue:43701", + "issue:43526|issue:45834", + "issue:43761|issue:45859", + "issue:29942|issue:36010", + "issue:43526|issue:44186", + "issue:29127|issue:33357", + "issue:43334|issue:43531", + "issue:29127|issue:35141", + "issue:44977|issue:45245" + ], + "prepared_review_unit_hash": "f6e29ee813731e0344287f0f77e4db6c31bfddf18ad95e72997196a0b1e0eedf", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36694, + "estimated_input_tokens": 9174, + "estimated_eval_tokens": 18604 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: the soft pairs mostly share only broad subsystem or model-family vocabulary, not the same concrete bug. I would not merge any of the suggested soft edges. If a single representative is needed, the broadest issue is the Qwen multi-model video regression report.", + "confidence": 0.88, + "canonical_issue_reason": "No pair here looks like a true duplicate set; if one issue must represent the cluster, issue 44479 is the broadest and clearest regression report because it covers multiple Qwen variants with one concrete video-input failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44479 is the best overall representative because it is specific enough to summarize a concrete regression while still spanning several affected models, making it the most cluster-like of the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "Unrelated failures: pipeline output shape vs Trainer broadcast TypeError." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "Different components and symptoms: SigLIP2 doc/example mismatch vs RT-DETR memory leak." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "Tokenizer padding/NER crash vs Wav2Vec2 save/tokenization failure; no shared code-path." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Feature extractor center-arg crash is unrelated to Qwen video-input regression." + }, + { + "left": "issue:44451", + "right": "issue:45520", + "accept": false, + "reason": "Model loading failure vs flash_attn import handling on Python 3.13; different bug classes." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "Different tokenizers and failure modes; no evidence of one underlying bug." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "Both involve quantization, but one is model/runtime failure and the other is saving quantized models." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "Eos-token config mismatch vs flash-attention generation regression; unrelated." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export exportability issue is not the same as a tokenizer/save_pretrained problem." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Both are model-implementation mismatches, but for different models and different defects." + }, + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Runtime special-token/post-processor update bug vs Gemma3 NaN embedding issue; unrelated." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "FlashAttention test failures are unrelated to void segmentation-map processing." + }, + { + "left": "issue:43257", + "right": "issue:43278", + "accept": false, + "reason": "Accelerate/deepspeed weight conversion bug vs training/eval dtype change; different root causes." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "Both mention CLIP-like vision models, but one is a bus error and the other is a documentation/example error bundle." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and TrainerState token-count updates are unrelated." + }, + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Trainer broadcast NoneType error and DeepSpeed OOM with Adafactor are different issues." + }, + { + "left": "issue:44610", + "right": "issue:45412", + "accept": false, + "reason": "Processor image-size mismatch is unrelated to RT-DETR memory reclamation." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "auto_docstring annotations crash and mistral_common import error are unrelated." + }, + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "Different model families and different behaviors: implementation divergence vs loss computation." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Specific Qwen2.5-VL import failure is not the same as a generic dependency issue." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "Tokenizer special-token registration bug vs AutoTokenizer loading failure for a different tokenizer." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Gradient checkpointing in eval mode and checkpoint key mismatch are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:45834", + "accept": false, + "reason": "BEiT label reduction bug is unrelated to Kosmos2.5 OCR index errors." + }, + { + "left": "issue:43761", + "right": "issue:45859", + "accept": false, + "reason": "CLIPVision hidden_states regression is unrelated to CPU-offload incompatibility in Qwen3_5Moe rotary embeddings." + }, + { + "left": "issue:29942", + "right": "issue:36010", + "accept": false, + "reason": "FlashAttention test failures are not the same as a GenerationMixin import regression." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "ImageProcessor label handling and LayoutLMv2 tokenizer crashes are different code paths." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "LayoutLMv3 error messaging and CLIP Mac bus error share no concrete failure mode." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Both are Qwen-family issues, but one is a missing config attribute and the other is a sliding-window behavior bug." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "LayoutLMv3 box-validation messaging vs token-embedding reinit in post_init are unrelated." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Flash-attention generation regression and category-cardinality runtime error are not the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 27, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:33290", + "issue:34567", + "issue:34689", + "issue:35141", + "issue:39692", + "issue:42907", + "issue:42915", + "issue:43232", + "issue:43262", + "issue:43316", + "issue:43381", + "issue:43388", + "issue:43441", + "issue:43525", + "issue:43526", + "issue:43575", + "issue:43611", + "issue:43646", + "issue:43701", + "issue:43723", + "issue:43761", + "issue:43873", + "issue:43931", + "issue:44062", + "issue:44077", + "issue:44079", + "issue:44188", + "issue:44206", + "issue:44387", + "issue:44442", + "issue:44464", + "issue:44479", + "issue:44743", + "issue:44811", + "issue:44857", + "issue:44898", + "issue:45003", + "issue:45072", + "issue:45216", + "issue:45290", + "issue:45440", + "issue:45479", + "issue:45491", + "issue:45588", + "issue:45698", + "issue:45803", + "issue:45907" + ], + "soft_pairs": [ + "issue:44464|issue:45491", + "issue:42915|issue:43931", + "issue:44442|issue:45072", + "issue:43262|issue:44206", + "issue:43526|issue:44857", + "issue:43232|issue:43388", + "issue:43441|issue:45588", + "issue:44743|issue:44811", + "issue:44062|issue:45907", + "issue:43575|issue:43701", + "issue:43646|issue:44077", + "issue:43575|issue:44387", + "issue:43761|issue:44079", + "issue:45003|issue:45440", + "issue:43723|issue:45479", + "issue:44442|issue:45803", + "issue:43761|issue:44188", + "issue:44479|issue:45290", + "issue:43316|issue:43525", + "issue:42907|issue:45698", + "issue:44442|issue:44898", + "issue:33290|issue:34567", + "issue:34689|issue:43611", + "issue:39692|issue:45491", + "issue:33290|issue:35141", + "issue:33290|issue:34689", + "issue:43381|issue:43873", + "issue:43931|issue:45216" + ], + "prepared_review_unit_hash": "efe8f8078e5e823f5b38202962c764b6464ab3046a368b8d44223996f0347cdf", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35639, + "estimated_input_tokens": 8910, + "estimated_eval_tokens": 18076 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The soft pairs are mostly superficial similarity matches; they span unrelated regressions across loading, generation, quantization, tokenizers, and model configs, so none look like true duplicates. No PRs are present.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43873 is the strongest anchor here: it is open, has the most discussion and inbound references, and describes a broad quantization/offloading runtime regression that seems more central than the other isolated reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best global representative because it is the most widely discussed and externally referenced issue in the set, making it the most useful canonical issue if one must be chosen.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44464", + "right": "issue:45491", + "accept": false, + "reason": "Both are generation-related, but one is about compiled chunked generation inconsistencies while the other is a Gemma3 NaN embedding bug from sliding-window attention; different failure modes." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "Both involve Qwen loading, but one is a FineGrainedFP8Config failure and the other is a model weight-shape mismatch; not the same underlying bug." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer loading failure and bfloat16 dtype mismatch are unrelated code paths, despite both affecting model startup/inference." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Audio processor default sampling-rate handling is different from an unsupported `center` argument crash in a feature extractor." + }, + { + "left": "issue:43526", + "right": "issue:44857", + "accept": false, + "reason": "One is a label-reduction logic bug, the other is an AMP/CUDA loss crash; they do not share the same root cause." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "`_update_model_kwargs_for_generation` after GPU sync and `gather_for_metrics` truncating tuple labels are separate bugs in different subsystems." + }, + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Both mention flash attention, but one is a model/regression issue for Ministral-3 and the other is a sink-less `s_aux=None` crash in the integration layer." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Recurrent state reset with cache and Whisper batch decode skipping special-token handling are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:45907", + "accept": false, + "reason": "`AddedToken(..., special=...)` argument duplication and `list_repo_files` not forwarding `revision` are unrelated API bugs." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "Tensor-parallel OOM on load and resume-from-checkpoint key mismatch are different problems." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "Custom model initialization breakage is not the same as the `patchtsmixer` `post_init` policy issue." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both mention OOM, but one is TP model loading OOM and the other is increased reserved memory under int4 quantization; different root causes." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "Missing `hidden_states` in CLIPVisionModel and incorrect `ModelOutput` key assignment are related only broadly; not the same concrete bug." + }, + { + "left": "issue:45003", + "right": "issue:45440", + "accept": false, + "reason": "Unsafe `sys.modules` access in `modeling_utils` and DeepseekV3MoE divergence are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer-loading regression and zero-loss sequence-classification behavior are separate issues." + }, + { + "left": "issue:44442", + "right": "issue:45803", + "accept": false, + "reason": "AutoTokenizer failing to load a tokenizer and a bare `except:` swallowing errors are unrelated bugs." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "CLIP hidden-state regression and attention-kernel divergence under `torch.compile` do not share the same code path." + }, + { + "left": "issue:44479", + "right": "issue:45290", + "accept": false, + "reason": "Video-input regression for Qwen VL models and chat-template crashes with tool calls are different feature paths." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "API discrepancy in `Gemma3TextConfig` and missing `pad_token_id` on `Llama4Config` are distinct config issues." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "Both are save/load related, but one is about saving dequantized models and the other about loading the wrong custom module after save/load; not the same bug." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer load failure and Perceiver non-default-resolution image classification failure are unrelated." + }, + { + "left": "issue:33290", + "right": "issue:34567", + "accept": false, + "reason": "OOM with Adafactor/DeepSpeed and `num_input_tokens_seen` not updating are unrelated trainer issues." + }, + { + "left": "issue:34689", + "right": "issue:43611", + "accept": false, + "reason": "Both are model-loading regressions, but one is specific to Llama 3.2 Vision while the other is a `base_model_prefix` loading issue; different causes." + }, + { + "left": "issue:39692", + "right": "issue:45491", + "accept": false, + "reason": "SigLIP2 documentation example errors and Gemma3 NaN embeddings are not the same underlying bug." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "OOM in training versus output embedding reinitialization during `post_init` are unrelated." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "Training OOM with DeepSpeed/Adafactor and model loading failure for Llama 3.2 Vision are different failures." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Gradient checkpointing being disallowed in eval mode is unrelated to quantization/offloading behavior." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Model weight-shape mismatch on load and a corrupted checkpoint saved by `save_pretrained` are distinct persistence bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 28, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:25251", + "issue:28282", + "issue:33290", + "issue:34689", + "issue:36010", + "issue:36331", + "issue:41628", + "issue:42175", + "issue:43065", + "issue:43295", + "issue:43316", + "issue:43381", + "issue:43388", + "issue:43404", + "issue:43441", + "issue:43493", + "issue:43540", + "issue:43550", + "issue:43643", + "issue:43673", + "issue:43688", + "issue:43720", + "issue:43749", + "issue:43825", + "issue:43827", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:44077", + "issue:44162", + "issue:44190", + "issue:44442", + "issue:44479", + "issue:44496", + "issue:44514", + "issue:44617", + "issue:44743", + "issue:44792", + "issue:44855", + "issue:44871", + "issue:44877", + "issue:44913", + "issue:45137", + "issue:45372", + "issue:45759" + ], + "soft_pairs": [ + "issue:43493|issue:43827", + "issue:43550|issue:44743", + "issue:44077|issue:44479", + "issue:43720|issue:44442", + "issue:43316|issue:45759", + "issue:41628|issue:45372", + "issue:43643|issue:44913", + "issue:43749|issue:43931", + "issue:33290|issue:36331", + "issue:34689|issue:36010", + "issue:43873|issue:44871", + "issue:44190|issue:44496", + "issue:43381|issue:43856", + "issue:44855|issue:45137", + "issue:43540|issue:44514", + "issue:43316|issue:44877", + "issue:43673|issue:44743", + "issue:42175|issue:43316", + "issue:43404|issue:43441", + "issue:44162|issue:44792", + "issue:43065|issue:44617", + "issue:25251|issue:28282", + "issue:43295|issue:43825", + "issue:43388|issue:43688", + "issue:16998|issue:22355" + ], + "prepared_review_unit_hash": "e202d11bf4c5a117f5ba32f6f4075bb455e5e3d27c8027a7782c62390b1269e5", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 25, + "serialized_chars": 34607, + "estimated_input_tokens": 8652, + "estimated_eval_tokens": 17560 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a set of largely unrelated Transformers issues. The soft-similarity edges are mostly superficial (shared words like import/config/video/model loading) rather than the same underlying bug, so all candidate duplicate links should be rejected.", + "confidence": 0.94, + "canonical_issue_reason": "No single canonical issue fits this set: the items cover distinct failures across model loading, processors, training, config, and docs, with no shared concrete code path.", + "canonical_pr_reason": null, + "best_issue_reason": "If one issue must be picked as the most representative, issue 43873 is the best fit because it is active, user-facing, and broad in impact; but it is not a true duplicate anchor for the rest of the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "One is a SigLIP2 implementation mismatch; the other is stale docs still referencing pipeline(). Different problem types and code paths." + }, + { + "left": "issue:43550", + "right": "issue:44743", + "accept": false, + "reason": "Torch.compile/SDPA failure in Bamba vs recurrent state reset in modular_qwen3_5.py. Different models and bugs." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "patchtsmixer post_init policy change is unrelated to Qwen video-input regression." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "BitNet packed-weight loading issue is unrelated to AutoTokenizer failing for FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy is a config-class consistency issue; Qwen 3.5/3.6 bug is about unwrapping text_config in AutoModelForCausalLM.from_config." + }, + { + "left": "issue:41628", + "right": "issue:45372", + "accept": false, + "reason": "AutoImageProcessor import failure and ReasoningEffort import failure are separate import regressions affecting different modules." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "trust_remote_code returning missing fields is not the same as GPTNeoXConfig rotary_pct not persisting on reload." + }, + { + "left": "issue:43749", + "right": "issue:43931", + "accept": false, + "reason": "FSDP CPU RAM efficient loading broken vs Qwen3-VL weight-shape mismatch; different loading paths and symptoms." + }, + { + "left": "issue:33290", + "right": "issue:36331", + "accept": false, + "reason": "AdaFactor+DeepSpeed OOM is a memory issue; CustomTrainer.compute_loss signature break is an API compatibility issue." + }, + { + "left": "issue:34689", + "right": "issue:36010", + "accept": false, + "reason": "Llama 3.2 Vision model loading regression and GenerationMixin import error are distinct regressions." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization offloading behavior and Gemma-3 eos_token_id mismatch are unrelated." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Local dataset loading in a script is unrelated to an unrecognized model/config.json model_type error." + }, + { + "left": "issue:43381", + "right": "issue:43856", + "accept": false, + "reason": "Gradient checkpointing eval-mode limitation is not the same as MoE memory inefficiency during training." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 IndentationError in DebertaV2 import is unrelated to DeepSpeed ZeRO3 deque pop error." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Qwen3OmniMoe video processing ValueError and Qwen2_5_VL batched chat-template crash are different processors and failure modes." + }, + { + "left": "issue:43316", + "right": "issue:44877", + "accept": false, + "reason": "Gemma3TextConfig API mismatch and strict config blocking granite_speech loading are separate config-handling issues." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "GenerationMixin cache missing in v5.0.0 during chunked_prefill is unrelated to Qwen recurrent state reset in modular_qwen3_5.py." + }, + { + "left": "issue:42175", + "right": "issue:43316", + "accept": false, + "reason": "TensorFlow missing from a torch-only install and Gemma3TextConfig inconsistency are unrelated packaging/config issues." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Mistral3 lm_head weight tying bug and Ministral-3 FlashAttention failure are different model-specific regressions." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 broken workflow issue and janus image-generation test failure do not share the same concrete bug." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder is a design/implementation question, while Sam3Video CUDA OOM is a runtime memory problem." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline top_k output nesting change and AutoModel missing PyTorch import error are unrelated API behaviors." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Processor.tokenizer regression and pipeline translation error-message regression are different v5 behavior breaks." + }, + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "gather_for_metrics dropping tuple labels and auxiliary loss normalization are separate training bugs." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "DeBERTa model_max_length question and missing transformers.onnx module are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 29, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:32090", + "issue:35141", + "issue:41720", + "issue:41950", + "issue:42617", + "issue:43065", + "issue:43295", + "issue:43452", + "issue:43475", + "issue:43504", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43701", + "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:44792", + "issue:44871", + "issue:44913", + "issue:44964", + "issue:44977", + "issue:45127", + "issue:45200", + "issue:45245", + "issue:45290", + "issue:45381", + "issue:45406", + "issue:45563", + "issue:45584", + "issue:45698" + ], + "soft_pairs": [ + "issue:41950|issue:43475", + "issue:44077|issue:44534", + "issue:44871|issue:45584", + "issue:43550|issue:43701", + "issue:45127|issue:45200", + "issue:43761|issue:44479", + "issue:43606|issue:43701", + "issue:43295|issue:44871", + "issue:42617|issue:43575", + "issue:43575|issue:43742", + "issue:43452|issue:44496", + "issue:44734|issue:45406", + "issue:43646|issue:44913", + "issue:43504|issue:45698", + "issue:43575|issue:43873", + "issue:44617|issue:45584", + "issue:43761|issue:44242", + "issue:43931|issue:44060", + "issue:32090|issue:35141", + "issue:45563|issue:45584", + "issue:44464|issue:44871", + "issue:44792|issue:44977", + "issue:43065|issue:44610", + "issue:44479|issue:44734", + "issue:44521|issue:45381", + "issue:43827|issue:43994", + "issue:43906|issue:45290", + "issue:44464|issue:45563", + "issue:44534|issue:44913", + "issue:44190|issue:44964", + "issue:41720|issue:45245" + ], + "prepared_review_unit_hash": "0e5afe1b5555a4f7d965b68abbb36abe5f9d4c98af485689c95c03a7f1a78ddb", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36957, + "estimated_input_tokens": 9240, + "estimated_eval_tokens": 18736 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: it mixes unrelated bugs in loading, multimodal processing, generation, offloading, and config handling. None of the soft-edge pairs appear to be true duplicates of the same underlying bug/change.", + "confidence": 0.97, + "canonical_issue_reason": "No single issue is a clear duplicate hub for this mixed cluster; if one must be chosen, issue 43761 is the broadest, most clearly scoped regression report.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43761 is the cleanest representative bug report in the set: concrete regression framing, specific failure mode, and broadly relevant model behavior.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Both touch multimodal/video paths, but one is a processor-selection bug and the other is a missing SAM3 video field; different code paths and symptoms." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "One is about optional post_init in patchtsmixer, the other about non-persistent buffers being corrupted in v5; unrelated bugs." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Both involve token/eos handling, but one is a Gemma config mismatch and the other is Whisper empty-transcription generation after align_special_tokens; not the same defect." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "torch.compile+SDPA on Bamba and resume_from_checkpoint key mismatch are unrelated failures in different subsystems." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "Both relate to model configuration, but one is LoRA/vocabulary merging collapse and the other is Gemma 4 multimodal token type defaults; different underlying issues." + }, + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "Both are v5 regressions, but one is CLIPVisionModel hidden_states being dropped and the other is Qwen VL video input behavior; different model families and code paths." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU offload device mismatch for bark-small is unrelated to checkpoint resume key mismatches." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Processor.tokenizer API breakage is a different bug from Gemma eos_token_id config inconsistency." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "A parallel script failure and a tp-induced OOM when loading a large Qwen model are not the same bug." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "Qwen2-57B tensor-parallel OOM and MobileLLM key error on load are distinct loading failures." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "gguf_file breaking auto classes is different from a model config missing model_type." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "Both affect transformers serve, but one crashes in KV-cache continuation indexing and the other lacks _tokenizer on Gemma4Processor." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "Custom model initialization breakage in v5 and GPTNeoX rotary_pct reload reset are unrelated configuration bugs." + }, + { + "left": "issue:43504", + "right": "issue:45698", + "accept": false, + "reason": "Legacy field loading for BEiT semantic segmentation is not the same as loading the wrong custom module after save_pretrained." + }, + { + "left": "issue:43575", + "right": "issue:43873", + "accept": false, + "reason": "Both concern memory/loading, but tp OOM on a specific model and offloading-with-quantization behavior are different problems." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "Sam3Video CUDA OOM and Whisper empty transcription generation are unrelated runtime failures." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "A CLIPVisionModel forward regression is not the same as missing load-balancing loss when output_router_logits=False." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch and Qwen3-Next tied-weights warning are separate model/config issues." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer GPU broadcast TypeError and embedding reinitialization after resizing are unrelated bugs." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "A stale warning in paged generate and Whisper failure on empty transcription are different generation-path issues." + }, + { + "left": "issue:44464", + "right": "issue:44871", + "accept": false, + "reason": "Compiled chunked generation inconsistency and Gemma eos_token_id mismatch are not the same bug." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Janus image-generation test failure and Qwen3.5 flash-attention generation bug are different model-specific issues." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "A dummy Conv2d in Sam3PixelDecoder and OmDet-Turbo's 640px/224px mismatch are unrelated architecture/preprocessing problems." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Both are server or multimodal regressions, but one is video-input behavior in Qwen VL models and the other is serve KV-cache continuation indexing." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "All-zero assistant masks in apply_chat_template and incorrect vision_position_ids in Qwen2.5-VL are different multimodal template bugs." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs still referencing pipeline() and SigLIP2 producing nonsense via AutoModel/pipeline are not the same underlying issue." + }, + { + "left": "issue:43906", + "right": "issue:45290", + "accept": false, + "reason": "An isolated reproduction issue and apply_chat_template crashing on tool-call messages are unrelated." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Compiled-forward inconsistency and a stale paged-generate warning do not share the same defect." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer junk in v5 and GPTNeoX rotary_pct reset on reload are different configuration/runtime problems." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Local dataset loading failure in a training script and a latest-transformers model load failure for Phi-4 multimodal are separate issues." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "Qwen3 auto device mapping cudaErrorAssert and category cardinality overflow are unrelated errors." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 30, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:36246", + "issue:38175", + "issue:38617", + "issue:39692", + "issue:41628", + "issue:41950", + "issue:43065", + "issue:43295", + "issue:43335", + "issue:43388", + "issue:43479", + "issue:43540", + "issue:43550", + "issue:43582", + "issue:43644", + "issue:43650", + "issue:43976", + "issue:44016", + "issue:44077", + "issue:44079", + "issue:44112", + "issue:44186", + "issue:44190", + "issue:44220", + "issue:44462", + "issue:44466", + "issue:44496", + "issue:44683", + "issue:44734", + "issue:44737", + "issue:44792", + "issue:44855", + "issue:44913", + "issue:44933", + "issue:44938", + "issue:45127", + "issue:45245", + "issue:45276", + "issue:45357", + "issue:45362", + "issue:45375", + "issue:45446", + "issue:45542", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45803", + "issue:45907" + ], + "soft_pairs": [ + "issue:44933|issue:45362", + "issue:44792|issue:45600", + "issue:44855|issue:45588", + "issue:44737|issue:45803", + "issue:43388|issue:43582", + "issue:45127|issue:45357", + "issue:44462|issue:45907", + "issue:43335|issue:44855", + "issue:44077|issue:44079", + "issue:44938|issue:45542", + "issue:43976|issue:44855", + "issue:43550|issue:44190", + "issue:43644|issue:44913", + "issue:43540|issue:44734", + "issue:44112|issue:45600", + "issue:44496|issue:45803", + "issue:43335|issue:44190", + "issue:43650|issue:44220", + "issue:45276|issue:45584", + "issue:39692|issue:44190", + "issue:45245|issue:45584", + "issue:43479|issue:45375", + "issue:43295|issue:45276", + "issue:41950|issue:44016", + "issue:36246|issue:41628", + "issue:36246|issue:38617", + "issue:44186|issue:44734", + "issue:38175|issue:43065", + "issue:44683|issue:45446", + "issue:44016|issue:44220", + "issue:44466|issue:44913" + ], + "prepared_review_unit_hash": "7c05a5fef0e6bc7d733d19b0018f5e8ccd2b1858f0a288e6104d7478909575d2", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37184, + "estimated_input_tokens": 9296, + "estimated_eval_tokens": 18848 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is noisy: most items are unrelated regressions, import/export issues, model-specific bugs, or CI/notebook problems. The soft-similar pairs mostly share broad vocabulary only, not the same concrete code path or fix.", + "confidence": 0.88, + "canonical_issue_reason": "issue:41628 is the broadest, most user-facing import/export regression in the set and best fits the loose theme of missing public symbols.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41628 is the clearest representative of the cluster\u2019s import-regression subset and is more general than the model-specific or highly niche reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "One is a missing import from image_utils; the other is a Qwen3.5 chat crash. Different failure modes and code paths." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "Janus image generation test failure vs. auto_mappings referencing removed Sam3LiteText configs. Unrelated subsystems." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Python 3.13 import parse error in DebertaV2 vs. flash_attention AttributeError on s_aux=None. No shared bug." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "XLNet device-placement bug vs. FuyuBatchFeature bare-except handling. Different models and fixes." + }, + { + "left": "issue:43388", + "right": "issue:43582", + "accept": false, + "reason": "Metric-gathering label truncation vs. Apple Silicon allocator TypeError. No overlap in code path." + }, + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "LoRA merge/tied-weights collapse vs. Qwen3.5 save_pretrained key serialization. Related only by broad model saving behavior." + }, + { + "left": "issue:44462", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer file selection vs. list_repo_files revision forwarding. Both repo-related, but not the same bug." + }, + { + "left": "issue:43335", + "right": "issue:44855", + "accept": false, + "reason": "SwitchTransformers config layer creation bug vs. DebertaV2 import parse failure. Unrelated." + }, + { + "left": "issue:44077", + "right": "issue:44079", + "accept": false, + "reason": "Optional post_init validation in patchtsmixer vs. ModelOutput key assignment semantics. Different objects and behavior." + }, + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Python 3.14 load failure vs. tensorboard-only backend detection. Not the same regression." + }, + { + "left": "issue:43976", + "right": "issue:44855", + "accept": false, + "reason": "Python version compatibility issue vs. a JIT parsing/import problem. Different layers of the stack." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "torch.compile SDPA failure in Bamba vs. local dataset loading in an example script. Unrelated." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer initialization vs. GPTNeoX config reload of rotary_pct. Different components." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Qwen3OmniMoe video input validation vs. serve KV-cache indexing. Distinct failures." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "GraniteSpeech CI test flake vs. removed Sam3LiteText auto-mapping references. Both CI-related, but not the same bug." + }, + { + "left": "issue:44496", + "right": "issue:45803", + "accept": false, + "reason": "Unrecognized model/config issue vs. bare except swallowing errors. No shared root cause." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug vs. dataset loading example failure. Unrelated." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "Placeholder issue title vs. fbank feature extraction bug. No meaningful match." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Gemma4 embedding resize issue vs. Whisper empty-transcription generation failure. Different models and code paths." + }, + { + "left": "issue:39692", + "right": "issue:44190", + "accept": false, + "reason": "SigLIP2 documentation example errors vs. local dataset loading in a training script. Separate issues." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "Category cardinality runtime error vs. Whisper generation bug. Unrelated." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Phi4Multimodal default config initialization vs. Qwen3_5MoeVisionConfig field dropping. Both config issues, but different models and semantics." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Processor/tokenizer regression vs. resize_token_embeddings not updating Gemma4 outputs. Different mechanisms." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "Video-classification pipeline processor lookup vs. notebook syntax error. No shared underlying bug." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and different modules; not the same concrete fix." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Qwen2_5_VLImageProcessor export missing vs. layer_type_validation import missing from configuration_utils. Different symbols and files." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "Tokenizer crashes on NER padding vs. server KV-cache continuation indexing. Different tokenizer/model-server paths." + }, + { + "left": "issue:38175", + "right": "issue:43065", + "accept": false, + "reason": "Zero probabilities in SigLIP2 vs. dummy Conv2d in Sam3PixelDecoder. No common bug." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Compiled flex_attention fails on torch>=2.9 vs. incorrect PyTorch version check for AuxRequest import. Related to attention imports only at a very high level." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error vs. audio feature extraction issue. Unrelated." + }, + { + "left": "issue:44466", + "right": "issue:44913", + "accept": false, + "reason": "Serialization of tied lm_head weights vs. GPTNeoX rotary_pct reload behavior. Different config/serialization bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 31, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:41950", + "issue:42175", + "issue:42907", + "issue:43295", + "issue:43299", + "issue:43381", + "issue:43404", + "issue:43441", + "issue:43526", + "issue:43531", + "issue:43643", + "issue:43673", + "issue:43824", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44075", + "issue:44112", + "issue:44242", + "issue:44246", + "issue:44279", + "issue:44336", + "issue:44351", + "issue:44368", + "issue:44496", + "issue:44521", + "issue:44561", + "issue:44734", + "issue:44933", + "issue:44964", + "issue:44987", + "issue:45003", + "issue:45020", + "issue:45042", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45325", + "issue:45335", + "issue:45357", + "issue:45406", + "issue:45584", + "issue:45593", + "issue:45600", + "issue:45632", + "issue:45759" + ], + "soft_pairs": [ + "issue:44734|issue:45632", + "issue:42175|issue:44279", + "issue:43404|issue:45216", + "issue:44561|issue:45632", + "issue:44246|issue:45003", + "issue:43643|issue:44496", + "issue:45406|issue:45632", + "issue:44964|issue:45600", + "issue:44112|issue:44964", + "issue:42907|issue:43441", + "issue:43854|issue:44964", + "issue:44016|issue:44062", + "issue:43295|issue:45335", + "issue:43976|issue:44987", + "issue:43673|issue:44336", + "issue:43531|issue:45600", + "issue:43824|issue:44933", + "issue:44496|issue:45042", + "issue:43526|issue:43901", + "issue:43299|issue:45216", + "issue:44075|issue:44368", + "issue:45335|issue:45584", + "issue:43976|issue:44279", + "issue:43901|issue:44496", + "issue:43381|issue:45593", + "issue:43299|issue:45357", + "issue:44496|issue:45759", + "issue:44351|issue:45278", + "issue:44521|issue:45325", + "issue:44242|issue:45200", + "issue:41950|issue:45020" + ], + "prepared_review_unit_hash": "6183d0bee9e580044c980e52bf1a321a5c0b6b4a9a7021d633f3604696d3bea3", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36957, + "estimated_input_tokens": 9240, + "estimated_eval_tokens": 18736 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The items are mostly heterogeneous bug reports with only loose lexical overlap around loading/imports/remote_code/multimodal behavior; none of the soft pairs look like true duplicates or mergeable PR-equivalents.", + "confidence": 0.68, + "canonical_issue_reason": "issue:45020 is the broadest open umbrella here: it covers recent-version breakage for remote_code models, which is the closest shared theme across the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the best single representative issue because it is broad, user-facing, and likely to subsume several of the version-regression/model-loading reports in this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44734", + "right": "issue:45632", + "accept": false, + "reason": "Serve KV-cache tensor indexing crash and trust_remote_code cache-path collision are different code paths and symptoms." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "One is missing TensorFlow in an extra install; the other is a generic dependency issue. Not the same bug." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Both relate to model saving/state, but they affect different models and failure modes: untied weights vs incorrect checkpoint keys." + }, + { + "left": "issue:44561", + "right": "issue:45632", + "accept": false, + "reason": "Both involve trust_remote_code, but one is a missing helper API and the other is a local cache-path collision." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Import slowness and unsafe sys.modules access are related at most by module import, not the same concrete defect." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "Missing fields from AutoConfig with trust_remote_code is not the same issue as failing to recognize a model type in config.json." + }, + { + "left": "issue:45406", + "right": "issue:45632", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer in serve is unrelated to remote-code cache path collisions." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "These are different model-loading regressions affecting different components and fixes." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "A stale CI test failure in GraniteSpeech is not the same as a model-loading failure for Phi-4 multimodal." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Both mention Ministral, but saving dequantized weights and FlashAttention loading are distinct bugs." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Different models and failure points; one is a GLM import/load issue, the other is Phi-4 multimodal loading." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "A notebook syntax error and an AddedToken argument TypeError are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "These affect different model families and different APIs: processor/tokenizer handling vs embedding resizing." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "Python-version compatibility and a model load failure are not the same underlying bug." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Generation cache behavior during chunked prefill and ANSI formatting in loading_report are unrelated." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and removed auto_mappings configs are different subsystems and failures." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Both are import-related, but they concern different missing symbols in different modules." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "Unrecognized model config and PIL backend image-processor torchvision dependency are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "A BeitImageProcessorFast label bug and a documentation/behavior mismatch in TextClassificationPipeline are not duplicates." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Both are Qwen3.5 regressions, but one is model loading and the other is incorrect save_pretrained output." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "Unused SGD arguments and a tie_word_embeddings warning concern different components and symptoms." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "Embedding resizing in t5gemma and Whisper empty-transcription generation are unrelated bugs." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "Python compatibility and a generic dependency issue do not describe the same defect." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "Docs mismatch for text classification is not the same as a config/model-type loading error." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Gradient checkpointing eval-mode behavior and D-FINE aux-loss handling are different training bugs." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Related model family, but one is load failure and the other is incorrect visual-encoder keys on save." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "An unrecognized model config and composite-config unwrapping in AutoModelForCausalLM are different code paths." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "A missing HybridCache export is just one symptom; the other is a broader import-error report, not the same bug." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "Assistant mask generation for multimodal chat templates and rope-index scaling are separate multimodal bugs." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "Router-logit load-balancing loss and Gemma4 mm_token_type_ids defaults are unrelated training issues." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "A video-classification pipeline processor lookup bug is unrelated to remote_code model breakage." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 32, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:33290", + "issue:38175", + "issue:39692", + "issue:41950", + "issue:42994", + "issue:43065", + "issue:43295", + "issue:43299", + "issue:43317", + "issue:43441", + "issue:43450", + "issue:43479", + "issue:43525", + "issue:43577", + "issue:43650", + "issue:43723", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44162", + "issue:44261", + "issue:44263", + "issue:44279", + "issue:44291", + "issue:44297", + "issue:44462", + "issue:44466", + "issue:44617", + "issue:44623", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:44898", + "issue:45081", + "issue:45216", + "issue:45276", + "issue:45290", + "issue:45397", + "issue:45405", + "issue:45542", + "issue:45600", + "issue:45656", + "issue:45712", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:45907" + ], + "soft_pairs": [ + "issue:43976|issue:45405", + "issue:43479|issue:43525", + "issue:44779|issue:45812", + "issue:41950|issue:45823", + "issue:43577|issue:45216", + "issue:38175|issue:44617", + "issue:44683|issue:45542", + "issue:43299|issue:43854", + "issue:39692|issue:43901", + "issue:44279|issue:45405", + "issue:43295|issue:44779", + "issue:41950|issue:43650", + "issue:44462|issue:45081", + "issue:33290|issue:45656", + "issue:44297|issue:44462", + "issue:45081|issue:45907", + "issue:44016|issue:44291", + "issue:45600|issue:45712", + "issue:43441|issue:45834", + "issue:43450|issue:44279", + "issue:43065|issue:44898", + "issue:42994|issue:43317", + "issue:44466|issue:45276", + "issue:44261|issue:44263", + "issue:43650|issue:45290", + "issue:43441|issue:44683", + "issue:44623|issue:45397", + "issue:44623|issue:44749", + "issue:43723|issue:44162" + ], + "prepared_review_unit_hash": "be6284bc14fbfa19ba5f38516efc0779780349f6374f88399ecba77678efeef2", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36357, + "estimated_input_tokens": 9090, + "estimated_eval_tokens": 18436 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft pairs look like true duplicates or the same fix; most are only loosely related by broad subsystems like tokenization, loading, or multimodal configs. I would keep them separate. The strongest issue to anchor triage is the open Granite token-ID regression.", + "confidence": 0.9, + "canonical_issue_reason": "issue:45812 is the strongest canonical issue candidate: it is open, clearly scoped, describes a concrete silent regression, and has multiple inbound references, making it the best anchor for dedupe/triage.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45812 is the most actionable and broadly impactful issue in this set, with a specific reproducible tokenizer regression and ongoing attention from other references.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Unrelated problems: Python version compatibility vs an unreleased dependency version bump." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Both are config-related, but one is multimodal default initialization and the other is a missing pad_token_id attribute; different bugs." + }, + { + "left": "issue:44779", + "right": "issue:45812", + "accept": false, + "reason": "Both involve tokenization regressions, but they affect different model families and likely different code paths/root causes." + }, + { + "left": "issue:41950", + "right": "issue:45823", + "accept": false, + "reason": "Different subsystems and failures: video-classification processor lookup vs Gemma4 PLE device mismatch." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "Both concern model loading/saving behavior, but the observed failures and affected code paths are not the same." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "Completely different symptoms and components: zero probabilities in SigLIP2 vs CUDA OOM in Sam3Video." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "Attention/torch-related on one side, backend installation issue on the other; no shared concrete bug." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Both are model-loading failures in v5, but they concern different models and distinct failure modes." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "Both are documentation-related, but one is a broken SigLIP2 example and the other is stale pipeline docs." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "Dependency complaints, but not the same dependency or failure mechanism." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Both are regressions around tokenizer/processor behavior, but the concrete bug is different: processor.tokenizer access vs wrong tokenization results." + }, + { + "left": "issue:41950", + "right": "issue:43650", + "accept": false, + "reason": "No shared bug: video pipeline processor lookup vs unrelated placeholder issue." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "Different tokenizer-loading failures: repository tokenizer.json being ignored vs a crash in Mistral regex patching." + }, + { + "left": "issue:33290", + "right": "issue:45656", + "accept": false, + "reason": "Both mention deepspeed/optimizer behavior, but one is OOM with Adafactor and the other is double optimizer stepping." + }, + { + "left": "issue:44297", + "right": "issue:44462", + "accept": false, + "reason": "Both touch tokenizer serialization/loading, but one is save_pretrained metadata mismatch and the other is AutoTokenizer ignoring tokenizer.json." + }, + { + "left": "issue:45081", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer patch crash vs list_repo_files not forwarding revision; unrelated APIs." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Notebook syntax error vs init_empty_weights TypeError; unrelated." + }, + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Both are repo maintenance/CI issues, but they target different files and problems." + }, + { + "left": "issue:43441", + "right": "issue:45834", + "accept": false, + "reason": "Both mention model/runtime failures, but one is FlashAttention on Ministral-3 and the other is an OCR index error in Kosmos2.5." + }, + { + "left": "issue:43450", + "right": "issue:44279", + "accept": false, + "reason": "Video processor shape bug vs generic dependency issue; no common underlying bug." + }, + { + "left": "issue:43065", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-model bugs: a dummy Conv2d in Sam3PixelDecoder vs Perceiver resize/interpolation failure." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Both involve quantization/loading, but saving a quantized model and loading a dequantized model with offload are different issues." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Both relate to embeddings/serialization, but one is device-dependent lm_head serialization and the other is resize_token_embeddings not propagating to Gemma4 components." + }, + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "One is a missing rms_norm_eps config precision issue; the other is torch.split return handling in GlmMoeDsaIndexer." + }, + { + "left": "issue:43650", + "right": "issue:45290", + "accept": false, + "reason": "Placeholder issue vs chat template crash on tool-call assistant messages; unrelated." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "Both are attention-related, but one is model-specific FlashAttention failure and the other is compiled flex_attention on newer torch." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "Both are save/load related, but they affect different APIs and failure modes." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "Processor save_pretrained file omission vs a performance regression when filtering data; unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "Tokenizer loading bug vs a separate model breakage in ESM2; no same underlying defect." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 33, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:41628", + "issue:41950", + "issue:42175", + "issue:42491", + "issue:43122", + "issue:43262", + "issue:43299", + "issue:43352", + "issue:43450", + "issue:43525", + "issue:43531", + "issue:43643", + "issue:43650", + "issue:43653", + "issue:43673", + "issue:43761", + "issue:43784", + "issue:43824", + "issue:43881", + "issue:43901", + "issue:43976", + "issue:43994", + "issue:44016", + "issue:44188", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44297", + "issue:44336", + "issue:44361", + "issue:44373", + "issue:44496", + "issue:44521", + "issue:44589", + "issue:44617", + "issue:44671", + "issue:44779", + "issue:44987", + "issue:45250", + "issue:45362", + "issue:45397", + "issue:45542", + "issue:45561", + "issue:45563", + "issue:45593", + "issue:45823" + ], + "soft_pairs": [ + "issue:43653|issue:45593", + "issue:43262|issue:44521", + "issue:43673|issue:43761", + "issue:44016|issue:45823", + "issue:44361|issue:45593", + "issue:43784|issue:45542", + "issue:42175|issue:43976", + "issue:44671|issue:44779", + "issue:43901|issue:43994", + "issue:43122|issue:43650", + "issue:44987|issue:45542", + "issue:43824|issue:43976", + "issue:44297|issue:45397", + "issue:43525|issue:43881", + "issue:44188|issue:44336", + "issue:43525|issue:43643", + "issue:43450|issue:44617", + "issue:42491|issue:43299", + "issue:44589|issue:45362", + "issue:41628|issue:43976", + "issue:45561|issue:45563", + "issue:43901|issue:44246", + "issue:44336|issue:44373", + "issue:44242|issue:45593", + "issue:44295|issue:45397", + "issue:41950|issue:44220", + "issue:44246|issue:44496", + "issue:43352|issue:45250", + "issue:43531|issue:45362" + ], + "prepared_review_unit_hash": "bd3825e2701f6905fc75a2257b3e0df9e763be5df3c1054498b6a0a27fb658be", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35403, + "estimated_input_tokens": 8851, + "estimated_eval_tokens": 17958 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is very heterogeneous: most items are unrelated regressions across tokenizers, vision/audio processors, model loading, packaging, and docs. I do not see any true duplicate pair among the soft edges.", + "confidence": 0.9, + "canonical_issue_reason": "issue:43122 is the broadest, most central user-facing regression in the set (tokenization changed between v4.57.3 and v5.0), so it is the best anchor if one issue must represent this mixed cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43122 is the most general and broadly applicable regression report here; it has the clearest duplicate potential compared with the more model-specific or documentation-only reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "BigBirdTokenizer special-token decoding vs D-FINE auxiliary-loss behavior; different models and code paths." + }, + { + "left": "issue:43262", + "right": "issue:44521", + "accept": false, + "reason": "Both mention apply_chat_template, but one is audio sampling rate and the other is multimodal assistant masks; not the same bug." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Generation cache/chunked_prefill regression vs CLIPVisionModel hidden_states bug; unrelated subsystems." + }, + { + "left": "issue:44016", + "right": "issue:45823", + "accept": false, + "reason": "Notebook syntax error vs Gemma4 device-map forward mismatch; unrelated." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "MLukeTokenizer task AttributeError vs D-FINE auxiliary losses; unrelated." + }, + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "sentence-transformers import NameError vs missing TensorFlow backend packaging; both are import-time issues but not the same one." + }, + { + "left": "issue:42175", + "right": "issue:43976", + "accept": false, + "reason": "Package extras/backends issue vs Python version compatibility regression; different failure modes." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "CamemBERT masked-LM predictions vs Deepseek tokenizer regression; both are model/tokenizer bugs but not the same code path or model." + }, + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "Docs mismatch for return_all_scores vs incorrect SigLIP2 outputs; one is documentation, the other is model behavior." + }, + { + "left": "issue:43122", + "right": "issue:43650", + "accept": false, + "reason": "Tokenizer regression across versions vs an unrelated low-information issue; no shared bug." + }, + { + "left": "issue:44987", + "right": "issue:45542", + "accept": false, + "reason": "Loading a specific model on >=5.1.0 vs TensorFlow backend installation error; unrelated." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Missing Qwen2_5_VL import vs Python 3.9/3.10 compatibility problem; unrelated." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch vs Gemma4 ZeRO3 from_pretrained failure; unrelated." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Llama4Config missing pad_token_id vs glm-4v loading failure; different models and causes." + }, + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "torch.compile attention-kernel divergence vs ANSI codes in loading_report; unrelated." + }, + { + "left": "issue:43525", + "right": "issue:43643", + "accept": false, + "reason": "Both involve missing config fields, but one is Llama4 pad_token_id and the other is trust_remote_code AutoConfig fields; not the same bug." + }, + { + "left": "issue:43450", + "right": "issue:44617", + "accept": false, + "reason": "Batched video processor shape bug vs Sam3Video CUDA OOM; unrelated." + }, + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Qwen3 MoE LoRA compatibility across hf versions vs Qwen3VL loading break on v5; related family, but different concrete failures." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Float8 storage deserialization error vs Qwen3.5 chat crash; unrelated." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "Cannot import AutoImageProcessor vs Python version support problem; unrelated." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "pytest-xdist file race vs stale generate() warning; different testing/runtime issues." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "Documentation mismatch vs slow import behavior; unrelated." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "ANSI output in loading_report vs wrong docstring for position_ids; both docs/UI-ish, but not the same bug." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "MoE load-balancing loss omission vs D-FINE auxiliary losses; both are loss-related but for different models and mechanisms." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "position_ids buffer access bug vs Gemma4 from_pretrained/ZeRO3 issue; unrelated." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification pipeline selecting image processors vs audio fbank extraction issue; different modalities and code paths." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Import latency regression vs unrecognized model/config error; unrelated." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "Model-specific Flash Attention 2 support error vs generic Flash Attention 2 discussion; not the same concrete bug." + }, + { + "left": "issue:43531", + "right": "issue:45362", + "accept": false, + "reason": "Qwen3-MoE sliding_window issue vs Qwen3.5-35B chat crash; different models and failure signatures." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 34, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:36246", + "issue:42757", + "issue:42886", + "issue:43065", + "issue:43232", + "issue:43257", + "issue:43262", + "issue:43408", + "issue:43475", + "issue:43526", + "issue:43540", + "issue:43596", + "issue:43653", + "issue:43673", + "issue:43901", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44336", + "issue:44351", + "issue:44368", + "issue:44466", + "issue:44485", + "issue:44496", + "issue:44561", + "issue:44749", + "issue:44779", + "issue:44857", + "issue:44871", + "issue:44877", + "issue:44987", + "issue:44995", + "issue:45003", + "issue:45137", + "issue:45200", + "issue:45229", + "issue:45292", + "issue:45335", + "issue:45375", + "issue:45459", + "issue:45561", + "issue:45563", + "issue:45584", + "issue:45701", + "issue:45759" + ], + "soft_pairs": [ + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:43526|issue:45003", + "issue:44336|issue:44485", + "issue:45003|issue:45561", + "issue:43673|issue:44561", + "issue:44246|issue:44871", + "issue:44242|issue:44857", + "issue:44162|issue:45200", + "issue:44295|issue:44749", + "issue:36246|issue:43596", + "issue:44877|issue:45375", + "issue:44779|issue:45292", + "issue:45292|issue:45584", + "issue:43901|issue:45003", + "issue:45292|issue:45701", + "issue:43232|issue:44336", + "issue:42886|issue:45229", + "issue:44496|issue:45003", + "issue:43065|issue:43408", + "issue:44987|issue:45459", + "issue:45375|issue:45759", + "issue:43257|issue:45137", + "issue:43262|issue:44246", + "issue:44995|issue:45200", + "issue:44060|issue:44077", + "issue:43475|issue:43540", + "issue:44466|issue:45335", + "issue:43653|issue:44162", + "issue:44368|issue:45563" + ], + "prepared_review_unit_hash": "5711006eb8754be780c969999188723f74ea9a1daa1b16a7b688643c4b5161cc", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36863, + "estimated_input_tokens": 9216, + "estimated_eval_tokens": 18688 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: mostly unrelated issue reports spanning import errors, model/config mismatches, generation bugs, and model-specific regressions. The soft matches are all superficial overlaps rather than the same underlying bug or change.", + "confidence": 0.89, + "canonical_issue_reason": "Issue 45003 is the broadest, most cross-cutting internal bug in the set; the rest are mostly narrower model- or feature-specific regressions, so it is the best representative if one must be chosen.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45003 is the most generally impactful and system-level item in an otherwise diverse cluster, making it the best single issue candidate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they involve different symbols and different packages; not the same defect." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Separate missing-import problems for unrelated names/modules." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "One is a BeitImageProcessorFast label-reduction bug; the other is unsafe sys.modules access in modeling_utils. Different code paths." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "ANSI-output handling and GLM-5 RoPE implementation are unrelated issues." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "sys.modules access in modeling_utils is unrelated to a pytest-xdist race in testing utils." + }, + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Generation cache behavior and removal of is_torch_fx_available are different regression classes." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Load-balancing loss logic and Gemma-3 eos_token_id config mismatch are unrelated." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Different model families and different failures: router-loss accounting vs AMP/CUDA crash." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "ESM2 breakage and Gemma 4 mm_token_type_ids defaulting are unrelated model-specific bugs." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "A position_ids buffer error and a tokenizer-speed regression are not the same issue." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "ImportError for Qwen2.5-VL image processor vs DeepSpeed ZeRO3 BertModel index error; unrelated." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Both are config-related, but they concern different model families and different missing-fields/loading behaviors." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "DeepSeek tokenization regression and resize_token_embeddings propagation are different bugs." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "Embedding resize propagation and Whisper empty-transcription handling are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Docs wording for TextClassificationPipeline is unrelated to sys.modules access in modeling_utils." + }, + { + "left": "issue:45292", + "right": "issue:45701", + "accept": false, + "reason": "Embedding resize behavior and general tokenization changes are not the same defect." + }, + { + "left": "issue:43232", + "right": "issue:44336", + "accept": false, + "reason": "Generation kwargs update after sync_gpus and ANSI code emission are unrelated." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Offline cache loading and Gemma4 multi-GPU OOM are distinct problems." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model config and unsafe sys.modules access are different failure modes." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "Both mention SAM3, but one is a pixel-decoder implementation issue and the other is a model-type warning from config mismatch." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "A loading failure on transformers>=5.1.0 and hidden tokenizer errors from protobuf exception handling are unrelated." + }, + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Missing Qwen3.5MoeVisionConfig field and AutoModelForCausalLM.from_config text_config unwrapping are different config-handling bugs." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed and a ZeRO3 deque index error are not the same code-path problem." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio sampling-rate defaults and MoE load-balancing loss are unrelated." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "Stale cache on second forward pass vs Gemma 4 token-type defaults are different bugs." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "A tied-weights warning in Qwen3-Next and a patchtsmixer post_init policy change are unrelated." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "Both are multimodal/video-related, but they affect different models and different failure points." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Both involve embedding/tied-weight behavior, but one is serialization consistency and the other is resize propagation into decoder embeddings; not the same underlying bug." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and ESM2 breakage are unrelated." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "A tied-embeddings warning and a stale num_return_sequences warning are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 35, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:41628", + "issue:42371", + "issue:42673", + "issue:42994", + "issue:43232", + "issue:43475", + "issue:43482", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43704", + "issue:43873", + "issue:43931", + "issue:44038", + "issue:44162", + "issue:44188", + "issue:44297", + "issue:44351", + "issue:44361", + "issue:44368", + "issue:44623", + "issue:44661", + "issue:44734", + "issue:44749", + "issue:44869", + "issue:44933", + "issue:44945", + "issue:44977", + "issue:44995", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45230", + "issue:45278", + "issue:45310", + "issue:45356", + "issue:45375", + "issue:45381", + "issue:45397", + "issue:45459", + "issue:45478", + "issue:45588", + "issue:45600", + "issue:45656" + ], + "soft_pairs": [ + "issue:44368|issue:44977", + "issue:44995|issue:45588", + "issue:45397|issue:45656", + "issue:43673|issue:44734", + "issue:45137|issue:45600", + "issue:43931|issue:44368", + "issue:44933|issue:45459", + "issue:44869|issue:45356", + "issue:43232|issue:43673", + "issue:44297|issue:44623", + "issue:43673|issue:45278", + "issue:43646|issue:44162", + "issue:44361|issue:44749", + "issue:44623|issue:45656", + "issue:44749|issue:45656", + "issue:44749|issue:45310", + "issue:28282|issue:30990", + "issue:42994|issue:43482", + "issue:42371|issue:45092", + "issue:43475|issue:45381", + "issue:43475|issue:43643", + "issue:45230|issue:45397", + "issue:44038|issue:44623", + "issue:41628|issue:44933", + "issue:45081|issue:45103", + "issue:43475|issue:44661", + "issue:42994|issue:43525", + "issue:45375|issue:45478", + "issue:42673|issue:43704", + "issue:43232|issue:44188", + "issue:43873|issue:44945", + "issue:45310|issue:45375", + "issue:41628|issue:44351" + ], + "prepared_review_unit_hash": "3a82fd340f8e2c0f553641ab7a1c66894debb5e51293a973f4da26bb0c565e69", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37434, + "estimated_input_tokens": 9359, + "estimated_eval_tokens": 18974 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a mix of superficially similar Transformers bug reports, but the only clear duplicate edge is the repeated Qwen3ForCausalLM VRAM leak report. The rest differ by model, subsystem, or concrete failure mode despite overlapping keywords like v5, quantization, generation, or import errors.", + "confidence": 0.92, + "canonical_issue_reason": "issue:42673 is the best canonical issue because it is the earlier, more discussed report of the exact same Qwen3ForCausalLM VRAM leak in multi-dataloader-thread usage as issue:43704.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42673 is the strongest representative issue in this cluster: it is an exact duplicate candidate with clear repro context and a specific code-path bug, unlike the many broader or unrelated reports here.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "Both mention Qwen3.5, but one is a tie_word_embeddings warning during LoRA fine-tuning while the other is a flash-attention generation failure. Different bugs and code paths." + }, + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "GlmMoeDsa stale indexer cache on second forward pass is unrelated to flash_attention.py crashing on s_aux=None for sink-less models." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Gemma-4 Zero3 from_pretrained failure and a double optimizer-step bug with deepspeed are different failures with no shared concrete code-path." + }, + { + "left": "issue:43673", + "right": "issue:44734", + "accept": false, + "reason": "One is a GenerationMixin cache regression in chunked_prefill; the other is a transformers serve tensor-indexing crash in /v1/responses. Not the same bug." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque underflow and stale auto_mappings references to removed Sam3LiteText configs are unrelated." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Qwen3-VL-30B weight-shape mismatch is a model-loading error; the other is a warning about tie_word_embeddings during fine-tuning." + }, + { + "left": "issue:44933", + "right": "issue:45459", + "accept": false, + "reason": "A missing import from image_utils and tokenizer error masking when protobuf is absent are separate import/exception-handling issues." + }, + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Whisper timestamp decoding on a trailing replacement character and a Kimi-K2.5 tokenizer regression are different tokenization/decoding bugs." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "sync_gpus generation kwargs handling and missing GenerationMixin cache during chunked_prefill are distinct generation regressions." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "tokenizer.save_pretrained writing the wrong tokenizer_class and processor.save_pretrained missing files are related only at a high level; not the same concrete bug." + }, + { + "left": "issue:43673", + "right": "issue:45278", + "accept": false, + "reason": "A generation cache regression and a broad report of many import errors after upgrading to v5.5.0 are not the same issue." + }, + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "Transformers 5 breaking custom model initialization is broader than the ESM2-specific breakage described here; different impacted code paths." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "MLukeTokenizer attribute errors on tasks and a performance slowdown after upgrading to 5.3.0 are unrelated." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "Missing files from processor.save_pretrained and optimizer.step being called twice under deepspeed are completely different behaviors." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "A throughput regression after upgrade and duplicate optimizer stepping with deepspeed do not describe the same bug." + }, + { + "left": "issue:44749", + "right": "issue:45310", + "accept": false, + "reason": "Performance regression after upgrade is not the same as Qwen3.5 Moe from_pretrained failing." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "AutoModel import error due to missing PyTorch and Sentence Transformers hanging while loading are unrelated installation/runtime problems." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure with v5 are different model I/O problems." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 API-setting guidance and old InternVL2 meta-initialization incompatibility are unrelated." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "SAM 3 video missing fpn_position_embeddings and qwen2.5-vl video vision_position_ids mismatch are separate video-model bugs." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "A missing field on Sam3VisionEncoderOutput is unrelated to AutoConfig returning missing fields under trust_remote_code=True." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "A generic bug report and a specific Gemma-4 Zero3 loading failure are not duplicates." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe model bug and processor.save_pretrained missing files are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Two import errors, but for different symbols and different subsystems: AutoImageProcessor versus image_utils." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "A Mistral regex patch crash and an auto_docstring future-annotations crash are unrelated exceptions." + }, + { + "left": "issue:43475", + "right": "issue:44661", + "accept": false, + "reason": "SAM 3 video output attribute loss and add-new-model-like failing on TOKENIZER_MAPPING_NAMES are different failures." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized saving failure and Llama4Config missing pad_token_id are separate model/config issues." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Missing deepstack_visual_indexes in Qwen3_5MoeVisionConfig and a Qwen3.5 Moe from_pretrained error are related by model family only, not the same concrete bug." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Exact same title and same failure mode: Qwen3ForCausalLM leaks VRAM when used in multiple dataloader threads. This is a clear duplicate." + }, + { + "left": "issue:43232", + "right": "issue:44188", + "accept": false, + "reason": "A generation kwargs update bug after sync_gpus and attention-kernel divergence under torch.compile are different issues." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Quantization/offloading behavior and incorrect LLM output under pipeline parallelism do not share the same root cause." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "One is a Qwen3.5 Moe from_pretrained error; the other is a missing config field silently dropped by strict serialization. Related ecosystem, different bugs." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "AutoImageProcessor missing import and HybridCache missing import are separate API export regressions." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet: issue:42673 and issue:43704 are indeed identical titles with a clear duplicate pattern, and the rest of the cluster is mostly separated by model or subsystem. The soft-edge rejections are conservative overall and do not overclaim duplicate status." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 36, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:33290", + "issue:35141", + "issue:36010", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42831", + "issue:42898", + "issue:42994", + "issue:43232", + "issue:43278", + "issue:43377", + "issue:43479", + "issue:43504", + "issue:43526", + "issue:43673", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43976", + "issue:44038", + "issue:44060", + "issue:44079", + "issue:44188", + "issue:44222", + "issue:44242", + "issue:44373", + "issue:44496", + "issue:44568", + "issue:44623", + "issue:44734", + "issue:44792", + "issue:44869", + "issue:44933", + "issue:44945", + "issue:44964", + "issue:45103", + "issue:45125", + "issue:45137", + "issue:45200", + "issue:45310", + "issue:45356", + "issue:45561", + "issue:45712", + "issue:45753", + "issue:45812" + ], + "soft_pairs": [ + "issue:44945|issue:45753", + "issue:42831|issue:44222", + "issue:42222|issue:45200", + "issue:44222|issue:45310", + "issue:36010|issue:41628", + "issue:43673|issue:44188", + "issue:42175|issue:45712", + "issue:42994|issue:43824", + "issue:43232|issue:44792", + "issue:43377|issue:44242", + "issue:44038|issue:45310", + "issue:43526|issue:45561", + "issue:43504|issue:45561", + "issue:44496|issue:45561", + "issue:44964|issue:45137", + "issue:42175|issue:44933", + "issue:42898|issue:44568", + "issue:43479|issue:44869", + "issue:30990|issue:35141", + "issue:43976|issue:44933", + "issue:43479|issue:44079", + "issue:43784|issue:43824", + "issue:44188|issue:44373", + "issue:44623|issue:45310", + "issue:43278|issue:45137", + "issue:42898|issue:45812", + "issue:44734|issue:45103", + "issue:30333|issue:33290", + "issue:43723|issue:44568", + "issue:44060|issue:45125", + "issue:42898|issue:45356" + ], + "prepared_review_unit_hash": "fa44a991afb32d81abb7feb027f8a9dd1da1d160227b175565e56a5be1f05365", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36918, + "estimated_input_tokens": 9230, + "estimated_eval_tokens": 18716 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated transformers bugs that were grouped by loose title similarity rather than shared root cause. No soft pair looks like a safe duplicate merge candidate.", + "confidence": 0.24, + "canonical_issue_reason": "issue:43824 is the strongest single anchor only because it has the most concrete user-facing import regression signal and the highest reference activity; however, the cluster is too heterogeneous to treat it as a true canonical duplicate target.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43824 is the most actionable representative of the set: it is a concrete import failure, has strong discussion/reference signals, and is self-contained enough for triage even though it does not unify the other issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Both mention Qwen models, but one is incorrect LLM output under pipeline parallelism and the other is an infinite loop on a specific image; different failure modes and likely different code paths." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "One is an accuracy regression in FineGrainedFP8, the other is a save_pretrained failure for FP8 moe models; related theme, but not the same bug." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "Vitpose model breakage and Gemma 4 mm_token_type_ids defaulting are unrelated model-specific issues." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "Both involve FP8/Qwen-related regressions, but one is moe save_pretrained and the other is a from_pretrained error on transformers>=5.4.0; not the same concrete bug." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they involve different symbols (GenerationMixin vs AutoImageProcessor) and different compatibility problems." + }, + { + "left": "issue:43673", + "right": "issue:44188", + "accept": false, + "reason": "Chunked prefill GenerationMixin cache issues and attention-kernel divergence under torch.compile are separate generation vs kernel behaviors." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Package backend dependency omission and leftover dummy classes leaking into dir(transformers) are unrelated maintenance issues." + }, + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Quantized model saving and missing Qwen2.5-VL import are different subsystems and failure surfaces." + }, + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "A generation kwargs update bug after sync_gpus is unrelated to a failed janus image-generation test." + }, + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "Missing padding-mask support in MIMI encoder and router-logits load-balancing loss are different model behaviors." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Qwen3-VL-Moe and Qwen3.5 Moe both concern Qwen, but the model variants and reported from_pretrained failures are not clearly the same bug." + }, + { + "left": "issue:43526", + "right": "issue:45561", + "accept": false, + "reason": "A BeitImageProcessorFast label-reduction bug and a pytest-xdist race on captured_info.txt are unrelated." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Beit semantic-segmentation preset loading and an xdist file race are different issues with no shared code path." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Unrecognized model loading in from_pretrained is unrelated to the testing-utils captured_info.txt race." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Phi-4 multimodal loading failure and DeepSpeed ZeRO3 deque underflow are distinct runtime problems." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Tensorflow backend packaging and a nonexistent import from image_utils are separate import/package issues." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "clean_up_tokenization_spaces behavior changes and add_special_tokens BOS/EOS regression are different tokenizer semantics." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Default vision/audio config initialization and Whisper timestamp decode crash are unrelated multimodal vs speech bugs." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "Sentence Transformers loading hang and token-embedding reinitialization after resize are unrelated issues." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python version compatibility for transformers 5.1.0 and a missing image_utils import are not the same defect." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Incorrect multimodal config defaults and ModelOutput key assignment when previously None are different object-initialization bugs." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Sentence-transformers import NameError and Qwen2.5-VL import failure are both import-related but affect different modules and causes." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "Attention-kernel divergence under compile and a wrong docstring for position_ids are unrelated; one is a runtime bug, the other documentation." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Missing processor.save_pretrained files and Qwen3.5 Moe from_pretrained regression are separate save/load problems." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Embedding dtype drift during evaluate and a ZeRO3 deque pop error are unrelated training/runtime failures." + }, + { + "left": "issue:42898", + "right": "issue:45812", + "accept": false, + "reason": "Tokenizer cleanup behavior change and Granite token ID regression are both tokenizer issues, but they concern different APIs and distinct regressions." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "KV-cache continuation crash in transformers serve and auto_docstring annotation handling are unrelated." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "MLflow reporting after failed training and OOM with Adafactor in DeepSpeed are different training/runtime problems." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer loading in v5 and add_special_tokens BOS/EOS regression are both tokenization-related, but not the same failure." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Both concern Qwen MoE tensor-related behavior, but one is tied-weights warning logic and the other is missing _tp_plan for tensor parallelism." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Tokenizer cleanup semantics and Kimi-K2.5 codec/regex regression are different tokenizer regressions." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 37, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:22355", + "issue:28282", + "issue:30333", + "issue:30990", + "issue:34689", + "issue:35141", + "issue:36331", + "issue:38175", + "issue:41628", + "issue:42757", + "issue:42831", + "issue:42898", + "issue:43116", + "issue:43232", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:43874", + "issue:44190", + "issue:44206", + "issue:44222", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44442", + "issue:44448", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44514", + "issue:44623", + "issue:44779", + "issue:44871", + "issue:44933", + "issue:44977", + "issue:44991", + "issue:45092", + "issue:45103", + "issue:45310", + "issue:45357", + "issue:45405", + "issue:45406", + "issue:45656", + "issue:45712", + "issue:45834" + ], + "soft_pairs": [ + "issue:43874|issue:45092", + "issue:42898|issue:44779", + "issue:45405|issue:45712", + "issue:42898|issue:44448", + "issue:44295|issue:45656", + "issue:44933|issue:45712", + "issue:44222|issue:44297", + "issue:43116|issue:44190", + "issue:44297|issue:45656", + "issue:43479|issue:43644", + "issue:44222|issue:44623", + "issue:44206|issue:45103", + "issue:42831|issue:44623", + "issue:41628|issue:42757", + "issue:22355|issue:28282", + "issue:30333|issue:36331", + "issue:30333|issue:34689", + "issue:43232|issue:45834", + "issue:44222|issue:44484", + "issue:44297|issue:45310", + "issue:43723|issue:44977", + "issue:28282|issue:30333", + "issue:30333|issue:35141", + "issue:44263|issue:45092", + "issue:44483|issue:45406", + "issue:44263|issue:44360", + "issue:44373|issue:44485", + "issue:44295|issue:44623", + "issue:44514|issue:45103", + "issue:44261|issue:45092", + "issue:45092|issue:45357", + "issue:43723|issue:44871", + "issue:30990|issue:38175", + "issue:43723|issue:44991", + "issue:43723|issue:44442" + ], + "prepared_review_unit_hash": "f876f7c2ff604678575108c31162c0dc0bb6a130babbf60e67c7e0a93844f81c", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38005, + "estimated_input_tokens": 9502, + "estimated_eval_tokens": 19260 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly noisy and spans several unrelated regressions. The only noticeable subcluster is around v5 tokenizer loading/import behavior, but none of the soft pairs are close enough to treat as the same underlying bug/change.", + "confidence": 0.81, + "canonical_issue_reason": "Best representative of the main subcluster: a broad v5 tokenizer-loading regression centered on AutoTokenizer.from_pretrained, which is the closest hub for the related tokenizer issues here.", + "canonical_pr_reason": null, + "best_issue_reason": "Most central and reusable issue for triage; it captures the broad tokenizer-loading failure mode without overfitting to one model or tokenizer implementation.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Different failures: one is a missing image-processor method, the other is a remote-code/meta-init incompatibility." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer-related, but one is about cleanup-space behavior and the other is a model-specific wrong-output regression." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "Unrelated: a version pin problem versus dummy class leakage in the import surface." + }, + { + "left": "issue:42898", + "right": "issue:44448", + "accept": false, + "reason": "Both mention v4/v5 output differences, but the concrete bugs are different and not clearly the same code path." + }, + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "Different subsystems: buffer access versus DeepSpeed optimizer stepping." + }, + { + "left": "issue:44933", + "right": "issue:45712", + "accept": false, + "reason": "Both touch imports, but one is a bad image_utils import and the other is dummy PT placeholders leaking into dir()." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "Both involve save_pretrained, but one is FP8 MoE weights and the other is tokenizer config metadata." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "Different example scripts and different failures: classification output handling versus local dataset loading." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer save metadata bug versus DeepSpeed double optimizer step; no shared concrete bug." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Both are initialization-related, but they affect different components and different failure modes." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "Both are save_pretrained-related, but one is model FP8 MoE serialization and the other is missing processor files." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "Different code paths: a feature extractor argument regression versus auto-docstring annotation handling." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "Accuracy/quantization issue versus missing processor save files; unrelated." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but from different packages and different causes." + }, + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Both are import failures, but one is a missing submodule and the other is an optional backend dependency check." + }, + { + "left": "issue:30333", + "right": "issue:36331", + "accept": false, + "reason": "MLflow reporting and trainer loss signature mismatch are unrelated." + }, + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "Different problems: training-job reporting versus model loading regression." + }, + { + "left": "issue:43232", + "right": "issue:45834", + "accept": false, + "reason": "Generation kwargs handling is unrelated to OCR index errors." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "Different save_pretrained topics: MoE FP8 serialization versus shard-size default behavior." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer config metadata mismatch is not the same bug as Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Both are tokenizer/generation regressions in v5, but they are model-specific and do not look like the same bug." + }, + { + "left": "issue:28282", + "right": "issue:30333", + "accept": false, + "reason": "Completely unrelated import error and MLflow reporting issue." + }, + { + "left": "issue:30333", + "right": "issue:35141", + "accept": false, + "reason": "Training-job status reporting versus embedding resize/post_init behavior." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "Unrelated GLM indexer bug versus InternVL2 meta-init incompatibility." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "Both involve serving, but one is request handling and the other is a processor attribute crash." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "Same module family, but different concrete bugs in the DSA indexer." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "A docstring issue is not the same as a RoPE implementation bug." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "Buffer registration/read bug versus missing processor save files." + }, + { + "left": "issue:44514", + "right": "issue:45103", + "accept": false, + "reason": "Batched chat-template crash and auto-docstring annotation crash are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "MLA layernorm epsilon precision issue versus remote-code/meta-init incompatibility." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "Both mention Qwen/vision-related v5 regressions, but the failure modes are different." + }, + { + "left": "issue:43723", + "right": "issue:44871", + "accept": false, + "reason": "Tokenizer loading regression versus tokenizer/model eos-token-id mismatch; not the same bug." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "Different models and different symptoms: loading stall versus zero probabilities." + }, + { + "left": "issue:43723", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading complaints, but they are specific to different tokenizers and lack a shared concrete failure path." + }, + { + "left": "issue:43723", + "right": "issue:44442", + "accept": false, + "reason": "Tokenizer-loading failure versus AutoTokenizer failing on a specific tokenizer class; related theme, not the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 38, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:22355", + "issue:28218", + "issue:30990", + "issue:33357", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43066", + "issue:43122", + "issue:43232", + "issue:43408", + "issue:43475", + "issue:43479", + "issue:43576", + "issue:43756", + "issue:43784", + "issue:43824", + "issue:43937", + "issue:43976", + "issue:44038", + "issue:44060", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44485", + "issue:44488", + "issue:44514", + "issue:44521", + "issue:44568", + "issue:44704", + "issue:44743", + "issue:44913", + "issue:44945", + "issue:45278", + "issue:45290", + "issue:45310", + "issue:45593", + "issue:45600", + "issue:45753" + ], + "soft_pairs": [ + "issue:41628|issue:43824", + "issue:43479|issue:44913", + "issue:42831|issue:45310", + "issue:39401|issue:44568", + "issue:44263|issue:44373", + "issue:42757|issue:45278", + "issue:42831|issue:44038", + "issue:43756|issue:44060", + "issue:42886|issue:42994", + "issue:43232|issue:45593", + "issue:43576|issue:43784", + "issue:42831|issue:45753", + "issue:42886|issue:43066", + "issue:22355|issue:28218", + "issue:43122|issue:44521", + "issue:42371|issue:44263", + "issue:44117|issue:44704", + "issue:43232|issue:44743", + "issue:30990|issue:33357", + "issue:42907|issue:44945", + "issue:44295|issue:44297", + "issue:44704|issue:45290", + "issue:43122|issue:44514", + "issue:41628|issue:45278", + "issue:43976|issue:44945", + "issue:43784|issue:45600", + "issue:39401|issue:44488", + "issue:43408|issue:43475", + "issue:42907|issue:43937", + "issue:44261|issue:44373", + "issue:44360|issue:44485" + ], + "prepared_review_unit_hash": "918337310c9add629896f2dc674f7768575998cd2fce99491e61ac6ead5d6fff", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36472, + "estimated_input_tokens": 9118, + "estimated_eval_tokens": 18492 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: it mixes import/API regressions, tokenizer behavior bugs, multimodal/model-specific issues, and unrelated configuration/documentation problems. None of the soft pairs look like true duplicates or the same underlying fix. There are no PRs in the cluster.", + "confidence": 0.89, + "canonical_issue_reason": "issue:45278 is the broadest and most representative issue here: it captures the general upgrade/import-regression theme that overlaps with several other reports in the set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best standalone issue to anchor the cluster because it is the most general, clearly user-facing, and could plausibly subsume multiple import-related reports without overfitting to a narrow model-specific bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Different missing symbols and different model areas: AutoImageProcessor import vs Qwen2_5_VLForConditionalGeneration import." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Unrelated config bugs: multimodal default config initialization vs GPTNeoX rotary_pct resetting on reload." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "Different failures: FineGrainedFP8 accuracy degradation vs Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer offset_mapping bug is not the same as add_special_tokens failing to add BOS/EOS tokens." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "DSA indexer torch.split behavior is unrelated to a wrong docstring for position_ids." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "One is a specific missing is_offline_mode import; the other is a broad collection of import errors after upgrade." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "Different model/tooling problems: FP8 accuracy issue vs Qwen3-VL-Moe bug." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Smollm3 RoPE-layer count discrepancy is unrelated to Qwen3-Next tied-weights warning." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Tokenizer offline cache loading and quantized model saving are different code paths and symptoms." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation kwargs update after sync_gpus is unrelated to D-FINE auxiliary-loss behavior." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "Broken transformers env command and sentence-transformers import NameError are distinct issues." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "FineGrainedFP8 accuracy issue is unrelated to Qwen3_5 infinite-loop behavior on a specific image." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Offline tokenizer cache loading and wrong tokenizer decoder type are separate tokenizer problems." + }, + { + "left": "issue:22355", + "right": "issue:28218", + "accept": false, + "reason": "Missing transformers.onnx module and an extra space after added token are unrelated bugs." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Version-to-version tokenization differences do not match all-zero assistant masks for multimodal inputs." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 settings warning is unrelated to the DSA indexer torch.split issue." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None and AutoProcessor kwargs propagation are different failure points." + }, + { + "left": "issue:43232", + "right": "issue:44743", + "accept": false, + "reason": "Generation sync_gpus behavior and recurrent-state reset in modular_qwen3_5 are not the same bug." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Sentence Transformers loading hang and MacOS bus error with CLIP are different runtime failures." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Dequantized model saving failure and incorrect pipeline-parallel LLM output are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "Position_ids buffer read error is unrelated to tokenizer_class mismatch in tokenizer_config.json." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "AutoProcessor kwargs passing and chat-template crashes with tool calls are different issues." + }, + { + "left": "issue:43122", + "right": "issue:44514", + "accept": false, + "reason": "General tokenization differences across versions are not the same as batched apply_chat_template crashes." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "Specific AutoImageProcessor import failure is only loosely related to a broad import-error report." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python version compatibility regression and pipeline-parallel output corruption are different problems." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "Import NameError in sentence-transformers is unrelated to stale auto_mappings references to removed configs." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Tokenizer offset_mapping bug is unrelated to a model loading failure for cjvt/sleng-bert." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Wrong model-type warning and missing fpn_position_embeddings attribute are distinct SAM 3 issues." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Quantized model saving failure and invalid GenerationConfig are different subsystems and symptoms." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "Precision error from missing rms_norm_eps is not the same as a wrong docstring." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer lacking ReLU and GLM-5 RoPE implementation are unrelated model-specific bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 39, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:30990", + "issue:36296", + "issue:38617", + "issue:39401", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:42757", + "issue:42907", + "issue:43064", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43278", + "issue:43441", + "issue:43452", + "issue:43493", + "issue:43673", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44222", + "issue:44295", + "issue:44297", + "issue:44351", + "issue:44485", + "issue:44521", + "issue:44625", + "issue:44704", + "issue:44779", + "issue:44829", + "issue:44945", + "issue:44964", + "issue:44977", + "issue:45072", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45446", + "issue:45478", + "issue:45753" + ], + "soft_pairs": [ + "issue:44625|issue:44829", + "issue:39401|issue:44521", + "issue:43116|issue:45200", + "issue:29127|issue:30990", + "issue:42907|issue:44964", + "issue:44038|issue:44297", + "issue:43122|issue:44779", + "issue:44038|issue:44295", + "issue:36296|issue:43116", + "issue:42907|issue:43452", + "issue:43824|issue:44351", + "issue:44038|issue:44222", + "issue:45216|issue:45478", + "issue:42757|issue:44351", + "issue:43278|issue:44945", + "issue:43493|issue:45753", + "issue:42907|issue:44222", + "issue:43122|issue:44977", + "issue:43992|issue:44704", + "issue:43066|issue:44779", + "issue:43784|issue:45446", + "issue:38617|issue:45278", + "issue:42673|issue:45072", + "issue:43937|issue:44485", + "issue:22355|issue:29127", + "issue:42371|issue:43874", + "issue:43976|issue:45446", + "issue:43064|issue:43673", + "issue:42222|issue:43441", + "issue:43116|issue:43493" + ], + "prepared_review_unit_hash": "cb23cd07c2291dd5f8ec298a4b030edaca47acfaabb2c9dd8af8b0b903656410", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 35753, + "estimated_input_tokens": 8939, + "estimated_eval_tokens": 18134 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly unrelated Transformers bug reports across different models, APIs, and failure modes. The only broadly representative issue is the open v4\u2192v5 import-regression report; none of the soft pairs look like true duplicates or mergeable together as one fix.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45278 is the broadest umbrella report here (multiple import errors after upgrading 4.57.0 to 5.5.0), so it best represents the small import-regression subset.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best issue to anchor triage because it is open, cross-cutting, and more general than the model-specific or single-symbol bug reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "Different bugs: config propagation for Qwen3.5 vs degenerate training with flash_attention_3." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Both involve tokenization-adjacent behavior, but one is offset_mapping and the other is assistant_mask generation for multimodal chat templates." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Example-script multi-label output bug is unrelated to Gemma 4 mm_token_type_ids defaulting." + }, + { + "left": "issue:29127", + "right": "issue:30990", + "accept": false, + "reason": "LayoutLMv3 error-message clarity is unrelated to Sentence Transformers loading stalls." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Saving dequantized Ministral/Devstral models is a different code path from loading Phi-4 multimodal instruct." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe failure is unrelated to tokenizer_config tokenizer_class mismatch during save_pretrained." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenization regressions, but they affect different model families and concrete tokenizer behavior." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Different failures: Qwen3-VL-Moe vs a position_ids buffer access error." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor-parallel optimizer-state bug is unrelated to the classification example returning empty results." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Saving dequantized models and gguf_file loading failures are separate issues." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they concern different symbols and different modules." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Different MoE/model bugs: Qwen3-VL-Moe vs FP8 save_pretrained for moe models." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Related to Qwen3.5 and version 5.4+, but one is save_pretrained checkpoint correctness and the other is from_pretrained loading; not the same concrete bug." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Separate missing imports from different packages and symbols." + }, + { + "left": "issue:43278", + "right": "issue:44945", + "accept": false, + "reason": "Embedding dtype drift in evaluate is unrelated to incorrect outputs under pipeline parallelism." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and a Qwen3_5 image-specific infinite loop are unrelated." + }, + { + "left": "issue:42907", + "right": "issue:44222", + "accept": false, + "reason": "Both involve saving, but they are different models and different quantization formats." + }, + { + "left": "issue:43122", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer regression vs flash-attention generation failure are different bugs." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Both use from_pretrained, but one misses a weight load and the other drops kwargs when resolving files." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Tokenizer decoder type bug is not the same as Deepseek tokenization changes in v5." + }, + { + "left": "issue:43784", + "right": "issue:45446", + "accept": false, + "reason": "Sentence-transformers import NameError and flex_attention version check are unrelated." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "A specific missing import is not the same underlying bug as a broad report of multiple import errors." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "VRAM leakage in multi-threaded loading is unrelated to bfloat16 dtype mismatches." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "GenerationConfig validation failure and GLM-5 RoPE implementation are distinct issues." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "onnx module import failure is unrelated to LayoutLMv3 error-message clarity." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 API guidance and missing get_number_of_image_patches are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python version support regression is unrelated to an AuxRequest flex_attention version-check bug." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "Optimizer-state mismatch under FSDP2/PEFT is unrelated to missing GenerationMixin cache during chunked prefill." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "Broken vitpose model code and Ministral-3 flash-attention failure are different model-specific regressions." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Classification example output bug is unrelated to SigLIP2 implementation discrepancy." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 40, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:33453", + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42222", + "issue:42371", + "issue:42617", + "issue:42673", + "issue:43116", + "issue:43408", + "issue:43421", + "issue:43450", + "issue:43577", + "issue:43582", + "issue:43742", + "issue:43825", + "issue:43827", + "issue:43906", + "issue:43937", + "issue:43986", + "issue:44038", + "issue:44075", + "issue:44117", + "issue:44261", + "issue:44291", + "issue:44410", + "issue:44484", + "issue:44496", + "issue:44779", + "issue:44829", + "issue:44933", + "issue:44936", + "issue:44977", + "issue:45071", + "issue:45081", + "issue:45083", + "issue:45216", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45357", + "issue:45405", + "issue:45446", + "issue:45478", + "issue:45542", + "issue:45561", + "issue:45563", + "issue:45588" + ], + "soft_pairs": [ + "issue:44075|issue:44936", + "issue:43116|issue:43825", + "issue:45357|issue:45478", + "issue:44038|issue:45478", + "issue:45083|issue:45446", + "issue:43937|issue:45563", + "issue:33453|issue:39401", + "issue:44117|issue:45588", + "issue:43827|issue:45446", + "issue:39401|issue:44779", + "issue:43937|issue:44977", + "issue:42673|issue:43906", + "issue:44291|issue:45071", + "issue:43450|issue:43986", + "issue:45405|issue:45446", + "issue:45341|issue:45561", + "issue:38617|issue:41628", + "issue:44117|issue:45081", + "issue:39401|issue:43421", + "issue:44410|issue:45310", + "issue:44484|issue:45310", + "issue:44829|issue:45216", + "issue:42673|issue:43582", + "issue:42371|issue:44261", + "issue:43408|issue:43742", + "issue:42617|issue:45341", + "issue:44933|issue:45278", + "issue:42222|issue:45542", + "issue:42673|issue:43577", + "issue:43937|issue:44496" + ], + "prepared_review_unit_hash": "b9014470ca55687855d80a8ed8a191530da0de530148f63721f6ac9b88e6582c", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36558, + "estimated_input_tokens": 9140, + "estimated_eval_tokens": 18536 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous: most pairs are only superficially similar (often same subsystem, model family, or release, but different concrete failures). I would not merge any of the soft edges as duplicates. If forced to pick a single issue anchor, 45278 is the broadest import-regression umbrella, but it is still not a true match for most of the set.", + "confidence": 0.23, + "canonical_issue_reason": "45278 is the broadest umbrella report in the set ('many import errors after update'), so it is the closest catch-all representative for the import-regression subset.", + "canonical_pr_reason": null, + "best_issue_reason": "45278 is the most general and discussion-rich issue here, making it the best single representative even though the overall cluster is mixed.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "Different bugs: optimizer argument handling vs Trainer.evaluate after train." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Different topics: example-script multi-label output vs v5 pipeline messaging/docs." + }, + { + "left": "issue:45357", + "right": "issue:45478", + "accept": false, + "reason": "Different code paths: save_pretrained serialization vs from_pretrained loading failure." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Different model families and failure modes; not the same bug." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "Unrelated helper behavior in qwen3_omni_moe vs a flex_attention version-check import issue." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "GenerationConfig validation error vs stale warning text; different problems." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Both tokenizer-related, but different models and symptoms (loading regression vs offset_mapping bug)." + }, + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "Tokenizer mapping fallback bug vs flash_attention crash; unrelated." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs still referencing pipeline() vs a runtime import/version check issue." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and likely different root causes." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "GenerationConfig invalid vs flash-attention generation behavior; not the same issue." + }, + { + "left": "issue:42673", + "right": "issue:43906", + "accept": false, + "reason": "VRAM leak in multi-threaded Qwen3ForCausalLM vs an isolated reproduction of a different issue." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "Different failures: init_empty_weights TypeError vs PretrainedConfig type-checking regression." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Batched video processor shape bug vs AutoProcessor crash without torchvision." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "Dependency version pinning issue vs PyTorch version gating in flex_attention." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "Both touch testing_utils, but one is a generic bug and the other is a specific xdist race." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Different missing imports/symbols; same broad symptom but not the same underlying export bug." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "Token mapping None vs Mistral regex patch crash; different tokenizer paths." + }, + { + "left": "issue:39401", + "right": "issue:43421", + "accept": false, + "reason": "Different tokenizer bugs: offset_mapping correctness vs runtime special-token post-processor sync." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Missing projections in qwen3next vs Qwen3.5 from_pretrained error; different concrete failures." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "save_pretrained shard-size question vs model loading error; unrelated." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "Training degeneration under flash_attention_3 vs incorrect saved checkpoint contents." + }, + { + "left": "issue:42673", + "right": "issue:43582", + "accept": false, + "reason": "VRAM leak in dataloader threads vs Apple Silicon warmup TypeError; unrelated." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 API guidance vs MLA precision/config epsilon bug." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "Model-type warning for sam3 vs key error loading MobileLLM; different loading problems." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "3d_parallel.py execution failure vs a testing_utils bug; no shared underlying defect." + }, + { + "left": "issue:44933", + "right": "issue:45278", + "accept": false, + "reason": "A specific missing import is not enough to equate it with a broad 'many import errors' report." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "Vitpose model breakage vs TensorFlow backend detection issue; unrelated." + }, + { + "left": "issue:42673", + "right": "issue:43577", + "accept": false, + "reason": "VRAM leak in Qwen3ForCausalLM threads vs dtype propagation in BLIP2; unrelated." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "Invalid GenerationConfig vs unrecognized model/config.json missing model_type; different load-time failures." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 41, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:38617", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42757", + "issue:42831", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43335", + "issue:43408", + "issue:43441", + "issue:43452", + "issue:43531", + "issue:43577", + "issue:43673", + "issue:43874", + "issue:43931", + "issue:43976", + "issue:43986", + "issue:44038", + "issue:44075", + "issue:44117", + "issue:44220", + "issue:44261", + "issue:44263", + "issue:44315", + "issue:44368", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44493", + "issue:44829", + "issue:44987", + "issue:45003", + "issue:45071", + "issue:45245", + "issue:45341", + "issue:45372", + "issue:45459", + "issue:45464", + "issue:45478", + "issue:45542", + "issue:45656" + ], + "soft_pairs": [ + "issue:42175|issue:42222", + "issue:44075|issue:45656", + "issue:44483|issue:45464", + "issue:38617|issue:44220", + "issue:42371|issue:42831", + "issue:43066|issue:43122", + "issue:43012|issue:43577", + "issue:43673|issue:45341", + "issue:44484|issue:45478", + "issue:43874|issue:44261", + "issue:44410|issue:45478", + "issue:43097|issue:44493", + "issue:43012|issue:43408", + "issue:44263|issue:44485", + "issue:41628|issue:45071", + "issue:42222|issue:45459", + "issue:43441|issue:44829", + "issue:45071|issue:45542", + "issue:43931|issue:44410", + "issue:43335|issue:45071", + "issue:45003|issue:45341", + "issue:43931|issue:44038", + "issue:43986|issue:45341", + "issue:43097|issue:43452", + "issue:43408|issue:44368", + "issue:44117|issue:44987", + "issue:43986|issue:44484", + "issue:43874|issue:44263", + "issue:43976|issue:45071", + "issue:43986|issue:45372", + "issue:44829|issue:45245", + "issue:43531|issue:45341", + "issue:38617|issue:42757", + "issue:44261|issue:44485", + "issue:43064|issue:44315" + ], + "prepared_review_unit_hash": "ce13f384107e6f20636fc52d687e60b09f94239aeb70bcb7fff005b4acf0dfa4", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38035, + "estimated_input_tokens": 9509, + "estimated_eval_tokens": 19274 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated single-bug reports across packaging, tokenizers, model loaders, and optimizer behavior. I found no soft edge that clearly describes the same underlying bug/change with enough specificity to merge the pair.", + "confidence": 0.68, + "canonical_issue_reason": "issue:43122 is the broadest clear regression report in the set, centered on tokenizer behavior changing between 4.57.3 and 5.0, which is a plausible umbrella for nearby tokenizer-related complaints.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43122 is the most suitable duplicate target because it describes a user-visible semantic regression across versions, not just a narrow implementation detail.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "Different problems: backend packaging for TensorFlow vs a broken VitPose model import path." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "Both involve optimizers, but one is about SGD arguments being ignored while the other is about DeepSpeed causing double optimizer steps." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Same API area, but one is a general chat/completions failure and the other is a streaming-inference bug on a specific Qwen model." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "Unrelated code paths: a missing import from configuration_utils vs an issue in _torch_extract_fbank_features()." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 settings behavior and FineGrainedFP8 accuracy are distinct numerical/configuration issues." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Both are tokenizer regressions, but one is about decoder type metadata and the other is a broader tokenization output change; not clearly the same bug." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "Both mention dtype/precision, but they affect different models and different loading behaviors." + }, + { + "left": "issue:43673", + "right": "issue:45341", + "accept": false, + "reason": "Generation cache behavior and a testing utility bug are unrelated." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "Saving shard size defaults and a Qwen3.5 MoE from_pretrained error are separate concerns." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Different model components and failures: image processor method missing vs MLA layernorm epsilon handling." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "A missing projection layer in Qwen3Next is not the same as a Qwen3.5 MoE loading failure." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "One is about a removed config helper; the other is about unexpected position-id keys during loading." + }, + { + "left": "issue:43012", + "right": "issue:43408", + "accept": false, + "reason": "TF32/bfloat16 warnings and a SAM3 model-type mismatch warning are not the same issue." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "Torch split return handling in one GLM MoE helper is distinct from a RoPE implementation bug." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "Importing AutoImageProcessor and PreTrainedConfig type checking are unrelated failures." + }, + { + "left": "issue:42222", + "right": "issue:45459", + "accept": false, + "reason": "A broken VitPose model and tokenizer error masking when protobuf is absent are different bugs." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "Both mention FlashAttention, but one is a model-specific inference failure and the other is degenerate training with FlashAttention 3." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Type-checking regression in PreTrainedConfig is unrelated to a missing TensorFlow backend when only tensorboard is installed." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Different model families and failure modes: Qwen3-VL weight mismatch vs Qwen3Next layer projection omission." + }, + { + "left": "issue:43335", + "right": "issue:45071", + "accept": false, + "reason": "SwitchTransformers sparse-layer construction and PreTrainedConfig type checking do not share the same code path." + }, + { + "left": "issue:45003", + "right": "issue:45341", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to a testing_utils bug." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "Both are model-loading problems, but they involve different Qwen variants and different root causes." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "AutoProcessor/video-model dependency handling is unrelated to testing_utils." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "A removed config helper and a gguf_file loading regression are different loader issues." + }, + { + "left": "issue:43408", + "right": "issue:44368", + "accept": false, + "reason": "The SAM3 model-type warning is not the same as the tie_word_embeddings warning during Qwen3.5 LoRA fine-tuning." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer mapping None handling and a specific physical-intelligence/fast loading failure are not the same bug." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Video-model auto-processing without torchvision is unrelated to max_shard_size in save_pretrained()." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Both touch GLM-related internals, but one is an image-processor method gap and the other is a torch.split indexing issue." + }, + { + "left": "issue:43976", + "right": "issue:45071", + "accept": false, + "reason": "Python version/package compatibility and config type checking are different concerns." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "Missing torchvision in AutoProcessor video loading is unrelated to the ReasoningEffort import error for Gemma 4." + }, + { + "left": "issue:44829", + "right": "issue:45245", + "accept": false, + "reason": "FlashAttention training degeneration and the 2^24 category limit are unrelated numerical/shape constraints." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "A sliding_window bug in Qwen3-MoE is unrelated to testing_utils." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they are about different missing symbols from different packages." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "Different GLM issues: missing rms_norm_eps in config vs a RoPE implementation discussion." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "FSDP2/PEFT optimizer-state corruption and Liger Kernel not applying in model_init are distinct training setup bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 42, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:36683", + "issue:38617", + "issue:42757", + "issue:42886", + "issue:42907", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43352", + "issue:43381", + "issue:43502", + "issue:43526", + "issue:43550", + "issue:43582", + "issue:43618", + "issue:43761", + "issue:43824", + "issue:43825", + "issue:43873", + "issue:43901", + "issue:43906", + "issue:43976", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44062", + "issue:44117", + "issue:44222", + "issue:44230", + "issue:44351", + "issue:44368", + "issue:44380", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44514", + "issue:44655", + "issue:44683", + "issue:44704", + "issue:45103", + "issue:45290", + "issue:45375", + "issue:45478", + "issue:45479", + "issue:45588", + "issue:45753" + ], + "soft_pairs": [ + "issue:38617|issue:44351", + "issue:44410|issue:45375", + "issue:44062|issue:44117", + "issue:43097|issue:45588", + "issue:44038|issue:44410", + "issue:44704|issue:45103", + "issue:43502|issue:43992", + "issue:43986|issue:44117", + "issue:43097|issue:43873", + "issue:43824|issue:43986", + "issue:43066|issue:43906", + "issue:43097|issue:43901", + "issue:43825|issue:44655", + "issue:43992|issue:44484", + "issue:43352|issue:44380", + "issue:43208|issue:43550", + "issue:43012|issue:43550", + "issue:43012|issue:44368", + "issue:42886|issue:43381", + "issue:42907|issue:45753", + "issue:42907|issue:44655", + "issue:44380|issue:45479", + "issue:43352|issue:43976", + "issue:43064|issue:44380", + "issue:43208|issue:43526", + "issue:36683|issue:43582", + "issue:44483|issue:44514", + "issue:44222|issue:45478", + "issue:43012|issue:44683", + "issue:43618|issue:43761", + "issue:43618|issue:44380", + "issue:42757|issue:43824", + "issue:38617|issue:43502", + "issue:44483|issue:45290", + "issue:44038|issue:44230" + ], + "prepared_review_unit_hash": "1c6a33e4672d25a3a752225d2fbb1ce7bec779e0c99d9fde8aacd394f3b0a7e3", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38062, + "estimated_input_tokens": 9516, + "estimated_eval_tokens": 19288 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is very broad and mostly contains unrelated Transformers regressions that only share a loose v5/compatibility/theme overlap. I do not see any soft pair that clearly describes the same underlying bug or change, so all soft edges should be rejected. The closest representative issue is the broad v5 API-breakage report around removed configuration behavior.", + "confidence": 0.89, + "canonical_issue_reason": "issue:43097 is the broadest v5 API-compatibility regression in the set and best matches the cluster\u2019s shared theme of breaking changes, even though the rest of the items are largely unrelated.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43097 is the best standalone representative because it captures a core, user-visible v5 breakage rather than a narrow model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and different failure modes." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Both concern Qwen-family configs, but they are different fields, models, and bugs." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "Tokenizer-related, but one is an AddedToken construction error and the other is a mapping/None-handling bug." + }, + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "One is a v5 config/API removal complaint; the other is a flash-attention AttributeError in a different code path." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Both touch Qwen models, but the concrete problem is different: model support vs missing config/projection fields." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "Both involve kwargs handling, but one is processor loading and the other is auto-docstring introspection." + }, + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "Offline/local-file behavior and missing state-dict weights are unrelated bugs." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "Different subsystems: video AutoProcessor dependency handling vs tokenizer mapping None-handling." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "A v5 API removal is not the same as quantization offloading misbehavior." + }, + { + "left": "issue:43824", + "right": "issue:43986", + "accept": false, + "reason": "One is a missing model import; the other is a torchvision dependency crash in AutoProcessor." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Both mention tokenizer behavior, but there is no evidence they describe the same bug." + }, + { + "left": "issue:43097", + "right": "issue:43901", + "accept": false, + "reason": "API removal and documentation drift for a pipeline argument are different issues." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "Both involve pipeline-related behavior, but one is an error message regression and the other is save_pretrained support." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Missing encoder weights and max_shard_size behavior are separate save/load concerns." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Both mention attention backends, but they affect different models and different concrete failures." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "Different model families and different training/compile failures." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "A precision warning during compile is not the same as a SDPA compile failure." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "Different symptoms: a PyTorch warning versus a tie_word_embeddings config warning." + }, + { + "left": "issue:42886", + "right": "issue:43381", + "accept": false, + "reason": "Offline cache loading and gradient checkpointing eval-mode behavior are unrelated." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Saving dequantized models and an infinite-loop generation bug are not the same issue." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Both are save-related, but one is dequantized-model saving and the other is pipeline serialization." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Different problem classes: attention backend scaling vs classification loss degeneracy." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "A model capability issue and a Python version compatibility issue are unrelated." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "Distributed optimizer-state corruption and attention scaling are unrelated bugs." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "Different models and different preprocessing/training failures." + }, + { + "left": "issue:36683", + "right": "issue:43582", + "accept": false, + "reason": "A missing config attribute and an Apple Silicon allocator TypeError are unrelated." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "Both involve chat-related APIs, but one is an endpoint acceptance regression and the other is processor batching behavior." + }, + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "Both concern Qwen3.5 MoE/FP8-ish areas, but one is save_pretrained and the other is from_pretrained; not the same bug." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "External dependency import failure and missing model-class export are different import problems." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "Different failures: an internal import symbol removal versus local_files_only still making network calls." + }, + { + "left": "issue:44483", + "right": "issue:45290", + "accept": false, + "reason": "Both are chat-template/request formatting issues, but the concrete crash conditions differ." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Same broad Qwen3/VL/MoE area, but one is a model bug and the other is explicit fp8 inference support." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 43, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:36010", + "issue:36683", + "issue:41669", + "issue:42757", + "issue:42890", + "issue:42898", + "issue:42994", + "issue:43010", + "issue:43064", + "issue:43066", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43525", + "issue:43611", + "issue:43618", + "issue:43756", + "issue:43867", + "issue:43901", + "issue:43931", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44077", + "issue:44188", + "issue:44242", + "issue:44246", + "issue:44263", + "issue:44355", + "issue:44380", + "issue:44483", + "issue:44485", + "issue:44492", + "issue:44655", + "issue:44855", + "issue:44938", + "issue:45071", + "issue:45216", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45375", + "issue:45381", + "issue:45440", + "issue:45561", + "issue:45593", + "issue:45753", + "issue:45850" + ], + "soft_pairs": [ + "issue:43299|issue:43867", + "issue:44483|issue:45381", + "issue:43756|issue:44485", + "issue:43299|issue:45753", + "issue:43901|issue:44655", + "issue:36683|issue:43475", + "issue:42994|issue:44655", + "issue:43352|issue:43937", + "issue:44077|issue:44380", + "issue:43867|issue:45216", + "issue:36010|issue:42757", + "issue:42898|issue:43066", + "issue:36683|issue:44077", + "issue:43611|issue:44655", + "issue:43976|issue:44938", + "issue:43352|issue:45593", + "issue:36683|issue:43525", + "issue:44263|issue:45440", + "issue:43618|issue:45216", + "issue:44188|issue:44380", + "issue:36683|issue:44242", + "issue:43064|issue:44492", + "issue:43352|issue:43931", + "issue:43010|issue:45278", + "issue:43992|issue:45310", + "issue:43867|issue:45375", + "issue:41669|issue:44246", + "issue:44355|issue:44855", + "issue:36683|issue:43064", + "issue:42890|issue:45561", + "issue:44938|issue:45071", + "issue:43867|issue:45357", + "issue:43010|issue:45850" + ], + "prepared_review_unit_hash": "c73576f4bb53809b99edaa0d2bf4d90964d08e6da71ab2bdd85420e5911cc5e2", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37161, + "estimated_input_tokens": 9291, + "estimated_eval_tokens": 18838 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose collection of unrelated Transformers regressions (imports, loading/saving, config mismatches, tokenizer behavior, training, and backend-specific bugs). I do not see any soft-edge pair that is safely the same underlying bug/change, so I would not merge any of the candidate pairs. If one issue must represent the cluster, issue 45278 is the broadest umbrella report.", + "confidence": 0.62, + "canonical_issue_reason": "issue:45278 is the most umbrella-like report: it describes broad import breakage after upgrading, which best matches the cluster\u2019s general upgrade/regression theme.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best single representative because it is broad, open, and central to the recurring \u2018upgrade to v5 breaks things\u2019 pattern seen across this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Both are model-loading failures, but one is a Qwen3VL MoE loading regression and the other is a state_dict ordering/load error; different code paths and fixes." + }, + { + "left": "issue:44483", + "right": "issue:45381", + "accept": false, + "reason": "One is an API request rejection in /v1/chat/completions, the other is a qwen2.5-vl video position-id bug; unrelated failures." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Both mention RoPE, but they concern different models and different implementation bugs; not the same concrete change." + }, + { + "left": "issue:43299", + "right": "issue:45753", + "accept": false, + "reason": "A loading regression for Qwen3VL MoE is not the same as an infinite-loop generation bug for a specific image in Qwen3_5." + }, + { + "left": "issue:43901", + "right": "issue:44655", + "accept": false, + "reason": "Docs mismatch for return_all_scores is unrelated to saving Pipeline objects." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Different missing-attribute errors on different model configs/outputs; no shared underlying bug is evident." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Quantized model save failure and Pipeline save_pretrained failure are distinct serialization issues." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "Flash-Attn support gating for Nemotron is unrelated to GLM-5 GenerationConfig validation." + }, + { + "left": "issue:44077", + "right": "issue:44380", + "accept": false, + "reason": "One is a patchtsmixer post_init constraint, the other is GPT2 attention scaling under SDPA/FlashAttention; different bugs." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "Both involve model persistence/loading, but one is a sorted-state_dict load error and the other is an incorrect Qwen3.5 save_pretrained checkpoint; not the same fix." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Different import errors from different modules/dependencies; same symptom, not same bug." + }, + { + "left": "issue:42898", + "right": "issue:43066", + "accept": false, + "reason": "Tokenizer cleanup-space behavior and wrong decoder type are separate v5 tokenizer regressions." + }, + { + "left": "issue:36683", + "right": "issue:44077", + "accept": false, + "reason": "A config attribute error and a patchtsmixer post_init policy issue are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:44655", + "accept": false, + "reason": "base_model_prefix loading breakage is not the same as saving Pipeline objects." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both are Python-version compatibility complaints, but they point to different releases and likely different root causes." + }, + { + "left": "issue:43352", + "right": "issue:45593", + "accept": false, + "reason": "Flash-Attn support for Nemotron has nothing to do with D-FINE auxiliary-loss handling." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Different models and different missing config attributes; not a shared code-path bug." + }, + { + "left": "issue:44263", + "right": "issue:45440", + "accept": false, + "reason": "A torch.split indexing bug and a DeepseekV3MoE divergence from remote implementation are different issues." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "Missing attentions assignment in CLIPOutput is unrelated to Qwen3.5 save_pretrained checkpoint corruption." + }, + { + "left": "issue:44188", + "right": "issue:44380", + "accept": false, + "reason": "Both touch attention backends, but one is a torch.compile branching divergence and the other is GPT2 attention scaling being ignored; distinct bugs." + }, + { + "left": "issue:36683", + "right": "issue:44242", + "accept": false, + "reason": "A missing config field and a missing load-balancing loss are separate model logic issues." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Distributed optimizer-state corruption and a cache-strategy typo are not the same underlying problem." + }, + { + "left": "issue:43352", + "right": "issue:43931", + "accept": false, + "reason": "FlashAttention support for Nemotron is unrelated to Qwen3-VL weight-shape mismatch on load." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "no_grad decoration for cache/layer updates is a code correctness change, while 45278 is a broad import-breakage report; different issues." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "Both are from_pretrained failures, but they concern different models and different missing/misrouted weights." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "State_dict loading error and a missing Qwen3_5MoeVisionConfig field are separate bugs." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": false, + "reason": "Import-time performance slowdown is related in theme, but the issue about removing import * usage is not clearly the same concrete bug as intermittent slow imports." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Compiled Python-file errors and a Python 3.13 torch.jit parsing indentation error are different import/parsing failures." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Python 3.14 load failure and PretrainedConfig type-checking regression are unrelated." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "Both are Qwen3.5 save/load regressions, but one is a load/state_dict ordering problem and the other is incorrect visual-encoder keys on save; not the same exact fix." + }, + { + "left": "issue:43010", + "right": "issue:45850", + "accept": false, + "reason": "A cache-update no_grad change does not match a broad \u2018too many models break since 5.0\u2019 complaint." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 44, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:28282", + "issue:30333", + "issue:30990", + "issue:31795", + "issue:33453", + "issue:36296", + "issue:42548", + "issue:42831", + "issue:42890", + "issue:43010", + "issue:43065", + "issue:43232", + "issue:43299", + "issue:43502", + "issue:43575", + "issue:43618", + "issue:43723", + "issue:43867", + "issue:43874", + "issue:43976", + "issue:43992", + "issue:44062", + "issue:44075", + "issue:44077", + "issue:44112", + "issue:44222", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44387", + "issue:44484", + "issue:44485", + "issue:44556", + "issue:44749", + "issue:44857", + "issue:44938", + "issue:44961", + "issue:44964", + "issue:44977", + "issue:45071", + "issue:45125", + "issue:45341", + "issue:45375", + "issue:45381", + "issue:45440", + "issue:45478", + "issue:45600" + ], + "soft_pairs": [ + "issue:43992|issue:45478", + "issue:43010|issue:43976", + "issue:44938|issue:45381", + "issue:43618|issue:44077", + "issue:36296|issue:43575", + "issue:44360|issue:45440", + "issue:28282|issue:33453", + "issue:43867|issue:44484", + "issue:43618|issue:44857", + "issue:43618|issue:44977", + "issue:42890|issue:44964", + "issue:42890|issue:44112", + "issue:42890|issue:45600", + "issue:43502|issue:44062", + "issue:43010|issue:44075", + "issue:43723|issue:44556", + "issue:42890|issue:45341", + "issue:44387|issue:44556", + "issue:44222|issue:45125", + "issue:43010|issue:43065", + "issue:42548|issue:44351", + "issue:43502|issue:44485", + "issue:44355|issue:44749", + "issue:43010|issue:43232", + "issue:30990|issue:31795", + "issue:45125|issue:45478", + "issue:45125|issue:45375", + "issue:30333|issue:33453", + "issue:43299|issue:44556", + "issue:44315|issue:44360", + "issue:43502|issue:44222", + "issue:42831|issue:43874", + "issue:43502|issue:44961", + "issue:30333|issue:36296", + "issue:44355|issue:45071" + ], + "prepared_review_unit_hash": "43610d24231281fba017f15c58177be24dce2da339c8223ce4ac321dbddee836", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 37521, + "estimated_input_tokens": 9381, + "estimated_eval_tokens": 19018 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is broad and mostly not duplicate-safe: many items are unrelated load/import/test/model-specific regressions that only share superficial wording. No soft pair looks like the same concrete bug or a PR that should be merged together. The closest common thread is v5-era loading regressions, with one issue serving as the best anchor.", + "confidence": 0.73, + "canonical_issue_reason": "issue:44556 is the clearest representative of the recurring v5 loading-regression theme: it explicitly describes a v4.57 -> v5.2/v5.3 checkpoint reload break with a concrete, user-facing failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44556 is the best global anchor because it is explicit about a versioned regression, concrete about the failing code path, and broad enough to represent the main loading-compatibility complaints in this cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Same model family and loading symptom, but one is missing embed_tokens.weight while the other is a generic v5.4.0 from_pretrained failure; not the same bug." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Cache no_grad behavior vs Python 3.9+ compatibility; unrelated." + }, + { + "left": "issue:44938", + "right": "issue:45381", + "accept": false, + "reason": "Python 3.14 load failure vs qwen2.5-vl video position ids bug; unrelated." + }, + { + "left": "issue:43618", + "right": "issue:44077", + "accept": false, + "reason": "CLIPOutput attentions regression vs patchtsmixer post_init validation; unrelated." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Tensor-parallel training bug vs tp OOM on a specific Qwen model; same area but different failure mode and likely different fix." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "DSA indexer ReLU discussion vs DeepSeekV3 divergence; unrelated." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "ImportError for AutoModel/PyTorch vs tokenizer loading regression; both loading-related but distinct symbols and code paths." + }, + { + "left": "issue:43867", + "right": "issue:44484", + "accept": false, + "reason": "State-dict sorting load error vs save_pretrained shard-size question; unrelated." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "CLIPOutput attentions regression vs LwDetrImageLoss AMP crash; unrelated." + }, + { + "left": "issue:43618", + "right": "issue:44977", + "accept": false, + "reason": "CLIPOutput attentions regression vs Qwen3.5 flash-attention generation issue; unrelated." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "Flaky integration tests vs model loading regression; unrelated." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "Both are CI/test issues, but different tests and failure causes." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "Test seed flakiness vs auto_mappings CI break; unrelated." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "local_files_only network leak vs AddedToken duplicate 'special' TypeError; unrelated." + }, + { + "left": "issue:43010", + "right": "issue:44075", + "accept": false, + "reason": "Cache no_grad update vs SGD args ignored; unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Both concern loading regressions after v5, but tokenizer loading and checkpoint reload are different concrete bugs." + }, + { + "left": "issue:42890", + "right": "issue:45341", + "accept": false, + "reason": "Seed-related test flakiness vs a testing_utils bug; unrelated." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "int4 quantization OOM vs checkpoint reload failure; unrelated." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 save_pretrained issue vs missing _tp_plan for tensor parallelism; different code paths." + }, + { + "left": "issue:43010", + "right": "issue:43065", + "accept": false, + "reason": "Generation cache update decorator vs dummy Conv2d in Sam3PixelDecoder; unrelated." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Both are import-name errors, but different exported symbols and likely different regressions." + }, + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "Offline-loading bug vs GLM-5 RoPE implementation; unrelated." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Compiled-file errors vs performance slowdown after upgrade; unrelated." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "Both touch generation helpers, but one is cache update/no_grad and the other sync_gpus state handling; not the same bug." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Loading hang vs documentation confusion; unrelated." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Same Qwen3.5 MoE area, but one is missing _tp_plan and the other is a generic from_pretrained error; not enough to merge." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Tensor-parallel plan issue vs missing vision config field; unrelated." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLflow reporting failure vs tokenizer loading regression; unrelated." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Version-related model loading failures, but different models and failure modes; not one duplicate." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger Kernel application vs DSA ReLU discussion; unrelated." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "Offline requests vs FP8 save_pretrained moe; unrelated." + }, + { + "left": "issue:42831", + "right": "issue:43874", + "accept": false, + "reason": "FP8 accuracy issue vs missing image processor method; unrelated." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "Offline API request bug vs placeholder racoon issue; unrelated." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLflow job tracking vs tensor parallel training bug; unrelated." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled Python load errors vs PretrainedConfig type checking; unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 45, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:38617", + "issue:41628", + "issue:41669", + "issue:42371", + "issue:42548", + "issue:42757", + "issue:42890", + "issue:43525", + "issue:43526", + "issue:43572", + "issue:43576", + "issue:43673", + "issue:43824", + "issue:43827", + "issue:43901", + "issue:44038", + "issue:44222", + "issue:44230", + "issue:44246", + "issue:44261", + "issue:44273", + "issue:44355", + "issue:44360", + "issue:44393", + "issue:44485", + "issue:44492", + "issue:44556", + "issue:44623", + "issue:44655", + "issue:44704", + "issue:44908", + "issue:44938", + "issue:44961", + "issue:45003", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45278", + "issue:45362", + "issue:45468", + "issue:45542", + "issue:45753", + "issue:45850" + ], + "soft_pairs": [ + "issue:43827|issue:44961", + "issue:44355|issue:44655", + "issue:33453|issue:34689", + "issue:42757|issue:43576", + "issue:44222|issue:44230", + "issue:33453|issue:36296", + "issue:43576|issue:43824", + "issue:43576|issue:45071", + "issue:31795|issue:38617", + "issue:44961|issue:45468", + "issue:41628|issue:42548", + "issue:38617|issue:43576", + "issue:43576|issue:44938", + "issue:43673|issue:44492", + "issue:44908|issue:44961", + "issue:44485|issue:44961", + "issue:36296|issue:43901", + "issue:41628|issue:44355", + "issue:36296|issue:43526", + "issue:45362|issue:45850", + "issue:44261|issue:44360", + "issue:44623|issue:45230", + "issue:44273|issue:45542", + "issue:44273|issue:45003", + "issue:44393|issue:45753", + "issue:44393|issue:45071", + "issue:43576|issue:44556", + "issue:44038|issue:45125", + "issue:43525|issue:45850", + "issue:41669|issue:43673", + "issue:44556|issue:45092", + "issue:44492|issue:45850", + "issue:42371|issue:44492", + "issue:43572|issue:45850", + "issue:44704|issue:44908", + "issue:42890|issue:44908", + "issue:45278|issue:45850", + "issue:44246|issue:44273" + ], + "prepared_review_unit_hash": "8f2030faa0c46c95b94dbaf3fa15d1c308e2e86b2969dd75f74341f7fcc57fde", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 38, + "serialized_chars": 38176, + "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 mix of unrelated Transformers v5 regressions, import/export failures, and model-specific bugs. The best umbrella issue is the broad post-upgrade import-error report.", + "confidence": 0.92, + "canonical_issue_reason": "#45278 is the most representative umbrella: it concretely describes broad import failures after upgrading to 5.5.0, which matches the general regression theme better than the other broad-but-vague reports.", + "canonical_pr_reason": null, + "best_issue_reason": "#45278 is the strongest canonical target for duplicate triage because it is broad enough to cover the shared import-regression pattern, while still being specific about the upgrade-triggered failure mode.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs about pipeline() references vs an unrelated issue titled \"racoon\"; no shared bug." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Compiled Python file errors and pipeline save_pretrained failures are different code paths." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Both are loading regressions, but one is tokenizer loading and the other is a specific model-loading breakage." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Missing import from huggingface_hub vs broken env command; different failures." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "Both mention FP8/moe, but one is save_pretrained and the other is inference support." + }, + { + "left": "issue:33453", + "right": "issue:36296", + "accept": false, + "reason": "Tokenizer loading regression and tensor-parallel training bug are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "Broken env command and missing Qwen2.5-VL export are different issues." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "Env command breakage is unrelated to PretrainedConfig type-checking." + }, + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Documentation confusion and an import error for layer_type_validation are not the same bug." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "Unrelated titles: no evidence of the same underlying defect." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Both are import errors, but for different symbols and likely different export regressions." + }, + { + "left": "issue:38617", + "right": "issue:43576", + "accept": false, + "reason": "ImportError for layer_type_validation and a broken env command are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:44938", + "accept": false, + "reason": "Env command failure is not the same as Python 3.14 load failure." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Cache missing during chunked prefill vs a typo in cache strategies are different problems." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs bug and unrelated \"racoon\" issue do not match." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "RoPE implementation discussion vs unrelated \"racoon\" issue." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor-parallel training bug and docs mentioning return_all_scores are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44355", + "accept": false, + "reason": "Different failure classes: import export vs compiled-file runtime errors." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor parallel training bug and BeitImageProcessorFast label reduction bug are unrelated." + }, + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "A specific Qwen3.5 chat crash is not the same as a broad complaint that v5 breaks too many models." + }, + { + "left": "issue:44261", + "right": "issue:44360", + "accept": false, + "reason": "MLA q_a_layernorm precision issue and DSA indexer ReLU issue are unrelated." + }, + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "Specific processor.save_pretrained missing files versus a generic bug report." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading problem and TensorFlow backend error are different issues." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": false, + "reason": "Lazy loading failure and unsafe sys.modules access are related at most broadly, not the same concrete bug." + }, + { + "left": "issue:44393", + "right": "issue:45753", + "accept": false, + "reason": "Different Qwen3_5VL behaviors: bounding-box output errors vs an infinite loop on one image." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "Model output error and PretrainedConfig type checking are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Broken env command and checkpoint reload incompatibility are different code paths." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Both touch Qwen/moe tensor-parallel space, but one is a model bug and the other is a missing tp plan; not the same fix." + }, + { + "left": "issue:43525", + "right": "issue:45850", + "accept": false, + "reason": "Missing pad_token_id in Llama4Config is a specific config bug, not the broad v5 complaint." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import-star performance cleanup and chunked-prefill cache missing are unrelated." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Both are checkpoint compatibility regressions, but they affect different models and different v5 mechanisms." + }, + { + "left": "issue:44492", + "right": "issue:45850", + "accept": false, + "reason": "A cache-strategy typo is not the same as a general v5 breakage report." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 API guidance and cache strategy typo are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:45850", + "accept": false, + "reason": "StableLmConfig pad_token_idx regression is specific and not the same as a vague broad breakage report." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "AutoProcessor kwargs forwarding and inverse_sqrt scheduler kwargs are different functions and bugs." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "A flaky test seeding issue is unrelated to scheduler kwargs handling." + }, + { + "left": "issue:45278", + "right": "issue:45850", + "accept": false, + "reason": "Both are broad complaints about v5 regressions, but they are not the same concrete bug report." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Slow imports and broken lazy loading may overlap thematically, but there is no evidence they stem from the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 46, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:31795", + "issue:33453", + "issue:34689", + "issue:37428", + "issue:41669", + "issue:42371", + "issue:42913", + "issue:43066", + "issue:43352", + "issue:43408", + "issue:43502", + "issue:43519", + "issue:43525", + "issue:43576", + "issue:43606", + "issue:43618", + "issue:43704", + "issue:43827", + "issue:43874", + "issue:43976", + "issue:44077", + "issue:44261", + "issue:44263", + "issue:44273", + "issue:44295", + "issue:44297", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44393", + "issue:44485", + "issue:44492", + "issue:44623", + "issue:44655", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:44945", + "issue:44998", + "issue:45071", + "issue:45092", + "issue:45230", + "issue:45290", + "issue:45310", + "issue:45341", + "issue:45468", + "issue:45478", + "issue:45850" + ], + "soft_pairs": [ + "issue:43408|issue:44315", + "issue:43976|issue:44393", + "issue:43352|issue:44393", + "issue:43576|issue:44908", + "issue:44297|issue:45230", + "issue:44908|issue:45071", + "issue:44295|issue:45230", + "issue:41669|issue:44351", + "issue:41669|issue:45341", + "issue:44492|issue:44945", + "issue:43874|issue:45850", + "issue:44393|issue:45478", + "issue:44077|issue:44861", + "issue:43827|issue:44908", + "issue:31795|issue:33453", + "issue:44315|issue:44829", + "issue:41669|issue:44908", + "issue:43576|issue:44273", + "issue:44655|issue:44861", + "issue:44360|issue:45468", + "issue:45230|issue:45310", + "issue:44273|issue:44908", + "issue:43519|issue:44485", + "issue:43519|issue:44908", + "issue:41669|issue:43704", + "issue:31795|issue:34689", + "issue:31795|issue:37428", + "issue:44315|issue:45092", + "issue:44485|issue:45468", + "issue:43704|issue:44485", + "issue: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:44492|issue:44998", + "issue:44998|issue:45850", + "issue:43618|issue:44998", + "issue:44998|issue:45341", + "issue:43525|issue:44861" + ], + "prepared_review_unit_hash": "ba1a4d3ecc1451cc45a577328a41faf5102d85b1c2d3dc3fd1d1849cb6cb6e23", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 46, + "serialized_chars": 40832, + "estimated_input_tokens": 10208, + "estimated_eval_tokens": 20672 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose collection of closed Transformer v5 regression reports, tokenizer/model-loading issues, and a few model-specific bugs. I don\u2019t see any strong duplicate pairs; the soft matches are mostly broad topical similarity rather than the same underlying defect.", + "confidence": 0.76, + "canonical_issue_reason": "Issue 45850 is the broadest umbrella report in the set: it captures the general theme of Transformers v5 breaking multiple models, which makes it the best top-level canonical bucket for this otherwise fragmented cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45850 is the most suitable global representative because it is broad enough to subsume many of the v5 regression-style reports, even though it is less concrete than some of the model-specific issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "Both mention model creation/configuration, but they are different problems: a SAM3 model-type mismatch warning vs. Liger Kernel application with model_init." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Unrelated failures: Python-version compatibility vs. Qwen3-VL 2D bounding-box output behavior." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Different model families and different failure modes; Flash Attention 2.0 support is not the same as Qwen3-VL output issues." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "CLI/env-command breakage and scheduler kwargs handling are unrelated code paths." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "One is a tokenizer_class metadata mismatch on save_pretrained; the other is only a generic bug report title." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwargs bug and PretrainedConfig type-checking regression are different issues." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "Position_ids buffer access and a generic bug report are not enough to establish a duplicate." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import * cleanup/performance and missing HybridCache import are different defects." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Broad import cleanup issue vs. a small testing_utils bug; not the same change or bug." + }, + { + "left": "issue:44492", + "right": "issue:44945", + "accept": false, + "reason": "Cache-strategy typo and pipeline-parallelism output corruption are unrelated." + }, + { + "left": "issue:43874", + "right": "issue:45850", + "accept": false, + "reason": "A specific GLM46V image-processor method is missing; 45850 is just a broad complaint about many models breaking." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Different Qwen model variants and different errors; no shared concrete code path is evident." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "Optional post_init enforcement and tied-weight-key list handling are separate model-init validations." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs still referencing pipeline() and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Documentation confusion for model.forward is unrelated to tokenizer loading regression." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel application with model_init is unrelated to flash_attention_3 causing degenerate training." + }, + { + "left": "issue:41669", + "right": "issue:44908", + "accept": false, + "reason": "Import cleanup/performance issue and scheduler kwargs bug do not share the same underlying defect." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Broken env command and lazy-loading behavior are unrelated." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Saving Pipeline objects and tied-weight-key handling are different save/load paths." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "DSA indexer ReLU discussion and Gemma-4 audio positional encoding are unrelated model internals." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "A generic bug report title is not enough to treat it as the same as the specific Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading and inverse_sqrt scheduler kwargs are unrelated." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Qwen3VL timestamp calculation and GLM-5 RoPE implementation are different multimodal/math bugs." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Timestamp calculation in a processor is unrelated to scheduler kwargs handling." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import * usage and VRAM leak under multiple dataloader threads are unrelated." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Model.forward docs confusion and Llama 3.2 Vision loading regression do not describe the same bug." + }, + { + "left": "issue:31795", + "right": "issue:37428", + "accept": false, + "reason": "Documentation confusion is unrelated to the flash_attention_utils import error." + }, + { + "left": "issue:44315", + "right": "issue:45092", + "accept": false, + "reason": "Both involve model instantiation, but one is Liger Kernel/model_init and the other is remote-code checkpoint/meta-init compatibility." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "Different model families and different algorithmic areas; no duplicate evidence." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "VRAM leak in dataloader threads is unrelated to GLM-5 RoPE behavior." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 API migration and VRAM leak are unrelated." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "local_files_only network leakage and timestamp calculation are different bugs." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Processor timestamp math and processor.save_pretrained missing files are different save/process flows." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Timestamp calculation bug and VRAM leak are unrelated." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "torch.split return handling in GlmMoeDsaIndexer is unrelated to Gemma-4 audio positional encoding." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch and Gemma-4 audio positional encoding are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "MLA q_a_layernorm epsilon precision discussion does not match Gemma-4 audio positional encoding." + }, + { + "left": "issue:43704", + "right": "issue:45341", + "accept": false, + "reason": "VRAM leak under multithreaded dataloaders and a testing_utils bug are not the same issue." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Broad v4-to-v5 tokenizer behavior change and a specific wrong tokenizer decoder type are related by theme but not clearly the same underlying bug." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Inverse_sqrt scheduler kwargs and an irrelevant 'Unemployment' issue are not comparable." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Tied-weight-key AttributeError and apply_chat_template crashes on tool-call assistant messages are different code paths." + }, + { + "left": "issue:44492", + "right": "issue:44998", + "accept": false, + "reason": "Cache-strategy typo and unrelated placeholder issue are not duplicates." + }, + { + "left": "issue:44998", + "right": "issue:45850", + "accept": false, + "reason": "The placeholder 'Unemployment' issue has no substantive relation to broad v5 model breakage." + }, + { + "left": "issue:43618", + "right": "issue:44998", + "accept": false, + "reason": "CLIPOutput attentions regression and the placeholder issue are unrelated." + }, + { + "left": "issue:44998", + "right": "issue:45341", + "accept": false, + "reason": "The placeholder issue is unrelated to the testing_utils bug." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Llama4Config pad_token_id AttributeError and tied-weight-key list handling are separate configuration bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 47, + "review_unit_count": 47, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-399", + "nodes": [ + "issue:30333", + "issue:31515", + "issue:31795", + "issue:33453", + "issue:36296", + "issue:38617", + "issue:43298", + "issue:43572", + "issue:44230", + "issue:44623", + "issue:44861", + "issue:45310" + ], + "soft_pairs": [ + "issue:43572|issue:44861", + "issue:31515|issue:38617", + "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": "a519aefdec659341cd4b0db535d79bb785e47f909078d7b5647dec77d6c8ed12", + "component_budget": { + "node_count": 399, + "item_count": 399, + "soft_pair_count": 1495, + "serialized_chars": 770922, + "estimated_input_tokens": 192731, + "estimated_eval_tokens": 385718 + }, + "budget": { + "node_count": 12, + "item_count": 12, + "soft_pair_count": 9, + "serialized_chars": 9480, + "estimated_input_tokens": 2370, + "estimated_eval_tokens": 4996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the issues cover MLflow reporting, checkpoint/tokenizer loading, docs, tensor-parallel training, import errors, audio batching, fp8 support, processor saving, and tied-weight handling. I don\u2019t see a single underlying bug/change that ties them together, so there is no true duplicate anchor here.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue #45310 is the strongest standalone bug report in the set: it names a concrete version regression (transformers>=5.4.0), a specific model family (Qwen3.5 MoE), and a reproducible from_pretrained failure.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "Different bugs: StableLmConfig pad_token_idx regression vs AttributeError in _get_tied_weight_keys. No shared code path or symptom." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "Checkpoint loading slowness and an ImportError for layer_type_validation are unrelated issues." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "fp8 inference support for Qwen models is a feature gap, while processor.save_pretrained missing files is a serialization bug." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLflow job-status reporting after failed training is unrelated to slow from_pretrained checkpoint loading." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "These concern different subsystems and failure modes: pretrained checkpoint load performance vs tensor-parallel training correctness." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Loading performance and documentation confusion are not the same underlying bug." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "Both mention Qwen3.5/MoE, but one is fp8 support for inference and the other is a from_pretrained regression; the concrete bugs differ." + }, + { + "left": "issue:31515", + "right": "issue:33453", + "accept": false, + "reason": "Slow checkpoint loading and tokenizer loading regression are distinct problems." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Continuous batching not supporting audio models is unrelated to fp8 Qwen model inference support." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-13244-3", + "nodes": [ + "issue:13244", + "issue:15980", + "issue:16618" + ], + "soft_pairs": [ + "issue:13244|issue:16618", + "issue:15980|issue:16618" + ], + "prepared_review_unit_hash": "2e82dc6d54a8e348c9dc5fb5e46180b2c8ebcf0b48f4de29a93bcbcee8d342e5", + "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": "These three issues do not appear to be duplicates as a cluster: one is about Tapas tokenization, and the other two are separate model-loading/authentication complaints. No soft edge is strong enough to merge.", + "confidence": 0.87, + "canonical_issue_reason": "Issue 16618 is the broadest and most central model-loading report in the set, while 15980 is narrowly about private-model login/error messaging and 13244 is unrelated tokenization.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 16618 is the best representative of the cluster\u2019s model-loading side because it is the most general user-facing failure report among the candidates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Different problem area: Tapas tokenization mismatch versus failure to load bert-base-uncased." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "Related broadly to model downloads, but one is about private-model login/error messaging and the other is a generic load failure for a public model; not the same concrete bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 3, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-26350-2", + "nodes": [ + "issue:26350", + "issue:33260" + ], + "soft_pairs": [ + "issue:26350|issue:33260" + ], + "prepared_review_unit_hash": "121743d6490668e44bab32e9eb9dbbb8fa72664684a10e99cf4764e18986d96d", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1590, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1590, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two separate feature requests with only superficial similarity: one asks for Flash Attention 2 support on more architectures, the other for GGUF support on more architectures. They do not appear to describe the same underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:26350 is the more established and discussion-heavy request, but it is not a duplicate of issue:33260.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Both are community contribution requests about broad architecture support, but they target different features and code paths (Flash Attention 2 vs GGUF). They are not the same underlying issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 4, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-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": "f5ca49f3633697e29c1b1e7a22157f7e05c0802598201785f03dc211be87f4b5", + "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 translation-doc request issues with very similar template text, but they appear to target different language requests rather than one identical underlying change. The Turkish issue is the most concrete and active representative.", + "confidence": 0.71, + "canonical_issue_reason": "Issue 27088 is the only concrete language-specific request ([i18n-TR] Turkish) and has the most discussion activity, so it is the best representative of this cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "27088 is the strongest issue candidate because it is specific, oldest, and has the most engagement; the others look like template-like variants rather than a clearer canonical target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Same translation-request theme, but the targets are not concretely the same; this is not enough to treat them as one underlying issue." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "The Turkish request and the generic placeholder request do not clearly refer to the same language change." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "Similar template wording, but no evidence they are the same exact translation request or language target." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": false, + "reason": "Very similar template text, but still only generic translation-request similarity; no concrete shared target to justify duplication." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": false, + "reason": "These look like near-template issues, but the available evidence does not show the same specific request." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": false, + "reason": "Close textual match, but not enough evidence that they are the same underlying translation 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": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-36817-15", + "nodes": [ + "issue:36817", + "issue:36879", + "issue:37015", + "issue:41115", + "issue:41211", + "issue:42503", + "issue:42738", + "issue:43671", + "issue:43837", + "issue:43909", + "issue:44121", + "issue:44205", + "issue:45027", + "issue:45177", + "issue:45322" + ], + "soft_pairs": [ + "issue:36817|issue:41211", + "issue:36817|issue:43671", + "issue:36817|issue:36879", + "issue:36879|issue:41211", + "issue:36879|issue:43909", + "issue:36879|issue:44121", + "issue:36879|issue:43837", + "issue:36879|issue:43671", + "issue:37015|issue:43837", + "issue:37015|issue:44121", + "issue:37015|issue:43909", + "issue:36879|issue:37015", + "issue:37015|issue:42503", + "issue:41115|issue:43837", + "issue:41115|issue:44121", + "issue:41115|issue:42503", + "issue:41115|issue:45027", + "issue:41211|issue:43671", + "issue:42503|issue:43837", + "issue:42503|issue:43671", + "issue:42503|issue:42738", + "issue:42738|issue:45027", + "issue:43671|issue:43837", + "issue:43671|issue:45027", + "issue:43671|issue:45177", + "issue:43671|issue:44121", + "issue:43671|issue:43909", + "issue:43837|issue:43909", + "issue:43837|issue:45322", + "issue:43837|issue:44121", + "issue:43909|issue:44205", + "issue:44121|issue:45322", + "issue:43837|issue:45027", + "issue:37015|issue:45027", + "issue:43837|issue:45177" + ], + "prepared_review_unit_hash": "38cf906479d70fe7c9213c436fabdb3fbdbb274ae9e7aa9a7031b7a4fce5f0f8", + "component_budget": { + "node_count": 15, + "item_count": 15, + "soft_pair_count": 35, + "serialized_chars": 20711, + "estimated_input_tokens": 5178, + "estimated_eval_tokens": 10612 + }, + "budget": { + "node_count": 15, + "item_count": 15, + "soft_pair_count": 35, + "serialized_chars": 20711, + "estimated_input_tokens": 5178, + "estimated_eval_tokens": 10612 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is a loose set of model-support request issues that only share a generic \"add model\" pattern. The titles span unrelated architectures and modalities (text, audio, vision, detection, OCR, sparse transformers), so they should not be treated as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "If a single issue must be chosen, issue #36817 is the broadest and most template-like open request ('Add EuroBert Model To Config'), but it is still only one specific model request and does not truly cover the rest of the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "#36817 is the best representative only in the sense that it is the most generic/open config-addition request; however, the cluster is too heterogeneous for any issue to serve as a real duplicate umbrella.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "EuroBert vs DEIMv2 are different model families with different support paths." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "EuroBert text model request is unrelated to Qwen3-TTS speech generation support." + }, + { + "left": "issue:36817", + "right": "issue:36879", + "accept": false, + "reason": "EuroBert NLP config addition and RF-DETR object detection are different underlying changes." + }, + { + "left": "issue:36879", + "right": "issue:41211", + "accept": false, + "reason": "RF-DETR and DEIMv2 are distinct detection models; not the same concrete fix." + }, + { + "left": "issue:36879", + "right": "issue:43909", + "accept": false, + "reason": "RF-DETR detection support is unrelated to LFM2.5 Audio model support." + }, + { + "left": "issue:36879", + "right": "issue:44121", + "accept": false, + "reason": "RF-DETR model support and weight-sparse transformer support are unrelated code paths." + }, + { + "left": "issue:36879", + "right": "issue:43837", + "accept": false, + "reason": "RF-DETR vision detection and Qwen3-ASR speech recognition are different tasks." + }, + { + "left": "issue:36879", + "right": "issue:43671", + "accept": false, + "reason": "RF-DETR and Qwen3-TTS are different modalities and model integrations." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT language-model support is unrelated to Qwen3-ASR audio transcription support." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT and the sparse-transformer request target different architectures and fixes." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "NeoBERT text model support is unrelated to LFM2.5 Audio support." + }, + { + "left": "issue:36879", + "right": "issue:37015", + "accept": false, + "reason": "RF-DETR and NeoBERT are different model families, not one bug/change." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT and ModernVBERT are distinct model additions even if both are BERT-like." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 architecture support is unrelated to Qwen3-ASR support." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 model architecture work does not match sparse-transformer support." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "MiniCPM3 and ModernVBERT are different model integrations." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "MiniCPM3 and Voxtral TTS support are different tasks/code paths." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 and Qwen3-TTS are unrelated model requests." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT and Qwen3-ASR are different models and modalities." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT text model support is unrelated to Qwen3-TTS audio generation." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "ModernVBERT and BERT-like models with RoPE are not the same concrete request." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "BERT-like RoPE support and Voxtral TTS support are unrelated changes." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "Qwen3-TTS and Qwen3-ASR share a name but fix different speech tasks and code paths." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Two different TTS model integrations; not enough to treat as one duplicate." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-TTS and DeepSeek-OCR2 are different model domains." + }, + { + "left": "issue:43671", + "right": "issue:44121", + "accept": false, + "reason": "Speech generation support is unrelated to the sparse-transformer request." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS and LFM2.5 Audio are distinct audio model integrations." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "ASR vs audio-model support are different concrete changes." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "Qwen3-ASR and EUPE are different model classes and implementations." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Speech recognition support is unrelated to the sparse-transformer model request." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "LFM2.5 Audio and SAM3-LiteText are different modalities and code paths." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Weight-sparse transformer support and EUPE support are unrelated architecture requests." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral-4B-TTS are different speech tasks." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT text-model support is unrelated to Voxtral TTS support." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "ASR support and OCR model support are different underlying changes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 6, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-37934-6", + "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": "72ac5eb4b655a5d95bf0045e0062b24c1045ca5c59c471d4e1041bd7cb42177c", + "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": "These issues are not duplicates of one another; they span unrelated bugs across different models/configs, a tokenizer regression, and a typo. The soft similarity is superficial rather than indicating the same underlying fix.", + "confidence": 0.97, + "canonical_issue_reason": "No true canonical issue: the set is heterogeneous (different models, different failure modes, and one pure typo/type-hint issue), so there isn\u2019t a single underlying bug to centralize.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43240 is the broadest functional bug and has inbound references, so it is the most representative placeholder if one must be chosen, though this cluster should not be treated as a duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "Different subjects: Llama4TextL2Norm/RMS norm question vs. ModernBERTConfig type-hint issue. No shared code-path or fix." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Both are model-related bugs, but one is about Mask2Former ignore_value handling and the other about kwargs not reaching loss computation. Distinct defects." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "Mask2Former ignore_value behavior is unrelated to a tokenizer regex warning regression." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "Model ignore_value handling and a typo in a quantization attribute are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "Mask2Former loss/ignore_value bug vs. a ModernBERT config type annotation problem; not the same bug." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Passing kwargs to loss calculation is a runtime behavior issue; incorrect norm_eps typing is a separate config metadata issue." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "ModernBERTConfig type-hint correction and a tokenizer warning regression are different problems in different areas." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "A config type-hint bug is not the same as a typo in a quantization_operation attribute." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 7, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41084-2", + "nodes": [ + "issue:41084", + "issue:43037" + ], + "soft_pairs": [ + "issue:41084|issue:43037" + ], + "prepared_review_unit_hash": "1a3ee235d7758d2ca8b665178cadcfb1fd068bba138fa1e52b07a45689b66c0f", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are unrelated: one is about block decoding, while the other is a feature request for DeepSeek v3.2 support. There is no duplicate relationship here and no sensible single canonical item for the cluster.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "Titles and likely scopes do not match: 'Set Block Decoding' suggests a decoding/configuration bug or feature, while 'DeepSeek v3.2 support' is an LLM compatibility request. No shared target or concrete code-path overlap is evident." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 8, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43598-2", + "nodes": [ + "issue:43598", + "issue:43600" + ], + "soft_pairs": [ + "issue:43598|issue:43600" + ], + "prepared_review_unit_hash": "60adfd38b83ae2ddf1b843b2a3648e16e34f2aa3557860055d88cdddf23b647f", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both items are about Trainer customizability, but they describe different problems: one is about `remove_unused_column`, the other about FLOPS estimation for custom trainers. Similar subsystem, not the same underlying bug.", + "confidence": 0.96, + "canonical_issue_reason": "issue #43598 is the more concrete and actionable problem statement, centered on a specific Trainer behavior (`remove_unused_column`) affecting custom trainers.", + "canonical_pr_reason": null, + "best_issue_reason": "issue #43598 is the stronger anchor for the cluster because it is narrower and more clearly tied to a single code path.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "These are related only at a high level through custom Trainer behavior. One concerns column pruning customizability; the other concerns FLOPS estimation. They would not be merged as 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": 9, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43684-2", + "nodes": [ + "issue:43684", + "issue:44405" + ], + "soft_pairs": [ + "issue:43684|issue:44405" + ], + "prepared_review_unit_hash": "7c42814c8935a3309a29d91343eb018190d36bbc0edcf471fbc72c6d13833df9", + "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 not duplicates: one requests broad AutoModel/AutoModelForConditionalGeneration registration for Qwen3-Omni, while the other requests AutoModelForSequenceClassification support for Qwen3.5. They concern different model families and different task heads.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Different underlying requests: general model registration for Qwen3-Omni versus sequence-classification support for Qwen3.5/Qwen3_5Config. They are not the same bug/change and would not plausibly be merged into one issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 10, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43835-4", + "nodes": [ + "issue:43835", + "issue:43845", + "issue:43846", + "issue:43859" + ], + "soft_pairs": [ + "issue:43835|issue:43845", + "issue:43835|issue:43846", + "issue:43835|issue:43859", + "issue:43845|issue:43846", + "issue:43845|issue:43859", + "issue:43846|issue:43859" + ], + "prepared_review_unit_hash": "104a0e5a3cf03a205c228facfa7e74dc1fe6491630323f8ea30909a2428551dd", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Three issues share the same generic title \"huggingface\" and appear to be duplicates of each other; issue 43835 ('Fett') looks unrelated. I would keep 43845 as the canonical issue among the huggingface duplicates.", + "confidence": 0.68, + "canonical_issue_reason": "43845 is the earliest of the near-identical \"huggingface\" issues and is the best representative of that duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "43845 is the strongest issue candidate because it matches the duplicate title pattern and predates the other matching issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "Different titles and no evidence they describe the same underlying problem." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "Different titles and no concrete shared bug description." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "Different titles and no concrete shared bug description." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": true, + "reason": "Same title, very close creation times, and no distinguishing details; they look like duplicate reports of the same issue." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": true, + "reason": "Same title and similar sparse content strongly suggest the same underlying report." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": true, + "reason": "Same title and similarly minimal content indicate a duplicate issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet: 43835 has a distinct title, while 43845/43846/43859 share the same near-identical title and are plausibly a duplicate set. The canonical choice of 43845 is conservative because it is the earliest of that trio. The soft-edge verdicts also stay within the evidence available." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 11, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44371-5", + "nodes": [ + "issue:44371", + "issue:44418", + "issue:44572", + "issue:44573", + "issue:44937" + ], + "soft_pairs": [ + "issue:44371|issue:44572", + "issue:44371|issue:44573", + "issue:44371|issue:44418", + "issue:44371|issue:44937", + "issue:44418|issue:44572", + "issue:44418|issue:44573", + "issue:44572|issue:44573", + "issue:44572|issue:44937", + "issue:44573|issue:44937" + ], + "prepared_review_unit_hash": "e22f90c55f83e5e3c40c265d1d9b89169994dd1607f7092fc9e6e6fd158be0f2", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5785, + "estimated_input_tokens": 1447, + "estimated_eval_tokens": 3150 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5785, + "estimated_input_tokens": 1447, + "estimated_eval_tokens": 3150 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "No convincing duplicate cluster here: the issues are mostly spam/generic placeholders, and the one specific title does not match the others on a concrete bug or change. I rejected all soft edges.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:44937 is the most specific and actionable title in the set, so it is the least-bad representative if one must be chosen.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44371", + "right": "issue:44572", + "accept": false, + "reason": "Both are generic spam placeholders; no evidence of the same underlying issue." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Both are generic spam placeholders; no shared concrete bug or change." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "Placeholder spam vs. a vague docs suggestion; too little evidence to be duplicates." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder and a specific 'Google LLC' topic are not the same issue." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "Vague documentation suggestion vs. spam placeholder; no concrete overlap." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "Vague documentation suggestion vs. spam placeholder; no concrete overlap." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Both are spam placeholders, but there is no evidence they describe the same underlying bug." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder and a specific title about Google LLC do not match on a concrete problem." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder and a specific title about Google LLC do not match on a concrete problem." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 12, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45145-2", + "nodes": [ + "issue:45145", + "issue:45151" + ], + "soft_pairs": [ + "issue:45145|issue:45151" + ], + "prepared_review_unit_hash": "312c1d76aa102015b6de5527a4e60ad1d7c93ab55ca8d00e8b4982a7430ba7bc", + "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 look like likely duplicates: they share an identical title, similar body sizes, and a soft-similarity signal. #45145 is the better canonical anchor because it has one inbound reference, but the evidence stops short of proving the bodies are effectively identical.", + "confidence": 0.83, + "canonical_issue_reason": "#45145 is the better anchor because it has the same title as #45151 and also has an inbound reference, which makes it the more established copy for deduplication.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the strongest issue representative since it has the same title as #45151 and slightly stronger linkage via inbound references.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": true, + "reason": "Same title and similar body size, plus a soft-similarity signal, make these plausibly the same underlying report, though the match is not definitive." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The pair looks plausibly duplicate, but the analyst is a bit too certain for the evidence provided. The packet only shows identical titles, similar body lengths, one inbound reference, and a soft-similarity signal; it does not directly establish that the bodies are \u201chighly similar,\u201d \u201ceffectively identical,\u201d or that both are definitively the same report. The canonical choice based on inbound reference is reasonable, but the confidence and duplicate wording are somewhat overstated." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45295-2", + "nodes": [ + "issue:45295", + "issue:45373" + ], + "soft_pairs": [], + "prepared_review_unit_hash": "f5fd95fada2f530ab0999e343c17c16266cde0c7182bb22509972ca44827cd49", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues describe the same missing Gemma 4 sequence-classification support; the later, more specific open issue is the better cluster representative.", + "confidence": 0.94, + "canonical_issue_reason": "issue #45373 is the clearest canonical tracker: it is open, explicitly names the missing class `Gemma4ForSequenceClassification`, and states the module-level gap relative to Gemma 2/3.", + "canonical_pr_reason": null, + "best_issue_reason": "issue #45373 best captures the concrete request and likely remaining work, while #45295 is a shorter earlier duplicate with less detail and is already closed.", + "best_pr_reason": null, + "soft_edge_verdicts": [] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 14, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45672-2", + "nodes": [ + "issue:45672", + "issue:45674" + ], + "soft_pairs": [ + "issue:45672|issue:45674" + ], + "prepared_review_unit_hash": "18c5850ebed6226e67fa8d1c49fe56e7e6326ace503057a887728e8ac41e3cc1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1668, + "estimated_input_tokens": 417, + "estimated_eval_tokens": 1090 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1668, + "estimated_input_tokens": 417, + "estimated_eval_tokens": 1090 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two separate 4-bit inference bugs. One is a Gemma4/audio-module dtype issue with `torch.finfo()` on `uint8` weights; the other is a BitsAndBytesConfig regression where `llm_int8_skip_modules` drops the default `lm_head` exclusion and triggers an assertion. They share the broad area of quantized inference but not the same underlying defect.", + "confidence": 0.97, + "canonical_issue_reason": "Issue #45674 is the better cluster representative because it describes a concrete, configuration-driven regression with a clear failure mode and likely wider applicability across models using BitsAndBytesConfig.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #45674 is more suitable as the primary issue for this cluster since it is a crisp, reproducible bug report about a specific regression path. Issue #45672 is a model-specific dtype failure and not the same problem.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45672", + "right": "issue:45674", + "accept": false, + "reason": "Different code paths and failure mechanisms: Gemma4/audio-module `torch.finfo()` on `uint8` vs. BitsAndBytesConfig `llm_int8_skip_modules` removing default `lm_head` exclusion. Shared 4-bit inference context is too broad to treat as a duplicate." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 15, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45762-2", + "nodes": [ + "issue:45762", + "issue:45763" + ], + "soft_pairs": [ + "issue:45762|issue:45763" + ], + "prepared_review_unit_hash": "a68e0224ec8852e86540007d82f8bf05fa84444146295384145b97c018669537", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are unrelated closed records with different personal-name titles and no substantive overlap; they should not be clustered as duplicates.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45762", + "right": "issue:45763", + "accept": false, + "reason": "Different people and dates in the titles, with no body text, links, targets, or shared context indicating the same underlying bug or request." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 16, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45764-2", + "nodes": [ + "issue:45764", + "issue:45765" + ], + "soft_pairs": [ + "issue:45764|issue:45765" + ], + "prepared_review_unit_hash": "3b7ce52105dcfbe9367fcef09c1789a9c13831341046ddf82ee6338636f760d7", + "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": "Two closed issues with only loosely similar names and dates in the titles, but no body, no shared targets, and no evidence of the same underlying bug or request. They do not look like duplicates beyond superficial name similarity.", + "confidence": 0.97, + "canonical_issue_reason": "Neither issue has substantive content; if one must be chosen as the cluster representative, issue #45764 is marginally earlier and just as content-free as #45765.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #45764 is the best representative only by timestamp; there is no technical signal to distinguish it from #45765.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45764", + "right": "issue:45765", + "accept": false, + "reason": "The titles are different personal names with dates, and there is no body, code target, or discussion evidence tying them to 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": 17, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45869-5", + "nodes": [ + "issue:45869", + "issue:45870", + "issue:45871", + "issue:45872", + "issue:45873" + ], + "soft_pairs": [ + "issue:45869|issue:45870", + "issue:45869|issue:45871", + "issue:45870|issue:45871", + "issue:45870|issue:45872", + "issue:45870|issue:45873", + "issue:45871|issue:45872", + "issue:45871|issue:45873", + "issue:45872|issue:45873" + ], + "prepared_review_unit_hash": "cc1ebd6e5a56fb6a8fdf8bb2139092a6529f99e85fdb299e9a7041af46685f4a", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5724, + "estimated_input_tokens": 1431, + "estimated_eval_tokens": 3118 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5724, + "estimated_input_tokens": 1431, + "estimated_eval_tokens": 3118 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster is not one single duplicate set: it contains two exact duplicate issue pairs (45870/45871 and 45872/45873) plus an unrelated dataset-promo issue (45869). The cross-pair matches are only broad topical similarity around financial data/DataForge, not the same underlying request.", + "confidence": 0.96, + "canonical_issue_reason": "There is no single canonical issue for the whole cluster because it splits into two separate duplicate pairs with different asks. If one representative must be chosen, 45872 is the most detailed and specific of the DataForge pair, but it should not absorb the other unrelated pair.", + "canonical_pr_reason": null, + "best_issue_reason": "45872 is the strongest representative within the cluster because it is a concrete, specific request and has an exact duplicate in 45873. However, it only represents one of the two duplicate subclusters, not the entire cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45869", + "right": "issue:45870", + "accept": false, + "reason": "Different artifacts: one is a financial dataset promo, the other is a DataForge example request. Similar keywords only, not the same issue." + }, + { + "left": "issue:45869", + "right": "issue:45871", + "accept": false, + "reason": "Different artifacts: one is a financial dataset promo, the other is a DataForge example request. Similar keywords only, not the same issue." + }, + { + "left": "issue:45870", + "right": "issue:45871", + "accept": true, + "reason": "Exact duplicate: same title and apparent same request, so they are the same underlying issue." + }, + { + "left": "issue:45870", + "right": "issue:45872", + "accept": false, + "reason": "Both mention financial time series/DataForge, but one is an example request and the other adds a scraping service component. Related topic, different underlying issue." + }, + { + "left": "issue:45870", + "right": "issue:45873", + "accept": false, + "reason": "Both mention financial time series/DataForge, but one is an example request and the other adds a scraping service component. Related topic, different underlying issue." + }, + { + "left": "issue:45871", + "right": "issue:45872", + "accept": false, + "reason": "Both mention financial time series/DataForge, but one is an example request and the other adds a scraping service component. Related topic, different underlying issue." + }, + { + "left": "issue:45871", + "right": "issue:45873", + "accept": false, + "reason": "Both mention financial time series/DataForge, but one is an example request and the other adds a scraping service component. Related topic, different underlying issue." + }, + { + "left": "issue:45872", + "right": "issue:45873", + "accept": true, + "reason": "Exact duplicate: same title and apparent same request, so they are the same underlying issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet: 45870/45871 and 45872/45873 are exact duplicate pairs, while 45869 is only softly related by topic. The cross-pair rejections are appropriately conservative, and the canonical-issue note correctly avoids collapsing two different asks into one cluster-wide representative." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 1, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-39895-16", + "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", + "pull_request:45899", + "pull_request:45919", + "pull_request:45922" + ], + "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:45919", + "pull_request:43424|pull_request:45899", + "pull_request:43424|pull_request:44827", + "pull_request:43424|pull_request:43823", + "pull_request:43823|pull_request:45655", + "pull_request:43823|pull_request:45922", + "pull_request:43823|pull_request:43995", + "pull_request:43995|pull_request:45213", + "pull_request:43995|pull_request:44827", + "pull_request:43995|pull_request:45899", + "pull_request:44827|pull_request:45899", + "pull_request:44827|pull_request:45170", + "pull_request:44827|pull_request:45795", + "pull_request:44827|pull_request:45655", + "pull_request:45170|pull_request:45899", + "pull_request:44827|pull_request:45213", + "pull_request:43823|pull_request:45213", + "pull_request:45213|pull_request:45899", + "pull_request:45795|pull_request:45899", + "pull_request:43823|pull_request:45795", + "pull_request:43823|pull_request:45899", + "pull_request:44827|pull_request:45922", + "pull_request:43251|pull_request:43254" + ], + "prepared_review_unit_hash": "8b6de8f398270997940292a09e62c61e2afdedf856a8210ece612a16d77998f8", + "component_budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 29, + "serialized_chars": 46661, + "estimated_input_tokens": 11666, + "estimated_eval_tokens": 23588 + }, + "budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 29, + "serialized_chars": 46661, + "estimated_input_tokens": 11666, + "estimated_eval_tokens": 23588 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a set of unrelated PRs that only overlap at a broad topic level (new-model additions, docs, tokenizer utilities, etc.). The only true duplicate-looking pair is the two `fixed_cross_entropy` PRs, which both change the same function to forward supported `cross_entropy` kwargs for the same issue.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43251 is the best canonical PR for the duplicate pair because it targets the shared issue `#43240`, patches the same `fixed_cross_entropy` code path, and appears to be the more issue-aligned version of the same fix.", + "best_issue_reason": null, + "best_pr_reason": "PR #43251 is the best representative PR in this cluster because it is the only clearly duplicated change and the one explicitly linked to the underlying bug report; the rest of the cluster is not duplicate-suitable.", + "soft_edge_verdicts": [ + { + "left": "pull_request:39895", + "right": "pull_request:43424", + "accept": false, + "reason": "Both are unrelated model/runtime PRs; one adds VideoPrism, the other adds an Executorch dynamic-shape test." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Tokenizer method additions for MistralCommon are unrelated to adding MobileLLM." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Mistral tokenizer cleanup and Executorch export testing fix different code paths." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor auto-loading robustness is not the same change as adding a new model implementation." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "Loss-function kwargs forwarding is unrelated to adding MobileLLM." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "A cross-entropy helper fix and Mistral4 test/model cleanup are different bugs." + }, + { + "left": "pull_request:43424", + "right": "pull_request:45919", + "accept": false, + "reason": "Executorch dynamic-shape testing is unrelated to adding Sapiens2." + }, + { + "left": "pull_request:43424", + "right": "pull_request:45899", + "accept": false, + "reason": "Executorch export tests and continuous-batching docs cover different concerns." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "Dynamic-shape export tests do not match the Mistral4 model/test fix." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "Executorch export testing is unrelated to MobileLLM model addition." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "Adding MobileLLM is unrelated to fixing `cls.config` resolution order." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45922", + "accept": false, + "reason": "MobileLLM model scaffolding and vision-model lru-cache leak fixes are unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "Adding a new model is not the same as refactoring Falcon output collection behavior." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon refactoring and a model-creation skill are unrelated changes." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Falcon output-interface refactoring does not match Mistral4 test fixes." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45899", + "accept": false, + "reason": "Falcon model internals and continuous-batching documentation are not the same underlying change." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45899", + "accept": false, + "reason": "Mistral4 test/model fixes do not overlap with a docs-only continuous-batching update." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 fixes are unrelated to the CLIP-like `layrnorm` spelling correction." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45795", + "accept": false, + "reason": "Mistral4 tests/model adjustments are unrelated to audio/video processor docs." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "Mistral4 changes are unrelated to fixing config resolution order." + }, + { + "left": "pull_request:45170", + "right": "pull_request:45899", + "accept": false, + "reason": "A weight-renaming/model spelling fix is unrelated to continuous-batching documentation." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 model/test fixes are unrelated to the model-creation skill PR." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "Adding MobileLLM is not the same as introducing a model creation skill." + }, + { + "left": "pull_request:45213", + "right": "pull_request:45899", + "accept": false, + "reason": "A model-creation skill and decode-fast-path docs are unrelated." + }, + { + "left": "pull_request:45795", + "right": "pull_request:45899", + "accept": false, + "reason": "Docs for audio/video processors do not duplicate decode-fast-path documentation." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45795", + "accept": false, + "reason": "MobileLLM model code and processor docs are different changes." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45899", + "accept": false, + "reason": "Adding a model and documenting a decode optimization are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45922", + "accept": false, + "reason": "Mistral4 fixes are unrelated to the BEiT/vision-model memory leak cache cleanup." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Both PRs modify `fixed_cross_entropy` in the same way to forward supported kwargs to `nn.functional.cross_entropy`, so they are the same concrete fix and could be merged into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted duplicate-like pair is the shared `fixed_cross_entropy` change (`43251`/`43254`), which is supported by the shared file and issue target `#43240`. The other pairwise rejections are consistent with the listed titles/diffs and do not overstate duplication." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 2, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-40546-2", + "nodes": [ + "pull_request:40546", + "pull_request:42781" + ], + "soft_pairs": [ + "pull_request:40546|pull_request:42781" + ], + "prepared_review_unit_hash": "c8eb94239f6eb8f3cfd36781ac9b2debfceedb6fe09310db64376120a7c0691f", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7225, + "estimated_input_tokens": 1807, + "estimated_eval_tokens": 3870 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7225, + "estimated_input_tokens": 1807, + "estimated_eval_tokens": 3870 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related VibeVoice pull requests, but they are not duplicates: PR 40546 is the main initial VibeVoice implementation, while PR 42781 adds a separate VibeVoice Realtime path plus an acoustic tokenizer and generation changes. Shared files are mostly the same model-family scaffolding, not the same concrete fix/change.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 40546 is the better canonical PR because it is the broader, more complete initial VibeVoice implementation and has much higher review activity; PR 42781 looks like a later extension for realtime support rather than the primary change.", + "best_issue_reason": null, + "best_pr_reason": "PR 40546 is the strongest representative of the cluster: it introduces the core VibeVoice integration across docs, auto-mapping, model code, processing, and pipeline hooks. PR 42781 is narrower and draft-status, focused on realtime/acoustic-tokenizer follow-up work.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Do not merge: the PRs share VibeVoice files and auto-registration, but 42781 adds a separate realtime/acoustic-tokenizer feature set and generation changes rather than fixing or implementing the same concrete code path as 40546." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 3, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-40755-2", + "nodes": [ + "pull_request:40755", + "pull_request:41224" + ], + "soft_pairs": [ + "pull_request:40755|pull_request:41224" + ], + "prepared_review_unit_hash": "77c2685f7696b2e48898c1fade88c87477d3dbde2820f4cb494c453f9109c025", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two pull requests are unrelated feature additions: one adds TimesFM covariate forecasting support, the other adds DINOv3 image classification support. They do not appear to fix the same underlying bug or implement the same change.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Different models and different capabilities; no shared code-path problem or mergeable duplicate change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 4, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41356-2", + "nodes": [ + "pull_request:41356", + "pull_request:44339" + ], + "soft_pairs": [ + "pull_request:41356|pull_request:44339" + ], + "prepared_review_unit_hash": "aead1308c167b0bae6ff11bf9358340baa5e76f816618383418b505a10a9cd01", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6050, + "estimated_input_tokens": 1513, + "estimated_eval_tokens": 3282 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6050, + "estimated_input_tokens": 1513, + "estimated_eval_tokens": 3282 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two DEIMv2 pull requests target the same model family, but they are not the same change: PR 41356 is an early/basic model + image processor add, while PR 44339 is a much broader follow-up that adds loss code, auto-mappings, conversion tooling, modular files, and more integration work.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44339 is the more complete and representative DEIMv2 implementation; PR 41356 looks like an earlier partial version of the same model bring-up.", + "best_issue_reason": null, + "best_pr_reason": "PR 44339 is the strongest single artifact for the cluster because it subsumes the basic model work and includes the broader integration needed for a full Transformers addition.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41356", + "right": "pull_request:44339", + "accept": false, + "reason": "Same model area, but not the same concrete change: 44339 expands the implementation substantially and adds new code paths beyond the basic PR 41356. They look like related iterations, not mergeable duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 5, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41797-3", + "nodes": [ + "pull_request:41797", + "pull_request:45186", + "pull_request:45630" + ], + "soft_pairs": [ + "pull_request:41797|pull_request:45186", + "pull_request:45186|pull_request:45630" + ], + "prepared_review_unit_hash": "6feb716c29ddcc88592b8f12d8208ba4891a135a91bc0f2a0d15cb689fd14647", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10409, + "estimated_input_tokens": 2603, + "estimated_eval_tokens": 5462 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10409, + "estimated_input_tokens": 2603, + "estimated_eval_tokens": 5462 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are three separate open PRs adding different models (DeepSeekOCR, Isaac, and Kimi2-6). They share the same broad scaffolding pattern for new model integration, but they are not the same underlying change.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: each PR adds a different model and touches its own model-specific code paths.", + "best_issue_reason": null, + "best_pr_reason": "No single PR is a good representative duplicate target because the changes are distinct model additions rather than one shared fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41797", + "right": "pull_request:45186", + "accept": false, + "reason": "Different models and different implementations; shared new-model boilerplate is not enough to treat them as the same change." + }, + { + "left": "pull_request:45186", + "right": "pull_request:45630", + "accept": false, + "reason": "Isaac and Kimi2-6 are separate model integrations with different files, configs, and code paths." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 6, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-42975-2", + "nodes": [ + "pull_request:42975", + "pull_request:42976" + ], + "soft_pairs": [ + "pull_request:42975|pull_request:42976" + ], + "prepared_review_unit_hash": "595a8e3e100876655e759346220735e4e2aa96afc32318d50673a0a50ac53524", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two PRs both touch GitHub Actions workflows, but they are different upgrade efforts: one focuses on checkout/cache/artifact actions for Node 24 compatibility, while the other bumps docker/build-push-action and a separate set of workflow actions. Shared files alone are not enough to treat them as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR for a duplicate cluster: the PRs address different action-version upgrade tracks and do not fix the same underlying change.", + "best_issue_reason": null, + "best_pr_reason": "If one PR must represent the cluster, 42975 is the broader GitHub Actions compatibility update, but it still should not absorb 42976 as a duplicate.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "They upgrade different GitHub Actions families and target different maintenance goals; overlapping workflow files do not make them the same change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 7, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-42978-7", + "nodes": [ + "pull_request:42978", + "pull_request:43448", + "pull_request:43451", + "pull_request:43973", + "pull_request:44601", + "pull_request:45490", + "pull_request:45838" + ], + "soft_pairs": [ + "pull_request:42978|pull_request:43448", + "pull_request:42978|pull_request:43451", + "pull_request:42978|pull_request:45490", + "pull_request:43448|pull_request:43451", + "pull_request:43448|pull_request:45490", + "pull_request:43451|pull_request:45490", + "pull_request:43973|pull_request:45838", + "pull_request:44601|pull_request:45838", + "pull_request:45490|pull_request:45838" + ], + "prepared_review_unit_hash": "eef849d5186c2f253d507ddd1180de42ad4402dde2758aae2dbf4e621aede737", + "component_budget": { + "node_count": 7, + "item_count": 7, + "soft_pair_count": 9, + "serialized_chars": 23541, + "estimated_input_tokens": 5886, + "estimated_eval_tokens": 12028 + }, + "budget": { + "node_count": 7, + "item_count": 7, + "soft_pair_count": 9, + "serialized_chars": 23541, + "estimated_input_tokens": 5886, + "estimated_eval_tokens": 12028 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly superficial similarity from model-addition PR boilerplate (docs, auto mappings, tests), plus one unrelated distributed-loading PR. None of the soft edges look like true duplicates or mergeable into a single change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: these are separate feature PRs for different model families or pipeline-parallel infrastructure, and the overlap is mostly docs/registry boilerplate.", + "best_issue_reason": null, + "best_pr_reason": "No PR is a true duplicate anchor here; if forced, 43451 is the most substantial standalone model-addition PR, but it does not subsume the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42978", + "right": "pull_request:43448", + "accept": false, + "reason": "Different model families (ViT NEPA vs Molmo) with distinct codepaths; shared docs/auto-registration files are boilerplate, not the same fix." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "ViT NEPA and Molmo2 are separate model integrations with different implementations and files; not the same underlying change." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "Different models (ViT NEPA vs CTSM) and different model code; overlap is only generic transformers registration/docs." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are related by name but are distinct model releases with different configs/modeling/processing code; not a duplicate PR." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo and CTSM are unrelated model additions; shared auto/model registry edits are broad subsystem boilerplate." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM are separate model integrations with different model paths; no concrete bug or change overlap beyond registration/docs." + }, + { + "left": "pull_request:43973", + "right": "pull_request:45838", + "accept": false, + "reason": "Audio model addition vs OCR text-recognition model support; different domains, code paths, and artifacts." + }, + { + "left": "pull_request:44601", + "right": "pull_request:45838", + "accept": false, + "reason": "Pipeline-parallel loading infrastructure is unrelated to PP-OCRv6 model support; not the same bug or change." + }, + { + "left": "pull_request:45490", + "right": "pull_request:45838", + "accept": false, + "reason": "CTSM model addition and PP-OCRv6 text recognition support are distinct features with different model implementations and purposes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 8, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43291-3", + "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": "356185cd66553e5d8185cfd6093f4b59457cfbf2293784af67f3df8b395e0efc", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7093, + "estimated_input_tokens": 1774, + "estimated_eval_tokens": 3804 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7093, + "estimated_input_tokens": 1774, + "estimated_eval_tokens": 3804 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three items are pull requests with unrelated fixes: Whisper tokenizer/test changes, CLIP/ViT test changes, and a dequantization/fine-grained FP8 fix. None are the same underlying change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44815 is the strongest canonical PR because it contains a substantive code-path fix across model loading and FP8 quantization, whereas 43291 and 43665 are narrow test-oriented changes with no clear shared bug.", + "best_issue_reason": null, + "best_pr_reason": "PR 44815 is the best representative artifact: it addresses a concrete implementation bug, touches production code, and has the most complete multi-file fix. The other PRs appear to be test-only or localized cleanup changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Both are test-related, but they target different models and different failures (Whisper tokenizer/tests vs CLIP/ViT test edits). They do not look like the same bug or one mergeable PR." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper tokenizer/test fixes and FP8 dequantization/loading fixes are unrelated code paths and subsystems. Shared 'fix' wording is too broad to treat as a duplicate." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "PR 43665 only changes CLIP/ViT tests, while PR 44815 changes model loading and FP8 quantization internals. These are not the same change and should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 9, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43488-2", + "nodes": [ + "pull_request:43488", + "pull_request:43532" + ], + "soft_pairs": [ + "pull_request:43488|pull_request:43532" + ], + "prepared_review_unit_hash": "d3ab2ef564602e14e357ccfde7525e4489bd7dc77d1335f263188828cb8d7be1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4561, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4561, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs are both marked as non-merging experiments, but they change different areas: one is a packaging/versioning and test-dependency tweak, while the other rewires CI/workflow and testing utilities. They do not look like duplicates.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "No clear canonical PR: the patches target different code paths and are not plausibly mergeable into one fix despite sharing a test filename.", + "best_issue_reason": null, + "best_pr_reason": "No single PR stands out as the representative duplicate target because the changes are unrelated beyond superficial overlap.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "Reject: shared vit test filename is incidental; one PR changes setup/versioning and pytest deps, the other changes workflow automation and testing utilities, 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": 10, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43612-2", + "nodes": [ + "pull_request:43612", + "pull_request:43613" + ], + "soft_pairs": [ + "pull_request:43612|pull_request:43613" + ], + "prepared_review_unit_hash": "5a4c7cf47869a9849de41c35f61ba0f8ae1249c3de1978a6cecaede877f08404", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6082, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6082, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related pipeline/docs additions, but they implement different features: promptable concept segmentation vs promptable visual segmentation. Shared infrastructure files are not enough to treat them as duplicates, so the soft edge 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:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Different underlying feature work and code paths: one adds PromptableConceptSegmentationPipeline, the other adds PromptableVisualSegmentationPipeline with different model/test coverage. They are related but not mergeable 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": 11, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43651-2", + "nodes": [ + "pull_request:43651", + "pull_request:43663" + ], + "soft_pairs": [ + "pull_request:43651|pull_request:43663" + ], + "prepared_review_unit_hash": "6b5a55abae0eab118852981dc3fc5e232dbc8d0033a4df319b2692a4a6449631", + "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": "Two separate trainer extensibility PRs: one adds a loss-scaling hook for gradient accumulation, the other adds an overridable signature-column helper for column filtering. Same file, but different code paths and behaviors; not duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43663 is the more general trainer API refactor: it extracts signature-column computation into an overridable helper and is the cleaner representative of the cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR 43663 is the strongest standalone change here because it introduces a reusable trainer hook with broad extensibility value and a self-contained code-path adjustment.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Both touch trainer.py, but one changes gradient-accumulation loss scaling and the other changes input-column filtering via signature inspection; they fix different problems and would not plausibly merge as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 12, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43656-3", + "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": "eede7eb815cf7f8b49b29ca3f8ee65f7b579dec35362cfc21762f445a0b8fc1d", + "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 modify `src/transformers/cli/serve.py` and are aimed at the same import-time `TypeAdapter` annotation NameError in the CLI. PR 43842 is the most focused fix; 43656 and 43836 look like broader/earlier variants of the same underlying problem.", + "confidence": 0.79, + "canonical_issue_reason": null, + "canonical_pr_reason": "43842 is the cleanest direct fix: it makes `TypeAdapter` a type-checking-only import and stringizes the annotation at the failing call site, with minimal unrelated churn.", + "best_issue_reason": null, + "best_pr_reason": "43842 is the best representative PR because it targets the exact runtime NameError in the most explicit and narrowly scoped way.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": true, + "reason": "Both address the same `TypeAdapter`-related NameError in `serve.py`; 43656 is a broader postponed-annotations variant of the same fix, while 43842 is the narrower follow-up." + }, + { + "left": "pull_request:43836", + "right": "pull_request:43842", + "accept": true, + "reason": "Same file, same import-time `TypeAdapter` bug, and both make the annotation safe at runtime; 43836 is just a broader refactor around the same concrete problem." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary and pair verdicts are grounded in the packet. All three PRs touch `src/transformers/cli/serve.py` and are clearly in the same TypeAdapter/annotation bug family. The soft-edge accepts are conservative enough: 43836 and 43842 share the same issue target, and 43656 is plausibly the same CLI NameError class of fix even though it is broader and only linked by soft similarity. No overstatement that would warrant rejection." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 13, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43779-2", + "nodes": [ + "pull_request:43779", + "pull_request:43816" + ], + "soft_pairs": [ + "pull_request:43779|pull_request:43816" + ], + "prepared_review_unit_hash": "e7c674adcc031aa8883109a5dfc678b4407fb20b10764302092e384d031ba83a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4362, + "estimated_input_tokens": 1091, + "estimated_eval_tokens": 2438 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4362, + "estimated_input_tokens": 1091, + "estimated_eval_tokens": 2438 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests appear to implement the same SwanLab integration change: adding run id/resume support in `integration_utils.py` for issue #43698, with one variant also expanding the docs. They are strong duplicate PR candidates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43816 is the cleaner representative: it carries the same SwanLab id/resume support plus the accompanying documentation updates, making it the more complete version of the shared change.", + "best_issue_reason": null, + "best_pr_reason": "PR #43816 is the best global representative because it appears to include the full feature change and docs for SwanLab run id/resume support, while #43779 is the narrower implementation-focused variant.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43779", + "right": "pull_request:43816", + "accept": true, + "reason": "Same underlying SwanLab callback enhancement for id/resume support in the same file and same issue target; the doc differences do not change the core code-path change, so these are duplicate PRs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both PRs target issue #43698, touch the same file, and the previews indicate the same SwanLab id/resume support with #43816 adding docs. The duplicate/mergeable verdict is conservative enough given the shared target and overlapping code change." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43996", + "pull_request:43997", + "pull_request:44007", + "pull_request:44010", + "pull_request:44013", + "pull_request:44017", + "pull_request:44018", + "pull_request:44019", + "pull_request: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:44018|pull_request:44116", + "pull_request:44071|pull_request:44722", + "pull_request:44018|pull_request:44086", + "pull_request:44018|pull_request:44085", + "pull_request:44071|pull_request:44085", + "pull_request:44068|pull_request:44086", + "pull_request:44059|pull_request:44086", + "pull_request:44085|pull_request:44086", + "pull_request:44098|pull_request:44101", + "pull_request:44059|pull_request:44722", + "pull_request:44018|pull_request:44071", + "pull_request:44017|pull_request:44161", + "pull_request:44017|pull_request:44722", + "pull_request:44017|pull_request:44073", + "pull_request:44068|pull_request:44071", + "pull_request: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:44018|pull_request:44059", + "pull_request:44056|pull_request:44101", + "pull_request:44073|pull_request:44161", + "pull_request:44101|pull_request:44161", + "pull_request:44017|pull_request:44018", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:44056|pull_request:44161", + "pull_request:43997|pull_request:44018", + "pull_request:43997|pull_request:44161", + "pull_request:44074|pull_request:44161", + "pull_request:44018|pull_request:44161", + "pull_request:43997|pull_request:44071", + "pull_request:44010|pull_request:44071", + "pull_request:44018|pull_request:44129", + "pull_request:44010|pull_request:44073", + "pull_request:43997|pull_request:44074", + "pull_request:43997|pull_request:44129", + "pull_request:44071|pull_request:44129", + "pull_request:44010|pull_request:44018", + "pull_request:43996|pull_request:44161", + "pull_request:43996|pull_request:44017", + "pull_request:44010|pull_request:44044", + "pull_request:44044|pull_request:44074", + "pull_request:44066|pull_request:44722", + "pull_request:44085|pull_request:44722", + "pull_request:44085|pull_request:44129", + "pull_request:43996|pull_request:44007", + "pull_request:43996|pull_request:44129", + "pull_request:44044|pull_request:44129", + "pull_request:44044|pull_request:44154" + ], + "prepared_review_unit_hash": "6281b41118f98e3a6c462158e3b164825d2cab748e960cf0764db645ced7ce7f", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84896, + "estimated_input_tokens": 21224, + "estimated_eval_tokens": 42704 + }, + "budget": { + "node_count": 27, + "item_count": 27, + "soft_pair_count": 72, + "serialized_chars": 75093, + "estimated_input_tokens": 18774, + "estimated_eval_tokens": 37804 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a broad wave of standardized output-tracing refactors across many unrelated models. The only clear duplicate-like overlaps are within the GPT-J/CodeGen subgroup and the GPT-Neo subgroup; the rest are just thematically similar, not the same underlying code path.", + "confidence": 0.84, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44066 is the strongest representative of the shared GPT-J/CodeGen output-tracing change: it is issue-linked, touches the core files in the duplicate set, and cleanly expresses the standardized decorator migration.", + "best_issue_reason": null, + "best_pr_reason": "PR 44066 is the best single PR to keep as the representative change because it is the most complete, issue-linked GPT-J/CodeGen refactor and has the clearest overlap with the accepted duplicate candidates.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Same general refactor theme, but different models and code paths (CVT/FNet vs GPT-J)." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "Both are output-tracing refactors, but they affect unrelated model implementations and are not the same change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": true, + "reason": "Both target GPT-J output tracing in the same file and are the same underlying decorator migration." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "Different vision models and separate implementations; only the refactor theme is shared." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "Different model families and no concrete code-path overlap." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen versus EfficientNet; same modernization pattern, not the same bug or change." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "MobileNetV2 and DeBERTa v2 are unrelated code paths." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "Different models and behavior; no shared concrete change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen refactor versus MPT refactor; unrelated implementation details." + }, + { + "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 are plausibly duplicate iterations of the same change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J versus GPT-Neo; same refactor family but not the same code path." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "Different model families and different concrete changes." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet refactor and GPT-J/CodeGen refactor are unrelated." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR and GPT-J/CodeGen are different model implementations." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44085", + "accept": false, + "reason": "ResNet versus GPT-J; only superficial similarity in refactor style." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet are unrelated code paths." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet and MPT are unrelated model changes." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo refactor versus Flaubert/XLM refactor; not the same underlying change." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo and GPT-J/CodeGen are different model families and changes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert/XLM are unrelated." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT and GPT-J/CodeGen are unrelated implementation changes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR do not share the same concrete code path." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo and GPT-J are separate model refactors." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT and Flaubert/XLM are unrelated model implementations." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-2 and GPT-J/CodeGen are different model-specific changes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are different code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "Segformer and LongT5 are unrelated models." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44722", + "accept": false, + "reason": "Segformer output tracing and GPT-J/CodeGen output tracing are not the same change." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44073", + "accept": false, + "reason": "Segformer and VisualBert are separate implementations." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are unrelated model changes." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44073", + "accept": false, + "reason": "RegNet and VisualBert are different code paths." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 are unrelated implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are different model-specific refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet and LongT5 are unrelated." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet and Flaubert/XLM are different code paths." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 are unrelated model changes." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet and GPT-Neo are unrelated." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBERT and MPT are different implementations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo and SpeechT5 are unrelated model changes." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBERT and VisualBert are unrelated." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet and TextNet are different code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44129", + "accept": false, + "reason": "RegNet and SpeechT5 are unrelated model refactors." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are unrelated." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44018", + "accept": false, + "reason": "SqueezeBERT and GPT-Neo are different model implementations." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44161", + "accept": false, + "reason": "CVT/FNet and LongT5 are unrelated." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44017", + "accept": false, + "reason": "Different models and different code paths." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBERT and DeBERTa v2 are unrelated." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa v2 and TextNet are different implementations." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44722", + "accept": true, + "reason": "Same GPT-J/CodeGen output-tracing refactor with substantial file overlap." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44722", + "accept": true, + "reason": "Both are GPT-J output-tracing changes to the same model code path, so they are duplicate-like." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-J and SpeechT5 are unrelated." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44007", + "accept": false, + "reason": "Different vision model families and separate implementations." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44129", + "accept": false, + "reason": "CVT/FNet and SpeechT5 are unrelated code paths." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44129", + "accept": false, + "reason": "DeBERTa v2 and SpeechT5 are different model-specific changes." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44154", + "accept": false, + "reason": "DeBERTa v2 and VITS are unrelated implementations." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall. The summary stays conservative about the non-overlapping model families, and the accepted pairs are limited to the clearly mergeable GPT-J/CodeGen and GPT-Neo iterations with shared file/code-path overlap." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 27, + "review_unit_index": 2, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43997", + "pull_request:44010", + "pull_request:44024", + "pull_request:44056", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44129" + ], + "soft_pairs": [ + "pull_request:44024|pull_request:44076", + "pull_request:44073|pull_request:44074", + "pull_request:44056|pull_request:44073", + "pull_request:44056|pull_request:44074", + "pull_request:43997|pull_request:44010", + "pull_request:44010|pull_request:44129" + ], + "prepared_review_unit_hash": "8b0dd9ed8cc612b756a51bd4a909c5d635a5f71b03d1adf05ac785f0888bafe5", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84896, + "estimated_input_tokens": 21224, + "estimated_eval_tokens": 42704 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 6, + "serialized_chars": 17124, + "estimated_input_tokens": 4281, + "estimated_eval_tokens": 8818 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "All items are separate model-specific pull requests that apply the same general output-tracing refactor pattern, but they target different architectures and code paths. They should not be deduplicated as the same change.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "44129 is the best representative PR because it has an explicit issue target and the most clearly scoped end-to-end refactor within a single model, but it is still not a duplicate of the others.", + "best_issue_reason": null, + "best_pr_reason": "44129 is the strongest standalone PR to keep: it is explicit about the target issue, has a clear model-specific change, and is more concretely described than the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": false, + "reason": "Both are standardized-output refactors, but they touch different models (FocalNet vs ImageGPT) and different forward paths, so they are not the same underlying change." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "These are separate refactors for VisualBert and TextNet. The shared theme is output capture modernization, not one concrete bug or mergeable PR." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "MPNet and VisualBert use different model code paths and output-recording hooks; this is only broad subsystem similarity." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "MPNet vs TextNet are distinct model implementations with different tracing changes, so they should not be merged as duplicates." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are unrelated model-specific output refactors; same refactor style, not the same concrete fix." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "SqueezeBert attention tracing and SpeechT5 output capture are different code paths in different models, so they are not duplicative." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 15, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43998-13", + "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": "18e007608d35ef73251ef95ed87e6fade473a9f2480d66db9bf2ced88c754da2", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40178, + "estimated_input_tokens": 10045, + "estimated_eval_tokens": 20346 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40178, + "estimated_input_tokens": 10045, + "estimated_eval_tokens": 20346 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is not a true duplicate set: all items are separate, model-specific output-tracing refactors that share a tooling pattern (`can_return_tuple` / output capturing) but not the same concrete bug or code path. I would keep them separate and, if forced to pick a representative, use the broadest mamba-related PR.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44003 is the broadest representative of the shared refactor pattern, touching two related model files and introducing the new tracing helpers, but it is still model-specific rather than a duplicate of the others.", + "best_issue_reason": null, + "best_pr_reason": "PR 44003 is the strongest single representative of the cluster\u2019s theme because it is the most substantial output-tracing change among the set and shows the new helper usage most clearly.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model wrappers (`timm_backbone` vs `upernet`) with separate forward paths; same refactor style, not the same change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "Different models (`timm_backbone` vs `vision_text_dual_encoder`) and different APIs; only the tracing pattern overlaps." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Unrelated model implementations; both adopt output-tracing helpers but do not fix the same code path." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "Separate model families (`timm_backbone` vs `mamba/falcon_mamba`) and distinct forward implementations." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Different architectures (`timm_backbone` vs `rwkv`); shared utility usage is not enough for duplication." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "`mobilenet_v1` and `univnet` are unrelated models; these are parallel refactors, not the same bug." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "Different subsystems (`mobilenet_v1` vs `upernet`) with separate output-shape/tracing code." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "`mobilenet_v1` and `depth_anything` are distinct models; they only share a common tracing migration." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "Different encoder-decoder wrapper families and unrelated forward paths." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "`mobilenet_v1` and `speech_encoder_decoder` are different code paths; not mergeable as one bug fix." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "Different models (`vision_text_dual_encoder` vs `univnet`); same refactor theme only." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "`vision_text_dual_encoder` and `upernet` are separate model wrappers with different outputs and call signatures." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "Different model stacks (`vision_text_dual_encoder` vs `mamba/falcon_mamba`) and different output-capture plumbing." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "Both are output-tracing refactors, but they target different models and cannot plausibly be one PR." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "Unrelated models (`univnet` vs `upernet`); no shared concrete defect." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Different model families and forward code paths; only broad helper adoption overlaps." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "`univnet` and `depth_anything` are separate model-specific changes, not one bug." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44029", + "accept": false, + "reason": "`upernet` and `rwkv` are unrelated implementations; shared output-tracing refactor is too broad." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "Different models and separate forward semantics; not the same underlying issue." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44028", + "accept": false, + "reason": "`mamba/falcon_mamba` vs `superpoint` are entirely different code paths; same refactor motif only." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "`mamba`/`falcon_mamba` and `codegen` are unrelated model implementations with distinct outputs." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "Different model families and APIs; the overlap is only the new output-tracing helper pattern." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "`univnet` and `codegen` touch different forward paths and output contracts." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "Separate models (`upernet` vs `codegen`); no shared concrete bug or change." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "`vision_text_dual_encoder` and `codegen` are unrelated; similar refactor language is not duplication." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "`depth_anything` and `vision_encoder_decoder` are different model wrappers and different code paths." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "`depth_anything` and `speech_encoder_decoder` are unrelated models; not mergeable as one fix." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "`depth_anything` and `superpoint` are separate vision models with distinct forward behavior." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "`depth_anything` and `rwkv` are unrelated model families; only the tracing migration matches." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "`depth_anything` and `dpr` are different architectures and output contracts." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "`vision_encoder_decoder` and `rwkv` are unrelated; same helper adoption does not imply duplicate fixes." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "Vision vs speech encoder-decoder wrappers are different code paths, not the same bug." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "`vision_encoder_decoder` and `superpoint` are distinct model implementations." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "`vision_encoder_decoder` and `dpr` do not share a concrete defect; only the refactor pattern overlaps." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "Speech encoder-decoder and SuperPoint are unrelated models with different output structures." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families and different forward signatures; not a duplicate." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "`speech_encoder_decoder` and `dpr` are separate wrappers with no shared concrete bug." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "`superpoint` and `rwkv` are unrelated code paths; the commonality is only output tracing." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "Different models (`superpoint` vs `dpr`) and different output contracts." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "`rwkv` and `dpr` are unrelated implementations; no evidence they fix the same underlying 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": 16, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44662-3", + "nodes": [ + "pull_request:44662", + "pull_request:45075", + "pull_request:45401" + ], + "soft_pairs": [ + "pull_request:44662|pull_request:45401", + "pull_request:44662|pull_request:45075", + "pull_request:45075|pull_request:45401" + ], + "prepared_review_unit_hash": "82528b93c5452b71a0585d099a0ef667a2130bfc4e37324d0d77b0243ecd5010", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 10913, + "estimated_input_tokens": 2729, + "estimated_eval_tokens": 5714 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 10913, + "estimated_input_tokens": 2729, + "estimated_eval_tokens": 5714 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are three separate model-addition PRs (PenguinVL, DeepSeek-OCR-2, Voxtral TTS) that share only common registry/doc boilerplate. They are not duplicate fixes or mergeable into one PR.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "pull_request:45401 is the strongest standalone representative because it has an explicit issue target and a focused, self-contained implementation for one model family.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Different models and different code paths; overlap is limited to shared Transformers registration/docs scaffolding, which is not enough to treat them as the same change." + }, + { + "left": "pull_request:44662", + "right": "pull_request:45075", + "accept": false, + "reason": "Both add distinct model implementations. Shared auto/registry files are generic model-onboarding changes, not evidence of the same underlying bug or feature." + }, + { + "left": "pull_request:45075", + "right": "pull_request:45401", + "accept": false, + "reason": "Separate model introductions for unrelated architectures; they do not fix the same concrete problem and would not plausibly be merged as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 17, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44979-2", + "nodes": [ + "pull_request:44979", + "pull_request:45363" + ], + "soft_pairs": [ + "pull_request:44979|pull_request:45363" + ], + "prepared_review_unit_hash": "52ccd303c16513f897ff3b645d1fb989edb836e7a54953a42f557edb5664e91b", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4406, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4406, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two PRs are related around module/kernel fusion, but they target different layers: #44979 introduces a general Module Fusion API, while #45363 wires that API into KernelConfig for n-to-1 kernel fusion. They do not appear to be the same concrete fix/change, so they should not be merged as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #44979 is the broader foundational change: it adds the new module_fusion abstraction and tests, which can underpin later integrations.", + "best_issue_reason": null, + "best_pr_reason": "PR #44979 is the best representative of the cluster because it captures the underlying API being introduced, rather than one specific consumer integration.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "#44979 adds a generic Module Fusion API; #45363 adds KernelConfig-based n-to-1 kernel fusion wiring and an example. They are adjacent but not the same concrete bug/change, and would not plausibly be merged into one PR without conflating API foundation with a specific integration." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 18, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45086-2", + "nodes": [ + "pull_request:45086", + "pull_request:45317" + ], + "soft_pairs": [ + "pull_request:45086|pull_request:45317" + ], + "prepared_review_unit_hash": "40e27eed167f0db272861bf2643d7355bfda449adca9e5a03af1fce8973cfaac", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests fix the same Mistral regex patching AttributeError in `src/transformers/tokenization_utils_tokenizers.py`; 45317 is the more complete candidate because it includes a test update in addition to the same code fix.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45317 is the better canonical PR because it implements the same bug fix as 45086 and also adds a regression test, making it the more complete representative of the change.", + "best_issue_reason": null, + "best_pr_reason": "PR 45317 is the best overall PR candidate: same underlying fix, same touched code path, and it includes tests, whereas 45086 is just the minimal code change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same concrete bug in `_patch_mistral_regex`, same file, same fix lines, and the second PR is a superset with tests; these are effectively 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": 19, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45115-2", + "nodes": [ + "pull_request:45115", + "pull_request:45569" + ], + "soft_pairs": [ + "pull_request:45115|pull_request:45569" + ], + "prepared_review_unit_hash": "9ea6e766baff0f1a7a71109bc1e794b286bf314bcbecb56e39ef747f464a31e6", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6065, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6065, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related NemotronH refactors, but they are not the same change: PR #45115 is a narrow conversion-mapping/inheritance refactor, while PR #45569 is a broader architecture split introducing separate dense/sparse model variants, docs, and tests.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45569 is the better canonical representative because it is the broader, more complete model-structure change and includes the user-facing docs/tests alongside code organization updates.", + "best_issue_reason": null, + "best_pr_reason": "PR #45569 is the strongest standalone PR in the cluster due to its wider scope and clearer end-state; PR #45115 is a smaller supporting refactor and not a substitute for the full change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "They share the NemotronH subsystem, but they address different concrete changes: one focuses on conversion/name-mapping and inheritance cleanup, the other splits the architecture into dense/sparse variants with documentation and test updates. They do not look mergeable as one PR fixing the same underlying bug or implementing 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": 20, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45332-2", + "nodes": [ + "pull_request:45332", + "pull_request:45333" + ], + "soft_pairs": [ + "pull_request:45332|pull_request:45333" + ], + "prepared_review_unit_hash": "1720418058ade51b413fce50d329a043c5017e8b8b103b89da3ff77f6f0503d1", + "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 heterogeneous per-layer config support, the other adds heterogeneous model/runtime support. They share helper files, but they are not the same concrete change.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45332 is the broader umbrella change and appears to subsume the model-side heterogeneous support along with the config plumbing.", + "best_issue_reason": null, + "best_pr_reason": "#45332 is the best representative PR for the cluster because it covers the larger end-to-end heterogeneous support work rather than only the config layer.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "Related infrastructure, but not the same underlying change: #45333 is focused on per-layer config support, while #45332 adds broader model/runtime heterogeneous support. They would not naturally collapse into one single PR without mixing separate scopes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 21, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45415-2", + "nodes": [ + "pull_request:45415", + "pull_request:45425" + ], + "soft_pairs": [ + "pull_request:45415|pull_request:45425" + ], + "prepared_review_unit_hash": "3b37c09dc8cf81299d8aec42538ec2a802e8a8c6f306a4c34b5e9873abb8ee0c", + "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 address different layers: #45415 is a broad repo-wide type-checking enablement pass, while #45425 adds specific typing aliases/protocols to satisfy a few code paths. They overlap in a few touched files, but not in the same concrete fix, so they should not be merged as duplicates.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "#45425 is the more focused, concrete typing-change PR with a clear implementation surface and narrower scope.", + "best_issue_reason": null, + "best_pr_reason": "#45425 is the best single PR in the cluster because it targets a specific typing compatibility problem, whereas #45415 is a broad type-checking cleanup across many files.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Shared typing theme and a few filenames, but #45415 is a broad type-checking pass and #45425 is a targeted typing-compatibility change; they do not look like the same underlying bug or one mergeable fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 22, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45487-6", + "nodes": [ + "pull_request:45487", + "pull_request:45817", + "pull_request:45843", + "pull_request:45858", + "pull_request:45890", + "pull_request:45926" + ], + "soft_pairs": [ + "pull_request:45487|pull_request:45843", + "pull_request:45487|pull_request:45926", + "pull_request:45817|pull_request:45843", + "pull_request:45858|pull_request:45890", + "pull_request:45858|pull_request:45926", + "pull_request:45890|pull_request:45926" + ], + "prepared_review_unit_hash": "c163ddb41c91c999a53df217577aaa1cb38c4082d9475705f53408369f6dca1b", + "component_budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 6, + "serialized_chars": 14704, + "estimated_input_tokens": 3676, + "estimated_eval_tokens": 7608 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 6, + "serialized_chars": 14704, + "estimated_input_tokens": 3676, + "estimated_eval_tokens": 7608 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly unrelated PRs that happen to mention parallelism or XPU. The only real overlap is the broad theme of model-parallel/device-placement fixes, but the concrete bugs and code paths are different enough that they should stay separate.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45487 is the strongest representative: it fixes an actual model-parallel device-placement bug in model code, spans multiple affected implementations, and includes tests. It is more substantive than the test-only expectation updates and broader than the single-model parallel fixes.", + "best_issue_reason": null, + "best_pr_reason": "PR 45487 is the best single PR for the cluster because it addresses a concrete runtime bug rather than only updating test expectations, and it has the clearest cross-model impact among the candidates.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45487", + "right": "pull_request:45843", + "accept": false, + "reason": "Both are model-parallel fixes, but for different underlying problems: altclip/chinese_clip device placement versus deimv2 backbone/_no_split_modules propagation. Not the same concrete bug." + }, + { + "left": "pull_request:45487", + "right": "pull_request:45926", + "accept": false, + "reason": "45487 is a runtime model-parallel bug fix; 45926 only adds XPU test expectations for a different model. Different artifact type and different underlying issue." + }, + { + "left": "pull_request:45817", + "right": "pull_request:45843", + "accept": false, + "reason": "Gemma4 fixes device handling and no-split modules in one model; deimv2 fixes wrapper/backbone split-module propagation in another. Same theme, not the same change or bug." + }, + { + "left": "pull_request:45858", + "right": "pull_request:45890", + "accept": false, + "reason": "Both are test expectation updates, but they cover different models and different outputs. They do not appear to fix the same underlying defect." + }, + { + "left": "pull_request:45858", + "right": "pull_request:45926", + "accept": false, + "reason": "These are separate XPU expectation updates for different model test files with different logits. Shared XPU context is too broad to merge as duplicates." + }, + { + "left": "pull_request:45890", + "right": "pull_request:45926", + "accept": false, + "reason": "Same rationale as the other test-only pairs: distinct model-specific expectation updates, not one shared 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": 23, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45548-2", + "nodes": [ + "pull_request:45548", + "pull_request:45662" + ], + "soft_pairs": [ + "pull_request:45548|pull_request:45662" + ], + "prepared_review_unit_hash": "d442628b16118ce14ecb8c5551fae8c80edfe64713ee5a8a7a509f0088abe73c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two related but distinct EP-loading fixes: one targets serialization/loading under accelerate + DeepSpeed ZeRO-3, the other targets expert parameters being overwritten during FSDP2/rank-0 broadcast. They overlap in modeling_utils but address different failure modes and code paths, so they should not be merged as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:45662 is the better representative PR because it fixes a specific concrete failure in the loading path and has the broader, more targeted EP/FSDP2 scope.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:45662 is the strongest standalone PR in the cluster: the bug description is concrete, the diff is tightly focused on the affected loading path, and the fix matches a clear runtime failure mode.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Same EP subsystem, but different underlying bugs: 45548 is about config serialization/loading with accelerate + ZeRO-3, while 45662 prevents expert tensors being overwritten during FSDP2 rank-0 broadcast. They are not the same code-path problem and should not be merged." + } + ] + }, + "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": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45639-2", + "nodes": [ + "pull_request:45639", + "pull_request:45645" + ], + "soft_pairs": [ + "pull_request:45639|pull_request:45645" + ], + "prepared_review_unit_hash": "a376a926c25f8ede7d60b3e3561461811140cdec4f978c345ceb1f0669f28392", + "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 underlying xdist/captured_info artifact collision problem in transformers testing utilities; PR 45645 is the broader, more complete variant because it also updates CI log aggregation for the new multi-file output scheme.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45645 is the better canonical PR because it contains the core xdist-safe testing_utils change plus the related CI workflow updates needed to consume multiple captured_info files, making it the more complete fix for the shared issue.", + "best_issue_reason": null, + "best_pr_reason": "PR 45645 is the best representative because it covers the full end-to-end change, not just the library-side file naming fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45639", + "right": "pull_request:45645", + "accept": true, + "reason": "Same concrete bug: xdist collisions in captured_info output. They share the same issue target and files, and 45645 looks like the superset that also updates CI handling for the new per-worker artifacts." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 25, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45683-2", + "nodes": [ + "pull_request:45683", + "pull_request:45734" + ], + "soft_pairs": [ + "pull_request:45683|pull_request:45734" + ], + "prepared_review_unit_hash": "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": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45737-2", + "nodes": [ + "pull_request:45737", + "pull_request:45738" + ], + "soft_pairs": [ + "pull_request:45737|pull_request:45738" + ], + "prepared_review_unit_hash": "cc2ade73a49ec6a8883a7a24c433988645588bcd049e8aad2bce56b1b24daa1c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same MusicgenMelody cache initialization bug in `forward()`: replacing `EncoderDecoderCache(DynamicCache, DynamicCache)` with a single `DynamicCache`. PR #45737 is the more complete duplicate because it bundles the fix with a regression test, while #45738 is the fix-only variant.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45737 is the best canonical PR because it contains the same code fix as #45738 plus a regression test, making it the most complete representation of the bug fix.", + "best_issue_reason": null, + "best_pr_reason": "PR #45737 is the strongest overall candidate: it fixes the concrete cache bug and adds a targeted regression test, whereas #45738 only changes the implementation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45737", + "right": "pull_request:45738", + "accept": true, + "reason": "Same underlying code-path bug in `musicgen_melody` cache setup; the diffs match on the core fix and the test-only addition in #45737 could plausibly live in the same PR as #45738." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet: both PRs make the same `musicgen_melody` cache initialization change, and #45737 only adds a regression test on top of that fix. The soft-edge accept is conservative enough because the diffs are mergeable into one PR for the same bugfix." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 27, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45861-2", + "nodes": [ + "pull_request:45861", + "pull_request:45903" + ], + "soft_pairs": [ + "pull_request:45861|pull_request:45903" + ], + "prepared_review_unit_hash": "798d0a6f2ef2b4ded29c5f297dd5c1b2da29fecc1d9b695a7a2963ad882829b9", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5546, + "estimated_input_tokens": 1387, + "estimated_eval_tokens": 3030 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5546, + "estimated_input_tokens": 1387, + "estimated_eval_tokens": 3030 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs target the same Qwen3-VL M-RoPE `inv_freq` device-handling path, with 45903 extending the narrower device-mismatch fix in 45861 to also cover meta/to_empty re-initialization and buffer lifecycle correctness.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "45903 is the more complete and robust change: it covers the same Qwen3-VL M-RoPE device mismatch plus the meta/to_empty re-init edge case, with explicit buffer state tracking and a forward-time reinitialization guard.", + "best_issue_reason": null, + "best_pr_reason": "45903 is the best representative because it addresses the underlying `inv_freq` placement problem more comprehensively and includes the additional initialization fix that 45861 does not.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45861", + "right": "pull_request:45903", + "accept": true, + "reason": "Same concrete M-RoPE `inv_freq` device-path bug in Qwen3-VL family; 45903 is a broader superset of the narrower CPU-offload device fix in 45861 and could plausibly be merged as one change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + } + ] +} diff --git a/snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/manifest.json b/snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..58063b46e822bcca22c3b7fa4e3bd124f1c66cf9 --- /dev/null +++ b/snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260513t060025z", + "artifacts": { + "hybrid": "snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-05-13T06:15:26Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260513T060025Z", + "variant": "hybrid" +} diff --git a/snapshots/20260513T060025Z/manifest.json b/snapshots/20260513T060025Z/manifest.json index a79f81cacb3a06b1acbfe6d8c86b4e4d02f100ca..e77b97004240b62e08e0aaeabb27e9c88da62fc7 100644 --- a/snapshots/20260513T060025Z/manifest.json +++ b/snapshots/20260513T060025Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 958 }, "extracted_at": "2026-05-13T06:00:25Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260513t060025z", + "runs": { + "hybrid-model-20260513t060025z": { + "analysis_id": "hybrid-model-20260513t060025z", + "artifacts": { + "hybrid": "snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260513T060025Z/analysis-runs/hybrid-model-20260513t060025z/manifest.json", + "model": null, + "published_at": "2026-05-13T06:15:26Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260513T060025Z", "watermark": {