diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index f8cf091455aae2c232ccf7ab465f7f9826282fb2..f423eb4fba86a2afa7d0f533b68750a0da036cb3 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": "20260629T060016Z", - "generated_at": "2026-06-29T06:10:08Z", + "snapshot_id": "20260629T120018Z", + "generated_at": "2026-06-29T12:11:18Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ @@ -1050,108 +1050,128 @@ ] }, { - "cluster_id": "cluster-46894-8", - "summary": "Cluster of 3 issues and 5 PRs centered on issue #46894.", + "cluster_id": "cluster-41211-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41211.", "status": "open", - "confidence": 0.8, - "canonical_issue_number": 46894, - "canonical_pr_number": 46903, + "confidence": 0.55, + "canonical_issue_number": 41211, + "canonical_pr_number": 44339, "issue_numbers": [ - 46894, - 46897, - 46901 + 41211 ], "pr_numbers": [ - 46895, - 46898, - 46903, - 46916, - 46947 + 41356, + 44339 ], "evidence_types": [ - "closing_reference", - "shared_issue_target", - "soft_similarity" + "closing_reference" ], "pr_comparisons": [ { - "left_pr_number": 46895, - "right_pr_number": 46898, - "code_similarity": 0.117, - "size_similarity": 0.467, - "file_overlap": 0.0, - "area_overlap": 0.0, - "patch_similarity": 0.156, - "shared_filenames": [], - "shared_file_areas": [] - }, - { - "left_pr_number": 46895, - "right_pr_number": 46903, - "code_similarity": 0.24, - "size_similarity": 0.844, - "file_overlap": 0.0, - "area_overlap": 0.0, - "patch_similarity": 0.476, - "shared_filenames": [], - "shared_file_areas": [] - }, - { - "left_pr_number": 46895, - "right_pr_number": 46916, - "code_similarity": 0.335, - "size_similarity": 0.622, - "file_overlap": 0.4, - "area_overlap": 0.194, - "patch_similarity": 0.154, + "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": [ - "src/transformers/models/cohere_asr/modeling_cohere_asr.py", - "src/transformers/models/cohere_asr/modular_cohere_asr.py" + "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": "src/transformers/models/cohere_asr/modeling_cohere_asr.py", + "filename": "docs/source/en/model_doc/deimv2.md", "left_ranges": [ [ - 615, - 627 - ], + 1, + 132 + ] + ], + "right_ranges": [ [ - 639, - 654 + 1, + 65 + ] + ] + }, + { + "filename": "src/transformers/models/deimv2/__init__.py", + "left_ranges": [ + [ + 1, + 15 ] ], "right_ranges": [ [ - 637, - 644 + 1, + 29 ] ] }, { - "filename": "src/transformers/models/cohere_asr/modular_cohere_asr.py", + "filename": "src/transformers/models/deimv2/configuration_deimv2.py", "left_ranges": [ [ - 478, - 490 - ], + 1, + 74 + ] + ], + "right_ranges": [ [ - 502, - 517 + 1, + 266 + ] + ] + }, + { + "filename": "tests/models/deimv2/test_modeling_deimv2.py", + "left_ranges": [ + [ + 1, + 15 ] ], "right_ranges": [ [ - 500, - 507 + 1, + 1734 ] ] } ] - }, + } + ] + }, + { + "cluster_id": "cluster-46894-4", + "summary": "Cluster of 2 issues and 2 PRs centered on issue #46894.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 46894, + "canonical_pr_number": 46895, + "issue_numbers": [ + 46894, + 46897 + ], + "pr_numbers": [ + 46895, + 46958 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ { "left_pr_number": 46895, - "right_pr_number": 46947, + "right_pr_number": 46958, "code_similarity": 0.715, "size_similarity": 0.833, "file_overlap": 1.0, @@ -1233,263 +1253,6 @@ ] } ] - }, - { - "left_pr_number": 46898, - "right_pr_number": 46903, - "code_similarity": 0.125, - "size_similarity": 0.553, - "file_overlap": 0.0, - "area_overlap": 0.0, - "patch_similarity": 0.099, - "shared_filenames": [], - "shared_file_areas": [] - }, - { - "left_pr_number": 46898, - "right_pr_number": 46916, - "code_similarity": 0.42, - "size_similarity": 0.75, - "file_overlap": 0.5, - "area_overlap": 0.194, - "patch_similarity": 0.35, - "shared_filenames": [ - "src/transformers/models/florence2/modeling_florence2.py", - "src/transformers/models/florence2/modular_florence2.py" - ], - "shared_file_areas": [ - { - "filename": "src/transformers/models/florence2/modeling_florence2.py", - "left_ranges": [ - [ - 872, - 878 - ], - [ - 881, - 887 - ], - [ - 904, - 918 - ] - ], - "right_ranges": [ - [ - 902, - 909 - ] - ] - }, - { - "filename": "src/transformers/models/florence2/modular_florence2.py", - "left_ranges": [ - [ - 1580, - 1586 - ], - [ - 1589, - 1595 - ], - [ - 1612, - 1626 - ] - ], - "right_ranges": [ - [ - 1610, - 1617 - ] - ] - } - ] - }, - { - "left_pr_number": 46898, - "right_pr_number": 46947, - "code_similarity": 0.126, - "size_similarity": 0.56, - "file_overlap": 0.0, - "area_overlap": 0.0, - "patch_similarity": 0.092, - "shared_filenames": [], - "shared_file_areas": [] - }, - { - "left_pr_number": 46903, - "right_pr_number": 46916, - "code_similarity": 0.17, - "size_similarity": 0.737, - "file_overlap": 0.0, - "area_overlap": 0.0, - "patch_similarity": 0.154, - "shared_filenames": [], - "shared_file_areas": [] - }, - { - "left_pr_number": 46903, - "right_pr_number": 46947, - "code_similarity": 0.269, - "size_similarity": 0.987, - "file_overlap": 0.0, - "area_overlap": 0.0, - "patch_similarity": 0.475, - "shared_filenames": [], - "shared_file_areas": [] - }, - { - "left_pr_number": 46916, - "right_pr_number": 46947, - "code_similarity": 0.447, - "size_similarity": 0.747, - "file_overlap": 0.4, - "area_overlap": 0.438, - "patch_similarity": 0.167, - "shared_filenames": [ - "src/transformers/models/cohere_asr/modeling_cohere_asr.py", - "src/transformers/models/cohere_asr/modular_cohere_asr.py" - ], - "shared_file_areas": [ - { - "filename": "src/transformers/models/cohere_asr/modeling_cohere_asr.py", - "left_ranges": [ - [ - 637, - 644 - ] - ], - "right_ranges": [ - [ - 22, - 28 - ], - [ - 638, - 645 - ] - ] - }, - { - "filename": "src/transformers/models/cohere_asr/modular_cohere_asr.py", - "left_ranges": [ - [ - 500, - 507 - ] - ], - "right_ranges": [ - [ - 16, - 22 - ], - [ - 501, - 508 - ] - ] - } - ] - } - ] - }, - { - "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 - ] - ] - } - ] } ] }, @@ -2244,143 +2007,214 @@ ] }, { - "cluster_id": "cluster-46504-3", - "summary": "Cluster of 3 related pull requests linked by soft_similarity.", + "cluster_id": "cluster-45332-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", "status": "open", "confidence": 0.5, "canonical_issue_number": null, - "canonical_pr_number": 46716, + "canonical_pr_number": 45333, "issue_numbers": [], "pr_numbers": [ - 46504, - 46563, - 46716 + 45332, + 45333 ], "evidence_types": [ "soft_similarity" ], "pr_comparisons": [ { - "left_pr_number": 46504, - "right_pr_number": 46563, - "code_similarity": 0.185, - "size_similarity": 0.923, - "file_overlap": 0.0, - "area_overlap": 0.0, - "patch_similarity": 0.0, - "shared_filenames": [], - "shared_file_areas": [] - }, - { - "left_pr_number": 46504, - "right_pr_number": 46716, - "code_similarity": 0.2, - "size_similarity": 1.0, - "file_overlap": 0.0, - "area_overlap": 0.0, - "patch_similarity": 0.0, - "shared_filenames": [], - "shared_file_areas": [] - }, - { - "left_pr_number": 46563, - "right_pr_number": 46716, - "code_similarity": 0.69, - "size_similarity": 0.923, - "file_overlap": 1.0, - "area_overlap": 0.571, - "patch_similarity": 0.033, + "left_pr_number": 45332, + "right_pr_number": 45333, + "code_similarity": 0.612, + "size_similarity": 0.35, + "file_overlap": 0.389, + "area_overlap": 0.988, + "patch_similarity": 0.531, "shared_filenames": [ - ".github/workflows/pr-ci-security-gate.yml" + "docs/source/en/_toctree.yml", + "docs/source/en/heterogeneous_configurations.md", + "src/transformers/configuration_utils.py", + "src/transformers/integrations/heterogeneity/__init__.py", + "src/transformers/integrations/heterogeneity/configuration_utils.py", + "tests/heterogeneity/__init__.py", + "tests/heterogeneity/test_configuration_utils.py" ], "shared_file_areas": [ { - "filename": ".github/workflows/pr-ci-security-gate.yml", + "filename": "docs/source/en/_toctree.yml", "left_ranges": [ [ - 13, - 19 + 16, + 23 ] ], "right_ranges": [ [ - 14, - 17 + 16, + 23 ] ] - } - ] - } - ] - }, - { - "cluster_id": "cluster-43656-2", - "summary": "Cluster of 2 related pull requests linked by soft_similarity.", - "status": "open", - "confidence": 0.5, - "canonical_issue_number": null, - "canonical_pr_number": 43656, - "issue_numbers": [], - "pr_numbers": [ - 43656, - 43842 - ], - "evidence_types": [ - "soft_similarity" - ], - "pr_comparisons": [ - { - "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", + "filename": "docs/source/en/heterogeneous_configurations.md", "left_ranges": [ [ - 11, - 18 - ], + 1, + 206 + ] + ], + "right_ranges": [ + [ + 1, + 206 + ] + ] + }, + { + "filename": "src/transformers/configuration_utils.py", + "left_ranges": [ [ 30, 36 ], [ - 315, - 323 + 123, + 129 + ], + [ + 191, + 198 + ], + [ + 306, + 315 ], [ - 665, - 671 + 320, + 328 ], [ - 931, - 937 + 447, + 454 ], [ - 1843, - 1849 + 857, + 865 + ], + [ + 963, + 971 + ], + [ + 989, + 1004 + ], + [ + 1027, + 1035 ], [ - 1868, - 1874 + 1080, + 1088 + ], + [ + 1367, + 1376 ] ], "right_ranges": [ [ - 54, - 61 + 30, + 36 + ], + [ + 123, + 129 + ], + [ + 191, + 198 + ], + [ + 306, + 315 + ], + [ + 320, + 328 + ], + [ + 447, + 454 + ], + [ + 857, + 865 + ], + [ + 963, + 971 + ], + [ + 989, + 1004 + ], + [ + 1027, + 1035 + ], + [ + 1080, + 1088 ], [ - 587, - 593 + 1367, + 1376 + ] + ] + }, + { + "filename": "src/transformers/integrations/heterogeneity/__init__.py", + "left_ranges": [ + [ + 1, + 35 + ] + ], + "right_ranges": [ + [ + 1, + 22 + ] + ] + }, + { + "filename": "src/transformers/integrations/heterogeneity/configuration_utils.py", + "left_ranges": [ + [ + 1, + 392 + ] + ], + "right_ranges": [ + [ + 1, + 392 + ] + ] + }, + { + "filename": "tests/heterogeneity/test_configuration_utils.py", + "left_ranges": [ + [ + 1, + 343 + ] + ], + "right_ranges": [ + [ + 1, + 343 ] ] } @@ -2666,13 +2500,12 @@ ], "duplicate_issues": [ { - "cluster_id": "cluster-46894-8", + "cluster_id": "cluster-46894-4", "canonical_issue_number": 46894, "duplicate_issue_numbers": [ - 46897, - 46901 + 46897 ], - "reason": "Issues in cluster-46894-8 are treated as duplicates because they share closing_reference, shared_issue_target, soft_similarity evidence." + "reason": "Issues in cluster-46894-4 are treated as duplicates because they share closing_reference, shared_issue_target, soft_similarity evidence." } ], "duplicate_prs": [ @@ -2694,15 +2527,6 @@ "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-43656-2", - "canonical_pr_number": 43656, - "duplicate_pr_numbers": [ - 43842 - ], - "target_issue_number": null, - "reason": "PRs in cluster-43656-2 are treated as duplicates because they share soft_similarity evidence." - }, { "cluster_id": "cluster-43698-3", "canonical_pr_number": 43779, @@ -2747,6 +2571,15 @@ "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-45332-2", + "canonical_pr_number": 45333, + "duplicate_pr_numbers": [ + 45332 + ], + "target_issue_number": null, + "reason": "PRs in cluster-45332-2 are treated as duplicates because they share soft_similarity evidence." + }, { "cluster_id": "cluster-45561-3", "canonical_pr_number": 45645, @@ -2783,16 +2616,6 @@ "target_issue_number": 46305, "reason": "PRs in cluster-46305-3 are treated as duplicates because they converge on issue #46305 with closing_reference, shared_issue_target, soft_similarity evidence." }, - { - "cluster_id": "cluster-46504-3", - "canonical_pr_number": 46716, - "duplicate_pr_numbers": [ - 46504, - 46563 - ], - "target_issue_number": null, - "reason": "PRs in cluster-46504-3 are treated as duplicates because they share soft_similarity evidence." - }, { "cluster_id": "cluster-46869-3", "canonical_pr_number": 46886, @@ -2803,23 +2626,20 @@ "reason": "PRs in cluster-46869-3 are treated as duplicates because they converge on issue #46869 with closing_reference, shared_issue_target evidence." }, { - "cluster_id": "cluster-46894-8", - "canonical_pr_number": 46903, + "cluster_id": "cluster-46894-4", + "canonical_pr_number": 46895, "duplicate_pr_numbers": [ - 46895, - 46898, - 46916, - 46947 + 46958 ], "target_issue_number": 46894, - "reason": "PRs in cluster-46894-8 are treated as duplicates because they converge on issue #46894 with closing_reference, shared_issue_target, soft_similarity evidence." + "reason": "PRs in cluster-46894-4 are treated as duplicates because they converge on issue #46894 with closing_reference, shared_issue_target, soft_similarity evidence." } ], "best_issue": { - "cluster_id": "cluster-46894-8", + "cluster_id": "cluster-46894-4", "issue_number": 46894, - "reason": "Issue #46894 is the strongest global issue candidate because it is open, belongs to a cluster with 8 artifacts, and carries 0 discussion comments plus 9 inbound references.", - "score": 189.0 + "reason": "Issue #46894 is the strongest global issue candidate because it is open, belongs to a cluster with 4 artifacts, and carries 0 discussion comments plus 6 inbound references.", + "score": 146.0 }, "best_pr": { "cluster_id": "cluster-43979-11", diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index 977b0d5158722538c14a735f46370b2238d6754d..ad0ec39cd94baaca5c0772d3e9f8d7f6a21f99bb 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": "20260629T060016Z", - "generated_at": "2026-06-29T06:10:08Z", + "snapshot_id": "20260629T120018Z", + "generated_at": "2026-06-29T12:11:18Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -10,7 +10,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 1, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -98,32 +98,32 @@ "issue:44292|issue:44457", "issue:44841|issue:46661", "issue:43716|issue:46869", - "issue:24643|issue:30277", "issue:34928|issue:44805", - "issue:45859|issue:45902", + "issue:24643|issue:30277", "issue:45685|issue:46869", + "issue:45859|issue:45902", "issue:43638|issue:45161", "issue:34928|issue:36331", "issue:43596|issue:45161", "issue:43716|issue:45507", - "issue:44292|issue:45237", "issue:45874|issue:46294", + "issue:44292|issue:45237", "issue:39290|issue:45081", - "issue:45800|issue:46515", "issue:45823|issue:45874", + "issue:45800|issue:46515", "issue:43425|issue:44292", "issue:34928|issue:41720", "issue:43828|issue:45161", - "issue:43541|issue:43716", "issue:43828|issue:44292", + "issue:43541|issue:43716", "issue:43716|issue:46468", "issue:44292|issue:44918", "issue:45507|issue:45685", "issue:43716|issue:44292", - "issue:43716|issue:45685", "issue:44918|issue:45161", - "issue:43039|issue:44918", + "issue:43716|issue:45685", "issue:41762|issue:43638", + "issue:43039|issue:44918", "issue:43827|issue:44512", "issue:43828|issue:45685", "issue:43039|issue:43828", @@ -132,370 +132,478 @@ "issue:45237|issue:46015", "issue:43039|issue:44805" ], - "prepared_review_unit_hash": "17c54a30dd9b71f88a9f9d4a15a652dc0759487242a7f8f28a1c4a3cff92f2f6", + "prepared_review_unit_hash": "0d38d2f8a2898a5e56f0042f457ce204b7262d5b8593c484ed6265c9a6f26983", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 63, - "serialized_chars": 46577, - "estimated_input_tokens": 11645, - "estimated_eval_tokens": 23546 + "serialized_chars": 46569, + "estimated_input_tokens": 11643, + "estimated_eval_tokens": 23542 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of unrelated model/runtime/docs bugs with a few loose similarity chains. I only see two clear duplicate pairs: the DeepSpeed ZeRO-3 Bert index-out-of-bounds reports, and the torch 2.4.1 incompatibility reports. Most other soft edges are broader subsystem overlap, not the same concrete bug.", - "confidence": 0.79, - "canonical_issue_reason": "Issue 43596 is the cleanest true duplicate anchor: it matches 43638 on the same DeepSpeed ZeRO-3 + BertModel index-out-of-bounds failure mode, with only a minor wording difference about pretrained vs non-pretrained Bert.", + "summary": "This cluster is mostly heterogeneous, with a few true duplicate subclusters: DeepSpeed ZeRO-3 index-0-out-of-bounds reports around BERT/Gemma3, a Qwen3.5 MoE rotary-embedding issue, and a torch 2.4.1 compatibility pair. Most other soft edges are only loosely theme-similar and should stay separate.", + "confidence": 0.72, + "canonical_issue_reason": "Best anchor is issue 43638: it sits at the center of the clearest duplicate subcluster (43596 and 41762) and captures the shared ZeRO-3 index-0-out-of-bounds failure more generically than the model-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43596 is the best overall issue to keep as the canonical duplicate target because it is a concrete, reproducible bug with a near-identical duplicate, while the rest of the cluster is largely heterogeneous.", + "best_issue_reason": "Issue 43638 is the strongest representative of the concrete ZeRO-3/index-error bug family in this set, with clean duplicate evidence from both 43596 and 41762.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44322", "right": "issue:45464", "accept": false, - "reason": "Different failures: Qwen3_5 config attribute mismatch vs streaming chat/completions inference error." + "reason": "Different failures: Qwen3.5 config attribute error vs streaming chat/completions API failure." }, { "left": "issue:45237", "right": "issue:45685", "accept": false, - "reason": "Both are hardware/runtime MoE issues, but not the same bug or code path." + "reason": "Unrelated hardware-specific bugs; AMD GPU failure is not the same as an MPS histogram-op limitation." }, { "left": "issue:43716", "right": "issue:45237", "accept": false, - "reason": "Different models and symptoms: Mistral-3 image dtype mismatch vs GPT-OSS AMD runtime failure." + "reason": "Both mention runtime problems, but the models, platforms, and concrete failures differ." }, { "left": "issue:43828", "right": "issue:45237", "accept": false, - "reason": "Both mention dtype issues, but these are separate model-specific bugs with different triggers." + "reason": "Only share a broad dtype/device theme; the actual model and error paths are different." }, { "left": "issue:45237", "right": "issue:46468", "accept": false, - "reason": "Different concrete problems: GPT-OSS on AMD GPUs vs a broken Nemotron example." + "reason": "GPT-OSS on AMD GPUs vs a broken Nemotron example are different issues." }, { "left": "issue:45237", "right": "issue:45507", "accept": false, - "reason": "MoE-related, but invalid method call is not the same as AMD GPU execution failure." + "reason": "Different model families and symptoms; not the same underlying bug." }, { "left": "issue:30277", "right": "issue:35545", "accept": false, - "reason": "DeepSpeed ZeRO-3 collective mismatch is unrelated to ModernBERT ONNX export." + "reason": "DeepSpeed collective mismatch and ONNX export failure are unrelated." }, { "left": "issue:35545", "right": "issue:42915", "accept": false, - "reason": "ModernBERT ONNX export and Qwen3Moe FP8 failure are unrelated model/export bugs." + "reason": "ModernBERT ONNX export and Qwen3Moe FP8 config failure are different bugs." }, { "left": "issue:35545", "right": "issue:41720", "accept": false, - "reason": "ONNX export issue is unrelated to Qwen3 auto device-map cuda assert." + "reason": "ONNX export error vs Qwen3 auto-device-map CUDA assert; same general area, different bug." }, { "left": "issue:45901", "right": "issue:46015", "accept": false, - "reason": "Table QA crash and outdated RAG docs are different issues." + "reason": "Table QA crash vs outdated RAG examples are not the same issue." }, { "left": "issue:44509", "right": "issue:44512", "accept": false, - "reason": "Both are v5 docs cleanup items, but they cover different removed doc content." + "reason": "Both are v5 docs cleanup, but they target different removed content and would not be one duplicate bug." }, { "left": "issue:35545", "right": "issue:36010", "accept": false, - "reason": "ONNX export bug is unrelated to GenerationMixin import breakage." + "reason": "ModernBERT export failure and GenerationMixin import error are unrelated." }, { "left": "issue:44918", "right": "issue:46165", "accept": false, - "reason": "TRL SFT embedding unpacking and resume-training/trust-remote-code are different failures." + "reason": "TRL SFT embedding unpacking and resume/trust_remote_code are different training issues." }, { "left": "issue:46165", "right": "issue:46869", "accept": false, - "reason": "Resume-training remote-code issue is unrelated to YOLO multi-GPU device-map failure." + "reason": "Remote-code resume failure and YOLO multi-GPU device-map failure are unrelated." }, { "left": "issue:44805", "right": "issue:45823", "accept": false, - "reason": "Mask shape IndexError and Gemma4 PLE device mismatch are different forward-path bugs." + "reason": "Mask-shape mismatch and Gemma4 PLE device-mapping failure are different forward-path bugs." }, { "left": "issue:43638", "right": "issue:44805", "accept": false, - "reason": "Zero-3 Bert index error is unrelated to mask/tensor shape mismatch." + "reason": "Both are runtime errors, but one is a ZeRO-3 init index error and the other is a mask/tensor shape mismatch." + }, + { + "left": "issue:43638", + "right": "issue:46165", + "accept": false, + "reason": "ZeRO-3 index error vs resume/trust_remote_code problem; no shared code-path." }, { "left": "issue:43596", "right": "issue:44805", "accept": false, - "reason": "Different errors and code paths: Bert zero-3 loading vs mask/indexing mismatch." + "reason": "Different failure modes: ZeRO-3 index error versus mask/tensor shape mismatch." + }, + { + "left": "issue:44805", + "right": "issue:46294", + "accept": false, + "reason": "These are different model-input bugs; one is mask indexing, the other is video feature/token mismatch." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Mask-shape mismatch is not the same as GPT-OSS TP failure." + }, + { + "left": "issue:46165", + "right": "issue:46468", + "accept": false, + "reason": "Remote-code resume failure and broken Nemotron example are unrelated." + }, + { + "left": "issue:44560", + "right": "issue:46165", + "accept": false, + "reason": "Qwen3-vl video StopIteration and training-resume remote-code issue are different." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "BarkProcessor voice_preset bug and ModernBERT ONNX export are unrelated." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Both involve DeepSpeed, but one is collective mismatch and the other is ZeRO-3 index error; different bugs." }, { "left": "issue:43596", "right": "issue:43638", "accept": true, - "reason": "Same DeepSpeed ZeRO-3 Bert loading IndexError; the pretrained/non-pretrained wording does not change the underlying bug." + "reason": "Same ZeRO-3 index-0-out-of-bounds bug during Bert model loading/init; only the model wording differs." }, { "left": "issue:43039", "right": "issue:45161", "accept": false, - "reason": "Liger Kernel cross-entropy usage is unrelated to GPT-OSS MoE tensor-parallel failure." + "reason": "Liger Kernel cross_entropy being called is not the same as GPT-OSS only-TP failure." }, { "left": "issue:43975", "right": "issue:44457", "accept": false, - "reason": "Incorrect detokenization and LoRA merge/save-load mismatch are unrelated bugs." + "reason": "Incorrect detokenization and LoRA merge/save inconsistency are different behaviors." }, { "left": "issue:44292", "right": "issue:44457", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime error and LoRA save/reload mismatch are different issues." + "reason": "Qwen-3-8B NVFP4 runtime failure vs LoRA merge/save mismatch are unrelated." }, { "left": "issue:44841", "right": "issue:46661", "accept": false, - "reason": "Voxtral processor failure is unrelated to remote-code model_class config_class AttributeError." + "reason": "Processor failure for Voxtral and remote-code config_class error are different problems." }, { "left": "issue:43716", "right": "issue:46869", "accept": false, - "reason": "Mistral-3 dtype mismatch is unrelated to YOLO multi-GPU device-map failure." - }, - { - "left": "issue:24643", - "right": "issue:30277", - "accept": false, - "reason": "DeepSpeed training weight dimensionality error and collective mismatch are different bugs." + "reason": "Mistral dtype mismatch and YOLO multi-GPU device-map failure are not the same bug." }, { "left": "issue:34928", "right": "issue:44805", "accept": false, - "reason": "Activation-checkpointing/FSDP recompute mismatch is unrelated to the mask shape IndexError." + "reason": "Activation-checkpointing/FSDP recomputation mismatch is not the same as a mask indexing error." }, { - "left": "issue:45859", - "right": "issue:45902", + "left": "issue:24643", + "right": "issue:30277", "accept": false, - "reason": "Same component family, but CPU offload incompatibility and uninitialized-memory materialization are different concrete bugs." + "reason": "DeepSpeed training error about weight shape and DeepSpeed collective mismatch are different issues." }, { "left": "issue:45685", "right": "issue:46869", "accept": false, - "reason": "MPS histogram error and YOLO multi-GPU device-map failure are not the same issue." + "reason": "MPS histogram-op limitation is unrelated to YOLO multi-GPU auto device mapping." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": true, + "reason": "Same Qwen3.5 MoE rotary-embedding bug family; both point to bad `inv_freq` handling across device/materialization/offload paths." }, { "left": "issue:43638", "right": "issue:45161", "accept": false, - "reason": "Bert zero-3 loading error is unrelated to GPT-OSS tensor-parallel failure." + "reason": "ZeRO-3 index error and GPT-OSS TP issue do not share the same underlying failure." }, { - "left": "issue:34928", - "right": "issue:36331", + "left": "issue:43828", + "right": "issue:44292", "accept": false, - "reason": "FSDP activation-checkpointing size mismatch and CustomTrainer.compute_loss signature break are different problems." + "reason": "Both are model runtime failures, but the actual mechanisms and models differ." }, { - "left": "issue:43596", - "right": "issue:45161", + "left": "issue:43541", + "right": "issue:43716", "accept": false, - "reason": "Bert zero-3 loading error is unrelated to GPT-OSS MoE tensor-parallel failure." + "reason": "Mixtral grouped_mm tracing failure and Mistral-3 dtype mismatch are different bugs." }, { "left": "issue:43716", - "right": "issue:45507", + "right": "issue:46468", "accept": false, - "reason": "Both involve MoE-ish models, but image dtype mismatch and invalid method call are not the same bug." + "reason": "Different model/example and different failure modes." }, { "left": "issue:44292", - "right": "issue:45237", + "right": "issue:44918", "accept": false, - "reason": "Qwen NVFP4 runtime failure and GPT-OSS AMD failure are different model/runtime issues." + "reason": "Qwen NVFP4 runtime failure and TRL SFT embedding unpacking are unrelated." }, { - "left": "issue:45874", - "right": "issue:46294", + "left": "issue:45507", + "right": "issue:45685", "accept": false, - "reason": "Inputs_embeds reverse-lookup cost and video token/template mismatch are unrelated multimodal issues." + "reason": "GraniteMoEHybrid invalid method and MPS histogram-op limitation are unrelated." }, { - "left": "issue:39290", - "right": "issue:45081", + "left": "issue:43716", + "right": "issue:44292", "accept": false, - "reason": "Gemma3 config attribute regression and Mistral tokenizer regex patch crash are unrelated." + "reason": "Mistral dtype mismatch vs Qwen NVFP4 failure; same broad area, different bugs." }, { - "left": "issue:45800", - "right": "issue:46515", + "left": "issue:44918", + "right": "issue:45161", + "accept": false, + "reason": "TRL SFT embedding issue and GPT-OSS tensor-parallel issue are not the same underlying bug." + }, + { + "left": "issue:43716", + "right": "issue:45685", + "accept": false, + "reason": "Mistral dtype mismatch and MPS histogram-op limitation are different failures." + }, + { + "left": "issue:41762", + "right": "issue:43638", "accept": true, - "reason": "Both report the same torch 2.4.1 incompatibility regression; the titles are effectively the same bug report." + "reason": "Same ZeRO-3 index-0-out-of-bounds loading/init failure; model names differ but the underlying bug is the same." }, { - "left": "issue:45823", - "right": "issue:45874", + "left": "issue:43039", + "right": "issue:44918", + "accept": false, + "reason": "Liger Kernel cross_entropy call and TRL SFT embedding unpacking are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs regressions after v5, but they refer to different removed docs content and should not be merged as duplicates." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Autocast dtype mismatch and MPS int histogram limitation are different bugs." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Liger Kernel cross_entropy being called and an autocast dtype mismatch are not the same code-path problem." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch and custom trainer compute_loss signature break are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP/checkpointing tensor-size mismatch and GenerationMixin import error are unrelated." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "AMD GPU failure and outdated RAG examples are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44805", "accept": false, - "reason": "Gemma4 device mismatch and expensive reverse embedding lookup are different issues." + "reason": "Liger Kernel cross_entropy call and mask-shape mismatch are different bugs." + }, + { + "left": "issue:45800", + "right": "issue:46515", + "accept": true, + "reason": "Same torch 2.4.1 incompatibility report; 45800 is the more specific versioned instance of the same issue." }, { "left": "issue:43425", "right": "issue:44292", "accept": false, - "reason": "Torch 2.10 incompatibility and Qwen-3-8B-NVFP4 runtime failure are different version/model issues." + "reason": "Torch 2.10 incompatibility and Qwen-3-8B NVFP4 failure are different compatibility/runtime issues." }, { "left": "issue:34928", "right": "issue:41720", "accept": false, - "reason": "FSDP recompute mismatch is unrelated to Qwen3 auto device-map cuda asserts." + "reason": "FSDP activation-checkpointing mismatch and Qwen3 auto-device-map CUDA assert are unrelated." }, { "left": "issue:43828", "right": "issue:45161", "accept": false, - "reason": "Phi autocast dtype mismatch is unrelated to GPT-OSS MoE tensor-parallel failure." + "reason": "Phi autocast dtype mismatch and GPT-OSS TP failure do not share the same bug." }, { "left": "issue:43828", "right": "issue:44292", "accept": false, - "reason": "Both are precision/runtime failures, but they affect different models and code paths." + "reason": "Different models and failure mechanisms; both are runtime errors only at a very broad level." + }, + { + "left": "issue:43541", + "right": "issue:43716", + "accept": false, + "reason": "Mixtral grouped_mm tracing and Mistral-3 dtype mismatch are distinct issues." }, { "left": "issue:43716", "right": "issue:46468", "accept": false, - "reason": "Mistral-3 image dtype mismatch is unrelated to the Nemotron example breakage." + "reason": "Mistral-3 dtype mismatch and broken Nemotron example are unrelated." }, { "left": "issue:44292", "right": "issue:44918", "accept": false, - "reason": "Different model families and different failure modes." + "reason": "Qwen NVFP4 runtime failure vs TRL SFT input-embedding unpacking are different bugs." }, { "left": "issue:45507", "right": "issue:45685", "accept": false, - "reason": "Invalid method call in GraniteMoEHybrid and MPS histogram failure are different bugs." + "reason": "GraniteMoEHybrid invalid method and MPS histogram-op error are unrelated." }, { "left": "issue:43716", "right": "issue:44292", "accept": false, - "reason": "Different models and symptoms; no shared concrete failure path." - }, - { - "left": "issue:43716", - "right": "issue:45685", - "accept": false, - "reason": "Mistral image dtype mismatch and MPS histogram not implemented are unrelated." + "reason": "Different model-specific failures; not the same underlying bug." }, { "left": "issue:44918", "right": "issue:45161", "accept": false, - "reason": "TRL SFT embedding unpacking issue is unrelated to GPT-OSS tensor-parallel failure." + "reason": "TRl SFT trainer bug and GPT-OSS tensor-parallel bug are unrelated." }, { - "left": "issue:43039", - "right": "issue:44918", + "left": "issue:43716", + "right": "issue:45685", "accept": false, - "reason": "Liger cross-entropy call and Qwen3.5 embedding unpacking are unrelated." + "reason": "Mistral dtype mismatch and MPS histogram-op limitation are unrelated." }, { "left": "issue:41762", "right": "issue:43638", + "accept": true, + "reason": "Same ZeRO-3 index-0-out-of-bounds loading/init failure; model names differ but the underlying bug is the same." + }, + { + "left": "issue:43039", + "right": "issue:44918", "accept": false, - "reason": "Both are ZeRO-3 index errors, but Gemma3 and BertModel loading failures are different concrete bugs." + "reason": "Liger Kernel cross_entropy call and TRL SFT embedding issue are unrelated." }, { "left": "issue:43827", "right": "issue:44512", "accept": false, - "reason": "Both are docs regressions from v5 removals, but they reference different removed content." + "reason": "Related v5 docs cleanup, but different removed references; not the same duplicate change." }, { "left": "issue:43828", "right": "issue:45685", "accept": false, - "reason": "Phi autocast dtype mismatch and MPS histogram failure are unrelated." + "reason": "Autocast dtype mismatch and MPS int histogram limitation are different bugs." }, { "left": "issue:43039", "right": "issue:43828", "accept": false, - "reason": "Cross-entropy call under Liger Kernel and Phi autocast dtype mismatch are different issues." + "reason": "Liger Kernel cross_entropy call and autocast dtype mismatch are different failure paths." }, { "left": "issue:30277", "right": "issue:36331", "accept": false, - "reason": "Collective mismatch under ZeRO-3 is unrelated to trainer loss-signature breakage." + "reason": "DeepSpeed collective mismatch and custom trainer signature error are unrelated." }, { "left": "issue:34928", "right": "issue:36010", "accept": false, - "reason": "FSDP activation-checkpointing mismatch is unrelated to GenerationMixin import failure." + "reason": "FSDP/checkpointing mismatch and GenerationMixin import error are unrelated." }, { "left": "issue:45237", "right": "issue:46015", "accept": false, - "reason": "GPT-OSS AMD runtime failure and outdated RAG docs are unrelated." + "reason": "AMD GPU runtime failure and outdated RAG docs are unrelated." }, { "left": "issue:43039", "right": "issue:44805", "accept": false, - "reason": "Liger Kernel cross-entropy issue is unrelated to the mask shape IndexError." + "reason": "Different bugs: Liger Kernel cross_entropy call versus mask indexing mismatch." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative: it only promotes the two strongest duplicate pairs supported by the titles, and the other soft-edge rejections are appropriately cautious given the limited evidence. Canonical issue selection for 43596 over 43638 is reasonable, and the torch 2.4.1 pair is also a solid duplicate match." + "feedback": "The cluster summary is grounded in the packet and stays conservative overall. The three accepted subclusters are supported by the titles/body patterns: ZeRO-3 index-0-out-of-bounds loading errors, the Qwen3.5 MoE rotary/inv_freq bug family, and the torch 2.4.1 incompatibility pair. The rejected soft edges are mostly well-justified as different models, platforms, or failure modes, without overclaiming duplication." }, "evaluator_used": true, "retried": false, @@ -508,7 +616,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 2, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -566,13 +674,13 @@ "issue:46869" ], "soft_pairs": [ - "issue:45800|issue:46304", "issue:43866|issue:46402", + "issue:45800|issue:46304", "issue:43296|issue:46304", "issue:41762|issue:43596", "issue:39290|issue:41720", - "issue:44589|issue:46304", "issue:43901|issue:44512", + "issue:44589|issue:46304", "issue:45237|issue:45901", "issue:43975|issue:46015", "issue:43039|issue:44560", @@ -582,205 +690,205 @@ "issue:44863|issue:46015", "issue:45020|issue:46304", "issue:45812|issue:45920", - "issue:43872|issue:46304", "issue:43866|issue:44863", "issue:44292|issue:46015", + "issue:43872|issue:46304", "issue:39290|issue:43531", "issue:41093|issue:43638", "issue:43792|issue:45901", "issue:43854|issue:43866", "issue:43425|issue:45070", "issue:43975|issue:45901", + "issue:45362|issue:45464", "issue:43792|issue:45070", "issue:43596|issue:45800", - "issue:45362|issue:45464", "issue:43541|issue:46869", "issue:44292|issue:45901", "issue:39290|issue:45823", - "issue:45464|issue:45685", "issue:36010|issue:39290", + "issue:45464|issue:45685", "issue:43541|issue:45685", - "issue:43541|issue:43828", "issue:44841|issue:45084", - "issue:45507|issue:46869", + "issue:43541|issue:43828", "issue:43404|issue:43866", + "issue:45507|issue:46869", "issue:41093|issue:41762", "issue:45399|issue:46661", "issue:43572|issue:45070", "issue:45800|issue:45823" ], - "prepared_review_unit_hash": "8745935334327a1b4f9c8731a74ae7203cb3211260a68bf6cc05476ad3466bdf", + "prepared_review_unit_hash": "ee4d1220f04e282440873d1fbdc953a3cfbf9d35f3d304cea6a909a19a2694dd", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 42, - "serialized_chars": 40045, - "estimated_input_tokens": 10012, - "estimated_eval_tokens": 20280 + "serialized_chars": 40040, + "estimated_input_tokens": 10010, + "estimated_eval_tokens": 20276 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous noise, with two near-miss issue subclusters: ZeRO-3 load-time IndexErrors and AutoTokenizer token-ID regressions. The broadest, most representative issue is the multi-model tokenizer regression at 45920. No PRs are present.", - "confidence": 0.81, - "canonical_issue_reason": "issue:45920 is the best canonical issue because it describes the broadest tokenizer regression across multiple model families, making it the most representative issue in that subcluster.", + "summary": "This cluster is mostly noisy and heterogeneous: it spans unrelated model-loading, device-mapping, docs, and runtime bugs. The only soft edge that looks like a likely true duplicate is the Granite/AutoTokenizer regression pair (45812 \u2194 45920).", + "confidence": 0.33, + "canonical_issue_reason": "Issue 45920 is the broadest representative of the tokenizer-ID regression subcluster, covering multiple affected models rather than just Granite.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45920 is the strongest overall issue candidate: it covers the widest impact surface and generalizes the shared AutoTokenizer token-ID regression better than the Granite-only report.", + "best_issue_reason": "Issue 45920 is the best single anchor for the cluster\u2019s only plausible duplicate theme because it describes the wider v5 tokenizer regression across several models.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:45800", - "right": "issue:46304", + "left": "issue:43866", + "right": "issue:46402", "accept": false, - "reason": "Different regressions: torch/transformers version incompatibility vs USE_DISTRIBUTED=0 support breakage." + "reason": "Different bugs: corrupted checkpoint vs. text-encoder weights not loading." }, { - "left": "issue:43866", - "right": "issue:46402", + "left": "issue:45800", + "right": "issue:46304", "accept": false, - "reason": "Unrelated failures: corrupted Ovis2 checkpoint vs Sam3Model weight-loading bug." + "reason": "Both are version-compatibility reports, but for different torch/transformers breakages." }, { "left": "issue:43296", "right": "issue:46304", "accept": false, - "reason": "Different code paths: PaddleOCR-VL/vLLM load failure vs distributed build/runtime support regression." + "reason": "PaddleOCR-VL load failure is unrelated to USE_DISTRIBUTED=0 support." }, { "left": "issue:41762", "right": "issue:43596", "accept": false, - "reason": "Both are ZeRO-3 IndexErrors, but the affected models and load paths differ; this is not strong enough to treat as the same bug." + "reason": "Same symptom shape, but different models and likely different zero3 code paths." }, { "left": "issue:39290", "right": "issue:41720", "accept": false, - "reason": "Both are model-specific runtime errors, but the concrete attributes and failure modes are different." + "reason": "Gemma3 config attribute error and Qwen3 device-map CUDA assert are unrelated." }, { - "left": "issue:44589", - "right": "issue:46304", + "left": "issue:43901", + "right": "issue:44512", "accept": false, - "reason": "Different incompatibilities: missing Float8 storage vs torch distributed support breakage." + "reason": "Both are docs issues, but they document different removed/deprecated behaviors." }, { - "left": "issue:43901", - "right": "issue:44512", + "left": "issue:44589", + "right": "issue:46304", "accept": false, - "reason": "Both are docs issues, but they refer to different documentation changes and behaviors." + "reason": "Float8 storage lookup failure is unrelated to distributed build support." }, { "left": "issue:45237", "right": "issue:45901", "accept": false, - "reason": "Different problem domains: AMD GPU execution vs table-question-answering crash." + "reason": "AMD GPU model failure and table-question-answering crash are different problems." }, { "left": "issue:43975", "right": "issue:46015", "accept": false, - "reason": "Different text-related issues: detokenization regression vs outdated RAG examples." + "reason": "Detokenization regression is unrelated to outdated RAG examples." }, { "left": "issue:43039", "right": "issue:44560", "accept": false, - "reason": "Different failures: Liger Kernel cross_entropy path vs Qwen3-vl video StopIteration." + "reason": "Liger Kernel cross-entropy fallback and Qwen3.5 video StopIteration are unrelated." }, { "left": "issue:45823", "right": "issue:46294", "accept": false, - "reason": "Both are multimodal model issues, but one is device placement and the other is video chat-template mismatch." + "reason": "Device-map PLE mismatch and chat-template video-token mismatch are different failures." }, { "left": "issue:44841", "right": "issue:45823", "accept": false, - "reason": "Processor failure for Voxtral is unrelated to Gemma4 device-map mismatch." + "reason": "Voxtral processor failure is unrelated to Gemma4 forward device mismatch." }, { "left": "issue:43866", "right": "issue:46772", "accept": false, - "reason": "Corrupted checkpoint and NVFP4 compatibility are not the same underlying bug." + "reason": "Corrupted checkpoint vs. ModelOpt/NVFP4 compatibility are different issues." }, { "left": "issue:44863", "right": "issue:46015", "accept": false, - "reason": "NemotronH checkpoint loading is unrelated to RAG documentation drift." + "reason": "NemotronH checkpoint loading is unrelated to RAG documentation examples." }, { "left": "issue:45020", "right": "issue:46304", "accept": false, - "reason": "Remote-code model breakage is different from USE_DISTRIBUTED=0 support regression." + "reason": "Remote-code model breakage and USE_DISTRIBUTED=0 support are separate regression classes." }, { "left": "issue:45812", "right": "issue:45920", - "accept": false, - "reason": "Both are tokenizer regressions, but 45812 is a Granite-only report while 45920 spans multiple model families; this is plausible but not strong enough to merge confidently." - }, - { - "left": "issue:43872", - "right": "issue:46304", - "accept": false, - "reason": "bitsandbytes Int8Params API incompatibility is unrelated to distributed build/runtime support." + "accept": true, + "reason": "Same underlying tokenizer-ID regression; 45920 is the broader follow-up report across multiple models." }, { "left": "issue:43866", "right": "issue:44863", "accept": false, - "reason": "Different failures: corrupted Ovis2 checkpoint vs NemotronH loading issue." + "reason": "Corrupted checkpoint and NemotronH loading failure are not the same bug." }, { "left": "issue:44292", "right": "issue:46015", "accept": false, - "reason": "Qwen-3 NVFP4 execution error is unrelated to RAG docs cleanup." + "reason": "Qwen-3 NVFP4 execution failure is unrelated to RAG docs." + }, + { + "left": "issue:43872", + "right": "issue:46304", + "accept": false, + "reason": "bitsandbytes Int8Params incompatibility is unrelated to distributed-build support." }, { "left": "issue:39290", "right": "issue:43531", "accept": false, - "reason": "Both mention sliding-window-related issues, but the model families and concrete failures differ." + "reason": "Both mention sliding_window, but Gemma3 vLLM config and Qwen3-MoE sliding_window are distinct." }, { "left": "issue:41093", "right": "issue:43638", "accept": false, - "reason": "Both are IndexError reports with ZeRO-3, but the model and circumstances are not the same bug." + "reason": "Both are zero3-related IndexErrors, but the error signatures and model contexts differ." }, { "left": "issue:43792", "right": "issue:45901", "accept": false, - "reason": "Whisper load/run failure is unrelated to table-question-answering crashes." + "reason": "Whisper run failure is unrelated to table-question-answering crashes." }, { "left": "issue:43854", "right": "issue:43866", "accept": false, - "reason": "Model test-loading failure for GLM-4.7 is not the same as a corrupted Ovis2 checkpoint." + "reason": "GLM-4.7 unit-test loading issue and corrupted Ovis2 checkpoint are unrelated." }, { "left": "issue:43425", "right": "issue:45070", "accept": false, - "reason": "Torch 2.10 incompatibility and v5.4.0 pydantic-field breakage are different regressions." + "reason": "Torch compatibility and PretrainedConfig pydantic regression are different breakages." }, { "left": "issue:43975", @@ -788,117 +896,117 @@ "accept": false, "reason": "Detokenization regression is unrelated to table-question-answering crashes." }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Qwen3.5 chat crash and streaming inference API failure are different concrete issues." + }, { "left": "issue:43792", "right": "issue:45070", "accept": false, - "reason": "Whisper loading failure and pydantic model-field breakage are unrelated." + "reason": "Whisper run failure and pydantic PretrainedConfig regression are unrelated." }, { "left": "issue:43596", "right": "issue:45800", "accept": false, - "reason": "ZeRO-3 BERT loading failure is unrelated to torch 2.4.1/transformers 5.8.0 incompatibility." - }, - { - "left": "issue:45362", - "right": "issue:45464", - "accept": false, - "reason": "Different Qwen3.5 failures: chat crash vs streaming chat/completions API failure." + "reason": "Zero3 BertModel IndexError and torch/transformers version incompatibility are not the same bug." }, { "left": "issue:43541", "right": "issue:46869", "accept": false, - "reason": "Different runtime contexts and symptoms: Mixtral torch-dynamo grouped_mm error vs multi-GPU device-map failure." + "reason": "Mixtral grouped_mm tracing error and YOLO multi-GPU device-map failure are unrelated." }, { "left": "issue:44292", "right": "issue:45901", "accept": false, - "reason": "Different tasks and failure modes: NVFP4 model run error vs table-question-answering crash." + "reason": "NVFP4 execution issue is unrelated to table QA crashes." }, { "left": "issue:39290", "right": "issue:45823", "accept": false, - "reason": "Gemma3TextConfig missing field in vLLM and Gemma4 device-map mismatch are distinct bugs." - }, - { - "left": "issue:45464", - "right": "issue:45685", - "accept": false, - "reason": "Qwen3.5 streaming API failure is unrelated to the MPS histogram Int limitation." + "reason": "Gemma3 sliding_window attribute error and Gemma4 device mismatch are different model bugs." }, { "left": "issue:36010", "right": "issue:39290", "accept": false, - "reason": "ImportError for GenerationMixin and Gemma3TextConfig attribute error are different compatibility issues." + "reason": "GenerationMixin import error and Gemma3 config attribute error are unrelated." }, { - "left": "issue:43541", + "left": "issue:45464", "right": "issue:45685", "accept": false, - "reason": "Mixtral torch-dynamo grouped_mm error and MPS histogram limitation are different code paths." + "reason": "Streaming chat API failure and MPS histogram kernel error are unrelated." }, { "left": "issue:43541", - "right": "issue:43828", + "right": "issue:45685", "accept": false, - "reason": "Both involve MoE/dtype-related failures, but the concrete triggers and code paths differ." + "reason": "Grouped_mm tracing error and MPS Int histogram error are different failures." }, { "left": "issue:44841", "right": "issue:45084", "accept": false, - "reason": "Processor load failure and template-node compilation error are unrelated." + "reason": "Processor failure and compile/non-template-node error are unrelated." }, { - "left": "issue:45507", - "right": "issue:46869", + "left": "issue:43541", + "right": "issue:43828", "accept": false, - "reason": "GraniteMoEHybrid invalid-method call and YOLO multi-GPU device-map failure are different issues." + "reason": "Both involve MoE/runtime dtype problems, but they are different model paths and symptoms." }, { "left": "issue:43404", "right": "issue:43866", "accept": false, - "reason": "Missing lm_head tying and corrupted checkpoint are unrelated." + "reason": "lm_head weight tying bug and corrupted checkpoint are unrelated." + }, + { + "left": "issue:45507", + "right": "issue:46869", + "accept": false, + "reason": "GraniteMoEHybrid invalid method and YOLO device-map bug are different issues." }, { "left": "issue:41093", "right": "issue:41762", "accept": false, - "reason": "Both are ZeRO-3 IndexErrors, but the model/load path differs enough that they do not clearly look like the same bug." + "reason": "Both are IndexErrors, but one is mask-shape mismatch and the other is zero-size tensor during loading." }, { "left": "issue:45399", "right": "issue:46661", "accept": false, - "reason": "Flash-attn fallback blocking and remote-code model_class/config_class errors are different mechanisms." + "reason": "flash-attn fallback blocking and remote-code config_class errors are unrelated." }, { "left": "issue:43572", "right": "issue:45070", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig is unrelated to the pydantic PretrainedConfig field breakage." + "reason": "Missing pad_token_idx and PretrainedConfig pydantic regression are different breakages." }, { "left": "issue:45800", "right": "issue:45823", "accept": false, - "reason": "Torch/transformers version incompatibility and Gemma4 device mismatch are distinct failures." + "reason": "Torch compatibility and Gemma4 device-map mismatch are unrelated." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The high-level clustering summary is mostly grounded, but the soft-edge judgments are not conservative enough. In particular, the analyst marked `issue:41762` \u2194 `issue:43596` and `issue:45812` \u2194 `issue:45920` as accepted duplicates even though the packet\u2019s own `soft_pairs` list marks both pairs as `deterministic_accept: false`. Those pairings may be plausible, but the evidence here is not strong enough to accept confidently. The rest of the verdicts are mostly conservative, but these two overstate the similarity." + "accept": true, + "feedback": "The assessment is grounded in the packet and stays conservative. The only accepted soft edge, 45812 \u2194 45920, is well supported by the titles and discussion: both are AutoTokenizer wrong-token-ID regressions, with 45920 clearly the broader v5 regression report. The rest of the rejected pairs are appropriately treated as different bugs, often with different models or failure modes. No obvious overstatement stands out." }, "evaluator_used": true, - "retried": true, - "accepted_nontrivial_soft_edge": false, + "retried": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -907,7 +1015,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 3, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -966,16 +1074,16 @@ ], "soft_pairs": [ "issue:43296|issue:43366", - "issue:44805|issue:44918", "issue:44387|issue:45005", + "issue:44805|issue:44918", "issue:43541|issue:45161", "issue:43828|issue:45464", "issue:43366|issue:45464", "issue:44291|issue:45070", "issue:43531|issue:45070", "issue:41093|issue:43596", - "issue:45020|issue:45800", "issue:43716|issue:46294", + "issue:45020|issue:45800", "issue:43054|issue:43994", "issue:44479|issue:46294", "issue:46468|issue:46869", @@ -994,23 +1102,23 @@ "issue:44360|issue:44512", "issue:44492|issue:44512", "issue:44863|issue:45005", - "issue:45183|issue:46463", - "issue:46489|issue:46491" + "issue:46489|issue:46491", + "issue:45183|issue:46463" ], - "prepared_review_unit_hash": "60f2fb9d01e34ffea834174c98c84d37189f3a5426fd8f1ab930db718ee5b9bc", + "prepared_review_unit_hash": "8902d863cadbb8d9fef1d7d81c2f341abfb4edefe256784d9a508cb40d75ccd1", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36889, + "serialized_chars": 36890, "estimated_input_tokens": 9223, "estimated_eval_tokens": 18702 }, @@ -1019,205 +1127,208 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly a grab-bag of independent Transformers regressions across different models and subsystems. None of the soft pairs are safe duplicate merges.", - "confidence": 0.94, - "canonical_issue_reason": "No single canonical issue stands out: the set is heterogeneous, and even the closest-looking pairs (e.g. trust_remote_code/custom-module loading or tokenizer regressions) do not clearly share one underlying bug.", + "summary": "Most pairs are broad topical overlaps rather than true duplicates. The clearest duplicate candidates are the Qwen3.5 video regression pair and the input-validation error-message pair; the rest are different bugs in the same subsystem.", + "confidence": 0.61, + "canonical_issue_reason": "Issue 44479 is the broadest, most concrete regression report here: it names the affected model families and the specific video-input failure mode, making it the best anchor if these reports are grouped.", "canonical_pr_reason": null, - "best_issue_reason": "If forced to pick a representative, issue 45632 is the clearest root-cause-style report because it names a concrete failure mode (cache-path collision), but it is still not safe to treat as a duplicate anchor for the other items.", + "best_issue_reason": "Issue 44479 is the strongest representative issue overall because it describes a specific, user-facing regression with multiple affected models and a clear failure path.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43296", "right": "issue:43366", "accept": false, - "reason": "Different problems: PaddleOCR-VL load failure vs gpt-oss GGUF architecture support." + "reason": "Different problems: one is a PaddleOCR-VL load failure, the other is GGUF support for gpt-oss." }, { - "left": "issue:44805", - "right": "issue:44918", + "left": "issue:44387", + "right": "issue:45005", "accept": false, - "reason": "Both mention IndexError, but the symptoms and code paths differ: mask length mismatch vs Qwen3.5 SFT embedding unpacking." + "reason": "Unrelated bugs; int4 memory/OOM behavior is not the same as tied-weights handling." }, { - "left": "issue:44387", - "right": "issue:45005", + "left": "issue:44805", + "right": "issue:44918", "accept": false, - "reason": "Int4 memory/OOM regression is unrelated to tied-weights behavior in translation models." + "reason": "Both are shape-related errors, but they occur in different code paths and model/training setups." }, { "left": "issue:43541", "right": "issue:45161", "accept": false, - "reason": "Grouped_mm tracing failure in Mixtral is a different code-path problem from TP-only GPT-OSS MoE support." + "reason": "Different MoE/Tensor Parallel failures; no clear shared underlying bug." }, { "left": "issue:43828", "right": "issue:45464", "accept": false, - "reason": "Autocast dtype mismatch is not the same bug as Qwen3.5 streaming inference/chat-completions failure." + "reason": "One is an autocast dtype mismatch, the other is a streaming chat/completions failure." }, { "left": "issue:43366", "right": "issue:45464", "accept": false, - "reason": "GGUF support for gpt-oss is unrelated to Qwen3.5 streaming chat API failures." + "reason": "GGUF gpt-oss support and Qwen3.5 streaming inference are unrelated." }, { "left": "issue:44291", "right": "issue:45070", "accept": false, - "reason": "init_empty_weights argument handling and PretrainedConfig/pydantic field breakage are separate regressions." + "reason": "Both involve v5 compatibility, but they are distinct issues affecting different APIs." }, { "left": "issue:43531", "right": "issue:45070", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior does not match the pydantic PretrainedConfig issue." + "reason": "Qwen3-MoE sliding-window behavior is not the same as the PretrainedConfig/pydantic breakage." }, { "left": "issue:41093", "right": "issue:43596", "accept": false, - "reason": "Both are IndexErrors, but one is a mask/index length mismatch and the other is a zero-size tensor during deepspeed zero3 init." - }, - { - "left": "issue:45020", - "right": "issue:45800", - "accept": false, - "reason": "Generic remote_code breakage vs torch 2.4.1/transformers 5.8.0 incompatibility are too broad and not the same bug." + "reason": "Both raise IndexError, but the tensor mismatch and zero-sized init failures are different bugs." }, { "left": "issue:43716", "right": "issue:46294", "accept": false, - "reason": "Image preprocessor/model dtype mismatch is a different multimodal bug from video token/feature count mismatch." + "reason": "Different multimodal failures: dtype mismatch in preprocessing versus video feature/token mismatch." }, { - "left": "issue:43054", + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "Remote-code breakage and torch/transformers version incompatibility are separate problems." + }, + { + "left": "issue:43054", "right": "issue:43994", "accept": false, - "reason": "Siglip2 underperformance and nonsensical AutoModel/pipeline output do not identify one shared defect." + "reason": "Same model family, but the reports describe different symptoms and likely different defects." }, { "left": "issue:44479", "right": "issue:46294", - "accept": false, - "reason": "Both involve video inputs, but one is a v5.3.0 regression across several models and the other is an apply_chat_template feature/token mismatch." + "accept": true, + "reason": "Same video-input regression family; 46294 looks like a specific instance of the broader Qwen3.5-related bug in 44479." }, { "left": "issue:46468", "right": "issue:46869", "accept": false, - "reason": "Official Nemotron example breakage is unrelated to YOLO multi-GPU device_map auto failures." + "reason": "Official Nemotron example breakage and YOLO multi-GPU device-map failure are unrelated." }, { "left": "issue:45759", "right": "issue:45902", "accept": false, - "reason": "Composite config unwrapping and uninitialized rotary inv_freq after to_empty are different Qwen3.5 issues." + "reason": "Different Qwen 3.5/3.6 issues: composite config unwrapping versus uninitialized rotary embedding memory." }, { "left": "issue:43596", "right": "issue:46165", "accept": false, - "reason": "Zero3/BertModel index failure and resume-training trust_remote_code problems are distinct failure modes." + "reason": "Different loading/training failures with different causes; no common code-path evidence." }, { "left": "issue:46640", "right": "issue:46661", "accept": false, - "reason": "Flash Attention 2 issue and remote-code config_class AttributeError are not the same code path." + "reason": "Flash Attention fallback behavior and remote-code config_class errors are unrelated." }, { "left": "issue:43716", "right": "issue:44918", "accept": false, - "reason": "Different multimodal/loader bugs: image dtype mismatch vs Qwen3.5 embeddings with SFT trainer." + "reason": "Different multimodal/training bugs with no concrete shared mechanism." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "FSDP weight-rank error and Qwen3.5 SFT embedding issue are unrelated." + "reason": "FSDP weight-shape error and TRL embedding unpacking failure are different issues." }, { "left": "issue:40990", "right": "issue:44960", "accept": false, - "reason": "High perplexity on gpt-oss-20b and GLM5 are separate model-quality reports." + "reason": "No evidence these perplexity/GLM5 reports share the same underlying defect." }, { "left": "issue:43645", "right": "issue:45399", "accept": false, - "reason": "Notebook custom-model initialization and flash-attn2 fallback blocking are unrelated." + "reason": "Custom-model notebook initialization and Flash Attention fallback checks are unrelated." }, { "left": "issue:45632", "right": "issue:45698", "accept": false, - "reason": "Both are about trust_remote_code/custom-module loading, but one is a cache-path collision by leaf name and the other is a wrong module after save_pretrained; not enough to prove one bug." + "reason": "Both involve remote-code/custom module loading, but the reported failure modes are not concrete enough to treat as the same bug." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Untied lm_head in Mistral3 and tied-weights translation-model regressions are different weight-tying bugs." + "reason": "Tied lm_head weights in Mistral3 and translation-model tied-weight issues are different model families and fixes." }, { "left": "issue:42915", "right": "issue:45399", "accept": false, - "reason": "Qwen3MoE fine-grained FP8 failure is unrelated to flash-attn2 fallback gating." + "reason": "FineGrainedFP8Config failure and Flash Attention fallback blocking are unrelated." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantization/offloading memory behavior is not the same as tied-weights regressions." + "reason": "Quantization offloading/OOM behavior is not the same as tied-weight handling." }, { "left": "issue:44530", "right": "issue:45005", "accept": false, - "reason": "PagedAttentionCache linear_attention crash is unrelated to tied-weight handling." + "reason": "PagedAttentionCache linear_attention crash and tied-weight translation models are unrelated." }, { "left": "issue:44360", "right": "issue:44512", "accept": false, - "reason": "A DSA indexer discussion is not the same as a docs typo about the removed run command." + "reason": "A DSA indexer discussion and a docs typo are not duplicate bugs." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "Cache-strategy typo and removed run-command docs are separate documentation issues." + "reason": "Both are documentation/naming issues, but they are different changes." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "NemotronH checkpoint loading failure is unrelated to tied-weight translation regressions." - }, - { - "left": "issue:45183", - "right": "issue:46463", - "accept": false, - "reason": "The generic input_ids/inputs_embeds XOR validation bug and the T5 decoder-input message issue are different validation paths." + "reason": "NemotronH checkpoint loading and tied-weights translation models are unrelated." }, { "left": "issue:46489", "right": "issue:46491", "accept": false, - "reason": "Both are tokenizer regressions, but for different models and different symptoms; not the same underlying bug." + "reason": "Both are tokenizer regressions, but they affect different models and symptoms." + }, + { + "left": "issue:45183", + "right": "issue:46463", + "accept": true, + "reason": "Both describe the same validation/error-message bug class; 46463 appears to be a T5-specific manifestation of the broader XOR/input validation issue." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet and stays appropriately conservative. The two accepted soft pairs are well-supported as likely same-bug clusters, while the rejections avoid overclaiming despite topical similarity. The overall summary is careful and consistent with the issue titles/evidence provided." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -1226,7 +1337,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 4, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -1284,8 +1395,8 @@ "issue:46932" ], "soft_pairs": [ - "issue:44918|issue:45507", "issue:43975|issue:44292", + "issue:44918|issue:45507", "issue:43950|issue:44534", "issue:39692|issue:43054", "issue:43644|issue:43950", @@ -1304,222 +1415,234 @@ "issue:42947|issue:44387", "issue:45663|issue:45902", "issue:44403|issue:44451", - "issue:41553|issue:43881", "issue:43975|issue:45520", - "issue:44403|issue:44488", - "issue:24643|issue:28530", + "issue:41553|issue:43881", "issue:42947|issue:43856", + "issue:24643|issue:28530", "issue:43643|issue:45020", + "issue:44403|issue:44488", "issue:41762|issue:42915", "issue:44991|issue:45399", - "issue:43824|issue:43872", - "issue:43782|issue:44960" + "issue:43782|issue:44960", + "issue:43824|issue:43872" ], - "prepared_review_unit_hash": "1b3980e567f5a7f84aac091a9cdf5615b47c10d5a78f667a0d02dcd5d63a11f9", + "prepared_review_unit_hash": "3bcee8f123458ecbff84421eba671b47d7c61d2bc9326f6f575e3d815458b721", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36792, - "estimated_input_tokens": 9198, - "estimated_eval_tokens": 18652 + "serialized_chars": 36794, + "estimated_input_tokens": 9199, + "estimated_eval_tokens": 18654 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most pairs are unrelated and should stay separate. The clearest duplicate cluster is the transformer 5.x non-persistent-buffer regression, with a second likely duplicate around cache length returning tensors instead of ints.", - "confidence": 0.83, - "canonical_issue_reason": "issue:43950 is the strongest canonical issue: it describes a concrete v5 regression in non-persistent buffer handling and has the clearest duplicate signal (#44534), with a closely related follow-up (#45902).", + "summary": "Mostly unrelated issue reports with a few small duplicate subgroups. The clearest cluster is the non-persistent-buffer regression anchored by #43950, with #43644 and #44534 as strong matches. #45902 is adjacent but not clearly the same bug, so it should not be treated as a confirmed duplicate of that subgroup. Smaller duplicate families appear around SigLIP2 discrepancies and static-cache seq-length behavior.", + "confidence": 0.66, + "canonical_issue_reason": "#43950 is still the best canonical issue for the non-persistent-buffer regression because it states the broad underlying bug clearly and matches the strongest confirmed duplicates (#43644, #44534). #45902 is related in theme but not safely the same failure mode, so it should not broaden the canonical cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43950 is the best global representative because it is specific, reproducible, and central to the same loading/materialization regression family seen in the cluster.", + "best_issue_reason": "#43950 is the most representative issue overall because it captures the broadest confirmed regression in the packet and has the strongest supporting duplicate links.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44918", - "right": "issue:45507", + "left": "issue:43975", + "right": "issue:44292", "accept": false, - "reason": "Different subsystems and failures: Qwen3.5 embedding unpacking vs GraniteMoEHybrid invalid method call." + "reason": "Different model-specific failures: tokenizer detokenization vs. Qwen3 NVFP4 loading error." }, { - "left": "issue:43975", - "right": "issue:44292", + "left": "issue:44918", + "right": "issue:45507", "accept": false, - "reason": "Tokenizer detokenization bug vs a separate Qwen NVFP4 runtime/load failure; not the same code-path." + "reason": "Unrelated code paths: TRL SFT embedding unpacking vs. GraniteMoEHybrid invalid method call." }, { "left": "issue:43950", "right": "issue:44534", "accept": true, - "reason": "Same underlying regression: v5 corrupts/fills non-persistent buffers with junk during loading/materialization." + "reason": "Same underlying regression: non-persistent buffers are being corrupted/junk-filled in Transformers v5." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "Docs example/model-processor mismatch and quantization errors are unrelated to a generic SigLIP2 embedding-quality report." + "reason": "Docs example errors/quantization failure is not the same bug as degraded SigLIP2 text embeddings." }, { - "left": "issue:43643", + "left": "issue:43644", "right": "issue:43950", - "accept": false, - "reason": "Remote-code config field loss is a different bug from non-persistent buffer corruption." + "accept": true, + "reason": "Both report the same non-persistent buffer corruption regression, with #43644 phrased as the broader symptom." }, { "left": "issue:36331", "right": "issue:41762", "accept": false, - "reason": "CustomTrainer compute_loss signature mismatch is unrelated to Gemma3 ZeRO-3 loading failure." + "reason": "Different failures: custom loss signature mismatch vs. DeepSpeed ZeRO-3 loading index error." }, { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "bitsandbytes parameter incompatibility is unrelated to missing Float8 storage registration." + "reason": "Both are low-level runtime compatibility issues, but they are different errors and different code paths." }, { "left": "issue:46899", "right": "issue:46932", "accept": true, - "reason": "Same Gemma4 quantized multimodal embedder dtype-casting bug; #46932 explicitly says it fixes the same issue." + "reason": "Same Gemma4 quantized vision/audio embedder dtype-casting bug; #46932 explicitly references the same fix." }, { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Both are dtype/precision problems, but they affect different models and different failure mechanisms." + "reason": "Different symptoms and model paths: autocast dtype mismatch vs. Qwen3.5 position_ids/Flash Attention crash." }, { "left": "issue:45920", "right": "issue:46489", - "accept": false, - "reason": "Both are tokenizer regressions, but 46489 is a DeepSeek-Coder-specific gap while 45920 is a broader multi-model token-ID bug." + "accept": true, + "reason": "Same tokenizer regression family: #46489 is a specific remaining DeepSeek-Coder v1 case of the broader v5 token-id bug." }, { "left": "issue:45987", "right": "issue:46858", "accept": true, - "reason": "Same core bug: get_seq_length is returning a tensor instead of a plain int, breaking downstream arithmetic." + "reason": "Same cache API contract bug: get_seq_length() returning a tensor instead of an int breaks downstream arithmetic." }, { "left": "issue:43054", "right": "issue:45701", "accept": false, - "reason": "SigLIP2 embedding quality issue and generic tokenization changes are not the same bug." + "reason": "SigLIP2 embedding quality complaints are not the same as a generic tokenizer-version change report." }, { "left": "issue:38175", "right": "issue:43054", "accept": false, - "reason": "Both mention SigLIP2, but one is zero probabilities and the other is worse text embeddings; different symptoms and likely different causes." + "reason": "Related model family, but the reported symptoms differ enough that this is not clearly the same underlying bug." }, { "left": "issue:43950", "right": "issue:45902", - "accept": true, - "reason": "Same non-persistent-buffer regression family; 45902 is a specific manifestation involving inv_freq after meta\u2192to_empty(cuda)." + "accept": false, + "reason": "#45902 is a Qwen3_5Moe rotary-embedding uninitialized-memory issue after meta -> to_empty(cuda) materialization; that is related in spirit but not clearly the same non-persistent-buffer corruption regression as #43950." }, { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "Bad AutoTokenizer error messaging for Voxtral is unrelated to Molmo tied-weights attribute failure." + "reason": "Bad AutoTokenizer error messaging for Voxtral is unrelated to Molmo missing tied-weights metadata." }, { "left": "issue:43054", "right": "issue:43493", - "accept": false, - "reason": "SigLIP2 quality mismatch vs HF/JAX implementation discrepancy are related only at a high level, not the same concrete bug." + "accept": true, + "reason": "Both point to the same SigLIP2 implementation discrepancy relative to the reference/JAX behavior." }, { "left": "issue:44910", "right": "issue:45859", "accept": false, - "reason": "Flash-Attention packed-sequence misinterpretation is unrelated to CPU-offload incompatibility in rotary embedding." + "reason": "One is a FlashAttention crash from 3D position_ids; the other is CPU-offload incompatibility in rotary embedding." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "Gradient checkpointing/LoRA behavior is unrelated to int4 quantization memory growth." + "reason": "Gradient checkpointing/LoRA behavior is unrelated to int4 quantization OOM from reserved CUDA memory growth." }, { "left": "issue:45663", "right": "issue:45902", "accept": false, - "reason": "Gemma4 FSDP2 shared_kv_states bug is unrelated to non-persistent buffer materialization corruption." + "reason": "Different model subsystems: Gemma4 FSDP2 shared_kv_states rebuild vs. Qwen3_5Moe rotary embedding materialization." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "Unnecessary loading noise and a model-loading failure are different problems." + "reason": "Generic loading noise is not the same bug as a model actually failing to load." }, { - "left": "issue:41553", - "right": "issue:43881", + "left": "issue:43975", + "right": "issue:45520", "accept": false, - "reason": "Different model-loading failures with no shared concrete code-path." + "reason": "Tokenizer detokenization regression is unrelated to the Python 3.13 flash_attn import error." }, { - "left": "issue:43975", - "right": "issue:45520", + "left": "issue:41553", + "right": "issue:43881", "accept": false, - "reason": "Tokenizer detokenization bug is unrelated to import_utils flash_attn KeyError on Python 3.13." + "reason": "Bad tokenizer error messaging for Voxtral is unrelated to glm-4v-9b loading failure." }, { - "left": "issue:44403", - "right": "issue:44488", + "left": "issue:42947", + "right": "issue:43856", "accept": false, - "reason": "Generic loading noise and inability to load a specific model are not the same issue." + "reason": "Different training inefficiency problems: PEFT gradient checkpointing vs. Qwen3 MoE memory usage." }, { "left": "issue:24643", "right": "issue:28530", "accept": false, - "reason": "DeepSpeed weight-dimension error is unrelated to early-stopping metric lookup." + "reason": "DeepSpeed weight-shape runtime error is unrelated to early-stopping metric selection." }, { - "left": "issue:42947", - "right": "issue:43856", + "left": "issue:43643", + "right": "issue:45020", "accept": false, - "reason": "LoRA gradient checkpointing inefficiency is unrelated to Qwen3 MoE memory usage." + "reason": "Missing fields from AutoConfig with trust_remote_code=True is a narrower config-loading bug, not the same as broad remote_code breakage." }, { - "left": "issue:43643", - "right": "issue:45020", + "left": "issue:44403", + "right": "issue:44488", "accept": false, - "reason": "Broad remote_code breakages and missing fields in AutoConfig are too coarse to treat as the same concrete bug here." + "reason": "Generic loading noise is not the same underlying problem as failing to load a specific model." }, { - "left": "issue:43824", - "right": "issue:43872", + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Different model-loading failures: Gemma3 ZeRO-3 index error vs. Qwen3Moe FineGrainedFP8Config issue." + }, + { + "left": "issue:44991", + "right": "issue:45399", "accept": false, - "reason": "Missing Qwen2.5-VL import and bitsandbytes _is_hf_initialized incompatibility are unrelated." + "reason": "Tokenizer load failure and flash-attn fallback logic are unrelated." }, { "left": "issue:43782", "right": "issue:44960", "accept": false, - "reason": "Qwen3VL weight_only loading failure and GLM5 are unrelated." + "reason": "Qwen3VL weight_only=True load error is unrelated to GLM5." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "ImportError for Qwen2.5 VL is unrelated to the bitsandbytes Int8Params incompatibility." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and the soft-edge calls stay conservative. The accepted duplicates are well supported by the issue titles (#43950/#44534, #46899/#46932, #45987/#46858), and the rejections mostly involve clearly different subsystems or failure modes." + "accept": false, + "feedback": "Most of the pair judgments are conservative and grounded, but the cluster summary overreaches on the non-persistent-buffer subgroup. In particular, the acceptance of #43950 with #45902 is not well supported by the packet alone: #45902 is a Qwen3_5Moe rotary-embedding uninitialized-memory issue after `meta -> to_empty(cuda)` materialization, which is not clearly the same bug as the explicit non-persistent-buffer regression in #43950/#43644/#44534. Because that link is too strong, the summary\u2019s claim that #45902 is a narrower manifestation of the same regression is overstated. The rest of the summary is mostly fine, but this edge should be downgraded or marked uncertain." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -1529,7 +1652,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 5, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -1597,11 +1720,11 @@ "issue:45715|issue:46315", "issue:45488|issue:46489", "issue:43827|issue:44509", - "issue:43872|issue:46869", "issue:46894|issue:46897", "issue:45759|issue:46661", - "issue:43872|issue:44291", + "issue:43872|issue:46869", "issue:43404|issue:43950", + "issue:43872|issue:44291", "issue:45020|issue:46661", "issue:43792|issue:44220", "issue:43257|issue:45305", @@ -1615,214 +1738,208 @@ "issue:44479|issue:44560", "issue:41553|issue:44488" ], - "prepared_review_unit_hash": "edd407c1d8ad65acf1e3716c333206fcf9be35779dfaea3a686e070af3df3da0", + "prepared_review_unit_hash": "bef3d89e437bd0880b77e6e30370eb8dd306e07fe3eb400677d778bf4839db8a", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 36185, - "estimated_input_tokens": 9047, - "estimated_eval_tokens": 18350 + "serialized_chars": 36189, + "estimated_input_tokens": 9048, + "estimated_eval_tokens": 18352 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "A few genuine duplicates/near-duplicates appear in this mostly mixed cluster: one exact mask-shape error pair, one docs cleanup pair, and several same-regression pairs around loading/video/initialization. Most other pairs are only loosely related by subsystem or broad symptom and should stay separate.", - "confidence": 0.78, - "canonical_issue_reason": "issue:41093 is the best canonical issue because it is a concrete, well-scoped bug report and is an exact duplicate of issue:44805\u2019s same mask-shape IndexError.", + "summary": "This cluster is mostly a grab-bag of unrelated tokenizer/loading/regression issues. The clear duplicate groups are the _is_hf_initialized / init_empty_weights loading bug, the docs pipeline-removal cleanup, and the mask-shape IndexError; most other pairs are only loosely similar and should stay separate.", + "confidence": 0.76, + "canonical_issue_reason": "issue:44291 is the best issue anchor because it states the root cause and affected path for the _is_hf_initialized loading regression, while issue:43872 is the more symptom-focused variant.", "canonical_pr_reason": null, - "best_issue_reason": "issue:41093 is the strongest representative for canonical issue triage: specific failure, clear reproduction class, and a one-to-one duplicate match with issue:44805.", + "best_issue_reason": "issue:44291 is the strongest representative issue in this cluster: concrete traceback, explicit version scope, and a clear fix target make it the most useful duplicate canonical.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45685", "right": "issue:45823", "accept": false, - "reason": "Different failures: MPS histogram-on-Int in MoE vs device placement mismatch in Gemma4 PLE." + "reason": "Different failures: MPS histogram/int handling vs device-map forward mismatch. Same broad area, not the same bug." }, { "left": "issue:43881", "right": "issue:43883", "accept": false, - "reason": "Both are model-loading errors, but the concrete causes differ: GLM-4V load failure vs missing all_tied_weights_keys on MolmoForCausalLM." + "reason": "Both are model-loading failures, but the models and failure modes differ; no shared concrete code-path is evident." }, { "left": "issue:45081", "right": "issue:45399", "accept": false, - "reason": "Tokenizer regex crash and flash-attention fallback blocking are unrelated code paths." + "reason": "Tokenizer regex crash vs flash-attn fallback gating. Unrelated fixes." }, { "left": "issue:45020", "right": "issue:45823", "accept": false, - "reason": "Remote-code loading regressions and Gemma4 device-map mismatch are different bugs." + "reason": "Broad remote_code regression vs Gemma4 device mismatch. Too different to merge." }, { "left": "issue:44661", "right": "issue:45020", "accept": false, - "reason": "Different root causes: tokenizer-mapping edge case vs broader remote_code breakage." + "reason": "One is an add-new-model-like tokenizer-mapping bug; the other is a broader remote_code regression. Related ecosystem, not same bug." }, { "left": "issue:42491", "right": "issue:44368", "accept": false, - "reason": "Both mention Qwen/LoRA, but one is a hf4.x\u2192hf5.x compatibility break and the other is just a warning about tie_word_embeddings." + "reason": "Both mention tied embeddings, but one is a LoRA usability regression and the other is a warning/config issue. Not the same underlying defect." }, { "left": "issue:43828", "right": "issue:45823", "accept": false, - "reason": "Autocast dtype mismatch is not the same as the Gemma4 device-map forwarding bug." + "reason": "Autocast dtype mismatch and device-map mismatch are different code paths." }, { "left": "issue:45020", "right": "issue:45356", "accept": false, - "reason": "Remote-code loading breakage and Kimi tokenizer codec handling are different problems." + "reason": "Remote_code breakage vs Kimi tokenizer codec regression. Different tokenizer/model subsystems." }, { "left": "issue:45715", "right": "issue:46315", "accept": false, - "reason": "Performance regressions in opposite tokenizer conversion directions are not the same bug." + "reason": "Separate performance regressions in different conversion directions; not a single bug." }, { "left": "issue:45488", "right": "issue:46489", "accept": false, - "reason": "Both involve DeepSeek tokenizers, but one is a LlamaTokenizer pre-tokenizer override regression and the other is a separate DeepSeek-Coder output bug." + "reason": "Both are tokenizer output regressions, but they affect different models and likely different tokenizer overrides/fixes." }, { "left": "issue:43827", "right": "issue:44509", "accept": true, - "reason": "Same documentation cleanup: both still reference pipeline tasks removed in v5." - }, - { - "left": "issue:43872", - "right": "issue:46869", - "accept": false, - "reason": "Unexpected _is_hf_initialized TypeError is unrelated to YOLO multi-GPU device-map failure." + "reason": "Same documentation cleanup: both complain that removed v5 pipeline task references still remain in docs." }, { "left": "issue:46894", "right": "issue:46897", "accept": true, - "reason": "Same training-loss double-shift regression pattern, just reported on different model families." + "reason": "Same training-loss double-shift bug pattern, reported for two model families and clearly pointing at the same underlying issue." }, { "left": "issue:45759", "right": "issue:46661", "accept": false, - "reason": "Composite Qwen config unwrap bug and remote-code config_class AttributeError are different model-construction failures." + "reason": "Composite multimodal config handling vs remote-code config_class AttributeError. Similar surface area only." }, { "left": "issue:43872", - "right": "issue:44291", - "accept": true, - "reason": "Both center on the same unexpected _is_hf_initialized constructor-argument TypeError during model initialization/loading." + "right": "issue:46869", + "accept": false, + "reason": "bitsandbytes init_empty_weights incompatibility vs YOLO multi-GPU device-map failure. Different problems." }, { "left": "issue:43404", "right": "issue:43950", "accept": false, - "reason": "Tied lm_head weights and non-persistent buffer corruption are unrelated regressions." + "reason": "lm_head tying bug vs non-persistent buffer corruption. No shared fix path." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "These describe the same loading incompatibility around unexpected _is_hf_initialized during init_empty_weights/bitsandbytes-style model loading." }, { "left": "issue:45020", "right": "issue:46661", "accept": false, - "reason": "Both affect remote-code models, but one is a loading regression and the other is a missing config_class attribute." + "reason": "The latter is a specific remote-code AttributeError, but the former is a broader regression report; not enough evidence they are the same bug." }, { "left": "issue:43792", "right": "issue:44220", "accept": false, - "reason": "Whisper cannot run is too broad to confirm it is the same issue as the _torch_extract_fbank_features bug." + "reason": "Whisper model failure vs a helper-function issue; could be related, but the underlying defect is not clearly the same." }, { "left": "issue:43257", "right": "issue:45305", "accept": false, - "reason": "Qwen3 MOE deepspeed weight conversion and GAS gradient averaging are different bugs." + "reason": "Both involve DeepSpeed, but one is missing weight conversion and the other is gradient averaging. Different defects." }, { "left": "issue:42947", "right": "issue:45923", "accept": false, - "reason": "Gradient checkpointing ineffective with PEFT LoRA is distinct from a missing supports_gradient_checkpointing flag on a remote-code variant." + "reason": "Gradient checkpointing ineffective with PEFT LoRA is not the same as a missing supports_gradient_checkpointing flag on a remote-code variant." }, { "left": "issue:36331", "right": "issue:41093", "accept": false, - "reason": "CustomTrainer loss-signature TypeError and mask-shape IndexError are unrelated failures." + "reason": "Unexpected compute_loss kwarg vs mask/tensor shape mismatch. Distinct failures." }, { "left": "issue:41553", "right": "issue:44991", "accept": false, - "reason": "Different tokenizer-loading problems with different models and symptoms." + "reason": "Different tokenizer-loading regressions for different models; only broad symptom overlap." }, { "left": "issue:43854", "right": "issue:44863", "accept": false, - "reason": "GLM-4.7 test loading failure and NemotronH checkpoint loading failure are different model implementations." + "reason": "Two separate model-loading/checkpoint problems with different model families and symptoms." }, { "left": "issue:43940", "right": "issue:45313", - "accept": true, - "reason": "Both describe the same ZeRO-3 Qwen-family weight-loading regression, with missing weights during distributed loading." + "accept": false, + "reason": "Both are ZeRO-3 loading issues, but the models and failure details differ; not enough to call them duplicates." }, { "left": "issue:44928", "right": "issue:45588", "accept": false, - "reason": "Gradient explosion under autocast is unrelated to the flash_attention s_aux=None crash." + "reason": "Gradient explosion from Qwen3.5 SDPA/BF16 behavior vs flash-attention sinkless-model crash. Different bugs." }, { "left": "issue:41093", "right": "issue:44805", "accept": true, - "reason": "Exact same mask-shape IndexError, just a more detailed later report." + "reason": "Same mask-shape IndexError, with one issue giving a fuller reproduction of the same underlying tensor indexing bug." }, { "left": "issue:44479", "right": "issue:44560", - "accept": true, - "reason": "Both report the same v5.3.0 video-input regression in the Qwen3/Qwen3.5 video pipeline." + "accept": false, + "reason": "Broad video-input regression across several Qwen models vs a specific StopIteration on Qwen3-vl-embedding. Likely related area, not the same bug." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "Voxtral error message quality and sleng-bert loading failure are unrelated." - }, - { - "left": "issue:43854", - "right": "issue:44863", - "accept": false, - "reason": "Different model-loading failures with no clear shared root cause." + "reason": "Both are model-loading complaints, but the models and likely root causes differ." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays mostly conservative. The accepted soft-edge pairs all have clear title-level evidence of the same bug class, and the rejected pairs are appropriately cautious. The canonical issue choice is also defensible given the near-identical mask-shape IndexError reports." + "feedback": "The summary is grounded in the packet, and the accepted duplicate pairs are reasonably supported by the issue titles/descriptions. The canonical choice for issue:44291 is conservative and well-justified, and the rejected soft-edge pairs mostly avoid overclaiming. No PR-mergeability concern applies here since this is an issue cluster." }, "evaluator_used": true, "retried": false, @@ -1835,7 +1952,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 6, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -1897,276 +2014,275 @@ "issue:44530|issue:44863", "issue:44863|issue:46402", "issue:43366|issue:43828", + "issue:45325|issue:45902", "issue:43596|issue:46515", "issue:43366|issue:43531", - "issue:45325|issue:45902", - "issue:43638|issue:46515", - "issue:45663|issue:45823", "issue:44155|issue:45584", + "issue:45663|issue:45823", + "issue:43638|issue:46515", "issue:43366|issue:44589", - "issue:43824|issue:43957", "issue:43950|issue:45834", + "issue:43824|issue:43957", "issue:42915|issue:43957", - "issue:43716|issue:44560", "issue:44910|issue:46858", + "issue:43716|issue:44560", "issue:43645|issue:43646", - "issue:44560|issue:45685", "issue:43550|issue:44910", + "issue:44560|issue:45685", "issue:42617|issue:43366", "issue:44683|issue:46515", - "issue:44560|issue:44918", "issue:36010|issue:42915", + "issue:44560|issue:44918", "issue:45812|issue:46489", "issue:45823|issue:46515", "issue:44661|issue:45823", - "issue:44560|issue:44805", "issue:38175|issue:43994", + "issue:44560|issue:44805", "issue:45672|issue:46899", - "issue:44877|issue:45030", "issue:43632|issue:45834", + "issue:44877|issue:45030", + "issue:42491|issue:43720", "issue:43638|issue:44661", - "issue:43596|issue:44661", - "issue:44661|issue:44805", - "issue:42491|issue:43720" + "issue:43596|issue:44661" ], - "prepared_review_unit_hash": "fcbf2885383943e0bad888abb8d130b7e1abeee244dd99e866fb2a2e3b653aed", + "prepared_review_unit_hash": "7ac4fa701a040373b3ae09262b43861e9fd51a9a3ebd46abd884b10f6914e145", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 35, - "serialized_chars": 38367, - "estimated_input_tokens": 9592, - "estimated_eval_tokens": 19440 + "soft_pair_count": 34, + "serialized_chars": 38070, + "estimated_input_tokens": 9518, + "estimated_eval_tokens": 19292 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous, with one main theme around Transformers v5/custom model loading regressions and a separate Gemma4 4-bit quantization subtheme. I only see clear duplicate-like pairs in the custom-init reports and the Gemma4 4-bit dtype-cast reports; the rest look like distinct bugs.", - "confidence": 0.58, - "canonical_issue_reason": "issue:43632 is the best anchor because it states the likely root regression (`_is_hf_initialized` in Transformers v5) and broadly explains several nearby custom-model/init/loading failures.", + "summary": "This cluster is mostly heterogeneous; most pairs are clearly unrelated. I only see three likely duplicate subgroups: custom model initialization (43645/43646), SigLIP2 bad outputs (38175/43994), and Gemma4 4-bit multimodal dtype-cast failures (45672/46899).", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue cleanly covers the whole cluster because it is largely mixed; the strongest duplicate subgroup is the SigLIP2 regression, where 43994 is the broadest representative.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43632 is the strongest representative issue: it is broad, root-cause-oriented, and closer to the shared loading/init regression theme than the narrower symptom reports.", + "best_issue_reason": "43994 is the best overall issue representative among the accepted duplicates because it states the SigLIP2 regression more generally than 38175 and is easier to map to the underlying bad-output bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:24643", "right": "issue:30064", "accept": false, - "reason": "Different failures: DeepSpeed training weight-shape error vs image processor void-segmentation-map handling." + "reason": "Different subsystems and failure modes: DeepSpeed training weight-shape error vs void segmentation map preprocessing." }, { "left": "issue:44530", "right": "issue:44863", "accept": false, - "reason": "Qwen3.5 PagedAttentionCache/linear_attention crash is unrelated to NemotronH checkpoint loading." + "reason": "Both are model-specific, but they describe different bugs and code paths: Qwen3.5 PagedAttentionCache crash vs NemotronH checkpoint loading." }, { "left": "issue:44863", "right": "issue:46402", "accept": false, - "reason": "Both are loading-related, but one is a NemotronH checkpoint load failure and the other is a Sam3Model weight-loading regression; different model/code paths." + "reason": "Both are loading issues, but for different models and different failures; not the same underlying bug." }, { "left": "issue:43366", "right": "issue:43828", "accept": false, - "reason": "GGUF gpt-oss support request is unrelated to Phi-tiny-MoE autocast dtype mismatch." + "reason": "Unrelated: GGUF gpt-oss support vs Phi-tiny-MoE autocast dtype mismatch." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "Both touch rotary/position logic, but the models and concrete bugs differ: Qwen2.5-VL temporal scaling vs Qwen3_5Moe uninitialized memory after meta materialization." }, { "left": "issue:43596", "right": "issue:46515", "accept": false, - "reason": "Bert/DeepSpeed ZeRO3 index error is a concrete model-init bug, while the other is a broad torch version incompatibility report." + "reason": "One is a DeepSpeed ZeRO3 Bert index error; the other is a broad torch version incompatibility claim." }, { "left": "issue:43366", "right": "issue:43531", "accept": false, - "reason": "Different Qwen/GGUF topics: gpt-oss support vs sliding_window behavior in Qwen3-MoE." - }, - { - "left": "issue:45325", - "right": "issue:45902", - "accept": false, - "reason": "Different rotary/position-id bugs in different model families; not the same concrete bug." + "reason": "Different topics: GGUF architecture support vs Qwen3-MoE sliding_window behavior." }, { - "left": "issue:43638", - "right": "issue:46515", + "left": "issue:44155", + "right": "issue:45584", "accept": false, - "reason": "Bert ZeRO3 initialization index error is not the same as a generic torch 2.4.1 incompatibility report." + "reason": "Different audio/text generation bugs with different symptoms and code paths." }, { "left": "issue:45663", "right": "issue:45823", "accept": false, - "reason": "Both are Gemma4, but one is an FSDP2 shared_kv_states rebuild bug and the other is a device-map mismatch; different failure paths." + "reason": "Same model family, but different failures: FSDP2 shared_kv_states logic vs device_map='auto' device mismatch." }, { - "left": "issue:44155", - "right": "issue:45584", + "left": "issue:43638", + "right": "issue:46515", "accept": false, - "reason": "AudioFlamingo3 track-leak bug is unrelated to Whisper empty-transcription generation failure." + "reason": "Different issues: ZeRO3 Bert index error vs general torch compatibility." }, { "left": "issue:43366", "right": "issue:44589", "accept": false, - "reason": "GGUF model support request is unrelated to Float8 storage import/type handling." - }, - { - "left": "issue:43824", - "right": "issue:43957", - "accept": false, - "reason": "Missing Qwen2.5-VL class export and meta-device loading regression are different bugs." + "reason": "Different models and failures; no shared underlying bug." }, { "left": "issue:43950", "right": "issue:45834", "accept": false, - "reason": "Non-persistent buffer corruption and Kosmos2.5 long-OCR index errors are unrelated." + "reason": "Non-persistent buffer loading regression vs long OCR input index error are unrelated." }, { - "left": "issue:42915", + "left": "issue:43824", "right": "issue:43957", "accept": false, - "reason": "Qwen3Moe FP8 failure and meta-device loading problems are distinct regressions." + "reason": "Different problems: missing Qwen2_5_VL import vs meta-device loading breakage." }, { - "left": "issue:43716", - "right": "issue:44560", + "left": "issue:42915", + "right": "issue:43957", "accept": false, - "reason": "Mistral-3 image dtype mismatch is unrelated to Qwen3-vl video StopIteration." + "reason": "Different code paths and symptoms; FP8 config failure is not the same as meta-loading breakage." }, { "left": "issue:44910", "right": "issue:46858", "accept": false, - "reason": "Flash-attention position_ids packing bug is different from static-cache seq_length device-tensor arithmetic on non-CUDA backends." + "reason": "Both involve cache/position arithmetic, but one is Qwen3.5 flash-attention illegal memory access and the other is a static-cache seq-length tensor issue on non-CUDA backends." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Different multimodal bugs in different models: Mistral-3 dtype mismatch vs Qwen3-vl video StopIteration." }, { "left": "issue:43645", "right": "issue:43646", "accept": true, - "reason": "Same underlying Transformers v5 custom-model initialization regression; 43645 is the notebook-specific manifestation of the broader 43646 failure." + "reason": "Same Transformers v5 custom-model initialization regression; the notebook mention is just a context-specific report of the same bug." }, { - "left": "issue:44560", - "right": "issue:45685", + "left": "issue:43550", + "right": "issue:44910", "accept": false, - "reason": "Qwen3-vl video StopIteration and MPS histogram-on-Int error are unrelated." + "reason": "Different attention-path failures in different models: SDPA/torch.compile on Bamba vs flash-attention position_id handling on Qwen3.5." }, { - "left": "issue:43550", - "right": "issue:44910", + "left": "issue:44560", + "right": "issue:45685", "accept": false, - "reason": "Bamba torch.compile+SDPA crash is not the same bug as Qwen3.5 flash-attention position_ids misinterpretation." + "reason": "Unrelated: Qwen3-vl video processing vs MPS histogram/int dtype limitation in moe code." }, { "left": "issue:42617", "right": "issue:43366", "accept": false, - "reason": "3d_parallel.py startup failure is unrelated to GGUF gpt-oss support." + "reason": "Different topics entirely: 3d_parallel.py runtime failure vs GGUF gpt-oss support." }, { "left": "issue:44683", "right": "issue:46515", "accept": false, - "reason": "Compiled flex_attention on torch>=2.9 is different from a torch 2.4.1 incompatibility report." + "reason": "Both mention torch versions, but the concrete problems are different and not mergeable as one fix." }, { - "left": "issue:44560", - "right": "issue:44918", + "left": "issue:36010", + "right": "issue:42915", "accept": false, - "reason": "Both touch Qwen3.x embedding handling, but one is a video StopIteration bug and the other is a TRL SFT embedding-unpacking failure; different paths." + "reason": "ImportError for GenerationMixin is unrelated to the Qwen3Moe FP8 failure." }, { - "left": "issue:36010", - "right": "issue:42915", + "left": "issue:44560", + "right": "issue:44918", "accept": false, - "reason": "GenerationMixin import failure is unrelated to Qwen3Moe FineGrainedFP8Config behavior." + "reason": "Both are Qwen3.x-related, but the failures differ: video StopIteration vs SFT trainer embedding unpacking." }, { "left": "issue:45812", "right": "issue:46489", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and do not clearly share the same concrete code path." + "reason": "Both are tokenizer regressions, but they affect different model families and appear to be separate bugs." }, { "left": "issue:45823", "right": "issue:46515", "accept": false, - "reason": "Gemma4 PLE device mismatch is a model/device-placement bug, not a generic torch-version incompatibility." + "reason": "Gemma4 device-mapping bug is unrelated to a generic torch 2.4.1 incompatibility report." }, { "left": "issue:44661", "right": "issue:45823", "accept": false, - "reason": "add-new-model-like/tokenizer-mapping failure is unrelated to Gemma4 PLE device mismatch." - }, - { - "left": "issue:44560", - "right": "issue:44805", - "accept": false, - "reason": "Qwen3-vl video StopIteration is unrelated to the mask/index shape mismatch error." + "reason": "Tokenizer-mapping validation failure is unrelated to Gemma4 device mismatch." }, { "left": "issue:38175", "right": "issue:43994", + "accept": true, + "reason": "Same SigLIP2 model regression causing bad outputs; 43994 is just the broader description of the same underlying bug." + }, + { + "left": "issue:44560", + "right": "issue:44805", "accept": false, - "reason": "Both mention siglip2, but one is zero probabilities and the other is nonsensical AutoModel/pipeline output; not enough evidence of the same concrete bug." + "reason": "Both are index-like failures, but they involve different models and tensor shapes; not the same bug." }, { "left": "issue:45672", "right": "issue:46899", "accept": true, - "reason": "Same Gemma4 4-bit quantization dtype-cast problem: 45672 is the audio-module torch.finfo() failure, and 46899 is the broader audio/vision embedder corruption from the same underlying issue." + "reason": "Same Gemma4 4-bit multimodal quantization problem: unsafe dtype casting in audio/vision embedders." + }, + { + "left": "issue:43632", + "right": "issue:45834", + "accept": false, + "reason": "Transformers v5 _is_hf_initialized regression is unrelated to Kosmos2.5 OCR indexing." }, { "left": "issue:44877", "right": "issue:45030", "accept": false, - "reason": "Different config-validation failures in different model families; not the same bug." + "reason": "Different config-validation issues on different models; not the same underlying change." }, { - "left": "issue:43632", - "right": "issue:45834", + "left": "issue:42491", + "right": "issue:43720", "accept": false, - "reason": "Transformers v5 `_is_hf_initialized` regression is unrelated to Kosmos2.5 OCR index errors." + "reason": "Different loading/compatibility failures with different model types and code paths." }, { "left": "issue:43638", "right": "issue:44661", "accept": false, - "reason": "Bert ZeRO3 training index error and add-new-model-like tokenizer mapping failure are different issues." + "reason": "DeepSpeed/Bert training index error is unrelated to tokenizer-mapping model registration failure." }, { "left": "issue:43596", "right": "issue:44661", "accept": false, - "reason": "DeepSpeed/Bert zero3 init crash is unrelated to tokenizer mapping behavior." - }, - { - "left": "issue:42491", - "right": "issue:43720", - "accept": false, - "reason": "LoRA/qwen3_moe hf4.x-to-hf5.x compatibility issue is unrelated to BitNet packed-weight loading in accelerate." + "reason": "Different failures: ZeRO3/Bert index out-of-bounds vs add-new-model-like tokenizer mapping logic." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is conservative and grounded in the packet. The main clustering around Transformers v5 custom-model/loading regressions and the Gemma4 4-bit dtype-cast issues is supported, and the two accepted soft pairs look plausibly duplicate-like from the issue titles/bodies. The false verdicts are generally cautious rather than overstated." + "feedback": "The summary is well grounded in the packet and stays conservative about the soft edges. The three accepted subgroups are plausible duplicates from the titles/body themes (43645/43646, 38175/43994, 45672/46899), and the rejected pairs are generally distinguished by model, subsystem, or failure mode without overclaiming. I don\u2019t see a mergeability or evidence-grounding problem here." }, "evaluator_used": true, "retried": false, @@ -2179,7 +2295,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 7, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -2206,11 +2322,12 @@ "issue:44164", "issue:44291", "issue:44387", - "issue:44402", "issue:44451", "issue:44488", "issue:44534", + "issue:44661", "issue:44756", + "issue:44805", "issue:44863", "issue:44910", "issue:44960", @@ -2237,250 +2354,250 @@ ], "soft_pairs": [ "issue:43317|issue:44756", + "issue:44661|issue:44805", "issue:38175|issue:39692", "issue:43644|issue:44534", "issue:41720|issue:42915", "issue:42491|issue:45276", "issue:42915|issue:44756", - "issue:44451|issue:44488", "issue:44960|issue:44977", - "issue:46459|issue:46512", + "issue:44451|issue:44488", "issue:44863|issue:46846", + "issue:46459|issue:46512", "issue:45313|issue:45357", - "issue:45507|issue:46468", "issue:44488|issue:44991", + "issue:45507|issue:46468", "issue:43644|issue:43645", - "issue:42491|issue:45292", "issue:39692|issue:43994", "issue:43257|issue:43854", "issue:43856|issue:43873", - "issue:43708|issue:44075", - "issue:43854|issue:43931", + "issue:42491|issue:45292", "issue:43856|issue:44387", + "issue:43854|issue:43931", + "issue:43708|issue:44075", "issue:42491|issue:45335", "issue:44960|issue:46050", "issue:46897|issue:46901", "issue:45488|issue:45812", - "issue:45276|issue:45335", "issue:45362|issue:45406", "issue:44164|issue:44291", - "issue:44910|issue:45588", - "issue:44402|issue:44488" + "issue:45276|issue:45335", + "issue:44910|issue:45588" ], - "prepared_review_unit_hash": "2619f9a3eb1ce79b551354bfd8c5a8f0056e1475328d4afade6b82f32eb672a1", + "prepared_review_unit_hash": "827488692e064c688ee9df435d20fb530ccbcb5151ca293110e5792be83db9c5", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36049, - "estimated_input_tokens": 9013, - "estimated_eval_tokens": 18282 + "serialized_chars": 36658, + "estimated_input_tokens": 9165, + "estimated_eval_tokens": 18586 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Only one soft pair is a clear duplicate: the Transformers v5 non-persistent-buffer regression. Most other pairs are related only at a broad subsystem level (tokenizers, loading, quantization, Qwen3) but look like different concrete bugs or model-specific reports.", - "confidence": 0.63, - "canonical_issue_reason": "issue:43644 is the clearest original report of the duplicated Transformers v5 non-persistent-buffer corruption, and it is broader/more referenced than the later reworded report.", + "summary": "This cluster is mostly a loose pile of Transformers regression reports that only superficially resemble one another. I found one clear duplicate pair around the v5 non-persistent-buffer corruption bug; the rest are different model- or feature-specific issues and should stay separate.", + "confidence": 0.87, + "canonical_issue_reason": "issue:43644 is the clearest, broadest report of the shared v5 non-persistent-buffer corruption regression, and it predates the near-identical follow-up report.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43644 is the best cluster anchor for the one confirmed duplicate family because it states the bug most generally and already has inbound reference traffic.", + "best_issue_reason": "Same reason as canonical: it is the most representative and broadly phrased issue in this set, with the strongest evidence of a real underlying regression rather than a model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43317", "right": "issue:44756", "accept": false, - "reason": "Both involve loading/offload problems, but one is a dequantized-model device_map failure and the other is a Strix Halo mmap OOM; different root causes." + "reason": "Both are offload/memory-related, but they describe different failures in different contexts (dequantized loading on GPU+CPU offload vs Strix Halo mmap OOM)." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Completely different bugs: tokenizer mapping failure vs tensor/mask shape mismatch." }, { "left": "issue:38175", "right": "issue:39692", "accept": false, - "reason": "SigLIP2 output/probability issue vs SigLIP2 doc example model/processor and quantization errors; not the same bug." + "reason": "Both mention SigLIP2, but one is inference probability weirdness and the other is documentation/example errors with quantization; not the same bug." }, { "left": "issue:43644", "right": "issue:44534", "accept": true, - "reason": "Same regression: Transformers v5 fills non-persistent buffers with junk." + "reason": "Near-identical wording and same underlying v5 bug: non-persistent buffers being filled with junk." }, { "left": "issue:41720", "right": "issue:42915", "accept": false, - "reason": "Different Qwen3 failures: auto device-map cuda assert vs FineGrainedFP8Config behavior." + "reason": "Different Qwen3 failure modes: CUDA assert under auto device map vs FineGrainedFP8Config failure." }, { "left": "issue:42491", "right": "issue:45276", "accept": false, - "reason": "Different model families and different symptoms; LoRA hf4.x/hf5.x compatibility is not the same as gemma4 resize_token_embeddings propagation." + "reason": "Different models and different symptoms; no shared code-path evidence." }, { "left": "issue:42915", "right": "issue:44756", "accept": false, - "reason": "Qwen3Moe FP8 config failure and Strix Halo mmap OOM are unrelated." - }, - { - "left": "issue:44451", - "right": "issue:44488", - "accept": false, - "reason": "Both are model-load failures, but the titles do not show the same concrete failure mode or code path." + "reason": "Unrelated bugs: Qwen3Moe FP8 failure vs mmap/OOM workaround on Strix Halo." }, { "left": "issue:44960", "right": "issue:44977", "accept": false, - "reason": "GLM5 is unrelated to the Qwen3.5 flash-attention generation crash." + "reason": "Different model families and different flash-attention failures; not the same change." }, { - "left": "issue:46459", - "right": "issue:46512", + "left": "issue:44451", + "right": "issue:44488", "accept": false, - "reason": "One is dtype being ignored on composite checkpoint loading; the other is from_config leaving model.config.dtype stale." + "reason": "Both are model-loading failures for different BERT repos, but there is not enough evidence they stem from the same root cause." }, { "left": "issue:44863", "right": "issue:46846", "accept": false, - "reason": "NemotronH checkpoint loading and Qwen3.5 TP-plan omissions are different issues." + "reason": "NemotronH checkpoint loading and Qwen3.5 tensor-parallel planning are separate issues." }, { - "left": "issue:45313", - "right": "issue:45357", + "left": "issue:46459", + "right": "issue:46512", "accept": false, - "reason": "DeepSpeed ZeRO-3 weight-loading failure and incorrect save_pretrained visual keys are different bugs." + "reason": "Both involve dtype handling, but one is composite-checkpoint loading and the other is from_config initialization/config staleness." }, { - "left": "issue:45507", - "right": "issue:46468", + "left": "issue:45313", + "right": "issue:45357", "accept": false, - "reason": "GraniteMoEHybrid invalid method call is unrelated to the broken Nemotron example." + "reason": "Different regressions: ZeRO-3 weight loading vs save_pretrained key handling for visual encoder weights." }, { "left": "issue:44488", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer/model-load regressions, but they affect different checkpoints and the titles do not establish one shared root cause." + "reason": "Both are tokenizer-load failures, but for different models and likely different tokenizer configs; too little to merge." }, { - "left": "issue:43644", - "right": "issue:43645", + "left": "issue:45507", + "right": "issue:46468", "accept": false, - "reason": "Non-persistent buffer junk regression vs notebook custom-model initialization regression; different failures." + "reason": "GraniteMoEHybrid invalid method call vs broken Nemotron example are unrelated." }, { - "left": "issue:42491", - "right": "issue:45292", + "left": "issue:43644", + "right": "issue:43645", "accept": false, - "reason": "LoRA model portability issue vs resize_token_embeddings not updating output_embeddings." + "reason": "Both are v5 regressions, but one corrupts non-persistent buffers and the other breaks custom model initialization in notebooks." }, { "left": "issue:39692", "right": "issue:43994", "accept": false, - "reason": "Doc/example mismatch and quantization failure versus nonsensical AutoModel/pipeline outputs are not the same bug." + "reason": "Same model family, but one is a doc/example mismatch and quantization failure while the other is bad inference output from AutoModel/pipeline." }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Qwen3 MoE deepspeed/accelerate conversion issue vs GLM-4.7-Flash test loading issue; different models and code paths." + "reason": "Different problems: Qwen3 MoE weight conversion vs GLM-4.7-Flash loading in tests." }, { "left": "issue:43856", "right": "issue:43873", "accept": false, - "reason": "Qwen3 MoE training memory usage and quantization offload behavior are different problems." + "reason": "Both mention memory/offload, but the causes and affected workflows are different." }, { - "left": "issue:43708", - "right": "issue:44075", + "left": "issue:42491", + "right": "issue:45292", "accept": false, - "reason": "Trainer resume/max_steps bug and SGD args not used are unrelated." + "reason": "LoRA/Qwen3-MoE branch incompatibility is unrelated to resize_token_embeddings not updating output embeddings." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both are memory regressions, but one is Qwen3 MoE training memory usage and the other is int4 quantization reserved-memory growth." }, { "left": "issue:43854", "right": "issue:43931", "accept": false, - "reason": "GLM-4.7-Flash test loading failure is unrelated to Qwen3-VL weight shape mismatch." + "reason": "Different model-loading failures with different symptoms and model families." }, { - "left": "issue:43856", - "right": "issue:44387", + "left": "issue:43708", + "right": "issue:44075", "accept": false, - "reason": "Training memory usage in Qwen3 MoE and int4 quantization reserved-memory/OOM are different regressions." + "reason": "Trainer resume/max_steps bug and SGD args not used are unrelated." }, { "left": "issue:42491", "right": "issue:45335", "accept": false, - "reason": "Different model families and different resize_token_embeddings targets; not the same code-path bug." + "reason": "Different model families and different resize_token_embeddings behavior." }, { "left": "issue:44960", "right": "issue:46050", "accept": false, - "reason": "GLM5 issue and sinq save/load breakage are unrelated." + "reason": "GLM5 issue and sinq save/load corruption are unrelated." }, { "left": "issue:46897", "right": "issue:46901", "accept": false, - "reason": "Both mention double-shift loss, but they are separate model implementations with no evidence of a shared code-path bug." + "reason": "Both describe training-loss double-shift symptoms, but for different models; not enough evidence they share the same implementation bug." }, { "left": "issue:45488", "right": "issue:45812", "accept": false, - "reason": "Different tokenizer regressions affecting different model families; not enough to treat as one duplicate." - }, - { - "left": "issue:45276", - "right": "issue:45335", - "accept": false, - "reason": "Both involve resize_token_embeddings, but they affect different model classes and different internal modules." + "reason": "Both are tokenizer regressions, but they affect different model families and different tokenizer behaviors." }, { "left": "issue:45362", "right": "issue:45406", "accept": false, - "reason": "Qwen3.5 chat crash and Gemma4Processor missing _tokenizer are unrelated." + "reason": "Qwen3.5 chat crash and Gemma4Processor missing _tokenizer are different code paths." }, { "left": "issue:44164", "right": "issue:44291", "accept": false, - "reason": "save/from_pretrained extra_state handling and init_empty_weights _is_hf_initialized argument error are different bugs." + "reason": "Different save/load bugs: extra_state handling vs unexpected _is_hf_initialized argument during init_empty_weights." }, { - "left": "issue:44910", - "right": "issue:45588", + "left": "issue:45276", + "right": "issue:45335", "accept": false, - "reason": "Flash-attention illegal memory access from 3D position_ids is unrelated to the s_aux=None AttributeError." + "reason": "Related resize_token_embeddings theme, but different models and different affected embeddings." }, { - "left": "issue:44402", - "right": "issue:44488", + "left": "issue:44910", + "right": "issue:45588", "accept": false, - "reason": "Token vocab-size mismatch on one model and a separate BERT load failure on another are not shown to share the same root cause." + "reason": "Flash-attention crashes in Qwen3.5 are not the same as the s_aux=None crash in sink-less models." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative on the soft pairs. The only accepted duplicate pair, issue:43644 \u2194 issue:44534, is well supported by the nearly identical titles and the fact that 43644 has inbound reference traffic. The remaining rejections do not overclaim shared root cause from the available evidence." + "feedback": "The summary is grounded in the packet and stays conservative. It identifies the one clearly supported duplicate pair around the v5 non-persistent-buffer corruption bug, and the remaining pairwise rejections avoid over-claiming sameness across different model/feature contexts. No soft-edge verdict appears overly aggressive." }, "evaluator_used": true, "retried": false, @@ -2493,7 +2610,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 8, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -2525,6 +2642,7 @@ "issue:44402", "issue:44451", "issue:44479", + "issue:44488", "issue:44756", "issue:44863", "issue:44877", @@ -2550,25 +2668,26 @@ "issue:46901" ], "soft_pairs": [ - "issue:43425|issue:43957", + "issue:44402|issue:44488", "issue:39692|issue:44960", "issue:41720|issue:44155", "issue:45663|issue:45759", + "issue:42947|issue:45704", "issue:43404|issue:43454", + "issue:43425|issue:43957", "issue:44291|issue:44756", - "issue:44079|issue:46829", - "issue:42947|issue:45704", "issue:43873|issue:44387", + "issue:44079|issue:46829", + "issue:36010|issue:39404", "issue:44292|issue:44912", "issue:45663|issue:45859", - "issue:36010|issue:39404", "issue:44387|issue:44756", - "issue:42947|issue:43381", "issue:44155|issue:45823", - "issue:43957|issue:44756", + "issue:42947|issue:43381", "issue:43335|issue:45663", - "issue:45276|issue:45292", + "issue:43957|issue:44756", "issue:44863|issue:46468", + "issue:45276|issue:45292", "issue:44479|issue:44877", "issue:43381|issue:44928", "issue:44164|issue:44756", @@ -2578,223 +2697,229 @@ "issue:43854|issue:45216", "issue:45292|issue:45335", "issue:46894|issue:46901", - "issue:43828|issue:43957", - "issue:44402|issue:44451" + "issue:44402|issue:44451", + "issue:43828|issue:43957" ], - "prepared_review_unit_hash": "abce9bbee48828d5ad17fe2a86d912bc82829e172de8718af2880f71dd435c46", + "prepared_review_unit_hash": "e5fb3c6b2ae5fc7c1fd987eb2c159aa3e3748d470beb68792dbb4067014dacca", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 30, - "serialized_chars": 36157, - "estimated_input_tokens": 9040, - "estimated_eval_tokens": 18336 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37011, + "estimated_input_tokens": 9253, + "estimated_eval_tokens": 18762 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of unrelated issues, but it contains two small duplicate chains: a resize_token_embeddings propagation bug and a training-loss double-shift bug.", - "confidence": 0.79, - "canonical_issue_reason": "issue:45292 is the best canonical issue because it is the most generic title in the strongest duplicate chain and can serve as the umbrella for the related resize_token_embeddings variants.", + "summary": "Most pairs are only thematically similar, not the same concrete bug. The only clear duplicate cluster is the `resize_token_embeddings` regression across model variants; everything else should stay separate.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 45292 is the cleanest general statement of the shared `resize_token_embeddings` propagation bug, and it sits between the more model-specific reports, making it the best representative issue.", "canonical_pr_reason": null, - "best_issue_reason": "Most general and central representative among the accepted duplicate issues.", + "best_issue_reason": "Issue 45292 best captures the underlying API regression without overfitting to one model family, so it is the strongest global issue representative here.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43425", - "right": "issue:43957", + "left": "issue:44402", + "right": "issue:44488", "accept": false, - "reason": "Different failures: torch 2.10 import incompatibility vs meta-device loading breakage." + "reason": "Both are model-loading failures for different BERT checkpoints, but the titles point to different concrete symptoms and no shared code path is evident." }, { "left": "issue:39692", "right": "issue:44960", "accept": false, - "reason": "Unrelated topics: SigLIP2 docs example errors vs GLM5 model-loading issue." + "reason": "SigLIP2 doc example errors and a GLM5 loading issue are unrelated beyond both being model-specific bug reports." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "Different models and code paths: Qwen3 auto device mapping vs AudioFlamingo3 batched inference leak." + "reason": "A Qwen3 device-mapping CUDA assert and an AudioFlamingo batched-inference token leak are different failures in different subsystems." }, { "left": "issue:45663", "right": "issue:45759", "accept": false, - "reason": "Different bugs: Gemma4 FSDP/shared_kv_states KeyError vs Qwen multimodal config unwrapping." + "reason": "One is an FSDP2 attention KeyError in Gemma4; the other is `from_config` not unwrapping `text_config` for Qwen multimodal configs." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing with LoRA and a T5 RMSNorm memory leak are unrelated bugs." }, { "left": "issue:43404", "right": "issue:43454", "accept": false, - "reason": "Same broad symptom area, but different model families and no clear shared code-path bug." + "reason": "Both mention `lm_head`, but they concern different model families and likely different weight-tying paths." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Torch version incompatibility and meta-device model loading failures are not the same bug." }, { "left": "issue:44291", "right": "issue:44756", "accept": false, - "reason": "Unrelated: init_empty_weights argument handling vs mmap OOM on Strix Halo." + "reason": "Unexpected `_is_hf_initialized` handling and Strix Halo mmap OOM are separate loading/runtime issues." }, { - "left": "issue:44079", - "right": "issue:46829", + "left": "issue:43873", + "right": "issue:44387", "accept": false, - "reason": "Generic ModelOutput key handling vs a VideoPrism output field mix-up are not the same bug." + "reason": "Quantization offloading behavior and int4 CUDA memory growth may overlap broadly, but they are not the same concrete defect." }, { - "left": "issue:42947", - "right": "issue:45704", + "left": "issue:44079", + "right": "issue:46829", "accept": false, - "reason": "Different problems: gradient checkpointing/LoRA vs T5 apex RMSNorm memory leak." + "reason": "Both involve output objects, but one is a `ModelOutput` key assignment bug and the other is a wrong tensor stored in `hidden_states`." }, { - "left": "issue:43873", - "right": "issue:44387", + "left": "issue:36010", + "right": "issue:39404", "accept": false, - "reason": "Both involve quantization memory, but one is offloading behavior and the other is int4 reserved-memory growth/OOM." + "reason": "An import error for `GenerationMixin` is unrelated to Whisper `return_language` pipeline behavior." }, { "left": "issue:44292", "right": "issue:44912", "accept": false, - "reason": "Different quantization formats and models (NVFP4 vs MXFP4) with different failure modes." + "reason": "Both are quantization-related, but one is a Qwen-3 NVFP4 runtime error and the other is MXFP4 fallback to bf16." }, { "left": "issue:45663", "right": "issue:45859", "accept": false, - "reason": "Different issues: Gemma4 FSDP KeyError vs Qwen3_5Moe CPU-offload incompatibility." + "reason": "Gemma4 FSDP2 attention state handling and Qwen3.5-MoE CPU-offload rotary embedding incompatibility are different code paths." }, { - "left": "issue:36010", - "right": "issue:39404", + "left": "issue:44387", + "right": "issue:44756", "accept": false, - "reason": "ImportError for GenerationMixin is unrelated to Whisper pipeline return_language regression." + "reason": "An int4 quantization memory regression is not the same as disabling mmap to avoid OOM on Strix Halo." }, { - "left": "issue:44387", - "right": "issue:44756", + "left": "issue:44155", + "right": "issue:45823", "accept": false, - "reason": "Different bugs: quantization OOM vs disabling mmap on Strix Halo." + "reason": "AudioFlamingo batching corruption and Gemma4 PLE device mismatch under `device_map=\"auto\"` are unrelated." }, { "left": "issue:42947", "right": "issue:43381", "accept": false, - "reason": "Both mention gradient checkpointing, but one is ineffective LoRA checkpointing and the other is eval-mode incompatibility." + "reason": "Both mention gradient checkpointing, but one is ineffective under PEFT LoRA while the other forbids eval-mode use." }, { - "left": "issue:44155", - "right": "issue:45823", + "left": "issue:43335", + "right": "issue:45663", "accept": false, - "reason": "Different model subsystems: batched inference token leak vs Gemma4 PLE device mismatch." + "reason": "A SwitchTransformers sparse-layer config bug is unrelated to Gemma4 shared KV state rebuilding." }, { "left": "issue:43957", "right": "issue:44756", "accept": false, - "reason": "Meta-device loading breakage is not the same as Strix Halo mmap/OOM behavior." + "reason": "Meta-device loading regressions and Strix Halo mmap OOM are different failure modes." }, { - "left": "issue:43335", - "right": "issue:45663", + "left": "issue:44863", + "right": "issue:46468", "accept": false, - "reason": "Unrelated model/config bug vs Gemma4 training FSDP KeyError." + "reason": "A NemotronH checkpoint-loading bug and a broken official Nemotron example are related by model family only, not by the same concrete defect." }, { "left": "issue:45276", "right": "issue:45292", "accept": true, - "reason": "Same resize_token_embeddings propagation bug; 45276 is the Gemma4-specific variant and 45292 is the generic report." - }, - { - "left": "issue:44863", - "right": "issue:46468", - "accept": false, - "reason": "Different Nemotron loading/example problems with no shared concrete fix evident." + "reason": "Both report the same `resize_token_embeddings` regression where updated embeddings are not propagated to output embeddings; 45276 is a model-specific manifestation of the same bug." }, { "left": "issue:44479", "right": "issue:44877", "accept": false, - "reason": "Different regressions: video-input handling vs strict config loading." + "reason": "A video-input regression for Qwen VL models is unrelated to a Granite speech config loading failure." }, { "left": "issue:43381", "right": "issue:44928", "accept": false, - "reason": "Gradient checkpointing eval-mode bug is unrelated to the RLHF NaN/SDPA fallback issue." + "reason": "Eval-mode gradient checkpointing restrictions and RLHF NaNs from 3D position IDs are different problems." }, { "left": "issue:44164", "right": "issue:44756", "accept": false, - "reason": "save/from_pretrained extra_state handling is unrelated to mmap OOM on Strix Halo." + "reason": "`extra_state` serialization and mmap/OOM behavior are separate save/load vs memory issues." }, { "left": "issue:45684", "right": "issue:45698", "accept": false, - "reason": "Both concern save/load custom models, but one is read-only permissions and the other is wrong module resolution." + "reason": "One is a permissions issue during save, the other is loading the wrong custom module after save; different bugs, same feature area." }, { "left": "issue:45588", "right": "issue:45663", "accept": false, - "reason": "Flash-attention s_aux=None crash is unrelated to Gemma4 FSDP shared_kv_states behavior." + "reason": "A flash-attention crash on `s_aux=None` and a Gemma4 FSDP2 `shared_kv_states` KeyError are unrelated." }, { "left": "issue:42796", "right": "issue:46326", "accept": false, - "reason": "Tokenizer bugs in a similar stack, but different concrete failures: BPE markers vs continue_final_message." + "reason": "Tokenizer BPE-marker handling and `continue_final_message` behavior are both tokenizer-related but not the same underlying bug." }, { "left": "issue:43854", "right": "issue:45216", "accept": false, - "reason": "Different problems: unit-test model loading failure vs Qwen3.5 save_pretrained regression." + "reason": "GLM-4.7-Flash test loading and Qwen3.5 checkpoint saving regression are different code paths." }, { "left": "issue:45292", "right": "issue:45335", "accept": true, - "reason": "Same underlying resize_token_embeddings bug, just affecting a different nested embedding target." + "reason": "Both describe the same `resize_token_embeddings` propagation bug, one for `output_embeddings` and the other for `decoder.embed_tokens`." }, { "left": "issue:46894", "right": "issue:46901", - "accept": true, - "reason": "Same training-loss double-shift bug: both compute loss against shifted labels instead of raw labels." - }, - { - "left": "issue:43828", - "right": "issue:43957", "accept": false, - "reason": "Autocast dtype mismatch is not the same as meta-device loading breakage." + "reason": "They share a generic double-shift pattern, but these are different model trainers and not clearly the same concrete implementation bug." }, { "left": "issue:44402", "right": "issue:44451", "accept": false, - "reason": "Different tokenizer/model-loading issues around vocab handling, not the same bug." + "reason": "Both are model-loading failures for BERT-like models, but the symptoms and likely root causes differ." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "A dtype mismatch under autocast and meta-device loading failures are unrelated runtime issues." } ] }, "evaluator_result": { "accept": true, - "feedback": "The clustering summary is conservative and grounded in the packet. The accepted duplicate pairs form two coherent chains backed by strong title overlap: the resize_token_embeddings variants centered on issue:45292, and the training-loss double-shift pair issue:46894/issue:46901. The non-duplicate verdicts are generally cautious and avoid overclaiming shared code paths where the evidence is only superficial." + "feedback": "Grounded overall. The summary stays conservative, and the only duplicated cluster called out is well supported by the packet: 45276/45292 and 45292/45335 have strong title-level overlap around `resize_token_embeddings` propagation. The other pair judgments avoid overclaiming shared root cause and are appropriately cautious." }, "evaluator_used": true, "retried": false, @@ -2807,7 +2932,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 9, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -2868,21 +2993,21 @@ "issue:43646|issue:43708", "issue:43645|issue:44164", "issue:44589|issue:44756", - "issue:43425|issue:43828", "issue:46315|issue:46491", "issue:45812|issue:46315", + "issue:43425|issue:43828", "issue:39404|issue:45701", "issue:44530|issue:45084", - "issue:43611|issue:43645", "issue:39404|issue:46665", - "issue:44683|issue:45823", + "issue:43611|issue:43645", "issue:43425|issue:46869", - "issue:43425|issue:43716", + "issue:44683|issue:45823", "issue:43425|issue:45823", - "issue:44589|issue:45084", + "issue:43425|issue:43716", "issue:44479|issue:44805", - "issue:43901|issue:44509", "issue:43906|issue:44530", + "issue:44589|issue:45084", + "issue:43901|issue:44509", "issue:44671|issue:44991", "issue:43994|issue:44079", "issue:36032|issue:43452", @@ -2892,246 +3017,240 @@ "issue:44164|issue:45759", "issue:43906|issue:46326", "issue:44991|issue:45081", - "issue:43299|issue:43931", "issue:45823|issue:46899", + "issue:43299|issue:43931", "issue:45479|issue:46894", "issue:43856|issue:44756", - "issue:43596|issue:44683", "issue:45479|issue:46897", - "issue:44164|issue:46821" + "issue:44164|issue:46821", + "issue:43596|issue:44683" ], - "prepared_review_unit_hash": "64c8a039b1ad27f0d7c3a1f7d4cc6e09118a3ba735d716147b6d8d37ebe5fc86", + "prepared_review_unit_hash": "b411695f0961b250623691bb8a8d68c8392a8c1d1d5f99b4b7b6e3c9d5aecdae", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 38123, - "estimated_input_tokens": 9531, - "estimated_eval_tokens": 19318 + "serialized_chars": 38130, + "estimated_input_tokens": 9533, + "estimated_eval_tokens": 19322 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated Transformers v5 regression reports with only superficial title overlap. I did not find any pair that clearly looks mergeable into one duplicate issue under a strict standard; most similarities are broad subsystem overlap (tokenizers, model loading, multimodal, training) rather than the same underlying bug.", - "confidence": 0.91, - "canonical_issue_reason": null, + "summary": "This is not a tight duplicate cluster; it\u2019s a loose set of Transformers v5 regression reports spanning tokenizer behavior, model loading, multimodal forwarding, training-loss bugs, and quantization/device issues. I picked the broad tokenization-regression issue as the best representative, but most soft edges are not true duplicates.", + "confidence": 0.72, + "canonical_issue_reason": "issue:45701 is the broadest umbrella in the set: it describes version-dependent tokenization changes, which is the closest common theme across several neighboring reports. It is still only a rough representative, not an exact duplicate of the whole cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43646 is the broadest representative of the custom-model initialization regression theme, but the overall cluster is too heterogeneous for a true canonical issue.", + "best_issue_reason": "issue:45701 is the best single anchor because it is broad, version-wide, and centered on a core regression class that recurs in this cluster. It is not a good substitute for the many unrelated loading, multimodal, and training bugs here, so this is a weak-best choice.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43646", "right": "issue:43708", "accept": false, - "reason": "Both are v5 regressions, but one is custom model init and the other is Trainer checkpoint math; different code paths and fixes." + "reason": "Custom model init in notebooks and Trainer checkpoint-resume max_steps are different code paths and failure modes." }, { "left": "issue:43645", "right": "issue:44164", "accept": false, - "reason": "Notebook custom-model init breakage is unrelated to save/load extra_state handling." + "reason": "Notebook custom-model initialization is unrelated to save/from_pretrained handling of extra_state." }, { "left": "issue:44589", "right": "issue:44756", "accept": false, - "reason": "Missing Float8 storage and mmap/OOM on Strix Halo are unrelated failures." - }, - { - "left": "issue:43425", - "right": "issue:43828", - "accept": false, - "reason": "Torch version incompatibility and autocast dtype mismatch are different problems." + "reason": "Different bugs: missing Float8 storage lookup vs mmap-related OOM on Strix Halo." }, { "left": "issue:46315", "right": "issue:46491", "accept": false, - "reason": "Tokenizer slowdown and CodeLlama space-stripping regression are distinct behaviors." + "reason": "One is tokenizer performance regression; the other is a CodeLlama round-trip spacing bug." }, { "left": "issue:45812", "right": "issue:46315", "accept": false, - "reason": "Wrong Granite token IDs and slow convert_tokens_to_ids are not the same bug." + "reason": "Wrong token IDs for Granite models and slow convert_tokens_to_ids are distinct tokenizer bugs." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Torch 2.10 incompatibility is broader and different from an autocast dtype mismatch in Phi-tiny-MoE." }, { "left": "issue:39404", "right": "issue:45701", "accept": false, - "reason": "Whisper pipeline return_language breakage is unrelated to tokenization changes across versions." + "reason": "Whisper pipeline return_language regression is unrelated to general tokenization changes." }, { "left": "issue:44530", "right": "issue:45084", "accept": false, - "reason": "Qwen3.5 PagedAttention group-type crash and template compilation error are different code paths." + "reason": "PagedAttentionCache linear_attention failure and a template-node compile error are unrelated." + }, + { + "left": "issue:39404", + "right": "issue:46665", + "accept": false, + "reason": "Whisper pipeline regression and from_pretrained-inside-classes behavior are different issues." }, { "left": "issue:43611", "right": "issue:43645", "accept": false, - "reason": "base_model_prefix loading breakage is not the same as notebook custom-model initialization." + "reason": "base_model_prefix loading regression is not the same as custom model initialization in notebooks." }, { - "left": "issue:39404", - "right": "issue:46665", + "left": "issue:43425", + "right": "issue:46869", "accept": false, - "reason": "Whisper pipeline regression and from_pretrained-inside-classes inconsistency are unrelated." + "reason": "Torch compatibility problem and YOLO multi-GPU device-map failure are not the same code-path bug." }, { "left": "issue:44683", "right": "issue:45823", "accept": false, - "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 device mismatch are different issues." + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 PLE device mismatch are unrelated." }, { "left": "issue:43425", - "right": "issue:46869", + "right": "issue:43716", "accept": false, - "reason": "Torch 2.10 incompatibility and YOLO multi-GPU device_map failure are not the same bug." + "reason": "General torch incompatibility does not match the Mistral-3 image-preprocessor dtype mismatch." }, { - "left": "issue:43425", - "right": "issue:43716", + "left": "issue:44479", + "right": "issue:44805", "accept": false, - "reason": "General torch incompatibility is distinct from Mistral-3 image-preprocessor dtype mismatch." + "reason": "A video-input regression for Qwen models is not the same as a mask-shape indexing error." }, { - "left": "issue:43425", - "right": "issue:45823", + "left": "issue:43906", + "right": "issue:44530", "accept": false, - "reason": "Torch 2.10 incompatibility does not match Gemma4 PLE device placement failure." + "reason": "Reproduction of a prior issue and Qwen3.5 PagedAttentionCache crash are different problems." }, { "left": "issue:44589", "right": "issue:45084", "accept": false, - "reason": "Float8 storage lookup and non-template compilation errors are unrelated." - }, - { - "left": "issue:44479", - "right": "issue:44805", - "accept": false, - "reason": "Qwen VL video-input regression and mask shape IndexError do not share the same underlying failure." + "reason": "Float8 storage lookup failure and template compile error are unrelated." }, { "left": "issue:43901", "right": "issue:44509", "accept": false, - "reason": "Both are docs-related, but one is about return_all_scores and the other about removed pipeline tasks." - }, - { - "left": "issue:43906", - "right": "issue:44530", - "accept": false, - "reason": "A reproduction of another issue is not the same as the Qwen3.5 cache crash here." + "reason": "Both are docs issues, but they cover different removed/deprecated pipeline APIs and should not be merged." }, { "left": "issue:44671", "right": "issue:44991", "accept": false, - "reason": "CamemBERT masked-LM prediction regression and est-roberta tokenizer loading failure are different symptoms." + "reason": "CamemBERT masked-LM prediction regression is not the same as tokenizer loading failure for est-roberta." }, { "left": "issue:43994", "right": "issue:44079", "accept": false, - "reason": "SigLIP2 bad outputs and ModelOutput key assignment when None are unrelated internals." + "reason": "Wrong SigLIP2 outputs and ModelOutput key assignment semantics are different bugs." }, { "left": "issue:36032", "right": "issue:43452", "accept": false, - "reason": "T5 tokenizer method conflict and gguf_file from_pretrained breakage are separate tokenizer/model-loading problems." + "reason": "T5 tokenizer method conflict and gguf_file loading breakage are unrelated loading failures." }, { "left": "issue:44265", "right": "issue:44898", "accept": false, - "reason": "torch.export/export with torch_compilable_check and Perceiver interpolation failure are different subsystems." + "reason": "torch.export + torch_compilable_check and Perceiver interpolation failure are different export/inference issues." }, { "left": "issue:39404", "right": "issue:44912", "accept": false, - "reason": "Whisper pipeline behavior and MXFP4 quantization fallback on git-oss-20b are unrelated." + "reason": "Whisper return_language regression does not match MXFP4 quantization fallback behavior." }, { "left": "issue:46050", "right": "issue:46326", "accept": false, - "reason": "Quantized save/load breakage and mistral_common continue_final_message behavior are different bugs." + "reason": "Quantized model save/load breakage and mistral_common continue_final_message behavior are unrelated." }, { "left": "issue:44164", "right": "issue:45759", "accept": false, - "reason": "extra_state persistence and unwrapping text_config in from_config are not the same issue." + "reason": "extra_state save/load handling is different from unwrapping text_config for composite multimodal configs." }, { "left": "issue:43906", "right": "issue:46326", "accept": false, - "reason": "A reproduction issue and a mistral_common tokenizer bug are not duplicates." + "reason": "A reproduction issue and a mistral_common tokenizer behavior bug are not the same underlying change." }, { "left": "issue:44991", "right": "issue:45081", "accept": false, - "reason": "Tokenizer loading for est-roberta and Mistral regex patch crash are unrelated." + "reason": "Tokenizer loading failure for est-roberta and Mistral regex patch crash are different tokenizer-path bugs." }, { - "left": "issue:43299", - "right": "issue:43931", + "left": "issue:45823", + "right": "issue:46899", "accept": false, - "reason": "Both involve Qwen3-VL loading, but they target different model variants and failure modes; not enough to treat as one bug." + "reason": "Device-map mismatch in Gemma4 PLE and 4-bit dtype-cast issues in Gemma4 multimodal embedders are different failures." }, { - "left": "issue:45823", - "right": "issue:46899", + "left": "issue:43299", + "right": "issue:43931", "accept": false, - "reason": "Gemma4 device_map mismatch and Gemma4 4-bit dtype-cast processing bugs affect different code paths." + "reason": "Both involve Qwen3-VL family loading, but one is a dev-version MOE breakage and the other is a weight-shape mismatch for a specific checkpoint." }, { "left": "issue:45479", "right": "issue:46894", "accept": false, - "reason": "Zero-loss classification and training-loss double-shift are different failure patterns." + "reason": "Single-label classification zero-loss and CohereASR double-shift training loss bugs affect different model families and code paths." }, { "left": "issue:43856", "right": "issue:44756", "accept": false, - "reason": "MoE training memory inefficiency and Strix Halo mmap OOM are unrelated." - }, - { - "left": "issue:43596", - "right": "issue:44683", - "accept": false, - "reason": "Deepspeed ZeRO3/BertModel index error is unrelated to compiled flex_attention on torch>=2.9." + "reason": "Qwen3 MoE training memory inefficiency and Strix Halo mmap OOM are unrelated." }, { "left": "issue:45479", "right": "issue:46897", "accept": false, - "reason": "Zero-loss classification and Florence2 double-shift bug are not the same underlying defect." + "reason": "Zero-loss sequence-classification bug and Florence2 double-shift training-loss bug are not the same issue." }, { "left": "issue:44164", "right": "issue:46821", "accept": false, - "reason": "extra_state save/load and TimesFM window_size AttributeError are unrelated." + "reason": "extra_state save/load handling and TimesFM window_size AttributeError are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "deepspeed zero3 BertModel indexing failure and compiled flex_attention breakage on torch>=2.9 are different bugs." } ] }, @@ -3147,7 +3266,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 10, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -3155,7 +3274,6 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-502", "nodes": [ - "issue:32090", "issue:34567", "issue:36032", "issue:39404", @@ -3188,6 +3306,7 @@ "issue:44488", "issue:44530", "issue:44661", + "issue:44671", "issue:44683", "issue:44737", "issue:44811", @@ -3207,25 +3326,25 @@ "soft_pairs": [ "issue:44488|issue:44987", "issue:44488|issue:45081", - "issue:44462|issue:46489", "issue:43611|issue:43644", + "issue:44462|issue:46489", "issue:42796|issue:43404", "issue:36032|issue:42915", "issue:46285|issue:46402", + "issue:39404|issue:44671", "issue:43716|issue:45084", - "issue:32090|issue:39404", - "issue:44661|issue:44683", - "issue:44737|issue:45859", "issue:43643|issue:44530", - "issue:44451|issue:44987", - "issue:40990|issue:43421", + "issue:44737|issue:45859", + "issue:44661|issue:44683", "issue:42915|issue:44530", - "issue:43541|issue:45084", + "issue:40990|issue:43421", + "issue:44451|issue:44987", "issue:43644|issue:43646", - "issue:42175|issue:46178", - "issue:43582|issue:43632", "issue:34567|issue:40444", + "issue:43541|issue:45084", "issue:40444|issue:46326", + "issue:42175|issue:46178", + "issue:43582|issue:43632", "issue:44206|issue:44220", "issue:44811|issue:45584", "issue:36032|issue:43645", @@ -3236,230 +3355,221 @@ "issue:45084|issue:45507", "issue:43262|issue:43377" ], - "prepared_review_unit_hash": "f89ac69837f033496d22ac9805fdf3b57d1235b89d2dfcaa2da76caf07392cb2", + "prepared_review_unit_hash": "13ba43acb0bfc7db64aec49cdace3097e58680a25347075dfa4c1e090c992770", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36780, - "estimated_input_tokens": 9195, - "estimated_eval_tokens": 18646 + "serialized_chars": 36732, + "estimated_input_tokens": 9183, + "estimated_eval_tokens": 18622 }, "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 issues, with the strongest recurring theme being tokenizer/model-loading regressions in v5+. Only one soft pair clearly looks like the same concrete bug class (GGUF loading).", - "confidence": 0.66, - "canonical_issue_reason": "issue:44462 is the broadest representative of the tokenizer-loading regression theme in this set; it captures the general failure mode behind several later model/tokenizer loading reports more cleanly than the narrower, model-specific issues.", + "summary": "The cluster is a loose mix of Transformers v5/v5.1 regressions and tokenizer/model-loading bugs, but the paired items describe different concrete failures and should not be merged as duplicates. No soft edge is strong enough to accept.", + "confidence": 0.81, + "canonical_issue_reason": "issue:44488 is the broadest representative of the tokenizer-loading regression reports and has the strongest discussion/inbound signal among the near-matches.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44462 is the best overall issue candidate because it is the most general and central to the repeated tokenizer-loading complaints, making it the most useful canonical anchor for this cluster.", + "best_issue_reason": "issue:44488 is the best single issue to anchor this cluster because it is a clear user-facing loading regression with broad applicability, while the other items are more model-specific or unrelated.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44488", "right": "issue:44987", "accept": false, - "reason": "Both are tokenizer/model load failures, but they name different repositories and no evidence shows the same underlying code-path bug." + "reason": "Both are tokenizer-loading failures, but they affect different models and describe different breakages; not the same underlying bug." }, { "left": "issue:44488", "right": "issue:45081", "accept": false, - "reason": "Different failure modes: one is a model load regression, the other is a Mistral regex-patching AttributeError." + "reason": "Both involve Mistral-related tokenizer loading, but one is a repository/model loading failure and the other is a specific regex-patching crash path." }, { - "left": "issue:44462", - "right": "issue:46489", + "left": "issue:43611", + "right": "issue:43644", "accept": false, - "reason": "Both concern tokenizer behavior on v5+, but one is generic tokenizer.json precedence while the other is a model-specific output regression; not enough to treat as the same bug." + "reason": "Both are Transformers v5 regressions, but one is about base_model_prefix loading and the other about non-persistent buffers; different code paths." }, { - "left": "issue:43611", - "right": "issue:43644", + "left": "issue:44462", + "right": "issue:46489", "accept": false, - "reason": "Base-model-prefix loading and non-persistent buffer corruption are unrelated breakages." + "reason": "Both touch tokenizer loading, but 44462 is about ignoring tokenizer.json while 46489 is an output regression for a specific tokenizer/model after a prior fix." }, { "left": "issue:42796", "right": "issue:43404", "accept": false, - "reason": "Tokenizer BPE-marker handling and lm_head weight tying are different subsystems and different bugs." + "reason": "A tokenizer BPE-marker bug and a tied-weights/model-head bug are unrelated failures." }, { "left": "issue:36032", "right": "issue:42915", "accept": false, - "reason": "A tokenizer method-name conflict is unrelated to a FineGrainedFP8 model failure." + "reason": "T5 tokenizer method-name conflict and Qwen3Moe FP8 config failure are unrelated." }, { "left": "issue:46285", "right": "issue:46402", "accept": false, - "reason": "Different model-loading issues affecting different architectures and code paths." + "reason": "Both concern model structure/loading, but they affect different model families and different missing components." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline return_language and CamemBERT masked-LM predictions are unrelated model-specific bugs." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Image preprocessor dtype mismatch and template compilation failure are unrelated." + "reason": "Image preprocessor dtype mismatch and template compilation failure are different subsystems and failure modes." }, { - "left": "issue:32090", - "right": "issue:39404", + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "Missing fields in AutoConfig with trust_remote_code is unrelated to a PagedAttentionCache linear_attention crash." + }, + { + "left": "issue:44737", + "right": "issue:45859", "accept": false, - "reason": "Trainer broadcast TypeError and Whisper return_language behavior are unrelated." + "reason": "Both are device/offload-related, but they concern different models and different tensor construction paths." }, { "left": "issue:44661", "right": "issue:44683", "accept": false, - "reason": "Auto-model registration failure and compiled flex_attention are unrelated." + "reason": "Tokenizer-mapping name handling and compiled flex_attention on torch >= 2.9 are unrelated." }, { - "left": "issue:44737", - "right": "issue:45859", + "left": "issue:42915", + "right": "issue:44530", "accept": false, - "reason": "Both mention device/offload concerns, but they affect different models and different code paths." + "reason": "Qwen3Moe FP8 config failure and Qwen3.5 linear_attention cache crash are different bugs in different model paths." }, { - "left": "issue:43643", - "right": "issue:44530", + "left": "issue:40990", + "right": "issue:43421", "accept": false, - "reason": "AutoConfig trust_remote_code field loss and PagedAttentionCache group-type failure are unrelated." + "reason": "High perplexity on a model and runtime special-token post-processor updates are not the same bug." }, { "left": "issue:44451", "right": "issue:44987", "accept": false, - "reason": "Both are load failures on newer transformers versions, but they are different repos and no shared concrete bug is evident." + "reason": "Both are model-loading complaints, but they target different repositories and likely different tokenizer assets." }, { - "left": "issue:40990", - "right": "issue:43421", + "left": "issue:43644", + "right": "issue:43646", "accept": false, - "reason": "Perplexity regression and runtime post-processor updates are unrelated." + "reason": "Both are v5 custom-model regressions, but one is about junk buffers and the other about initialization itself; too distinct to merge." }, { - "left": "issue:42915", - "right": "issue:44530", + "left": "issue:34567", + "right": "issue:40444", "accept": false, - "reason": "Qwen3Moe FP8 config failure and PagedAttentionCache linear_attention crash are unrelated." + "reason": "TrainerState token counter and multi-image IterableDataset finetuning failure are unrelated." }, { "left": "issue:43541", "right": "issue:45084", "accept": false, - "reason": "Torch dynamo grouped_mm tracing failure and template-node compilation error are unrelated." + "reason": "Mixtral grouped_mm tracing and non-template-node compilation errors are different runtime/compiler issues." }, { - "left": "issue:43644", - "right": "issue:43646", + "left": "issue:40444", + "right": "issue:46326", "accept": false, - "reason": "Junk-filled non-persistent buffers and custom model initialization issues are different regressions." + "reason": "Multi-image finetuning and Whisper continue_final_message behavior are unrelated." }, { "left": "issue:42175", "right": "issue:46178", "accept": false, - "reason": "Missing tensorflow in extras and an unsupported backend name are related to packaging/backends broadly, but not the same concrete bug." + "reason": "Both are backend-installation issues, but one is missing tensorflow in the torch extra and the other is an unsupported tensorflow_text backend entry." }, { "left": "issue:43582", "right": "issue:43632", "accept": false, - "reason": "AppleSilicon allocator type error and _is_hf_initialized flag breakage are unrelated." - }, - { - "left": "issue:34567", - "right": "issue:40444", - "accept": false, - "reason": "TrainerState token counting and multimodal iterable-dataset failures are unrelated." - }, - { - "left": "issue:40444", - "right": "issue:46326", - "accept": false, - "reason": "Different multimodal/audio tokenizer behaviors with no shared bug path." + "reason": "Apple Silicon warmup TypeError and _is_hf_initialized flag breakage are unrelated." }, { "left": "issue:44206", "right": "issue:44220", "accept": false, - "reason": "Both are audio feature-extraction related, but one is a specific unsupported-arg crash and the other is too generic to confirm the same underlying defect." + "reason": "Both are audio-feature related, but one is about an unsupported center argument and the other about a different fbank extraction issue." }, { "left": "issue:44811", "right": "issue:45584", "accept": false, - "reason": "Both are Whisper special-token issues, but they affect different operations and do not clearly share the same bug." + "reason": "Both involve Whisper, but decode skip_special_tokens handling and empty-transcription generation after align_special_tokens are different bugs." }, { - "left": "issue:44987", - "right": "issue:44991", + "left": "issue:36032", + "right": "issue:43645", "accept": false, - "reason": "Both are tokenizer-loading regressions on v5+, but they are different models and there is no evidence they stem from the same concrete defect." + "reason": "Tokenizer method conflict and Jupyter custom-model initialization failure are not the same issue." }, { - "left": "issue:43317", + "left": "issue:43452", "right": "issue:43482", "accept": false, - "reason": "GPU/CPU offload dequantization failure and GGUF loading failure are different model-loading paths." + "reason": "Both are GGUF/loading regressions, but one is a generic from_pretrained gguf_file breakage and the other is a specific Qwen2.5-GGUF failure; not enough evidence of one underlying bug." }, { - "left": "issue:45084", - "right": "issue:45507", + "left": "issue:43883", + "right": "issue:44488", "accept": false, - "reason": "Template compilation and GraniteMoEHybrid invalid-method calls are unrelated." + "reason": "Molmo tied-weights attribute error and tokenizer loading failures are unrelated." }, { - "left": "issue:43262", - "right": "issue:43377", + "left": "issue:44987", + "right": "issue:44991", "accept": false, - "reason": "Audio chat-template sampling-rate defaults and MIMI batch-vs-single padding-mask behavior are unrelated." + "reason": "Both are tokenizer load failures, but for different models and likely different tokenizer implementations." }, { - "left": "issue:43452", + "left": "issue:43317", "right": "issue:43482", - "accept": true, - "reason": "Both are GGUF loading regressions in transformers v5, and they plausibly reflect the same underlying loading-path breakage." - }, - { - "left": "issue:43883", - "right": "issue:44488", "accept": false, - "reason": "Missing all_tied_weights_keys and a model-loading failure for a specific BERT repo are different issues." + "reason": "Dequantized model offload loading and Qwen2.5-GGUF loading are different loading paths." }, { - "left": "issue:44987", - "right": "issue:44991", + "left": "issue:45084", + "right": "issue:45507", "accept": false, - "reason": "Same broad symptom class, but no concrete evidence they are the same bug or fix." + "reason": "Template compilation failure and GraniteMoEHybrid invalid method calls are unrelated." }, { - "left": "issue:43317", - "right": "issue:43482", + "left": "issue:43262", + "right": "issue:43377", "accept": false, - "reason": "Dequantized device-map offload failure and GGUF loading failure are unrelated." + "reason": "Audio chat-template sample-rate defaults and MIMI batched-vs-single padding-mask behavior are different bugs." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative: it correctly treats the cluster as mostly unrelated, picks a reasonable tokenizer-loading canonical issue, and does not overclaim most soft pairs. The one accepted soft edge (43452/43482) is plausibly supported by the titles as the same GGUF loading regression class." - }, - "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 }, @@ -3468,7 +3578,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 11, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -3485,7 +3595,6 @@ "issue:43322", "issue:43377", "issue:43404", - "issue:43425", "issue:43531", "issue:43540", "issue:43632", @@ -3521,256 +3630,250 @@ "issue:45987", "issue:46315", "issue:46326", + "issue:46661", "issue:46821", "issue:46869", "issue:46899" ], "soft_pairs": [ "issue:43746|issue:46326", - "issue:43792|issue:46869", - "issue:45084|issue:46869", "issue:43646|issue:44534", - "issue:44403|issue:45081", - "issue:43645|issue:44534", - "issue:43883|issue:44991", "issue:44849|issue:45325", + "issue:44403|issue:45081", + "issue:43645|issue:46661", + "issue:42915|issue:45084", "issue:43377|issue:43819", + "issue:43883|issue:44991", + "issue:43792|issue:46869", "issue:43404|issue:44466", "issue:44464|issue:45987", - "issue:44987|issue:45081", "issue:43994|issue:44610", - "issue:44451|issue:45081", - "issue:42915|issue:44291", + "issue:44987|issue:45081", "issue:28530|issue:34567", "issue:44008|issue:46899", + "issue:42915|issue:44291", + "issue:44451|issue:45081", "issue:36032|issue:43742", "issue:44062|issue:44987", "issue:44265|issue:44610", - "issue:42915|issue:44589", "issue:43089|issue:45987", "issue:43322|issue:43540", - "issue:43632|issue:46821", "issue:44479|issue:45325", + "issue:42915|issue:44589", "issue:36032|issue:41720", - "issue:43531|issue:44877", - "issue:45216|issue:45357", + "issue:43632|issue:46821", "issue:44991|issue:46315", - "issue:43425|issue:44403" + "issue:43531|issue:44877", + "issue:45216|issue:45357" ], - "prepared_review_unit_hash": "176bba81bf22b90ee2d99e129457fe22c60514353e592f347b29d6ea85610d29", + "prepared_review_unit_hash": "45eff05b7d86399e88a8518a718273b14d5796df77ea02bfb0e64974bf6ec6f4", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36822, - "estimated_input_tokens": 9206, - "estimated_eval_tokens": 18668 + "soft_pair_count": 29, + "serialized_chars": 36615, + "estimated_input_tokens": 9154, + "estimated_eval_tokens": 18564 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous set of Transformers issues with one clear duplicate pair around Qwen3.5 save_pretrained checkpoint serialization. The rest are broad-topic collisions (tokenizers, multimodal models, generation, loading, and v5 regressions) but not the same concrete bug.", - "confidence": 0.79, - "canonical_issue_reason": "issue:45216 is the earliest and broadest report for the Qwen3.5 save_pretrained checkpoint corruption, and issue:45357 reads as the same visual-encoder-key serialization bug resurfacing later.", + "summary": "The cluster is mostly heterogeneous; nearly all pairs are adjacent-sounding but clearly different bugs. Only the two Qwen3.5 `save_pretrained` regressions look like a plausible duplicate pair.", + "confidence": 0.77, + "canonical_issue_reason": "issue:45216 is the clearest representative of the only strong duplicate-like area here: Qwen3.5 `save_pretrained` checkpoint serialization regressions.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45216 is the best canonical anchor for the only strong duplicate-like subcluster here: same API, same model family, and a concrete save_pretrained serialization fault.", + "best_issue_reason": "issue:45216 is the best standalone anchor because it describes a concrete, user-visible regression and overlaps directly with the closest soft match in the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43746", "right": "issue:46326", "accept": false, - "reason": "Different tokenizer subsystems and failure modes: PEFT checkpoint loading vs Mistral-common continue_final_message behavior." - }, - { - "left": "issue:43792", - "right": "issue:46869", - "accept": false, - "reason": "Whisper runtime failure and YOLO multi-GPU device-map failure are unrelated model paths." + "reason": "Different features: GraniteSpeech local checkpoint loading vs mistral_common tokenizer message handling." }, { - "left": "issue:45084", - "right": "issue:46869", + "left": "issue:43646", + "right": "issue:44534", "accept": false, - "reason": "Template compilation error vs YOLO device-map/autodispatch failure; no shared code-path." + "reason": "Custom model init breakage is unrelated to non-persistent buffer corruption." }, { - "left": "issue:43646", - "right": "issue:44534", + "left": "issue:44849", + "right": "issue:45325", "accept": false, - "reason": "Custom model initialization breakage vs non-persistent buffer serialization junk; different regressions." + "reason": "Different models and failures: Qwen3.5 hidden states vs Qwen2.5-VL rope/position-id scaling." }, { "left": "issue:44403", "right": "issue:45081", "accept": false, - "reason": "One is logging/noise during load, the other is a Mistral regex patch crash; not the same bug." + "reason": "Different code paths: generic loading noise vs Mistral tokenizer regex patch crash." }, { "left": "issue:43645", - "right": "issue:44534", + "right": "issue:46661", "accept": false, - "reason": "Notebook custom-model init regression is unrelated to buffer initialization corruption." + "reason": "Both touch model loading, but one is notebook/custom-model init and the other is remote-code `config_class` handling." }, { - "left": "issue:43883", - "right": "issue:44991", + "left": "issue:42915", + "right": "issue:45084", "accept": false, - "reason": "Different model-loading failures: missing tied-weights attribute vs tokenizer load failure for a specific model." + "reason": "FineGrainedFP8 Qwen3Moe failure is unrelated to template-node compile errors." }, { - "left": "issue:44849", - "right": "issue:45325", + "left": "issue:43377", + "right": "issue:43819", "accept": false, - "reason": "Output-hidden-states bug in Qwen3.5 is not the same as Qwen2.5-VL rope-index scaling." + "reason": "MIMI padding-mask batching bug is unrelated to DAC latent reconstruction mismatch." }, { - "left": "issue:43377", - "right": "issue:43819", + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Missing `all_tied_weights_keys` is unrelated to tokenizer loading for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43792", + "right": "issue:46869", "accept": false, - "reason": "MIMI batched-padding-mask mismatch and DAC from_latents STE mismatch are separate model bugs." + "reason": "Whisper model loading failure is unrelated to YOLO auto device-map multi-GPU failure." }, { "left": "issue:43404", "right": "issue:44466", "accept": false, - "reason": "Weight-tying defect in one model is not the same as device-dependent save_pretrained serialization inconsistency." + "reason": "Both involve lm_head/tied weights, but one is a missing tie and the other is device-dependent serialization inconsistency." }, { "left": "issue:44464", "right": "issue:45987", "accept": false, - "reason": "Chunked generation/compiled forward inconsistency is unrelated to StaticCache.get_seq_length return type." - }, - { - "left": "issue:44987", - "right": "issue:45081", - "accept": false, - "reason": "Tokenizer load regression for a specific model does not match the Mistral regex patch crash." + "reason": "Chunked generation/compiled forward mismatch is unrelated to `StaticCache.get_seq_length()` returning a tensor." }, { "left": "issue:43994", "right": "issue:44610", "accept": false, - "reason": "Siglip2 inference nonsense and OmDet-Turbo processor size mismatch are different model/processor bugs." + "reason": "SigLIP2 bad outputs and OmDet processor size mismatch are separate preprocessing/model-shape bugs." }, { - "left": "issue:44451", + "left": "issue:44987", "right": "issue:45081", "accept": false, - "reason": "Generic model load failure and Mistral tokenizer patch crash are distinct issues." - }, - { - "left": "issue:42915", - "right": "issue:44291", - "accept": false, - "reason": "FineGrainedFP8Config failure is unrelated to init_empty_weights / _is_hf_initialized handling." + "reason": "Tokenizer loading regression vs Mistral regex patch crash; different failure modes and targets." }, { "left": "issue:28530", "right": "issue:34567", "accept": false, - "reason": "Early stopping metric selection and num_input_tokens_seen not updating are unrelated trainer issues." + "reason": "Early stopping metric selection is unrelated to `num_input_tokens_seen` not updating." }, { "left": "issue:44008", "right": "issue:46899", "accept": false, - "reason": "Gemma3n variable collision and Gemma4 4-bit dtype cast are separate multimodal bugs." + "reason": "Gemma3n name collision bug is unrelated to Gemma4 4-bit multimodal dtype casting." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "Both are init/quantization related at a high level, but the concrete failures are different and not mergeable." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "Different models and root causes: ScandiBERT loading vs Mistral tokenizer regex patching." }, { "left": "issue:36032", "right": "issue:43742", "accept": false, - "reason": "T5 tokenizer special-token conflict is unrelated to MobileLLM key error." + "reason": "Tokenizer special-token conflict is unrelated to MobileLLM key error on load." }, { "left": "issue:44062", "right": "issue:44987", "accept": false, - "reason": "AddedToken special-argument duplication is not the same as loading physical-intelligence/fast." + "reason": "AddedToken argument duplication is not the same bug as the physical-intelligence loading failure." }, { "left": "issue:44265", "right": "issue:44610", "accept": false, - "reason": "torch.export torch_compilable_check failure is unrelated to OmDet-Turbo image-size mismatch." - }, - { - "left": "issue:42915", - "right": "issue:44589", - "accept": false, - "reason": "FP8 config failure and float8 storage lookup failure are separate quantization/runtime issues." + "reason": "`torch.export`/`torch_compilable_check` failure is unrelated to processor image-size mismatch." }, { "left": "issue:43089", "right": "issue:45987", "accept": false, - "reason": "Generation performance overhead and cache seq-length return-type mismatch are not the same bug." + "reason": "Generation performance overhead is unrelated to cache length return type." }, { "left": "issue:43322", "right": "issue:43540", "accept": false, - "reason": "Llava Next loading segfault and Qwen3OmniMoe video-input ValueError are different multimodal failures." + "reason": "Llava Next loading crash and Qwen3OmniMoe video-input error are different model pipelines." }, { - "left": "issue:43632", - "right": "issue:46821", + "left": "issue:44479", + "right": "issue:45325", "accept": false, - "reason": "_is_hf_initialized regression is unrelated to TimesFM window_size AttributeError." + "reason": "Both involve Qwen VL variants, but one is a broad video regression and the other is a still-image RoPE scaling bug." }, { - "left": "issue:44479", - "right": "issue:45325", + "left": "issue:42915", + "right": "issue:44589", "accept": false, - "reason": "Both are Qwen VL-family regressions, but one is a broad video-input failure and the other is a specific still-image rope-index bug; not clearly the same concrete change." + "reason": "FP8 config failure is unrelated to missing storage-object resolution for float8 tensors." }, { "left": "issue:36032", "right": "issue:41720", "accept": false, - "reason": "Tokenizer construction error and Qwen3 auto-device-map CUDA assert are unrelated." + "reason": "Tokenizer method-name conflict is unrelated to Qwen3 device-map CUDA assert." }, { - "left": "issue:43531", - "right": "issue:44877", + "left": "issue:43632", + "right": "issue:46821", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and granite_speech config loading are unrelated." - }, - { - "left": "issue:45216", - "right": "issue:45357", - "accept": true, - "reason": "Same Qwen3.5 save_pretrained serialization regression: both report incorrect checkpoint contents/visual encoder keys from the same API." + "reason": "`_is_hf_initialized` breakage is unrelated to TimesFM `window_size` handling." }, { "left": "issue:44991", "right": "issue:46315", "accept": false, - "reason": "Tokenizer load failure for one model is not the same as slow convert_tokens_to_ids regression." + "reason": "Tokenizer load failure and slow `convert_tokens_to_ids` are different tokenizer problems." }, { - "left": "issue:43425", - "right": "issue:44403", + "left": "issue:43531", + "right": "issue:44877", "accept": false, - "reason": "Torch version compatibility complaint and generic loading noise are not the same underlying bug." + "reason": "Qwen3-MoE sliding-window behavior is unrelated to strict GraniteSpeech config loading." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Both report the same Qwen3.5 `save_pretrained` regression: incorrect checkpoint serialization, with the visual-encoder-key problem being a narrower manifestation." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet. The Qwen3.5 pair (issue:45216 / issue:45357) is the only clearly supported duplicate-like cluster from the titles and chronology, and the remaining soft-edge rejections are conservative rather than overstated. No PR-mergeability concerns apply here." + "feedback": "The evaluation is grounded in the packet and stays conservative. The only accepted soft pair, issue:45216 vs issue:45357, is plausibly the same Qwen3.5 `save_pretrained` checkpoint-serialization bug with a narrower visual-encoder-key manifestation. The rejected soft pairs are explained with model/feature-level differences rather than overstated evidence. No clear precision issue stands out." }, "evaluator_used": true, "retried": false, @@ -3783,7 +3886,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 12, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -3803,6 +3906,7 @@ "issue:43317", "issue:43381", "issue:43421", + "issue:43425", "issue:43482", "issue:43525", "issue:43531", @@ -3835,250 +3939,249 @@ "issue:45923", "issue:46491", "issue:46710", - "issue:46821", "issue:46899", "issue:46909" ], "soft_pairs": [ - "issue:43296|issue:43531", - "issue:44220|issue:44683", - "issue:43611|issue:44534", "issue:40444|issue:45325", + "issue:43611|issue:44534", + "issue:43425|issue:44403", + "issue:43296|issue:43531", "issue:43716|issue:46899", "issue:45812|issue:46491", - "issue:44403|issue:44991", + "issue:44220|issue:44683", + "issue:42796|issue:43746", "issue:43381|issue:45923", + "issue:44403|issue:44991", "issue:43632|issue:44756", - "issue:42796|issue:43746", - "issue:43792|issue:44589", "issue:33666|issue:40444", "issue:40990|issue:44811", + "issue:43792|issue:44589", "issue:44291|issue:44589", "issue:42796|issue:46710", - "issue:43482|issue:43525", - "issue:35707|issue:40444", "issue:44079|issue:45663", - "issue:43482|issue:45538", + "issue:35707|issue:40444", + "issue:43482|issue:43525", "issue:43421|issue:44568", - "issue:43531|issue:44291", + "issue:43482|issue:45538", "issue:43632|issue:44534", - "issue:42175|issue:45542", + "issue:43531|issue:44291", "issue:41553|issue:43582", + "issue:42175|issue:45542", "issue:41720|issue:45081", - "issue:44462|issue:45812", "issue:43317|issue:43856", + "issue:44462|issue:45812", "issue:44079|issue:44534", - "issue:44509|issue:46909", - "issue:43296|issue:46821" + "issue:44509|issue:46909" ], - "prepared_review_unit_hash": "e3b5456eb387971c297dbd2065ecca422326b353b17a05f88903e69194a460e5", + "prepared_review_unit_hash": "9e56bb5744a78f3b817f16213efd36f1d43d274d8a7c59fe72fa9de33e7f06be", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36038, - "estimated_input_tokens": 9010, - "estimated_eval_tokens": 18276 + "serialized_chars": 36000, + "estimated_input_tokens": 9000, + "estimated_eval_tokens": 18256 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: it mixes unrelated tokenizer regressions, multimodal model-loading bugs, training/runtime failures, and docs/backend issues. None of the soft-edge pairs look like the same underlying bug or change, so I would not merge any of them as duplicates.", - "confidence": 0.31, - "canonical_issue_reason": null, + "summary": "No true duplicates here: the soft matches are mostly superficial overlaps within Transformers/v5, tokenizer, and multimodal areas, but each item describes a different concrete bug or regression. No PRs are present.", + "confidence": 0.93, + "canonical_issue_reason": "issue:45812 is the closest cluster representative: it describes a broad, concrete AutoTokenizer regression affecting all Granite models and has follow-up references; however, the set is still heterogeneous, so it is only a representative, not a duplicate target.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45812 is the broadest and most user-impacting regression here (token IDs wrong across Granite models after v4\u2192v5), so it is the best representative issue if one must be chosen.", + "best_issue_reason": "issue:45812 is the best single issue to anchor the cluster because it is specific, user-impacting, and broadly scoped compared with the others; the rest are separate bugs across different models/code paths.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43296", - "right": "issue:43531", + "left": "issue:40444", + "right": "issue:45325", "accept": false, - "reason": "Different problems: PaddleOCR-VL load failure in vLLM vs Qwen3-MoE sliding_window behavior." + "reason": "Both are Qwen2.5-VL-related, but one is a finetuning failure with multiple images per prompt while the other is a rope-index scaling bug for still-image position_ids; different failure modes and fixes." }, { - "left": "issue:44220", - "right": "issue:44683", + "left": "issue:43611", + "right": "issue:44534", "accept": false, - "reason": "Different code paths: audio feature extraction failure vs compiled flex_attention breakage on torch>=2.9." + "reason": "Both mention Transformers v5, but one is about loading models with base_model_prefix and the other about non-persistent buffers being filled with junk; unrelated code paths." }, { - "left": "issue:43611", - "right": "issue:44534", + "left": "issue:43425", + "right": "issue:44403", "accept": false, - "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is non-persistent buffer initialization." + "reason": "One is a Torch 2.10 compatibility break, the other is a logging/noise problem when loading a transformer; not the same bug." }, { - "left": "issue:40444", - "right": "issue:45325", + "left": "issue:43296", + "right": "issue:43531", "accept": false, - "reason": "Both involve Qwen2.5-VL, but one is multi-image finetuning failure and the other is temporal position_id scaling for still images." + "reason": "PaddleOCR-VL model loading in vLLM and Qwen3-MoE sliding_window behavior are separate model-specific issues." }, { "left": "issue:43716", "right": "issue:46899", "accept": false, - "reason": "Both are dtype-related multimodal issues, but they affect different models and different failure modes." + "reason": "Both involve multimodal dtype handling, but one is a Mistral-3 preprocessor/model dtype mismatch and the other is unsafe 4-bit cast in Gemma4 embedders; distinct paths." }, { "left": "issue:45812", "right": "issue:46491", "accept": false, - "reason": "Both are tokenizer regressions, but Granite wrong token IDs and CodeLlama leading-space round-trip loss are distinct bugs." + "reason": "Granite tokenizer wrong IDs and CodeLlama leading-space round-trip are both tokenizer regressions, but they affect different tokenizers and different semantics." }, { - "left": "issue:44403", - "right": "issue:44991", + "left": "issue:44220", + "right": "issue:44683", "accept": false, - "reason": "Unnecessary loading noise is not the same as a tokenizer-loading failure for a specific model." + "reason": "_torch_extract_fbank_features() and compiled flex_attention on torch>=2.9 are different low-level failures with no shared code-path evidence." }, { - "left": "issue:43381", - "right": "issue:45923", + "left": "issue:42796", + "right": "issue:43746", "accept": false, - "reason": "One is gradient checkpointing being unusable in eval mode; the other is a missing supports_gradient_checkpointing flag." + "reason": "Ministral tokenizer BPE-marker handling and GraniteSpeech PEFT local checkpoint loading are unrelated tokenizer vs model-loading bugs." }, { - "left": "issue:43632", - "right": "issue:44756", + "left": "issue:43381", + "right": "issue:45923", "accept": false, - "reason": "Unrelated failures: _is_hf_initialized breakage vs mmap OOM on Strix Halo." + "reason": "Eval-mode gradient checkpointing and a missing supports_gradient_checkpointing flag are related only at a high level; one is a runtime mode restriction, the other is a config flag omission." }, { - "left": "issue:42796", - "right": "issue:43746", + "left": "issue:44403", + "right": "issue:44991", "accept": false, - "reason": "Tokenizer BPE-marker handling and PEFT checkpoint loading are unrelated." + "reason": "Unnecessary loading noise and tokenizer loading failure for EMBEDDIA/est-roberta are different issues; one is cosmetic, the other blocks loading." }, { - "left": "issue:43792", - "right": "issue:44589", + "left": "issue:43632", + "right": "issue:44756", "accept": false, - "reason": "Whisper runtime failure and missing Float8 storage type are different loading/runtime issues." + "reason": "_is_hf_initialized regression and mmap OOM on Strix Halo are unrelated v5/runtime environment problems." }, { "left": "issue:33666", "right": "issue:40444", "accept": false, - "reason": "Both are Qwen2-VL/Qwen2.5-VL training-related, but multi-GPU training and multi-image dataset failure are not the same bug." + "reason": "Qwen2-VL multi-GPU training and Qwen2.5-VL multi-image finetuning are both training topics, but they concern different models and different bottlenecks." }, { "left": "issue:40990", "right": "issue:44811", "accept": false, - "reason": "Perplexity regression on GPT-OSS is unrelated to Whisper batch_decode special-token handling." + "reason": "High perplexity on gpt-oss-20b and Whisper batch_decode skip_special_tokens behavior are unrelated symptoms and code paths." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Whisper-large-v2 loading failure and missing Float8 storage support are distinct backend/runtime errors." }, { "left": "issue:44291", "right": "issue:44589", "accept": false, - "reason": "Different root causes: unexpected _is_hf_initialized arg versus missing Float8 storage object." + "reason": "Unexpected _is_hf_initialized argument during init_empty_weights and missing Float8 storage are unrelated transformer-v5 breakages." }, { "left": "issue:42796", "right": "issue:46710", "accept": false, - "reason": "Tokenizer marker handling for Ministral is unrelated to DeepSeek accuracy regression after upgrade." + "reason": "Tokenizer BPE-marker handling for Ministral and accuracy regression for DeepSeek-R1-Distill-Llama are different model-specific regressions." }, { - "left": "issue:43482", - "right": "issue:43525", + "left": "issue:44079", + "right": "issue:45663", "accept": false, - "reason": "Both are v5 loading issues, but Qwen2.5-GGUF failure and missing pad_token_id on Llama4Config are distinct." + "reason": "ModelOutput key assignment bug and Gemma4 FSDP2 shared_kv_states KeyError are unrelated internals." }, { "left": "issue:35707", "right": "issue:40444", "accept": false, - "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to Qwen2.5-VL multi-image finetuning." - }, - { - "left": "issue:44079", - "right": "issue:45663", - "accept": false, - "reason": "ModelOutput key assignment bug and Gemma4 FSDP2 shared_kv_states KeyError are different problems." + "reason": "Progressive generation with inputs_embeds/past_key_values and Qwen2.5-VL multi-image finetuning are different generation vs training bugs." }, { "left": "issue:43482", - "right": "issue:45538", + "right": "issue:43525", "accept": false, - "reason": "Qwen2.5-GGUF loading failure is unrelated to CLIPTokenizer model_max_length behavior." + "reason": "Qwen2.5-GGUF loading in v5 and Llama4Config missing pad_token_id are separate config/loading issues." }, { "left": "issue:43421", "right": "issue:44568", "accept": false, - "reason": "Runtime post-processor update semantics are not the same as add_special_tokens failing to insert BOS/EOS." + "reason": "Runtime post-processor updates in TokenizersBackend and mdeberta-v3 add_special_tokens behavior are different tokenizer-stack bugs." }, { - "left": "issue:43531", - "right": "issue:44291", + "left": "issue:43482", + "right": "issue:45538", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior is unrelated to init_empty_weights argument handling." + "reason": "Qwen2.5-GGUF loading failure and CLIPTokenizer model_max_length are unrelated model-loading vs tokenizer-constant issues." }, { "left": "issue:43632", "right": "issue:44534", "accept": false, - "reason": "_is_hf_initialized flag breakage and non-persistent buffer junk are different v5 regressions." + "reason": "_is_hf_initialized flag breakage and non-persistent buffers junk are both v5 regressions, but they affect different internals and fixes." }, { - "left": "issue:42175", - "right": "issue:45542", + "left": "issue:43531", + "right": "issue:44291", "accept": false, - "reason": "Both mention TensorFlow backend packaging, but missing TF in torch extras and undefined backend with tensorboard-only install are different issues." + "reason": "Qwen3-MoE sliding_window and unexpected _is_hf_initialized argument are not the same bug; one is model config behavior, the other is init-time argument compatibility." }, { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "AutoTokenizer error messaging for Voxtral is unrelated to AppleSilicon caching_allocator_warmup TypeError." + "reason": "Bad AutoTokenizer error for Voxtral and caching_allocator_warmup TypeError on Apple Silicon are completely different issues." }, { - "left": "issue:41720", - "right": "issue:45081", + "left": "issue:42175", + "right": "issue:45542", "accept": false, - "reason": "Qwen3 auto device mapping cudaErrorAssert and Mistral tokenizer regex patch crash are unrelated." + "reason": "TensorFlow missing from `.[torch]` install and tensorboard-only-without-TensorFlow backend error are related packaging issues but not the same underlying fix." }, { - "left": "issue:44462", - "right": "issue:45812", + "left": "issue:41720", + "right": "issue:45081", "accept": false, - "reason": "Ignoring tokenizer.json and producing wrong Granite token IDs are different AutoTokenizer regressions." + "reason": "Qwen3 auto device mapping cudaErrorAssert and Mistral regex patch crash are distinct model/device vs tokenizer-patch failures." }, { "left": "issue:43317", "right": "issue:43856", "accept": false, - "reason": "Dequantized model offload loading failure and Qwen3 MoE memory inefficiency are different load/training issues." + "reason": "Dequantized model offload failure with device_map=auto and Qwen3 MoE training memory inefficiency are different load-time vs training-memory problems." }, { - "left": "issue:44079", - "right": "issue:44534", + "left": "issue:44462", + "right": "issue:45812", "accept": false, - "reason": "ModelOutput key bookkeeping and v5 non-persistent buffer initialization are unrelated." + "reason": "Ignoring tokenizer.json from repo and wrong Granite token IDs both involve AutoTokenizer, but they are different repository-selection vs token-construction bugs." }, { - "left": "issue:44509", - "right": "issue:46909", + "left": "issue:44079", + "right": "issue:44534", "accept": false, - "reason": "Both are documentation issues, but they concern different docs and different content errors." + "reason": "ModelOutput key handling and non-persistent buffer junk are unrelated internal state-management regressions." }, { - "left": "issue:43296", - "right": "issue:46821", + "left": "issue:44509", + "right": "issue:46909", "accept": false, - "reason": "PaddleOCR-VL loading failure is unrelated to TimesFM window_size AttributeError." + "reason": "Removed pipeline-task docs and a broken DIA documentation example are both docs issues, but they concern different pages and content." } ] }, @@ -4094,7 +4197,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 13, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -4153,236 +4256,243 @@ "soft_pairs": [ "issue:43377|issue:45491", "issue:43611|issue:43646", + "issue:43296|issue:46821", "issue:43818|issue:46402", - "issue:42175|issue:43296", "issue:44829|issue:45479", "issue:43644|issue:44079", "issue:43818|issue:45357", - "issue:44265|issue:45072", + "issue:42175|issue:43296", "issue:43493|issue:43994", + "issue:44265|issue:45072", "issue:40444|issue:45381", "issue:45325|issue:45834", "issue:45325|issue:45381", "issue:43454|issue:44466", "issue:42915|issue:43482", - "issue:44291|issue:44877", "issue:43873|issue:45701", + "issue:44291|issue:44877", "issue:43701|issue:45325", "issue:42796|issue:45381", "issue:43606|issue:44265", "issue:40990|issue:45584", "issue:43262|issue:43746", "issue:44479|issue:45823", - "issue:43828|issue:46821", "issue:35707|issue:41720", "issue:34634|issue:35707", + "issue:43828|issue:46821", "issue:39692|issue:40444", - "issue:43454|issue:43550", "issue:44186|issue:44265", + "issue:43454|issue:43550", "issue:43606|issue:43819" ], - "prepared_review_unit_hash": "1ceb374f4786504636c349a26116e56ecac230e5acbe0cf8e4ab13d17ad570ac", + "prepared_review_unit_hash": "df6d9f9670d1d6349e0ddcb6ad9496ba555abcd5019504bcbff5617db1ec3a21", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 29, - "serialized_chars": 36078, - "estimated_input_tokens": 9020, - "estimated_eval_tokens": 18296 + "soft_pair_count": 30, + "serialized_chars": 36374, + "estimated_input_tokens": 9094, + "estimated_eval_tokens": 18444 }, "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 across different models and subsystems. There is no strong duplicate hub; at best it has a loose theme of v5/v5.x regressions and model-specific correctness issues.", + "summary": "This is a loose collection of mostly unrelated model/runtime issues, not a tight duplicate cluster. All soft pairs look like superficial similarity only, so none should be merged as the same underlying bug.", "confidence": 0.84, - "canonical_issue_reason": "issue:43611 is the broadest, most central regression report in the set: a Transformers 5 loading break that can affect many model families. It is more representative of the cluster than the highly model-specific bugs.", + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43611 is the best representative issue for triage because it captures the widest cross-model failure mode. That said, the cluster is too heterogeneous to treat as a true duplicate set.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43377", "right": "issue:45491", "accept": false, - "reason": "Both are encoder/attention correctness bugs, but they affect different models and symptoms (padding mask vs sliding-window NaNs)." + "reason": "Different bugs: MIMI padding-mask batching mismatch vs Gemma3 sliding-window/all-padding NaNs." }, { "left": "issue:43611", "right": "issue:43646", "accept": false, - "reason": "Both involve Transformers 5 regressions, but one is model loading/base_model_prefix and the other is custom initialization; not the same bug." + "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is custom model initialization." }, { - "left": "issue:43818", - "right": "issue:46402", + "left": "issue:43296", + "right": "issue:46821", "accept": false, - "reason": "Different models and failures: missing temporal attention in Video-LLaVA vs Sam3 text encoder weights not loaded." + "reason": "Unrelated models and failure modes: PaddleOCR-VL load error vs TimesFM window_size AttributeError." }, { - "left": "issue:42175", - "right": "issue:43296", + "left": "issue:43818", + "right": "issue:46402", "accept": false, - "reason": "Unrelated packaging/backend issue versus a vLLM model-load failure." + "reason": "Different code paths: Video-LLaVA architecture/weights issue vs Sam3 text encoder weights not loading." }, { "left": "issue:44829", "right": "issue:45479", "accept": false, - "reason": "Both are training degeneracy reports, but the root causes differ: FA3 attention regression vs single-label num_labels=1 loss behavior." + "reason": "Different training regressions: flash_attention_3 degenerate training vs num_labels=1 zero-loss classification behavior." }, { "left": "issue:43644", "right": "issue:44079", "accept": false, - "reason": "Different code paths: non-persistent buffer serialization junk versus ModelOutput key assignment." + "reason": "Non-persistent buffer serialization bug is unrelated to ModelOutput key assignment." }, { "left": "issue:43818", "right": "issue:45357", "accept": false, - "reason": "Different failures in different models: Video-LLaVA architecture issue versus Qwen3.5 save_pretrained key regression." + "reason": "Video-LLaVA tower bug is unrelated to Qwen3.5 save_pretrained visual-key regression." }, { - "left": "issue:44265", - "right": "issue:45072", + "left": "issue:42175", + "right": "issue:43296", "accept": false, - "reason": "One is torch.export failing on torch_compilable_check; the other is bfloat16 dtype mismatch in inference." + "reason": "Packaging/backend dependency issue is unrelated to a specific model load failure." }, { "left": "issue:43493", "right": "issue:43994", "accept": false, - "reason": "Both are SigLIP2 correctness complaints, but one is HF-vs-JAX parity and the other is AutoModel/pipeline behavior; not clearly the same bug." + "reason": "Both touch SigLIP2, but one is HF-vs-JAX implementation mismatch and the other is bad AutoModel/pipeline outputs." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "torch.export export/check failure is a different bug from bfloat16 inference dtype mismatches." }, { "left": "issue:40444", "right": "issue:45381", "accept": false, - "reason": "Different Qwen2.5-VL failures: iterable-dataset multi-image finetuning versus video position-id regression." + "reason": "Multiple-images finetuning failure is not the same as Qwen2.5-VL video position-id regression." }, { "left": "issue:45325", "right": "issue:45834", "accept": false, - "reason": "Both concern Qwen-style multimodal indexing, but one is temporal position scaling and the other is OCR long-input index errors." + "reason": "Different models and defects: Qwen2.5-VL rope scaling vs Kosmos2.5 long-OCR index error." }, { "left": "issue:45325", "right": "issue:45381", "accept": false, - "reason": "Related video-position bugs, but the reported symptoms and affected code paths are distinct enough to keep separate." + "reason": "Same subsystem, but distinct bugs: still-image temporal scaling vs video vision_position_ids." }, { "left": "issue:43454", "right": "issue:44466", "accept": false, - "reason": "Weight tying / lm_head loading issue versus inconsistent serialization behavior; related area, not the same defect." + "reason": "Tied-weight serialization inconsistency is not the same as failing to tie lm_head weights." }, { "left": "issue:42915", "right": "issue:43482", "accept": false, - "reason": "Different model families and failures: FineGrainedFP8Config versus GGUF loading under v5." + "reason": "FP8 config failure and GGUF loading failure are unrelated issues." }, { - "left": "issue:44291", - "right": "issue:44877", + "left": "issue:43873", + "right": "issue:45701", "accept": false, - "reason": "Both are config/init compatibility issues, but one is init_empty_weights argument handling and the other is strict config loading for granite_speech." + "reason": "Quantization/offloading behavior is unrelated to tokenization-version changes." }, { - "left": "issue:43873", - "right": "issue:45701", + "left": "issue:44291", + "right": "issue:44877", "accept": false, - "reason": "Quantization/offloading behavior is unrelated to tokenizer changes across versions." + "reason": "Unexpected init argument during init_empty_weights is unrelated to strict config loading for granite_speech." }, { "left": "issue:43701", "right": "issue:45325", "accept": false, - "reason": "resume_from_checkpoint key mismatch is unrelated to Qwen2.5-VL temporal position-id scaling." + "reason": "Checkpoint key mismatch is unrelated to Qwen2.5-VL rope-index scaling." }, { "left": "issue:42796", "right": "issue:45381", "accept": false, - "reason": "Tokenizer BPE-marker handling and Qwen2.5-VL video position ids are different bugs." + "reason": "Tokenizer BPE-marker handling is unrelated to video vision_position_ids." }, { "left": "issue:43606", "right": "issue:44265", "accept": false, - "reason": "CPU-offload device mismatch for Bark-small is unrelated to torch.export/torch_compilable_check failures." + "reason": "CPU offload device mismatch is unrelated to torch_compilable_check export failures." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "Perplexity anomaly on gpt-oss-20b is not the same as Whisper empty-transcription generation failure." + "reason": "High perplexity on WikiText-2 and Whisper empty-transcription generation failure are unrelated." }, { "left": "issue:43262", "right": "issue:43746", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting and PEFT checkpoint loading are unrelated." + "reason": "Audio chat-template sample-rate defaulting is unrelated to GraniteSpeech PEFT checkpoint loading." }, { "left": "issue:44479", "right": "issue:45823", "accept": false, - "reason": "Both are device/vision-related, but one is a Qwen video regression and the other is a Gemma4 PLE device_map forward mismatch." - }, - { - "left": "issue:43828", - "right": "issue:46821", - "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to a TimesFM window_size AttributeError." + "reason": "Qwen video-input regression and Gemma4 device-mismatch are different concrete bugs." }, { "left": "issue:35707", "right": "issue:41720", "accept": false, - "reason": "Progressive generation with inputs_embeds/past_key_values is a different code path from Qwen3 auto device mapping cudaErrorAssert." + "reason": "inputs_embeds/past_key_values generation bug is unrelated to Qwen3 auto device-map CUDA asserts." }, { "left": "issue:34634", "right": "issue:35707", "accept": false, - "reason": "Bark voice_preset and generation with inputs_embeds/past_key_values are unrelated failures." + "reason": "BarkProcessor voice_preset and progressive generation are unrelated features." }, { - "left": "issue:39692", - "right": "issue:40444", + "left": "issue:43828", + "right": "issue:46821", "accept": false, - "reason": "SigLIP2 docs/model mismatch and Qwen2.5-VL multi-image finetuning failure are different issues." + "reason": "Autocast dtype mismatch and TimesFM window_size AttributeError are different failures." }, { - "left": "issue:43454", - "right": "issue:43550", + "left": "issue:39692", + "right": "issue:40444", "accept": false, - "reason": "Weight tying/text-generation garbage is unrelated to Bamba torch.compile SDPA failure." + "reason": "SigLIP2 doc/example issues are unrelated to Qwen2.5-VL multi-image finetuning failures." }, { "left": "issue:44186", "right": "issue:44265", "accept": false, - "reason": "LayoutLMv2 tokenizer crashes are unrelated to torch.export/torch_compilable_check export failures." + "reason": "Tokenizer padding/truncation crash is unrelated to torch.export/export-check behavior." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "AyaVision weight-tying bug is unrelated to Bamba torch.compile SDPA failure." }, { "left": "issue:43606", "right": "issue:43819", "accept": false, - "reason": "Both are model correctness bugs, but one is device offload mismatch and the other is DAC latents/STE mismatch; not mergeable as one fix." + "reason": "CPU offload device mismatch is unrelated to DAC.from_latents vs forward mismatch." } ] }, @@ -4398,7 +4508,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 14, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -4469,246 +4579,246 @@ "issue:43377|issue:44898", "issue:43454|issue:43720", "issue:42617|issue:42915", + "issue:43749|issue:44534", "issue:43317|issue:43873", "issue:43572|issue:45759", - "issue:43749|issue:44534", "issue:43322|issue:46050", "issue:41720|issue:43322", "issue:36331|issue:42617", "issue:44387|issue:45704", "issue:41720|issue:44977", - "issue:36331|issue:42915", "issue:44466|issue:45357", + "issue:36331|issue:42915", "issue:43526|issue:43883", + "issue:40990|issue:45701", "issue:43450|issue:45701", + "issue:43881|issue:44991", "issue:44877|issue:44991", "issue:43065|issue:46402", - "issue:43881|issue:44991", - "issue:40990|issue:45701", - "issue:41720|issue:42617", "issue:26413|issue:28530", + "issue:41720|issue:42617", "issue:44186|issue:45072" ], - "prepared_review_unit_hash": "b876dcedb78b7b119fc45c360ad1df54612d08a3e563ec490f41770f9b7bbfd9", + "prepared_review_unit_hash": "2051762cde6eb64631f75b775900e1e6cc740f0786d2b76112847237f1540eea", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37302, - "estimated_input_tokens": 9326, - "estimated_eval_tokens": 18908 + "serialized_chars": 37300, + "estimated_input_tokens": 9325, + "estimated_eval_tokens": 18906 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a heterogeneous issue cluster with no true duplicate pairings; the soft-similarity matches are mostly broad-topic overlaps (tokenization, loading, device_map, offloading) but not the same concrete bug.", - "confidence": 0.95, - "canonical_issue_reason": "No single issue cleanly represents the whole cluster; if one must be chosen, issue 43873 is the closest fit because it is the most concrete and high-signal offloading/quantization regression with multiple inbound refs.", + "summary": "This cluster is largely heterogeneous: it spans tokenizer regressions, model-loading failures, device_map/offload issues, quantization bugs, and a few trainer/model-configuration problems. There is no strong duplicate pair; the closest theme is broad loading/offload/quantization regressions.", + "confidence": 0.39, + "canonical_issue_reason": "Issue 43873 is the broadest and most central report in the cluster, covering the shared offloading/quantization theme better than the more model-specific incidents.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43873 is the strongest standalone representative here: open, user-facing, and backed by multiple references/discussion, unlike the mostly unrelated neighboring reports.", + "best_issue_reason": "43873 is the best anchor for the cluster because it is general enough to represent the loading/offload/quantization subgroup, even though the overall set is not a clean duplicate cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43881", "right": "issue:44488", "accept": false, - "reason": "Both are model loading failures, but for different models and likely different root causes; not the same bug." + "reason": "Both are loading failures, but they involve different models and likely different failure modes; too little evidence they are the same bug." }, { "left": "issue:41720", "right": "issue:45823", "accept": false, - "reason": "Both involve device_map/placement issues, but one is a CUDA assert on A800 and the other is a forward-time device mismatch in a different model." + "reason": "Both mention device_map/auto, but one is a cudaErrorAssert on Qwen3 and the other is a device mismatch in Gemma4; not the same concrete bug." }, { "left": "issue:43377", "right": "issue:44610", "accept": false, - "reason": "Both mention batched behavior, but one is an encoder output mismatch and the other is a processor shape mismatch in a different pipeline." + "reason": "Different subsystems entirely: MIMI encoder batching/padding vs OmDet-Turbo processor image sizing." }, { "left": "issue:43257", "right": "issue:45923", "accept": false, - "reason": "Different problems: MoE weight conversion with accelerate/deepspeed vs a missing gradient-checkpointing flag on a trust_remote_code model." + "reason": "Qwen3 MoE loading/accelerate-deepspeed conversion issue is unrelated to a missing gradient_checkpointing flag." }, { "left": "issue:46305", "right": "issue:46315", "accept": false, - "reason": "Completely unrelated code paths: a sliding-window mask off-by-one vs tokenizer conversion performance regression." + "reason": "One is a sliding-window mask off-by-one bug, the other is a tokenizer performance regression; unrelated." }, { "left": "issue:44568", "right": "issue:45920", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different APIs and symptoms; not the same concrete defect." + "reason": "Both are tokenizer regressions, but one is specific BOS/EOS insertion for a single tokenizer while the other is a broader AutoTokenizer ID mapping problem across several models." }, { "left": "issue:42796", "right": "issue:46209", "accept": false, - "reason": "Tokenizer BPE-marker handling and FP8 inference on B200 are unrelated bugs." + "reason": "Tokenizer BPE-marker handling and FP8 inference failure on B200 are unrelated code paths." }, { "left": "issue:43454", "right": "issue:45072", "accept": false, - "reason": "lm_head weight tying and bfloat16 dtype mismatches are different model/inference issues." + "reason": "Weight-tying/generation corruption and bfloat16 dtype mismatches are different bugs." }, { "left": "issue:45823", "right": "issue:46821", "accept": false, - "reason": "Different subsystems entirely: device_map forward mismatch vs a TimesFM window_size AttributeError." + "reason": "Device-mapping forward mismatch and a TimesFM window_size AttributeError do not appear to share a code path." }, { "left": "issue:40990", "right": "issue:42947", "accept": false, - "reason": "Perplexity on a model/dataset and ineffective gradient checkpointing are not the same bug." + "reason": "Perplexity regression on a model is unrelated to gradient checkpointing ineffective with PEFT LoRA." }, { "left": "issue:43377", "right": "issue:44898", "accept": false, - "reason": "Both are input-shape related, but one is batched-vs-single encoder output drift and the other is resolution handling in Perceiver classification." + "reason": "Missing padding-mask support in MIMI encoder is not the same as Perceiver non-default-resolution failures." }, { "left": "issue:43454", "right": "issue:43720", "accept": false, - "reason": "Weight tying in AyaVision and unpacking packed weights in BitNet are different loading paths and failures." + "reason": "Tied-weight generation issues and BitNet packed-weight loading failures are distinct problems." }, { "left": "issue:42617", "right": "issue:42915", "accept": false, - "reason": "No shared bug: one is a 3d_parallel.py runtime failure, the other is a FineGrainedFP8Config issue." + "reason": "Unable to run a parallel script and a FineGrainedFP8Config failure are unrelated." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "Both involve loading behavior, but one is FSDP CPU RAM efficient loading and the other is non-persistent buffer initialization; not the same bug." }, { "left": "issue:43317", "right": "issue:43873", "accept": false, - "reason": "Both touch offloading/quantization, but one is a dequantized-model load failure and the other is broader offloading not working; root cause is not established as the same." + "reason": "These are the closest pair, but one is a specific dequantized model offload failure and the other is a broader offloading/quantization report; insufficient evidence to merge as the same bug." }, { "left": "issue:43572", "right": "issue:45759", "accept": false, - "reason": "Config-field absence and text_config unwrapping are different config-loading problems." - }, - { - "left": "issue:43749", - "right": "issue:44534", - "accept": false, - "reason": "Both involve loading/state behavior in v5, but one is FSDP CPU RAM-efficient loading and the other is non-persistent buffer corruption." + "reason": "Missing pad_token_idx after a v5 update and failing to unwrap text_config in composite Qwen configs are different configuration issues." }, { "left": "issue:43322", "right": "issue:46050", "accept": false, - "reason": "Segmentation fault while loading Llava Next is unrelated to sinq quantization save/load breakage." + "reason": "Segmentation fault loading Llava Next and broken save_pretrained/from_pretrained after sinq quantization are unrelated." }, { "left": "issue:41720", "right": "issue:43322", "accept": false, - "reason": "Different failures and code paths: auto device-map CUDA assert vs loading crash/segfault for Llava Next." + "reason": "Different symptoms and models: cudaErrorAssert with auto device mapping vs a segmentation fault while loading Llava Next." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "Unexpected compute_loss kwargs and 3d_parallel execution failure are unrelated training issues." + "reason": "CustomTrainer compute_loss signature change is unrelated to a 3d_parallel.py runtime failure." }, { "left": "issue:44387", "right": "issue:45704", "accept": false, - "reason": "Both are memory-related, but one is int4 CUDA reserved-memory growth and the other is an apex RMSNorm leak." + "reason": "Int4 CUDA reserved memory/OOM and an apex.FusedRMSNorm memory leak are distinct quantization/memory issues." }, { "left": "issue:41720", "right": "issue:44977", "accept": false, - "reason": "Device-map CUDA assert and flash-attention generation failure are different bugs." + "reason": "Both involve Qwen-family models, but one is a device-map assert and the other is a flash-attention generation issue." }, { - "left": "issue:36331", - "right": "issue:42915", + "left": "issue:44466", + "right": "issue:45357", "accept": false, - "reason": "Unexpected compute_loss keyword handling and Qwen3Moe FP8 failure are unrelated." + "reason": "Both are serialization-related, but one is general tied-weight serialization and the other is incorrect visual encoder keys for Qwen3.5; not the same concrete bug." }, { - "left": "issue:44466", - "right": "issue:45357", + "left": "issue:36331", + "right": "issue:42915", "accept": false, - "reason": "Both concern serialization, but one is tied lm_head weight serialization and the other is visual encoder keys in Qwen3.5 save_pretrained." + "reason": "Unexpected compute_loss keyword argument and FineGrainedFP8Config failure are unrelated trainer/model-loading issues." }, { "left": "issue:43526", "right": "issue:43883", "accept": false, - "reason": "reduce_labels output shape and missing all_tied_weights_keys are different model utility issues." + "reason": "BeitImageProcessorFast label reduction and Molmo tied-weights attribute errors do not share a code path." }, { - "left": "issue:43450", + "left": "issue:40990", "right": "issue:45701", "accept": false, - "reason": "Video processor batch-shape bug and tokenization changes are unrelated." - }, - { - "left": "issue:44877", - "right": "issue:44991", - "accept": false, - "reason": "Strict config loading and tokenizer loading regression are different failure modes." + "reason": "High perplexity on a benchmark and tokenizer behavior changing across versions are not the same bug." }, { - "left": "issue:43065", - "right": "issue:46402", + "left": "issue:43450", + "right": "issue:45701", "accept": false, - "reason": "Both are Sam3-related, but one is a dummy Conv2d concern and the other is missing text encoder weight loading." + "reason": "Video processor batched-shape bug and tokenization changes are different subsystems." }, { "left": "issue:43881", "right": "issue:44991", "accept": false, - "reason": "A model loading failure and a tokenizer loading regression are not the same bug." + "reason": "Both are loading/tokenizer failures, but they concern different models and different failure modes." }, { - "left": "issue:40990", - "right": "issue:45701", + "left": "issue:44877", + "right": "issue:44991", "accept": false, - "reason": "Different symptoms and code paths: perplexity regression vs tokenization changes." + "reason": "Strict config blocking granite_speech and tokenizer loading failure for est-roberta are unrelated." }, { - "left": "issue:41720", - "right": "issue:42617", + "left": "issue:43065", + "right": "issue:46402", "accept": false, - "reason": "Auto device mapping CUDA assert and a parallel script failure are unrelated." + "reason": "Dummy Conv2d in Sam3PixelDecoder and Sam3Model text encoder weights not loading are different model components and bugs." }, { "left": "issue:26413", "right": "issue:28530", "accept": false, - "reason": "Checkpoint resume epoch_iterator failure and early stopping metric lookup are unrelated trainer issues." + "reason": "Resume-from-checkpoint epoch iterator failure and early-stopping metric handling are separate trainer issues." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Device-map assert on a model and failure to run 3d_parallel.py do not appear related." }, { "left": "issue:44186", "right": "issue:45072", "accept": false, - "reason": "Tokenizer crash on NER/batched padding and dtype mismatches in inference are different bugs." + "reason": "LayoutLMv2Tokenizer batching/padding crashes and bf16 dtype mismatches in inference are unrelated." } ] }, @@ -4724,7 +4834,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 15, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -4783,237 +4893,237 @@ ], "soft_pairs": [ "issue:43873|issue:44756", - "issue:43550|issue:43720", "issue:46612|issue:46858", + "issue:43550|issue:43720", "issue:43577|issue:43708", "issue:43295|issue:45701", - "issue:45685|issue:46821", "issue:43550|issue:43606", - "issue:43646|issue:46665", "issue:44186|issue:44610", + "issue:45685|issue:46821", + "issue:43646|issue:46665", "issue:43404|issue:45357", - "issue:42915|issue:43317", "issue:44610|issue:44898", "issue:36010|issue:42617", "issue:44734|issue:46858", + "issue:42915|issue:43317", "issue:44423|issue:45406", - "issue:46821|issue:46869", "issue:44008|issue:45823", + "issue:46821|issue:46869", "issue:45632|issue:45684", - "issue:44464|issue:46858", "issue:41720|issue:45161", + "issue:44464|issue:46858", "issue:28530|issue:33666", - "issue:43531|issue:43572", "issue:44671|issue:44743", - "issue:42915|issue:46291", + "issue:43531|issue:43572", "issue:36331|issue:41720", + "issue:42915|issue:46291", "issue:44186|issue:44361", "issue:43316|issue:44913", - "issue:43425|issue:46821", - "issue:42915|issue:43322" + "issue:42915|issue:43322", + "issue:43425|issue:46821" ], - "prepared_review_unit_hash": "ac54a9700144e2390f411743057bbac3cccedcc5747190ab01518ef825202cbe", + "prepared_review_unit_hash": "287755d9279118edf189fd7a9bda4fa03ef5db46b6a5eab67e84f6b27ac8fd27", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36449, - "estimated_input_tokens": 9113, - "estimated_eval_tokens": 18482 + "serialized_chars": 36448, + "estimated_input_tokens": 9112, + "estimated_eval_tokens": 18480 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a heterogeneous mix of unrelated Transformers issues (offloading/quantization, generation cache bugs, multimodal processor regressions, config persistence, trainer/API changes). I don\u2019t see any soft pair that looks like the same underlying bug or a PR-worthy consolidation.", - "confidence": 0.94, - "canonical_issue_reason": "issue:43873 is the most central representative by discussion and inbound references, and it sits in the broad offloading/quantization/device-map theme that overlaps most with the noisy cluster.", + "summary": "These issues span many unrelated bugs across offloading, generation/cache handling, tokenization, config serialization, and multimodal/model-loading paths. None of the suggested soft pairs look like safe duplicates of the same underlying bug or change.", + "confidence": 0.86, + "canonical_issue_reason": "issue:43873 is the broadest open issue in the set and has the strongest cross-references, making it the best umbrella candidate for the offloading/quantization-related cluster even though the other items are not true duplicates.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the strongest issue candidate as a representative anchor for the cluster, though only loosely so; it has the highest apparent impact/activity and broadest topic overlap.", + "best_issue_reason": "issue:43873 is the best single issue to anchor this set because it is open, broadly framed, and most likely to attract related reports about offloading behavior.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43873", "right": "issue:44756", "accept": false, - "reason": "Both are offload/memory-related, but one is quantization offloading and the other is Strix Halo mmap OOM; different root causes." + "reason": "Both mention memory/offloading, but one is about quantization offload behavior and the other is a Strix Halo mmap/OOM workaround; different root causes and fixes." }, { - "left": "issue:43550", - "right": "issue:43720", + "left": "issue:46612", + "right": "issue:46858", "accept": false, - "reason": "One is torch.compile+SDPA on Bamba, the other is packed-weight unpacking during accelerate loading; not the same bug." + "reason": "Both involve cache logic, but one is a beam-search cache reorder bug across specific models while the other is a static-cache prefill arithmetic bug on non-CUDA backends under torch.compile." }, { - "left": "issue:46612", - "right": "issue:46858", + "left": "issue:43550", + "right": "issue:43720", "accept": false, - "reason": "Both involve cache/generation, but one is beam-search cache reorder omission and the other is static-cache slice arithmetic on non-CUDA backends." + "reason": "Different failures: one is torch.compile with SDPA on Bamba-9B-v2, the other is accelerate loading of packed BitNet weights." }, { "left": "issue:43577", "right": "issue:43708", "accept": false, - "reason": "Model dtype propagation in BLIP2 vs trainer max_steps on checkpoint resume are unrelated." + "reason": "Unrelated problems: dtype propagation for Blip2 versus Trainer checkpoint-resume max_steps calculation." }, { "left": "issue:43295", "right": "issue:45701", "accept": false, - "reason": "Both mention version/regression behavior around tokenization, but one is processor/tokenizer API breakage and the other is tokenization output changes." - }, - { - "left": "issue:45685", - "right": "issue:46821", - "accept": false, - "reason": "MPS histogram kernel error in MoE vs TimesFM window_size AttributeError are unrelated." + "reason": "Both touch tokenization behavior, but one is a regression in processor/tokenizer access for custom multimodal code and the other is a broader version-dependent tokenization change; not the same concrete bug." }, { "left": "issue:43550", "right": "issue:43606", "accept": false, - "reason": "Different failure modes: SDPA/torch.compile compile path vs CPU offload device mismatch." + "reason": "Both are model-specific runtime failures, but the affected models and failure mechanisms differ (torch.compile/SDPA vs CPU-offload device mismatch)." }, { - "left": "issue:43646", - "right": "issue:46665", + "left": "issue:44186", + "right": "issue:44610", "accept": false, - "reason": "Both concern custom model loading, but one is Transformers 5.0 init breakage and the other is inconsistent from_pretrained behavior in Python classes." + "reason": "Different components and bugs: LayoutLMv2 tokenizer crashes on NER/batched padding versus OmDet-Turbo processor output-size mismatch." }, { - "left": "issue:44186", - "right": "issue:44610", + "left": "issue:45685", + "right": "issue:46821", "accept": false, - "reason": "Tokenizer NER/padding crash vs processor image size mismatch; different components and symptoms." + "reason": "No shared underlying bug: an MPS histogram kernel limitation for MoE versus a TimesFM window_size AttributeError." }, { - "left": "issue:43404", - "right": "issue:45357", + "left": "issue:43646", + "right": "issue:46665", "accept": false, - "reason": "Weight tying in Mistral3 conditional generation vs save_pretrained visual-key regression in Qwen3.5 are different problems." + "reason": "Both concern model loading, but one is Transformers 5.0 custom-model initialization breakage and the other is inconsistent from_pretrained behavior inside Python classes; not the same code path." }, { - "left": "issue:42915", - "right": "issue:43317", + "left": "issue:43404", + "right": "issue:45357", "accept": false, - "reason": "Qwen3Moe FP8 failure and dequantized device_map offload failure are separate loading/quantization issues." + "reason": "Different serialization/model-head issues: untied lm_head in Mistral3ForConditionalGeneration versus incorrect visual encoder keys saved by Qwen3.5." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "Processor output size mismatch vs Perceiver interpolation failure; not the same code path." + "reason": "Both are vision-related, but one is a processor resize/output mismatch and the other is a non-default-resolution interpolation failure in Perceiver classification." }, { "left": "issue:36010", "right": "issue:42617", "accept": false, - "reason": "GenerationMixin import regression and inability to run 3d_parallel.py are unrelated." + "reason": "Completely different failures: an import regression for GenerationMixin versus inability to run a 3D parallel script." }, { "left": "issue:44734", "right": "issue:46858", "accept": false, - "reason": "Both are cache-related, but one is serve KV-cache continuation indexing and the other is a device-tensor length bug in static cache prefill." + "reason": "Both involve cache handling, but the first is an API serving KV-cache indexing bug while the second is a static-cache prefill tensor/device arithmetic bug under compilation." }, { - "left": "issue:44423", - "right": "issue:45406", + "left": "issue:42915", + "right": "issue:43317", "accept": false, - "reason": "Multimodal continuous-batching crash from string/.to handling vs Gemma4Processor missing _tokenizer; different bugs." + "reason": "Both are model-loading/runtime failures around quantization/offload, but one is FineGrainedFP8 on Qwen3Moe and the other is dequantized model loading with device_map auto; different concrete problems." }, { - "left": "issue:46821", - "right": "issue:46869", + "left": "issue:44423", + "right": "issue:45406", "accept": false, - "reason": "TimesFM window_size AttributeError and YOLO multi-GPU device_map failure are unrelated." + "reason": "Both are serve/multimodal processor crashes, but one is a string-to-device conversion failure in continuous batching and the other is a missing _tokenizer attribute on Gemma4Processor." }, { "left": "issue:44008", "right": "issue:45823", "accept": false, - "reason": "Gemma3n forward variable collision vs Gemma4 PLE device mismatch are different model bugs." + "reason": "Different Gemma-family bugs: a variable-name collision in forward versus a device mismatch during forward with device_map=auto." }, { - "left": "issue:45632", - "right": "issue:45684", + "left": "issue:46821", + "right": "issue:46869", "accept": false, - "reason": "Cache-path collision for trust_remote_code vs permission propagation in save_pretrained are unrelated file-handling issues." + "reason": "Different models and symptoms: a TimesFM window_size AttributeError versus YOLO multi-GPU device_map auto failure." }, { - "left": "issue:44464", - "right": "issue:46858", + "left": "issue:45632", + "right": "issue:45684", "accept": false, - "reason": "Both involve compiled generation/cache, but one is inconsistent chunked generation outputs and the other is a static-cache length arithmetic bug." + "reason": "Unrelated filesystem/permissions issue versus save_pretrained read-only permission propagation." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "Auto device mapping cudaErrorAssert on Qwen3 vs tensor-parallel-only failure on GPT-OSS MoE; different models and failure modes." + "reason": "Both mention distributed/device mapping style concerns, but one is a Qwen3 auto-device-map CUDA assert and the other is GPT-OSS MoE tensor-parallel support; not the same bug." }, { - "left": "issue:28530", - "right": "issue:33666", + "left": "issue:44464", + "right": "issue:46858", "accept": false, - "reason": "Early stopping metric resolution and Qwen2-VL multi-GPU training are unrelated." + "reason": "Both mention compiled generation/cache behavior, but one reports inconsistent chunked generation outputs while the other is a specific cache length tensor arithmetic bug." }, { - "left": "issue:43531", - "right": "issue:43572", + "left": "issue:28530", + "right": "issue:33666", "accept": false, - "reason": "Qwen3-MoE sliding window bug vs StableLmConfig missing pad_token_idx are separate config/model issues." + "reason": "No overlap in underlying bug: early stopping metric selection versus Qwen2-VL multi-GPU training." }, { "left": "issue:44671", "right": "issue:44743", "accept": false, - "reason": "CamemBERT masked-LM regression and Qwen3.5 recurrent-state reset under cache are unrelated." + "reason": "Both are model-specific correctness bugs, but one is masked-LM prediction regression in CamemBERT and the other is Qwen3.5 recurrent state reset with cache." }, { - "left": "issue:42915", - "right": "issue:46291", + "left": "issue:43531", + "right": "issue:43572", "accept": false, - "reason": "FP8 model loading failure vs kernels package import/version requirement mismatch are unrelated." + "reason": "Different config/model issues: sliding_window behavior in Qwen3-MoE versus missing pad_token_idx in StableLmConfig." }, { "left": "issue:36331", "right": "issue:41720", "accept": false, - "reason": "Trainer compute_loss signature breakage and Qwen3 auto device-map CUDA assert are different problems." + "reason": "Trainer custom compute_loss signature change versus Qwen3 device_map CUDA assert are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:46291", + "accept": false, + "reason": "Different subsystems: Qwen3Moe FP8 loading failure versus import-time pipeline failure due to kernels package requirements." }, { "left": "issue:44186", "right": "issue:44361", "accept": false, - "reason": "Different tokenizers and different crash conditions; too broad to be a duplicate." + "reason": "Both tokenizer-related, but on different tokenizers and different failure modes; not a single duplicate bug." }, { "left": "issue:43316", "right": "issue:44913", "accept": false, - "reason": "Gemma3TextConfig API inconsistency vs GPTNeoX rotary_pct reload bug are different config issues." + "reason": "Both are config consistency problems, but one is a Gemma3TextConfig API discrepancy and the other is GPTNeoX rotary_pct not persisting on reload." }, { - "left": "issue:43425", - "right": "issue:46821", + "left": "issue:42915", + "right": "issue:43322", "accept": false, - "reason": "Torch 2.10 compatibility issue vs TimesFM window_size AttributeError are unrelated." + "reason": "Different model families and runtime errors: Qwen3Moe FP8 failure versus Llava Next segmentation fault on load." }, { - "left": "issue:42915", - "right": "issue:43322", + "left": "issue:43425", + "right": "issue:46821", "accept": false, - "reason": "Qwen3Moe FP8 config failure and Llava Next segmentation fault are different model-specific bugs." + "reason": "Torch version compatibility and a TimesFM attribute error are not the same underlying bug." } ] }, @@ -5029,7 +5139,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 16, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -5091,241 +5201,241 @@ "issue:43526|issue:43881", "issue:33357|issue:35707", "issue:29159|issue:36032", - "issue:43540|issue:43782", - "issue:43873|issue:45704", "issue:44387|issue:45701", - "issue:43550|issue:43653", + "issue:43873|issue:45704", + "issue:43540|issue:43782", "issue:30064|issue:34634", + "issue:43550|issue:43653", "issue:43606|issue:43720", - "issue:44492|issue:44509", "issue:44534|issue:46665", + "issue:44492|issue:44509", "issue:43531|issue:45759", "issue:41720|issue:43746", + "issue:45406|issue:45464", "issue:42490|issue:43329", "issue:43388|issue:45059", - "issue:45406|issue:45464", "issue:43606|issue:43653", "issue:34567|issue:35707", - "issue:43531|issue:46821", "issue:34634|issue:41720", + "issue:43531|issue:46821", "issue:43540|issue:46858", - "issue:44451|issue:44991", "issue:45759|issue:45859", "issue:45357|issue:46402", + "issue:44451|issue:44991", "issue:43531|issue:44589", "issue:44077|issue:44625", "issue:43644|issue:46665", "issue:44062|issue:44451", "issue:44898|issue:45072" ], - "prepared_review_unit_hash": "e773804cd27075e851b6691d1c07de970df206efd7dafb7642fb7bdf37cc61ec", + "prepared_review_unit_hash": "2d9b333363498906fb14e087f90c844685041e5730a2b5aebc217e995d3bb48c", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36791, - "estimated_input_tokens": 9198, - "estimated_eval_tokens": 18652 + "serialized_chars": 36783, + "estimated_input_tokens": 9196, + "estimated_eval_tokens": 18648 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pair looks like the same underlying bug or change. Most are only loosely related by subsystem or broad regression themes, but the concrete failures differ too much to merge as duplicates.", - "confidence": 0.97, - "canonical_issue_reason": "Issue 44534 is the broadest representative of the repeated v5 buffer-regression theme, but the overall set is too heterogeneous to treat it as the canonical duplicate target.", - "canonical_pr_reason": null, - "best_issue_reason": "Issue 44534 is the best general anchor among the issues because it captures the most broadly framed regression theme in this set, though it should not subsume the others.", - "best_pr_reason": null, + "summary": "The proposed soft edges are all false positives. The items span unrelated bugs across tokenizers, multimodal processors, loading/offload, training state, docs, and model-specific regressions, with no clear duplicate cluster.", + "confidence": 0.95, + "canonical_issue_reason": "No canonical issue: the set is too heterogeneous and the soft pairs do not describe one shared underlying bug.", + "canonical_pr_reason": "No PRs are present in this cluster.", + "best_issue_reason": "No best issue stands out; none of the issues is representative of the whole set.", + "best_pr_reason": "No PRs are present in this cluster.", "soft_edge_verdicts": [ { "left": "issue:28530", "right": "issue:30064", "accept": false, - "reason": "Early stopping metric handling vs void segmentation map processing; unrelated bugs." + "reason": "Early-stopping metric lookup and void segmentation-map processing are unrelated bugs." }, { "left": "issue:43526", "right": "issue:43881", "accept": false, - "reason": "BEiT label reduction bug vs GLM loading failure; different models and code paths." + "reason": "BeitImageProcessorFast label reduction and glm-4v loading failure affect different code paths." }, { "left": "issue:33357", "right": "issue:35707", "accept": false, - "reason": "MacOS CLIP bus error vs generation with inputs_embeds/past_key_values; not the same issue." + "reason": "MacOS CLIP bus error and progressive generation with inputs_embeds/past_key_values are unrelated." }, { "left": "issue:29159", "right": "issue:36032", "accept": false, - "reason": "Tokenizer inconsistency report vs T5Tokenizer load-time special-token conflict; distinct failures." + "reason": "Slow/fast tokenizer inconsistency and T5Tokenizer add_special_tokens conflict are different tokenizer issues." }, { - "left": "issue:43540", - "right": "issue:43782", + "left": "issue:44387", + "right": "issue:45701", "accept": false, - "reason": "Qwen3OmniMoe video input ValueError vs Qwen3VL weight_only load error; different problems." + "reason": "Int4 quantization OOM and tokenization changes are not the same problem." }, { "left": "issue:43873", "right": "issue:45704", "accept": false, - "reason": "Quantization/offloading behavior vs T5 apex RMSNorm memory leak; not mergeable as one bug." + "reason": "Quantization/offloading behavior and T5 apex RMSNorm memory leak are unrelated." }, { - "left": "issue:44387", - "right": "issue:45701", + "left": "issue:43540", + "right": "issue:43782", "accept": false, - "reason": "int4 OOM regression vs tokenization changes; same release area, different symptom and cause." + "reason": "Qwen3OmniMoe video processing error and Qwen3VL weight_only load error are separate failures." }, { - "left": "issue:43550", - "right": "issue:43653", + "left": "issue:30064", + "right": "issue:34634", "accept": false, - "reason": "Bamba torch.compile SDPA issue vs BigBirdTokenizer special-token decode bug; unrelated." + "reason": "Void segmentation maps and BarkProcessor voice_preset are different processor bugs." }, { - "left": "issue:30064", - "right": "issue:34634", + "left": "issue:43550", + "right": "issue:43653", "accept": false, - "reason": "Void segmentation maps vs BarkProcessor voice_preset handling; different components." + "reason": "Bamba torch.compile SDPA failure and BigBirdTokenizer special-token registration are unrelated." }, { "left": "issue:43606", "right": "issue:43720", "accept": false, - "reason": "CPU offload device mismatch vs BitNet packed-weight unpacking on load; distinct code paths." + "reason": "CPU offload device mismatch and BitNet packed-weight unpacking are different loading/offload issues." }, { - "left": "issue:44492", - "right": "issue:44509", + "left": "issue:44534", + "right": "issue:46665", "accept": false, - "reason": "Cache strategy typo vs docs mentioning removed pipeline tasks; not the same change." + "reason": "Non-persistent buffer junk and from_pretrained behavior inside Python classes are not the same bug." }, { - "left": "issue:44534", - "right": "issue:46665", + "left": "issue:44492", + "right": "issue:44509", "accept": false, - "reason": "Non-persistent buffer initialization regression vs from_pretrained behavior inside Python classes; unrelated." + "reason": "A typo in cache strategies and removed pipeline-task docs are unrelated documentation issues." }, { "left": "issue:43531", "right": "issue:45759", "accept": false, - "reason": "Qwen3-MoE sliding_window bug vs composite config unwrapping for AutoModelForCausalLM; different fixes." + "reason": "Qwen3-MoE sliding_window and multimodal text_config unwrapping are distinct configuration bugs." }, { "left": "issue:41720", "right": "issue:43746", "accept": false, - "reason": "Qwen3 auto device mapping cuda assert vs GraniteSpeech PEFT local checkpoint loading; separate issues." + "reason": "Qwen3 auto device mapping cudaErrorAssert and GraniteSpeech PEFT local checkpoint loading are unrelated." }, { - "left": "issue:42490", - "right": "issue:43329", + "left": "issue:45406", + "right": "issue:45464", "accept": false, - "reason": "Trailing-slash local path warning vs undefined video-token helper variables; different failure modes." + "reason": "Gemma4Processor missing _tokenizer in transformers serve and Qwen3.5 streaming chat/completions failure are different code paths." }, { - "left": "issue:43388", - "right": "issue:45059", + "left": "issue:42490", + "right": "issue:43329", "accept": false, - "reason": "gather_for_metrics label truncation vs SAM3 text/bbox behavior; unrelated functionality." + "reason": "Trailing-slash local model import warning and VLM video token counting are unrelated." }, { - "left": "issue:45406", - "right": "issue:45464", + "left": "issue:43388", + "right": "issue:45059", "accept": false, - "reason": "Gemma4Processor missing _tokenizer vs Qwen3.5 streaming API failure; distinct code paths." + "reason": "gather_for_metrics label truncation and SAM3 text+box behavior are different bugs." }, { "left": "issue:43606", "right": "issue:43653", "accept": false, - "reason": "CPU offload device mismatch vs tokenizer special-token registration; no shared underlying bug." + "reason": "CPU offload device mismatch and tokenizer special-token decoding are not the same issue." }, { "left": "issue:34567", "right": "issue:35707", "accept": false, - "reason": "TrainerState token counter not updating vs progressive generation with caches; unrelated." - }, - { - "left": "issue:43531", - "right": "issue:46821", - "accept": false, - "reason": "Qwen3-MoE sliding_window bug vs TimesFM window_size AttributeError; different models and causes." + "reason": "TrainerState token counter not updating and progressive generation are unrelated." }, { "left": "issue:34634", "right": "issue:41720", "accept": false, - "reason": "Bark voice_preset issue vs Qwen3 device-mapping assert; no duplicate relation." + "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping affect different models and code paths." }, { - "left": "issue:43540", - "right": "issue:46858", + "left": "issue:43531", + "right": "issue:46821", "accept": false, - "reason": "Qwen3OmniMoe video input error vs cache seq_length device-tensor arithmetic bug; different code paths." + "reason": "Qwen3-MoE sliding_window and TimesFM window_size AttributeError are only superficially similar." }, { - "left": "issue:44451", - "right": "issue:44991", + "left": "issue:43540", + "right": "issue:46858", "accept": false, - "reason": "Specific model loading failure vs tokenizer loading regression for a different model; not the same bug." + "reason": "Qwen3OmniMoe video input failure and cache seq-length device tensor arithmetic are unrelated." }, { "left": "issue:45759", "right": "issue:45859", "accept": false, - "reason": "Composite Qwen config unwrapping vs CPU-offload rotary embedding incompatibility; unrelated." + "reason": "Composite Qwen config unwrapping and CPU-offload rotary embedding behavior are different bugs." }, { "left": "issue:45357", "right": "issue:46402", "accept": false, - "reason": "Incorrect save_pretrained visual encoder keys vs text encoder weights not loaded; distinct load/save issues." + "reason": "Incorrect visual encoder keys on save and Sam3 text encoder weights not loaded are separate serialization/loading issues." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "ScandiBERT model loading and est-roberta tokenizer loading are not the same bug." }, { "left": "issue:43531", "right": "issue:44589", "accept": false, - "reason": "Sliding-window support bug vs missing Float8 storage object during loading; different failures." + "reason": "Qwen3-MoE sliding_window and missing Float8 storage are unrelated." }, { "left": "issue:44077", "right": "issue:44625", "accept": false, - "reason": "patchtsmixer post_init policy vs Qwen3.5 num_labels propagation; unrelated." + "reason": "patchtsmixer post_init policy and Qwen3.5 num_labels propagation are separate model/config issues." }, { "left": "issue:43644", "right": "issue:46665", "accept": false, - "reason": "Non-persistent buffer junk in v5 vs inconsistent from_pretrained inside classes; not the same bug." + "reason": "Non-persistent buffers filled with junk and from_pretrained inconsistency inside Python classes are not the same underlying bug." }, { "left": "issue:44062", "right": "issue:44451", "accept": false, - "reason": "AddedToken special-arg conflict vs model load failure for ScandiBERT; different tokenizer/model problems." + "reason": "AddedToken special-argument conflict and ScandiBERT loading failure are unrelated tokenizer/model issues." }, { "left": "issue:44898", "right": "issue:45072", "accept": false, - "reason": "Perceiver resolution handling vs bfloat16 dtype mismatch in inference; separate issues." + "reason": "Perceiver interpolation failure and bfloat16 dtype mismatches are different model/runtime problems." } ] }, @@ -5341,7 +5451,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 17, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -5404,253 +5514,247 @@ "issue:43381|issue:43844", "issue:44368|issue:45357", "issue:39692|issue:44610", - "issue:46655|issue:46693", "issue:41720|issue:45704", - "issue:43421|issue:43927", "issue:43746|issue:46050", + "issue:43421|issue:43927", + "issue:46655|issue:46693", "issue:43381|issue:45704", - "issue:42886|issue:44843", "issue:45704|issue:45923", + "issue:42886|issue:44843", "issue:44918|issue:44936", - "issue:44514|issue:44545", "issue:45759|issue:45923", - "issue:43421|issue:43653", "issue:43746|issue:43873", + "issue:44514|issue:44545", "issue:43994|issue:44671", + "issue:43421|issue:43653", "issue:39692|issue:43493", "issue:46032|issue:46463", + "issue:43421|issue:44186", "issue:43782|issue:45381", + "issue:45081|issue:45356", "issue:30064|issue:36010", - "issue:43421|issue:44186", "issue:44995|issue:46858", - "issue:45081|issue:45356", "issue:45356|issue:45907", - "issue:43720|issue:43819", "issue:39692|issue:43873", + "issue:43720|issue:43819", "issue:43653|issue:43720", "issue:45987|issue:46829", "issue:46829|issue:46832" ], - "prepared_review_unit_hash": "8d5888b57b98e40ea1fcdc0c73488c5f30fe5ddc92c29f9cd93027c40bd0f1d6", + "prepared_review_unit_hash": "d5974906f2726f0612a983d5b847c69aabd01d6a4fbd5030e1061291ad5fc09f", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 37541, - "estimated_input_tokens": 9386, - "estimated_eval_tokens": 19028 + "serialized_chars": 37548, + "estimated_input_tokens": 9387, + "estimated_eval_tokens": 19030 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a grab bag of unrelated HF bug reports; only the Qwen2_5_VLProcessor batched-padding pair (44514/44545) is a true duplicate.", - "confidence": 0.96, - "canonical_issue_reason": "Pick 44545 as the representative issue because it is the cleanest title for the only true duplicate pair with 44514; the rest of the cluster is unrelated.", + "summary": "This set is mostly heterogeneous and does not form one duplicate cluster. The only clear duplicate pair is the Qwen2.5-VL batched `apply_chat_template` crash; the rest are separate issues that only share broad subsystem similarity.", + "confidence": 0.86, + "canonical_issue_reason": "Issue 44545 is the cleanest canonical for the only true duplicate pair (44514/44545) and has slightly more inbound/reference activity.", "canonical_pr_reason": null, - "best_issue_reason": "44545 is the best global issue representative for the same reason: it cleanly captures the only clear duplicate bug in the set.", + "best_issue_reason": "44545 is the best representative issue in this set, but only for the one exact duplicate pair; it is not a good global representative of the full mixed cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44987", "right": "issue:45042", "accept": false, - "reason": "Different bugs: local model loading failure versus a PIL/torchvision dependency regression." + "reason": "Different bugs: one is a model-loading failure for `physical-intelligence/fast`, the other is an image processor dependency regression requiring `torchvision`." }, { "left": "issue:44625", "right": "issue:45479", "accept": false, - "reason": "Unrelated config-propagation bug versus a single-label classification loss edge case." + "reason": "Both are classification-related, but one is `num_labels` propagation in Qwen3.5 configs and the other is a generic single-label zero-loss bug across models." }, { "left": "issue:43381", "right": "issue:43844", "accept": false, - "reason": "Eval-mode checkpointing restriction is not the same as ZeRO-3 gradient inflation." + "reason": "Both involve training behavior, but eval-mode gradient checkpointing and ZeRO-3 gradient growth are different code paths and failure modes." }, { "left": "issue:44368", "right": "issue:45357", "accept": false, - "reason": "LoRA tie_word_embeddings warning is unrelated to incorrect visual-encoder keys on save." + "reason": "Different Qwen issues: one is a LoRA warning about `tie_word_embeddings`, the other is incorrect visual encoder keys on `save_pretrained`." }, { "left": "issue:39692", "right": "issue:44610", "accept": false, - "reason": "SigLIP2 doc/model mismatch and OmDet-Turbo input-size mismatch are different bugs." + "reason": "Different models and symptoms: SigLIP2 doc example mistakes vs OmDet-Turbo processor output size mismatch." }, { - "left": "issue:46655", - "right": "issue:46693", + "left": "issue:41720", + "right": "issue:45704", "accept": false, - "reason": "Video padding bug and FA2 long-context performance regression are different code paths." + "reason": "Unrelated failures: Qwen3 auto device mapping CUDA assert vs a T5 Apex RMSNorm memory leak." }, { - "left": "issue:41720", - "right": "issue:45704", + "left": "issue:43746", + "right": "issue:46050", "accept": false, - "reason": "Qwen3 auto device-map CUDA assert is unrelated to the T5 apex RMSNorm leak." + "reason": "Different loading/saving paths and different model families; PEFT local checkpoint loading is not the same as sinq quantized save/load breakage." }, { "left": "issue:43421", "right": "issue:43927", "accept": false, - "reason": "Runtime post-processor refresh and custom token-ID persistence are different tokenizer bugs." + "reason": "Both are token-related, but one is runtime special-token post-processing and the other is config persistence of custom token IDs." }, { - "left": "issue:43746", - "right": "issue:46050", + "left": "issue:46655", + "right": "issue:46693", "accept": false, - "reason": "Local PEFT checkpoint loading and quantized sinq save/load breakage are different failures." + "reason": "Different performance/processing bugs: video processor padding calculation vs FA2 long-context generation regression." }, { "left": "issue:43381", "right": "issue:45704", "accept": false, - "reason": "Checkpointing-in-eval and T5 memory leak are unrelated issues." + "reason": "No shared bug: gradient checkpointing eval-mode failure is unrelated to the T5 fused RMSNorm leak." }, { - "left": "issue:42886", - "right": "issue:44843", + "left": "issue:45704", + "right": "issue:45923", "accept": false, - "reason": "Offline cache loading and mistral-regex patching are related in area but not the same failure." + "reason": "Different concerns entirely: Apex memory leak vs a missing `supports_gradient_checkpointing` flag on a remote-code model." }, { - "left": "issue:45704", - "right": "issue:45923", + "left": "issue:42886", + "right": "issue:44843", "accept": false, - "reason": "T5 apex memory leak and missing gradient-checkpointing metadata are unrelated." + "reason": "Both mention offline tokenizer behavior, but one is generic cache loading and the other is a specific `_patch_mistral_regex` unconditional `model_info()` call; not enough to treat as the same bug." }, { "left": "issue:44918", "right": "issue:44936", "accept": false, - "reason": "Embedding-unpacking failure in SFT training is different from train-then-evaluate failure." - }, - { - "left": "issue:44514", - "right": "issue:44545", - "accept": true, - "reason": "Same batched apply_chat_template padding=False crash; the titles describe the same bug." + "reason": "Distinct trainer issues: Qwen3.5 embedding unpacking with TRL SFT vs `trainer.evaluate()` failing after `trainer.train()`." }, { "left": "issue:45759", "right": "issue:45923", "accept": false, - "reason": "Composite Qwen text_config unwrapping is unrelated to a missing gradient-checkpointing flag." - }, - { - "left": "issue:43421", - "right": "issue:43653", - "accept": false, - "reason": "Special-token post-processor updates and BigBird mask-token registration are different bugs." + "reason": "Different fixes: unwrapping `text_config` for composite Qwen multimodal configs is unrelated to a checkpointing flag on Nemotron." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "PEFT checkpoint loading and quantization offloading are different code paths." + "reason": "Both touch loading/offload, but PEFT checkpoint loading and quantization offloading behavior are different bugs." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Same concrete bug: identical Qwen2.5-VL batched `apply_chat_template` crash when `padding=False`." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "SigLIP2 pipeline misbehavior and CamemBERT masked-LM regression are unrelated." + "reason": "Different model-specific correctness bugs: SigLIP2 bad outputs vs CamemBERT masked LM predictions in v5." }, { - "left": "issue:39692", - "right": "issue:43493", + "left": "issue:43421", + "right": "issue:43653", "accept": false, - "reason": "SigLIP2 doc mismatch and SigLIP2 implementation discrepancy are related by model but not the same concrete bug." + "reason": "Both involve special tokens, but one is a tokenizer backend update bug and the other is BigBirdTokenizer mask-token registration/decoding behavior." }, { "left": "issue:46032", "right": "issue:46463", "accept": false, - "reason": "Quantized use_cache correctness and decoder error-message wording are unrelated." + "reason": "Unrelated: Mamba2 cache/use_cache correctness is not the same as a T5 decoder-input error message issue." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Different tokenizer problems with different triggers; runtime special-token updates are not the same as LayoutLMv2 NER padding/truncation crashes." }, { "left": "issue:43782", "right": "issue:45381", "accept": false, - "reason": "Qwen3VL weight_only loading and Qwen2.5-VL video position ids are different multimodal issues." + "reason": "Different Qwen/VL bugs: weight_only loading failure vs wrong vision_position_ids for video input." }, { - "left": "issue:30064", - "right": "issue:36010", + "left": "issue:45081", + "right": "issue:45356", "accept": false, - "reason": "Void segmentation-map processing and GenerationMixin import failure are unrelated." + "reason": "Both mention `fix_mistral_regex`, but one is a crash in `_patch_mistral_regex` and the other is a Kimi-K2.5 tokenizer codec regression with a misleading warning; too different to merge." }, { - "left": "issue:43421", - "right": "issue:44186", + "left": "issue:30064", + "right": "issue:36010", "accept": false, - "reason": "Tokenizer special-token updates and LayoutLMv2 NER/padding crashes are different bugs." + "reason": "Completely unrelated: image processor void segmentation maps vs GenerationMixin import failure." }, { "left": "issue:44995", "right": "issue:46858", "accept": false, - "reason": "Stale second-forward cache indexing and non-CUDA seq-length tensor arithmetic are different cache bugs." - }, - { - "left": "issue:45081", - "right": "issue:45356", - "accept": false, - "reason": "Mistral regex AttributeError and Kimi tokenizer regression share a theme but not the same concrete failure." + "reason": "Both mention cache/indexing, but the stale indexer cache on second forward is not the same bug as `get_seq_length()` returning a device tensor in `_prefill`." }, { "left": "issue:45356", "right": "issue:45907", "accept": false, - "reason": "Tokenizer regression and list_repo_files revision forwarding are unrelated." + "reason": "Kimi tokenizer codec/fix_mistral_regex regression is unrelated to `list_repo_files` not forwarding `revision`." }, { - "left": "issue:43720", - "right": "issue:43819", + "left": "issue:39692", + "right": "issue:43873", "accept": false, - "reason": "Packed-weight loading and missing STE in DAC are different model internals." + "reason": "Different areas and symptoms: SigLIP2 documentation/model mismatch vs quantization offloading not working as expected." }, { - "left": "issue:39692", - "right": "issue:43873", + "left": "issue:43720", + "right": "issue:43819", "accept": false, - "reason": "SigLIP2 doc/model issue and quantization offloading bug are unrelated." + "reason": "No shared underlying change: packed-weight loading failure in BitNet vs DAC latent reconstruction mismatch from missing STE." }, { "left": "issue:43653", "right": "issue:43720", "accept": false, - "reason": "BigBird special-token registration and BitNet packed-weight unpacking are different bugs." + "reason": "Different models and bugs: BigBird special-token decoding vs BitNet accelerate-loading unpacking." }, { "left": "issue:45987", "right": "issue:46829", "accept": false, - "reason": "StaticCache seq_length shape bug and VideoPrism hidden_states field mix-up are unrelated outputs." + "reason": "Different cache/output bugs: static cache sequence length type mismatch vs VideoPrism output field wiring." }, { "left": "issue:46829", "right": "issue:46832", "accept": false, - "reason": "VideoPrism output-field bug and RT-DETR feature-level mismatch are different model issues." + "reason": "Different model heads and failure modes: VideoPrism hidden_states misassignment vs RT-DETR feature-level mismatch in forward." } ] }, "evaluator_result": { "accept": true, - "feedback": "The analyst stays conservative and the conclusions are grounded in the packet. The only accepted soft pair, 44514/44545, is a clear title-level duplicate, and the rest are plausibly distinct bugs despite some topical overlap. The representative issue choice and the summary are supported by the issue titles and pair list." + "feedback": "The analysis is grounded in the packet and stays conservative on the soft edges. The only clearly duplicate pair in the provided titles is 44514/44545, and the choice of 44545 as canonical is supported by the slightly higher inbound/reference activity. The remaining pairwise rejections are appropriately cautious given the limited evidence." }, "evaluator_used": true, "retried": false, @@ -5663,7 +5767,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 18, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -5686,10 +5790,8 @@ "issue:43611", "issue:43618", "issue:43643", - "issue:43644", "issue:43653", "issue:43697", - "issue:43749", "issue:43756", "issue:43819", "issue:43927", @@ -5710,8 +5812,9 @@ "issue:44991", "issue:45081", "issue:45084", + "issue:45183", "issue:45381", - "issue:45584", + "issue:45479", "issue:45834", "issue:45987", "issue:46294", @@ -5722,237 +5825,230 @@ ], "soft_pairs": [ "issue:43295|issue:45834", - "issue:43550|issue:44361", "issue:43550|issue:43927", + "issue:43550|issue:44361", "issue:44062|issue:44488", - "issue:43596|issue:44220", - "issue:45381|issue:46693", "issue:43540|issue:46294", + "issue:45381|issue:46693", "issue:43531|issue:43643", + "issue:43596|issue:44220", "issue:44008|issue:46858", "issue:43611|issue:46665", - "issue:44220|issue:44661", "issue:43653|issue:43819", "issue:44186|issue:44898", - "issue:45987|issue:46932", "issue:41720|issue:44991", + "issue:45987|issue:46932", "issue:30064|issue:36331", - "issue:43618|issue:43697", + "issue:44220|issue:44661", "issue:43643|issue:44877", + "issue:43618|issue:43697", "issue:43335|issue:43756", "issue:43550|issue:43819", "issue:44843|issue:45081", - "issue:44821|issue:45084", - "issue:44186|issue:44442", - "issue:39692|issue:45834", "issue:44448|issue:44671", + "issue:39692|issue:45834", + "issue:44186|issue:44442", + "issue:44821|issue:45084", "issue:43540|issue:44008", "issue:34634|issue:36010", - "issue:43644|issue:43749", - "issue:45584|issue:45834" + "issue:45183|issue:45479" ], - "prepared_review_unit_hash": "8aff6f8e57097e0a1f2158045dc9f8a41442122b42eea731e14ee86c5caca066", + "prepared_review_unit_hash": "5444e746e36ab4e8cd575036d9e382c930d584f3a43f77a3a996706a749ec46e", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36479, - "estimated_input_tokens": 9120, - "estimated_eval_tokens": 18496 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35781, + "estimated_input_tokens": 8946, + "estimated_eval_tokens": 18148 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is heterogeneous: tokenizers, multimodal processors, cache/compile behavior, config loading, and model-specific regressions. None of the soft pairs look like true duplicates under strict triage.", - "confidence": 0.93, - "canonical_issue_reason": "Issue 43644 is the broadest standalone regression report (v5 non-persistent buffer initialization) and the closest thing to a cluster anchor, but the other items are not duplicate variants of it.", + "summary": "No duplicate cluster here: the items span unrelated regressions across tokenizers, configs, vision/audio models, generation caches, and docs. All soft edges are rejected.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue fits this heterogeneous set; the titles indicate many distinct code paths and failure modes with little or no overlap.", "canonical_pr_reason": null, - "best_issue_reason": "43644 is a clear, reproducible Transformers v5 regression with broad potential impact, making it the best representative issue in this otherwise mixed cluster.", + "best_issue_reason": "If forced to choose one representative, issue 43611 is the broadest core regression around model loading, but it still does not meaningfully represent the rest of the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43295", "right": "issue:45834", "accept": false, - "reason": "Different bugs: processor.tokenizer regression vs long OCR index overflow; no shared code-path." + "reason": "Different subsystems and failures: processor/tokenizer regression vs static cache sequence-length shape bug." }, { "left": "issue:43550", - "right": "issue:44361", + "right": "issue:43927", "accept": false, - "reason": "Bamba torch.compile/SDPA failure is unrelated to MLukeTokenizer AttributeError." + "reason": "Unrelated models and code paths: torch.compile/SDPA issue vs config save/load token-ID regression." }, { "left": "issue:43550", - "right": "issue:43927", + "right": "issue:44361", "accept": false, - "reason": "Different models and failure modes: SDPA compile bug vs DiaConfig token ID persistence." + "reason": "Different model-specific runtime errors with no shared bug evidence." }, { "left": "issue:44062", "right": "issue:44488", "accept": false, - "reason": "Different tokenizer bugs: AddedToken keyword collision vs model/tokenizer load failure for sleng-bert." + "reason": "Token construction error vs tokenizer-loading regression for a specific model; not the same bug." }, { - "left": "issue:43596", - "right": "issue:44220", + "left": "issue:43540", + "right": "issue:46294", "accept": false, - "reason": "BertModel/Deepspeed zero3 init index error is unrelated to _torch_extract_fbank_features()." + "reason": "Both involve video inputs, but the model families and concrete failures differ too much to be one fix." }, { "left": "issue:45381", "right": "issue:46693", "accept": false, - "reason": "Qwen2.5-VL video position-id bug is unrelated to FA2 long-context performance regression." + "reason": "Vision position IDs bug vs performance regression in FA2 long-context generation; unrelated." }, { - "left": "issue:43540", - "right": "issue:46294", + "left": "issue:43531", + "right": "issue:43643", "accept": false, - "reason": "Both are video-input issues, but they hit different Qwen models and different code paths." + "reason": "MoE sliding-window bug vs AutoConfig trust_remote_code field-loss issue; different paths." }, { - "left": "issue:43531", - "right": "issue:43643", + "left": "issue:43596", + "right": "issue:44220", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior is unrelated to AutoConfig missing fields with trust_remote_code." + "reason": "Deepspeed/BertModel init error vs audio fbank extraction issue; no overlap." }, { "left": "issue:44008", "right": "issue:46858", "accept": false, - "reason": "Gemma3n variable-collision AttributeError is unrelated to StaticCache seq-length tensor semantics." + "reason": "Gemma3n variable collision vs cache.get_seq_length tensor-return bug; distinct code paths." }, { "left": "issue:43611", "right": "issue:46665", "accept": false, - "reason": "base_model_prefix load regression is unrelated to from_pretrained behavior inside Python classes." - }, - { - "left": "issue:44220", - "right": "issue:44661", - "accept": false, - "reason": "Audio fbank extraction bug is unrelated to add-new-model-like / TOKENIZER_MAPPING_NAMES handling." + "reason": "Model-loading regression in v5 vs from_pretrained behavior inside Python classes; not the same defect." }, { "left": "issue:43653", "right": "issue:43819", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug is unrelated to DAC.from_latents forward mismatch." + "reason": "BigBird tokenizer special-token registration bug vs DAC latents/STE mismatch; unrelated." }, { "left": "issue:44186", "right": "issue:44898", "accept": false, - "reason": "LayoutLMv2 tokenizer crash and Perceiver image classification failure are separate model/tokenizer paths." + "reason": "LayoutLMv2 tokenizer crashes vs Perceiver image classification failure; different models and causes." }, { - "left": "issue:45987", - "right": "issue:46932", + "left": "issue:41720", + "right": "issue:44991", "accept": false, - "reason": "StaticCache get_seq_length contract bug is unrelated to Gemma4 quantized embedder dtype casting." + "reason": "CUDA/device-mapping crash vs tokenizer loading failure for a different model; no shared bug." }, { - "left": "issue:41720", - "right": "issue:44991", + "left": "issue:45987", + "right": "issue:46932", "accept": false, - "reason": "Qwen3 auto device mapping CUDA assert is unrelated to est-roberta tokenizer loading." + "reason": "Static cache sequence-length contract bug vs Gemma4 quantized embedder dtype casting; unrelated." }, { "left": "issue:30064", "right": "issue:36331", "accept": false, - "reason": "Void segmentation map processing and CustomTrainer.compute_loss signature mismatch are unrelated." + "reason": "Image processor segmentation-map bug vs CustomTrainer compute_loss signature mismatch; different areas." }, { - "left": "issue:43618", - "right": "issue:43697", + "left": "issue:44220", + "right": "issue:44661", "accept": false, - "reason": "CLIPOutput attentions regression and RTDetrV2 output drift are different model-specific issues." + "reason": "Audio feature extraction bug vs tokenizer mapping-name model-addition bug; no common code path." }, { "left": "issue:43643", "right": "issue:44877", "accept": false, - "reason": "Missing AutoConfig fields with trust_remote_code and strict granite_speech config loading are separate config-loading failures." + "reason": "Both are config-loading related, but one is missing fields with trust_remote_code and the other is strict config rejection; not the same issue." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "CLIPOutput attentions regression vs RTDetrV2 output mismatch; different models and symptoms." }, { "left": "issue:43335", "right": "issue:43756", "accept": false, - "reason": "SwitchTransformers sparse-layer construction bug is unrelated to Smollm3 RoPE-layer dropping." + "reason": "SwitchTransformers sparse-layer config bug vs Smollm3 RoPE-layer count bug; unrelated." }, { "left": "issue:43550", "right": "issue:43819", "accept": false, - "reason": "Bamba torch.compile SDPA failure is unrelated to DAC.from_latents / STE mismatch." + "reason": "torch.compile/SDPA failure vs DAC forward/from_latents mismatch; no shared concrete defect." }, { "left": "issue:44843", "right": "issue:45081", "accept": false, - "reason": "Same helper area (_patch_mistral_regex), but different root causes: offline model_info call vs missing backend_tokenizer attribute." + "reason": "Both touch _patch_mistral_regex, but one is offline model_info access and the other is a backend_tokenizer attribute crash; different bugs." }, { - "left": "issue:44821", - "right": "issue:45084", + "left": "issue:44448", + "right": "issue:44671", "accept": false, - "reason": "AutoImageProcessor URL loading and template-node compilation errors are unrelated." + "reason": "Different model regressions with separate outputs/predictions issues; not mergeable as one fix." }, { - "left": "issue:44186", - "right": "issue:44442", + "left": "issue:39692", + "right": "issue:45834", "accept": false, - "reason": "LayoutLMv2 NER padding/truncation crash is unrelated to AutoTokenizer loading FastSpeech2ConformerTokenizer." + "reason": "Docs example errors vs OCR long-input index error; unrelated." }, { - "left": "issue:39692", - "right": "issue:45834", + "left": "issue:44186", + "right": "issue:44442", "accept": false, - "reason": "SigLIP2 doc/model mismatch and Kosmos2.5 long OCR index error are distinct documentation/model bugs." + "reason": "Tokenizer crash vs AutoTokenizer loading failure for a different tokenizer family." }, { - "left": "issue:44448", - "right": "issue:44671", + "left": "issue:44821", + "right": "issue:45084", "accept": false, - "reason": "Pegasus v4/v5 output drift and CamemBERT masked-LM prediction errors are different model regressions." + "reason": "ImageProcessor-from-URL loading issue vs compilation/template-node error; unrelated." }, { "left": "issue:43540", "right": "issue:44008", "accept": false, - "reason": "Qwen3OmniMoe video-input processing and Gemma3n audio_mel_mask name collision are different multimodal bugs." + "reason": "Two different multimodal model bugs with different failure modes and code paths." }, { "left": "issue:34634", "right": "issue:36010", "accept": false, - "reason": "BarkProcessor voice_preset and GenerationMixin import regression are unrelated." - }, - { - "left": "issue:43644", - "right": "issue:43749", - "accept": false, - "reason": "Non-persistent buffer initialization junk and FSDP CPU RAM efficient loading are related to loading, but not the same concrete bug." + "reason": "Voice preset bug in BarkProcessor vs GenerationMixin import error; unrelated." }, { - "left": "issue:45584", - "right": "issue:45834", + "left": "issue:45183", + "right": "issue:45479", "accept": false, - "reason": "Whisper empty-transcription failure after align_special_tokens is unrelated to Kosmos2.5 OCR indexing." + "reason": "Input validation error messaging vs zero-loss classification bug; different model logic." } ] }, @@ -5968,7 +6064,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 19, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -5978,7 +6074,8 @@ "nodes": [ "issue:28218", "issue:29159", - "issue:42175", + "issue:34634", + "issue:36331", "issue:43322", "issue:43388", "issue:43421", @@ -5987,6 +6084,7 @@ "issue:43577", "issue:43582", "issue:43606", + "issue:43644", "issue:43645", "issue:43716", "issue:43720", @@ -6010,10 +6108,8 @@ "issue:44991", "issue:45020", "issue:45072", - "issue:45081", - "issue:45183", "issue:45381", - "issue:45479", + "issue:45584", "issue:45685", "issue:45701", "issue:45834", @@ -6026,228 +6122,238 @@ "issue:46846" ], "soft_pairs": [ - "issue:45183|issue:45479", - "issue:43606|issue:45072", + "issue:43644|issue:43749", + "issue:45584|issue:45834", "issue:43322|issue:45685", - "issue:43550|issue:45072", - "issue:43782|issue:44977", + "issue:43606|issue:45072", "issue:43388|issue:46032", - "issue:44479|issue:45834", "issue:43825|issue:43994", - "issue:44991|issue:46291", - "issue:43582|issue:46512", + "issue:44479|issue:45834", + "issue:43550|issue:45072", + "issue:43782|issue:44977", "issue:43421|issue:44811", - "issue:43577|issue:43818", "issue:44561|issue:45020", + "issue:43577|issue:43818", + "issue:44991|issue:46291", + "issue:46294|issue:46463", + "issue:43582|issue:46512", "issue:43716|issue:44008", "issue:43931|issue:46846", - "issue:46294|issue:46463", "issue:28218|issue:29159", - "issue:44805|issue:44936", "issue:44610|issue:45072", "issue:43540|issue:45381", + "issue:44805|issue:44936", + "issue:43749|issue:44291", "issue:43720|issue:45072", "issue:43825|issue:43827", - "issue:43749|issue:44291", - "issue:43645|issue:46665", "issue:44671|issue:45701", - "issue:43577|issue:43582", - "issue:42175|issue:45081" + "issue:43645|issue:46665", + "issue:34634|issue:36331", + "issue:43577|issue:43582" ], - "prepared_review_unit_hash": "a4e31c8563701e26236c915596e5495c65334bd297784c436d0717ce72dc0919", + "prepared_review_unit_hash": "78c65142b2062abfd77938b6cdabeca50f51400daa3bf74fed04fb4a98ba7654", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 27, - "serialized_chars": 36274, - "estimated_input_tokens": 9069, - "estimated_eval_tokens": 18394 + "soft_pair_count": 28, + "serialized_chars": 36324, + "estimated_input_tokens": 9081, + "estimated_eval_tokens": 18418 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous cluster with one plausible duplicate pair around tokenizer spacing / slow-vs-fast tokenizer behavior. The rest of the soft edges appear to be different bugs, subsystems, or only loosely related regressions.", - "confidence": 0.41, - "canonical_issue_reason": "issue:28218 is the most concrete, symptom-level tokenizer bug in the set (\u201cadditional space after the added token\u201d) and best serves as the anchor for the one plausible tokenizer duplicate pair.", + "summary": "This cluster is highly heterogeneous: most items are unrelated regressions across tokenizers, video/image models, loading, training, and dtype/device issues. None of the soft pairs looks like the same concrete bug/change, so I would not merge any of them as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:28218 is the strongest global issue candidate because it states a specific reproducible bug rather than a broad umbrella regression title.", + "best_issue_reason": "issue:45020 is the broadest representative of the version/regression theme, but it is still too generic to serve as a true duplicate canonical for this cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:45183", - "right": "issue:45479", + "left": "issue:43644", + "right": "issue:43749", "accept": false, - "reason": "One is input_ids/inputs_embeds validation; the other is a classification loss bug. Same general area, different failure mode and code path." + "reason": "Different loading/initialization failures: non-persistent buffer junk vs FSDP CPU RAM efficient loading." }, { - "left": "issue:43606", - "right": "issue:45072", + "left": "issue:45584", + "right": "issue:45834", "accept": false, - "reason": "Different bugs: CPU offload device mismatch vs bfloat16 dtype mismatch. Not the same concrete issue." + "reason": "Different models and bugs: Whisper empty transcription after align_special_tokens vs Kosmos2.5 long OCR index error." }, { "left": "issue:43322", "right": "issue:45685", "accept": false, - "reason": "Llava Next segfault loading is unrelated to an MPS histogram error in MoE code." + "reason": "Unrelated paths: Llava Next load segfault vs MPS histogram Int error in moe." }, { - "left": "issue:43550", + "left": "issue:43606", "right": "issue:45072", "accept": false, - "reason": "torch.compile/SDPA failure in Bamba is not the same as a dtype mismatch regression." - }, - { - "left": "issue:43782", - "right": "issue:44977", - "accept": false, - "reason": "Qwen3VL weight_only loading error and Qwen3.5 flash-attention generation bug are different model/runtime problems." + "reason": "Both mention mismatches, but they are different concrete problems in different models/code paths." }, { "left": "issue:43388", "right": "issue:46032", "accept": false, - "reason": "Metric gathering label truncation and Mamba2 cache correctness are unrelated." + "reason": "Different bugs: last-batch label dropping vs Mamba2Mixer use_cache correctness." }, { - "left": "issue:44479", - "right": "issue:45834", + "left": "issue:43825", + "right": "issue:43994", "accept": false, - "reason": "Both are video/OCR related broadly, but they describe different models and different failures." + "reason": "One is an error-message/docs issue; the other is a model output quality regression." }, { - "left": "issue:43825", - "right": "issue:43994", + "left": "issue:44479", + "right": "issue:45834", "accept": false, - "reason": "Error-message wording about translation support is not the same as SigLIP2 bad outputs." + "reason": "Different models and failure modes: Qwen video regression vs Kosmos2.5 OCR long-input error." }, { - "left": "issue:44991", - "right": "issue:46291", + "left": "issue:43550", + "right": "issue:45072", "accept": false, - "reason": "Tokenizer loading failure and pipelines import failure with kernels v0.15 are different dependency/code-path issues." + "reason": "Different concrete issues: torch.compile/SDPA failure vs bfloat16 dtype mismatches." }, { - "left": "issue:43582", - "right": "issue:46512", + "left": "issue:43782", + "right": "issue:44977", "accept": false, - "reason": "Apple Silicon warmup TypeError is unrelated to stale config.dtype after from_config." + "reason": "Different failure surfaces: weight_only loading error vs flash-attention generation bug." }, { "left": "issue:43421", "right": "issue:44811", "accept": false, - "reason": "Runtime special-token post-processor updates and Whisper batch_decode skip_special_tokens are different tokenizer behaviors." + "reason": "Both are tokenizer-related, but one is post-processor runtime updates and the other is batch_decode skip_special_tokens behavior." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "Related theme (remote_code breakage) but not the same concrete underlying bug or code path." }, { "left": "issue:43577", "right": "issue:43818", "accept": false, - "reason": "BLIP2 dtype preservation and Video-LLaVA temporal-attention/weight-sharing issues are unrelated." + "reason": "Different model-specific issues: Blip2 dtype state vs Video-LLaVA temporal-attention/weight-sharing bug." }, { - "left": "issue:44561", - "right": "issue:45020", + "left": "issue:44991", + "right": "issue:46291", + "accept": false, + "reason": "Different breakages: tokenizer loading for a specific repo vs pipelines import failing due to kernels dependency requirements." + }, + { + "left": "issue:46294", + "right": "issue:46463", "accept": false, - "reason": "Broad remote_code breakage and removal of is_torch_fx_available are too loosely related to treat as the same bug." + "reason": "Different domains: Qwen3.5 video/chat-template mismatch vs T5 decoder input validation message." + }, + { + "left": "issue:43582", + "right": "issue:46512", + "accept": false, + "reason": "Different dtype/config issues on unrelated code paths and platforms." }, { "left": "issue:43716", "right": "issue:44008", "accept": false, - "reason": "Image preprocessor dtype mismatch in Mistral-3 and Gemma3n variable collision are different code paths." + "reason": "Different causes: image preprocessor/model dtype mismatch vs Gemma3n variable-name collision AttributeError." }, { "left": "issue:43931", "right": "issue:46846", "accept": false, - "reason": "Qwen3-VL weight-shape loading mismatch is distinct from tensor-parallel plan omissions causing OOM/shape errors." + "reason": "Both are Qwen/VL shape problems, but one is checkpoint weight-shape mismatch and the other is TP plan coverage/shape error." }, { - "left": "issue:46294", - "right": "issue:46463", + "left": "issue:28218", + "right": "issue:29159", "accept": false, - "reason": "Video chat-template feature/token mismatch is not the same as T5 decoder-input validation messaging." + "reason": "The second issue is much broader; the titles do not clearly identify the same tokenizer-space bug." }, { - "left": "issue:28218", - "right": "issue:29159", - "accept": true, - "reason": "Both point to tokenizer inconsistency around added tokens / slow-vs-fast behavior; these plausibly describe the same underlying tokenizer spacing bug." + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Different model/input-size issue vs dtype mismatch issue." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Both concern video inputs, but they affect different Qwen model variants and different failure modes." }, { "left": "issue:44805", "right": "issue:44936", "accept": false, - "reason": "Mask shape IndexError and trainer.evaluate() failing after train() are different symptoms and likely different causes." + "reason": "Unrelated: mask/index shape mismatch vs trainer lifecycle failure after train()." }, { - "left": "issue:44610", - "right": "issue:45072", + "left": "issue:43749", + "right": "issue:44291", "accept": false, - "reason": "Processor image-size mismatch is unrelated to dtype mismatch issues." + "reason": "Both involve loading, but the error mechanics are different: FSDP RAM-efficient loading vs unexpected _is_hf_initialized argument." }, { "left": "issue:43720", "right": "issue:45072", "accept": false, - "reason": "BitNet packed-weight unpacking on accelerate loading is a different failure than dtype mismatches." + "reason": "Different failures: packed-weight unpacking during accelerate loading vs bfloat16 dtype mismatches." }, { "left": "issue:43825", "right": "issue:43827", "accept": false, - "reason": "One is an error message regression; the other is stale documentation after pipeline removals." + "reason": "Same general pipeline-removal theme, but one is an error message and the other is stale docs." }, { - "left": "issue:43749", - "right": "issue:44291", + "left": "issue:44671", + "right": "issue:45701", "accept": false, - "reason": "FSDP CPU RAM efficient loading and init_empty_weights argument handling are different initialization/loading bugs." + "reason": "Different problems: CamemBERT masked LM regression vs tokenizer version-dependent tokenization changes." }, { "left": "issue:43645", "right": "issue:46665", "accept": false, - "reason": "Both involve custom/from_pretrained initialization, but the notebook-specific custom-model issue is not clearly the same as class-based from_pretrained inconsistency." + "reason": "Both mention model initialization, but they are distinct context-specific behaviors with no shared concrete bug." }, { - "left": "issue:44671", - "right": "issue:45701", + "left": "issue:34634", + "right": "issue:36331", "accept": false, - "reason": "CamemBERT masked LM prediction regression and tokenization changes are not the same bug." + "reason": "Completely different subsystems: BarkProcessor voice_preset vs Trainer compute_loss signature change." }, { "left": "issue:43577", "right": "issue:43582", "accept": false, - "reason": "Model dtype persistence and Apple Silicon allocator warmup TypeError are unrelated." - }, - { - "left": "issue:42175", - "right": "issue:45081", - "accept": false, - "reason": "Missing TensorFlow backend in an install extra is unrelated to a Mistral tokenizer regex crash." + "reason": "Unrelated bugs: Blip2 dtype state vs Apple Silicon caching_allocator_warmup TypeError." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary stays grounded in the packet and is appropriately cautious. Most rejections are conservative and match the titles\u2019 different failure modes/code paths. The only accepted pair, issue:28218 with issue:29159, is plausibly the same tokenizer spacing/slow-vs-fast behavior bug based on the titles alone, and the summary correctly phrases it as a plausible pair rather than a certainty." - }, - "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 }, @@ -6256,7 +6362,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 20, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -6264,7 +6370,6 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-502", "nodes": [ - "issue:34634", "issue:36010", "issue:36331", "issue:42175", @@ -6314,231 +6419,231 @@ "issue:46932" ], "soft_pairs": [ - "issue:34634|issue:36331", + "issue:42175|issue:45081", "issue:46175|issue:46932", "issue:45003|issue:46285", - "issue:43653|issue:44912", "issue:46697|issue:46728", + "issue:43653|issue:44912", "issue:43782|issue:45325", "issue:43262|issue:43819", - "issue:43582|issue:44857", - "issue:45081|issue:45406", "issue:45372|issue:46326", + "issue:45081|issue:45406", + "issue:43582|issue:44857", "issue:44589|issue:45042", "issue:43742|issue:45020", "issue:43550|issue:44912", "issue:43452|issue:44991", - "issue:44060|issue:44368", "issue:43329|issue:45600", - "issue:45084|issue:45538", "issue:43257|issue:43329", + "issue:44060|issue:44368", "issue:43526|issue:43697", + "issue:36010|issue:36331", "issue:43906|issue:46209", + "issue:45084|issue:45538", + "issue:36010|issue:43645", "issue:43526|issue:44554", - "issue:36010|issue:36331", "issue:43819|issue:43927", - "issue:36010|issue:43645", "issue:43531|issue:45042", - "issue:42175|issue:46291", "issue:43638|issue:44936", + "issue:42175|issue:46291", "issue:42490|issue:43257" ], - "prepared_review_unit_hash": "a3e514c079d798295676488d1c27df026db4b72a06bea03481e422bd5beadfe7", + "prepared_review_unit_hash": "6642d6a9f31d0e0e5e06642f456175eccadc0ca46e13571d7b4b474cd31458c8", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 36566, - "estimated_input_tokens": 9142, - "estimated_eval_tokens": 18540 + "serialized_chars": 36030, + "estimated_input_tokens": 9008, + "estimated_eval_tokens": 18272 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is a loose set of unrelated Transformers bug reports across tokenizers, vision/audio processors, training, quantization, and model-loading paths. No pair looks like the same concrete bug, so I reject all soft duplicate candidates.", - "confidence": 0.72, - "canonical_issue_reason": "Issue #45020 is the broadest, most central regression report in the set: recent Transformers versions breaking models that rely on `remote_code`. It best serves as the umbrella representative for this otherwise heterogeneous cluster.", + "summary": "This cluster is mostly a grab bag of unrelated closed issues from Transformers. None of the soft pairs looks like the same underlying bug or change; the closest thing to an umbrella topic is loading/regression behavior, but it is still too broad for duplicate merging.", + "confidence": 0.93, + "canonical_issue_reason": "issue:45020 is the broadest umbrella-style report in the set and the closest thing to a hub for loading/regression breakages, though it is not a true duplicate canonical.", "canonical_pr_reason": null, - "best_issue_reason": "#45020 has the widest user impact and the most general framing among the issues, making it the most useful representative of the cluster.", + "best_issue_reason": "issue:45020 is the best overall representative of the cluster because it has the widest scope and strongest \u201crecent versions break models\u201d framing, but the cluster is too heterogeneous to treat it as a duplicate anchor.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:34634", - "right": "issue:36331", + "left": "issue:42175", + "right": "issue:45081", "accept": false, - "reason": "Different subsystems and failure modes: Bark voice presets vs. Trainer loss signature change." + "reason": "Different bugs: backend dependency packaging vs Mistral regex/tokenizer attribute handling." }, { "left": "issue:46175", "right": "issue:46932", "accept": false, - "reason": "Both are Gemma4-related, but one is a large position-embedding memory issue and the other is dtype casting for quantized embedders; different code paths." + "reason": "Both are Gemma4-related, but one is a huge one-hot position embedding issue and the other is quantized embedder dtype casting; not the same fix." }, { "left": "issue:45003", "right": "issue:46285", "accept": false, - "reason": "`sys.modules` access in loading logic is unrelated to CLIP `.text_model` structure changes." + "reason": "Unrelated: unsafe sys.modules access vs CLIP text model structural inconsistency." }, { - "left": "issue:43653", - "right": "issue:44912", + "left": "issue:46697", + "right": "issue:46728", "accept": false, - "reason": "BigBird special-token registration and MXFP4 quantization fallback are unrelated bugs." + "reason": "Both concern image sizing/padding, but they affect different models and different code paths." }, { - "left": "issue:46697", - "right": "issue:46728", + "left": "issue:43653", + "right": "issue:44912", "accept": false, - "reason": "Different image processors and distinct padding vs. patch-count logic." + "reason": "Tokenizer special-token registration vs quantization fallback behavior are unrelated." }, { "left": "issue:43782", "right": "issue:45325", "accept": false, - "reason": "Qwen2.5-VL weight loading error is unrelated to Qwen2.5-VL RoPE index scaling." + "reason": "Both are VL model issues, but the failure modes are different and not mergeable as one concrete bug." }, { "left": "issue:43262", "right": "issue:43819", "accept": false, - "reason": "Audio chat-template sample-rate handling is unrelated to DAC latent/forward mismatch." + "reason": "Audio chat-template sampling-rate defaulting is unrelated to DAC latent reconstruction/STE mismatch." }, { - "left": "issue:43582", - "right": "issue:44857", + "left": "issue:45372", + "right": "issue:46326", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError is unrelated to CUDA AMP loss crashes." + "reason": "Both mention mistral_common, but one is an import failure for Gemma4 processor loading and the other is continue_final_message behavior." }, { "left": "issue:45081", "right": "issue:45406", "accept": false, - "reason": "Both involve model loading, but Mistral tokenizer regex patching and Gemma4Processor initialization are different bugs." + "reason": "Different attribute errors in different components: tokenizer backend patching vs Gemma4Processor state." }, { - "left": "issue:45372", - "right": "issue:46326", + "left": "issue:43582", + "right": "issue:44857", "accept": false, - "reason": "Both mention `mistral_common`, but one is an import error and the other is `continue_final_message` behavior; not the same defect." + "reason": "Apple Silicon warmup TypeError and CUDA AMP loss crash are unrelated runtime failures." }, { "left": "issue:44589", "right": "issue:45042", "accept": false, - "reason": "Float8 storage lookup failure is unrelated to torchvision being required by PIL image processors." + "reason": "Storage-object lookup failure is unrelated to PIL backend torchvision dependency handling." }, { "left": "issue:43742", "right": "issue:45020", "accept": false, - "reason": "MobileLLM key error is a specific model-loading failure, not the broader remote_code regression." + "reason": "A specific model load KeyError is not the same concrete bug as the broader remote_code regression report." }, { "left": "issue:43550", "right": "issue:44912", "accept": false, - "reason": "torch.compile/SDPA failure in Bamba is unrelated to MXFP4 quantization fallback." + "reason": "torch.compile SDPA failure and MXFP4 quantization fallback are different code paths." }, { "left": "issue:43452", "right": "issue:44991", "accept": false, - "reason": "gguf_file handling for AutoTokenizer/AutoModel is not the same bug as EMBEDDIA tokenizer loading regression." - }, - { - "left": "issue:44060", - "right": "issue:44368", - "accept": false, - "reason": "Both are warning-related, but one is an incorrect weight-tying warning and the other is a config warning during LoRA fine-tuning." + "reason": "Both are loading failures, but for different tokenizer/model formats and different root causes." }, { "left": "issue:43329", "right": "issue:45600", "accept": false, - "reason": "Undefined video multimodal helpers are unrelated to stale auto_mappings references." + "reason": "Multimodal token counting bug is unrelated to auto_mappings referencing removed configs." }, { - "left": "issue:45084", - "right": "issue:45538", + "left": "issue:43257", + "right": "issue:43329", "accept": false, - "reason": "Template-compilation TypeError is unrelated to CLIPTokenizer max-length behavior." + "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is unrelated to a multimodal token-counting undefined-name bug." }, { - "left": "issue:43257", - "right": "issue:43329", + "left": "issue:44060", + "right": "issue:44368", "accept": false, - "reason": "Qwen3 MoE weight conversion and multimodal token counting are different problems." + "reason": "Both involve tied-weights warnings in Qwen variants, but they describe different warnings and different root causes." }, { "left": "issue:43526", "right": "issue:43697", "accept": false, - "reason": "BEiT label reduction and RTDetrV2 output drift are separate model bugs." + "reason": "Beit label reduction bug and RTDetr output divergence are unrelated." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "ImportError for GenerationMixin and compute_loss signature mismatch are unrelated trainer/import issues." }, { "left": "issue:43906", "right": "issue:46209", "accept": false, - "reason": "A reproduction of an older issue is unrelated to FP8 inference failing on B200." + "reason": "An isolated reproduction of an old issue is unrelated to FP8 inference on B200." }, { - "left": "issue:43526", - "right": "issue:44554", + "left": "issue:45084", + "right": "issue:45538", "accept": false, - "reason": "BEiT preprocessing bug is unrelated to the MPS attention correctness issue." + "reason": "Template compilation error and CLIPTokenizer max_length bug are unrelated." }, { "left": "issue:36010", - "right": "issue:36331", + "right": "issue:43645", "accept": false, - "reason": "Importing `GenerationMixin` and `compute_loss` signature mismatch are unrelated breakages." + "reason": "Import path breakage and Jupyter custom-model initialization are different bugs." }, { - "left": "issue:43819", - "right": "issue:43927", + "left": "issue:43526", + "right": "issue:44554", "accept": false, - "reason": "DAC latent reconstruction mismatch and DiaConfig token-ID loss are different model/config bugs." + "reason": "Label reduction in BeitImageProcessorFast is unrelated to MPS attention correctness." }, { - "left": "issue:36010", - "right": "issue:43645", + "left": "issue:43819", + "right": "issue:43927", "accept": false, - "reason": "`GenerationMixin` import failure is not the same as notebook custom-model initialization breakage." + "reason": "DAC forward/latent mismatch and DiaConfig token ID loss are different model/config problems." }, { "left": "issue:43531", "right": "issue:45042", "accept": false, - "reason": "Qwen3-MoE sliding-window behavior and PIL/torchvision requirements are unrelated." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to PIL backend torchvision requirements." }, { - "left": "issue:42175", - "right": "issue:46291", + "left": "issue:43638", + "right": "issue:44936", "accept": false, - "reason": "Package install backend selection and pipeline import dependency validation are different issues." + "reason": "Both involve training/eval, but the failures are different and not the same underlying bug." }, { - "left": "issue:43638", - "right": "issue:44936", + "left": "issue:42175", + "right": "issue:46291", "accept": false, - "reason": "DeepSpeed ZeRO-3 index error is unrelated to evaluate() failing after train()." + "reason": "TensorFlow extra packaging and kernels version/revision validation are unrelated dependency issues." }, { "left": "issue:42490", "right": "issue:43257", "accept": false, - "reason": "Trailing-slash local path warnings and Qwen3 MoE weight conversion are unrelated." + "reason": "Trailing-slash local model path warning is unrelated to Qwen3 MoE weight conversion." } ] }, @@ -6554,7 +6659,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 21, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -6566,6 +6671,7 @@ "issue:29159", "issue:29870", "issue:34567", + "issue:34689", "issue:38175", "issue:39692", "issue:40990", @@ -6574,7 +6680,6 @@ "issue:43262", "issue:43329", "issue:43450", - "issue:43525", "issue:43540", "issue:43606", "issue:43643", @@ -6599,246 +6704,251 @@ "issue:44821", "issue:44912", "issue:44977", + "issue:44987", "issue:45042", "issue:45356", "issue:45459", + "issue:45538", "issue:45759", "issue:45803", "issue:45859", - "issue:46291", "issue:46421", "issue:46424", "issue:46701" ], "soft_pairs": [ + "issue:43746|issue:44811", "issue:43720|issue:44610", - "issue:39692|issue:44912", "issue:43854|issue:44977", - "issue:43746|issue:44811", + "issue:39692|issue:44912", "issue:43643|issue:45759", "issue:38175|issue:44610", "issue:43257|issue:43696", "issue:43262|issue:44811", - "issue:44387|issue:44617", "issue:43257|issue:43931", - "issue:45459|issue:45803", + "issue:44387|issue:44617", "issue:43329|issue:43450", + "issue:43329|issue:44479", "issue:46421|issue:46424", + "issue:45459|issue:45803", "issue:42915|issue:43742", - "issue:43329|issue:44479", "issue:34567|issue:40990", - "issue:44291|issue:45042", - "issue:43650|issue:43906", "issue:43749|issue:44756", - "issue:42915|issue:44821", + "issue:44291|issue:45042", "issue:29159|issue:34567", + "issue:43650|issue:43906", "issue:43606|issue:45859", + "issue:42915|issue:44821", "issue:43927|issue:44912", - "issue:28282|issue:29870", "issue:43540|issue:46701", - "issue:43525|issue:46291", - "issue:44568|issue:45356" + "issue:28282|issue:29870", + "issue:44568|issue:45356", + "issue:44987|issue:45538", + "issue:28282|issue:34689" ], - "prepared_review_unit_hash": "3f27502981b1a0f04237a91de5fa417035eb9ad2ceb44a474c9b227bd30f6623", + "prepared_review_unit_hash": "eb9a1b81900132798ba1ff1e882d18c8ab88c7e3eb0c8796558969108eb30183", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 27, - "serialized_chars": 35411, - "estimated_input_tokens": 8853, - "estimated_eval_tokens": 17962 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36222, + "estimated_input_tokens": 9056, + "estimated_eval_tokens": 18368 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is mostly heterogeneous; nearly all pairs are superficial matches on broad subsystems. The only likely duplicate edge is the Qwen video-input regression pair, which appears to describe the same underlying multimodal video-token handling bug.", - "confidence": 0.83, - "canonical_issue_reason": "issue:44479 is the better canonical issue because it is the broader, user-facing regression report covering multiple affected Qwen VL/MoE models in v5.3.0, while issue:43329 is a narrower root-cause/debugging description of the same video-token path.", + "summary": "This cluster is heterogeneous: it spans unrelated issues in loading, tokenization, multimodal/video processing, cache behavior, and memory/offload problems. None of the soft-edge pairs look like the same underlying bug or change, so I would not consolidate them.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue is suitable: the items are from different subsystems and failure modes, with no clear shared underlying defect across the set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44479 best represents the duplicate cluster: it clearly states the regression, affected models, and release scope, making it the most actionable single issue to keep.", + "best_issue_reason": "No global best issue: there isn\u2019t one representative bug that covers the rest of the cluster without losing important distinctions.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43720", - "right": "issue:44610", + "left": "issue:43746", + "right": "issue:44811", "accept": false, - "reason": "Different bugs: packed-weight loading in BitNet vs a processor input-size mismatch in OmDet-Turbo." + "reason": "Different code paths: local PEFT adapter loading versus Whisper batch decoding behavior." }, { - "left": "issue:39692", - "right": "issue:44912", + "left": "issue:43720", + "right": "issue:44610", "accept": false, - "reason": "Docs example errors are unrelated to MXFP4 quantization fallback behavior." + "reason": "Unrelated problems: BitNet packed-weight loading on accelerate versus OmDet-Turbo processor image-size mismatch." }, { "left": "issue:43854", "right": "issue:44977", "accept": false, - "reason": "Unrelated model-load failure in tests vs a flash-attention generation issue." + "reason": "Both are model-specific, but one is a unit-test loading failure and the other is a flash-attention generation regression; not the same bug." }, { - "left": "issue:43746", - "right": "issue:44811", + "left": "issue:39692", + "right": "issue:44912", "accept": false, - "reason": "PEFT local-checkpoint loading and Whisper batch_decode token skipping are different code paths." + "reason": "Documentation/example errors are distinct from MXFP4 quantization fallback/loading behavior." }, { "left": "issue:43643", "right": "issue:45759", "accept": false, - "reason": "Both involve config handling, but the failures are different: missing fields from remote config load vs not unwrapping text_config in from_config." + "reason": "Both concern config loading, but they affect different APIs and different missing-field/unwrap behaviors." }, { "left": "issue:38175", "right": "issue:44610", "accept": false, - "reason": "Different symptom and subsystem: SigLIP2 zero probabilities vs OmDet-Turbo processor/model size mismatch." + "reason": "Different symptoms and subsystems: zero probabilities in SigLIP2 versus processor output size mismatch in OmDet-Turbo." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Qwen3 MoE weight conversion/loading bug is unrelated to a GPT-oss CUDA OOM." + "reason": "Qwen3 MoE weight-conversion failure is unrelated to a generic CUDA OOM in GPT-oss." }, { "left": "issue:43262", "right": "issue:44811", "accept": false, - "reason": "Audio chat-template sample-rate defaulting and Whisper special-token decoding are unrelated." - }, - { - "left": "issue:44387", - "right": "issue:44617", - "accept": false, - "reason": "Both are OOMs, but one is an int4 memory-regression issue and the other is a model-specific video OOM." + "reason": "Audio chat-template sampling-rate defaulting is unrelated to Whisper's skip_special_tokens handling." }, { "left": "issue:43257", "right": "issue:43931", "accept": false, - "reason": "Both concern Qwen loading, but one is accelerate/deepspeed conversion and the other is a shape mismatch for a VL model." + "reason": "Both are Qwen3-related, but one is MoE weight conversion and the other is a Qwen3-VL weight-shape mismatch." }, { - "left": "issue:45459", - "right": "issue:45803", + "left": "issue:44387", + "right": "issue:44617", "accept": false, - "reason": "Both mention exception handling, but they are different bugs in different converters/tokenizer paths." + "reason": "Both mention OOM, but the causes are different: int4 quantization memory regression versus a model-specific video OOM." }, { "left": "issue:43329", "right": "issue:43450", "accept": false, - "reason": "Video-token counting undefined symbols vs batched video processor shape are different failures." + "reason": "Video-related, but one is an undefined helper in multimodal token counting and the other is a batched video processor shape bug." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "One is a broken internal video token-count path; the other is a broader v5.3.0 video-input regression across models." }, { "left": "issue:46421", "right": "issue:46424", "accept": false, - "reason": "Both are cache-related, but one is lazy-init recompilation and the other is max_cache_length handling." + "reason": "Both are cache-related, but one is lazy-init recompilation and the other is max_cache_length being honored only for quantized cache." }, { - "left": "issue:42915", - "right": "issue:43742", + "left": "issue:45459", + "right": "issue:45803", "accept": false, - "reason": "Different loading failures: FineGrainedFP8Config support vs a key error on MobileLLM-125M." + "reason": "Both involve exception handling, but they are different swallows: protobuf decode errors versus a bare except in FuyuBatchFeature." }, { - "left": "issue:43329", - "right": "issue:44479", - "accept": true, - "reason": "Both point to the same Qwen video-input regression in the multimodal/video path; one gives the concrete undefined-helper root cause and the other reports the broader model-level regression." + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Different failures: FineGrainedFP8Config support in Qwen3Moe versus a key error while loading MobileLLM." }, { "left": "issue:34567", "right": "issue:40990", "accept": false, - "reason": "Trainer token accounting and GPT-oss perplexity are unrelated." + "reason": "Trainer token-count bookkeeping is unrelated to GPT-oss perplexity quality issues." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "Both touch loading/memory, but one is FSDP CPU RAM efficient loading and the other is mmap behavior on Strix Halo." }, { "left": "issue:44291", "right": "issue:45042", "accept": false, - "reason": "init_empty_weights argument handling and PIL backend torchvision dependency are unrelated bugs." + "reason": "Different APIs and failure modes: init_empty_weights argument handling versus PIL image processors requiring torchvision." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency is unrelated to TrainerState token counting." }, { "left": "issue:43650", "right": "issue:43906", "accept": false, - "reason": "Generic title-only issue vs an isolated reproduction of a specific prior bug; no shared underlying failure." + "reason": "No substantive overlap; one is a nonsensical placeholder title and the other is a reproduction issue." }, { - "left": "issue:43749", - "right": "issue:44756", + "left": "issue:43606", + "right": "issue:45859", "accept": false, - "reason": "Both reference memory/offload concerns, but one is FSDP RAM-efficient loading and the other is mmap on Strix Halo." + "reason": "Both involve CPU offload, but the failing code paths and models are different." }, { "left": "issue:42915", "right": "issue:44821", "accept": false, - "reason": "Different problems: FP8 config loading vs AutoImageProcessor URL loading." + "reason": "Qwen3Moe FP8 config failure is unrelated to AutoImageProcessor loading from URL." }, { - "left": "issue:29159", - "right": "issue:34567", + "left": "issue:43927", + "right": "issue:44912", "accept": false, - "reason": "Tokenizer inconsistency and TrainerState token counters are unrelated." + "reason": "Config save/load token IDs bug is unrelated to MXFP4 quantization fallback." }, { - "left": "issue:43606", - "right": "issue:45859", + "left": "issue:43540", + "right": "issue:46701", "accept": false, - "reason": "Both involve CPU offload, but they affect different models and different failure modes." - }, - { - "left": "issue:43927", - "right": "issue:44912", - "accept": false, - "reason": "DiaConfig token-ID persistence is unrelated to gpt-oss MXFP4 quantization loading." + "reason": "Both are multimodal, but they concern different models, modalities, and failure points." }, { "left": "issue:28282", "right": "issue:29870", "accept": false, - "reason": "ImportError for missing PyTorch and Bart generate() decoder mask initialization are unrelated." + "reason": "ImportError for missing PyTorch is unrelated to Bart generate's decoder_attention_mask assignment bug." }, { - "left": "issue:43540", - "right": "issue:46701", + "left": "issue:44568", + "right": "issue:45356", "accept": false, - "reason": "Video-processing ValueError in Qwen3OmniMoe vs a DiffusionGemma image-text pipeline issue are different bugs." + "reason": "Both are tokenizer regressions, but they affect different tokenizers and different mechanisms." }, { - "left": "issue:43525", - "right": "issue:46291", + "left": "issue:44987", + "right": "issue:45538", "accept": false, - "reason": "Llama4Config missing pad_token_id is unrelated to kernels import-time version/revision requirements." + "reason": "Transformers loading failure for a model is unrelated to CLIPTokenizer's model_max_length setting." }, { - "left": "issue:44568", - "right": "issue:45356", + "left": "issue:28282", + "right": "issue:34689", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different tokenizer behaviors; no shared concrete code-path bug." + "reason": "Missing PyTorch ImportError and Llama 3.2 vision model-loading regression are distinct problems." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet, and the one accepted soft-edge pair (43329 vs 44479) is a conservative duplicate call: both concern the same Qwen video-input/video-token path, with 43329 describing the lower-level undefined-helper failure and 44479 describing the broader v5.3.0 regression across affected Qwen VL/MoE models. The remaining pairwise rejects are also appropriately cautious given the titles provided." - }, - "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 }, @@ -6847,7 +6957,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 22, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -6862,7 +6972,6 @@ "issue:29159", "issue:33357", "issue:34567", - "issue:34689", "issue:41720", "issue:42175", "issue:42915", @@ -6870,6 +6979,7 @@ "issue:43425", "issue:43452", "issue:43454", + "issue:43525", "issue:43540", "issue:43577", "issue:43582", @@ -6880,18 +6990,17 @@ "issue:44062", "issue:44291", "issue:44479", - "issue:44488", + "issue:44560", "issue:44568", - "issue:44792", "issue:44849", "issue:44857", "issue:44936", - "issue:44964", "issue:44987", "issue:44991", "issue:45127", "issue:45137", "issue:45357", + "issue:45381", "issue:45405", "issue:45488", "issue:45538", @@ -6905,249 +7014,224 @@ "issue:46869" ], "soft_pairs": [ - "issue:44987|issue:45538", "issue:43454|issue:43577", - "issue:28282|issue:34689", - "issue:45538|issue:46869", - "issue:43425|issue:46291", - "issue:44857|issue:46832", + "issue:43525|issue:46291", "issue:43697|issue:45357", + "issue:44857|issue:46832", "issue:33357|issue:34567", "issue:43404|issue:45127", + "issue:45538|issue:46869", + "issue:43425|issue:46291", "issue:45137|issue:45405", - "issue:42175|issue:43824", "issue:42915|issue:45538", "issue:28282|issue:29127", + "issue:42175|issue:43824", + "issue:44479|issue:46693", "issue:44062|issue:44991", "issue:43742|issue:44291", - "issue:44479|issue:46693", - "issue:44936|issue:46821", - "issue:43577|issue:46512", "issue:42915|issue:43749", + "issue:44936|issue:46821", "issue:44849|issue:45701", - "issue:43452|issue:43582", - "issue:27957|issue:28282", "issue:24643|issue:29159", + "issue:43452|issue:43582", + "issue:43577|issue:46512", "issue:44987|issue:45907", + "issue:27957|issue:28282", "issue:41720|issue:43540", - "issue:42175|issue:44488", - "issue:44568|issue:45488", - "issue:44792|issue:44964" + "issue:43540|issue:44560", + "issue:44479|issue:45381", + "issue:44568|issue:45488" ], - "prepared_review_unit_hash": "9183cde81faacd2e54ed06b6a7d49b2142be546d6592d3d852a793d507cab9c5", + "prepared_review_unit_hash": "6c25d5bac050358ee0772fb10288da196e0eae916f4d87e8fabf481ab80b085a", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 28, - "serialized_chars": 36028, - "estimated_input_tokens": 9007, - "estimated_eval_tokens": 18270 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35181, + "estimated_input_tokens": 8796, + "estimated_eval_tokens": 17848 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the soft pairs mostly share broad themes like tokenization, loading, dtype, or training, but they do not look like the same concrete bug. I would not merge any of the suggested pairs.", + "summary": "This cluster is not a duplicate set: the issues span unrelated subsystems (tokenizers, multimodal models, training, device mapping, packaging, and version compatibility). The soft pairs are only superficially similar and do not describe the same underlying bug or change.", "confidence": 0.96, - "canonical_issue_reason": "No single issue is a safe canonical representative because the cluster spans many unrelated failure modes across tokenizer behavior, model loading, training, multimodal models, and backend-specific regressions.", + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "If forced to pick a single issue for visibility, issue 45488 is the broadest/highest-impact tokenizer regression, but it is not representative of the cluster as a whole.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44987", - "right": "issue:45538", - "accept": false, - "reason": "Different bugs: model loading regression vs CLIPTokenizer max-length behavior." - }, { "left": "issue:43454", "right": "issue:43577", "accept": false, - "reason": "Different code paths: missing lm_head tying vs dtype propagation in BLIP2." - }, - { - "left": "issue:28282", - "right": "issue:34689", - "accept": false, - "reason": "Unrelated failures: missing PyTorch dependency vs Llama 3.2 vision loading breakage." + "reason": "Both mention weight tying, but one is about Mistral3 lm_head tying and the other about BLIP2 dtype propagation; different models and different failure modes." }, { - "left": "issue:45538", - "right": "issue:46869", + "left": "issue:43525", + "right": "issue:46291", "accept": false, - "reason": "Unrelated subsystems: tokenizer max length vs YOLO multi-GPU device-map failure." + "reason": "Unrelated errors: missing pad_token_id on Llama4Config vs kernels package requiring version/revision metadata during import." }, { - "left": "issue:43425", - "right": "issue:46291", + "left": "issue:43697", + "right": "issue:45357", "accept": false, - "reason": "Different incompatibilities: Torch 2.10 support vs kernels revision/version requirement." + "reason": "RTDetrV2 output regression is unrelated to Qwen3.5 save_pretrained writing wrong visual encoder keys." }, { "left": "issue:44857", "right": "issue:46832", "accept": false, - "reason": "Both DETR-related, but different bugs and different failing paths/models." - }, - { - "left": "issue:43697", - "right": "issue:45357", - "accept": false, - "reason": "Different issues: RTDetr output mismatch vs incorrect save_pretrained key mapping." + "reason": "Different RT-DETR bugs: AMP/CUDA crash in LwDetrImageLoss vs forward TypeError when feature levels exceed backbone outputs." }, { "left": "issue:33357", "right": "issue:34567", "accept": false, - "reason": "Unrelated: Mac bus error on CLIP model vs TrainerState token counter not updating." + "reason": "MacOS bus error in CLIP community model loading is unrelated to TrainerState.num_input_tokens_seen not updating." }, { "left": "issue:43404", "right": "issue:45127", "accept": false, - "reason": "Both involve tied weights, but different models and different failure mechanisms." + "reason": "Both involve tied weights, but one is a Mistral3 generation bug and the other is LoRA merge with extended vocabulary causing model collapse; not the same code path." }, { - "left": "issue:45137", - "right": "issue:45405", + "left": "issue:45538", + "right": "issue:46869", "accept": false, - "reason": "Completely unrelated: DeepSpeed deque error vs PEFT version pin problem." + "reason": "CLIPTokenizer model_max_length constant is unrelated to YOLO multi-GPU auto device mapping failures." }, { - "left": "issue:42175", - "right": "issue:43824", + "left": "issue:43425", + "right": "issue:46291", + "accept": false, + "reason": "Torch 2.10 compatibility and kernels import requirements are different dependency issues with no shared bug mechanism." + }, + { + "left": "issue:45137", + "right": "issue:45405", "accept": false, - "reason": "Different failures: backend packaging issue vs missing model import symbol." + "reason": "DeepSpeed ZeRO3 deque underflow is unrelated to an unreleased PEFT version bump." }, { "left": "issue:42915", "right": "issue:45538", "accept": false, - "reason": "Different bugs: FineGrainedFP8 loading failure vs tokenizer max-length setting." + "reason": "Qwen3Moe FineGrainedFP8Config failure is unrelated to CLIPTokenizer's max length constant." }, { "left": "issue:28282", "right": "issue:29127", "accept": false, - "reason": "Unrelated: missing PyTorch dependency vs LayoutLMv3 error-message clarity." - }, - { - "left": "issue:44062", - "right": "issue:44991", - "accept": false, - "reason": "Both tokenizer-related, but one is AddedToken argument handling and the other is a model-specific tokenizer loading failure." + "reason": "AutoModel ImportError for missing PyTorch is unrelated to LayoutLMv3 box-validation error messaging." }, { - "left": "issue:43742", - "right": "issue:44291", + "left": "issue:42175", + "right": "issue:43824", "accept": false, - "reason": "Different loading bugs: missing key on MobileLLM vs unexpected init_empty_weights argument." + "reason": "TensorFlow missing from a torch-only install is unrelated to the Qwen2_5_VL import/export issue." }, { "left": "issue:44479", "right": "issue:46693", "accept": false, - "reason": "Different regressions: video-input breakage vs long-context performance regression." + "reason": "Both are video/FA2 related at a high level, but one is a v5 video input regression across Qwen variants and the other is a long-context performance regression with FA2; not the same defect." }, { - "left": "issue:44936", - "right": "issue:46821", + "left": "issue:44062", + "right": "issue:44991", "accept": false, - "reason": "Unrelated: trainer lifecycle bug vs TimesFM window_size attribute error." + "reason": "AddedToken special-argument duplication is a tokenizer construction bug, while EMBEDDIA/est-roberta loading failure is a separate tokenizer compatibility issue." }, { - "left": "issue:43577", - "right": "issue:46512", + "left": "issue:43742", + "right": "issue:44291", "accept": false, - "reason": "Both dtype-related, but one is stale submodule dtypes and the other is stale config dtype after from_config." + "reason": "MobileLLM key mismatch and init_empty_weights unexpected _is_hf_initialized argument are different loading-time failures." }, { "left": "issue:42915", "right": "issue:43749", "accept": false, - "reason": "Different areas: FP8 config failure vs FSDP CPU RAM efficient loading." - }, - { - "left": "issue:44849", - "right": "issue:45701", - "accept": false, - "reason": "Different bugs: hidden_states handling in Qwen3.5 vs version-dependent tokenization changes." + "reason": "Qwen3Moe FP8 config failure and FSDP_CPU_RAM_EFFICIENT_LOADING breakage are unrelated model/config vs distributed-loading issues." }, { - "left": "issue:43452", - "right": "issue:43582", + "left": "issue:44936", + "right": "issue:46821", "accept": false, - "reason": "Unrelated: gguf_file loading path vs Apple Silicon allocator warmup TypeError." + "reason": "Trainer evaluate-after-train failure is unrelated to TimesFM 2.5 window_size AttributeError." }, { - "left": "issue:27957", - "right": "issue:28282", + "left": "issue:44849", + "right": "issue:45701", "accept": false, - "reason": "Different failures: Flash Attention 2 issue vs missing PyTorch dependency." + "reason": "Qwen3.5 output_hidden_states bug is a model-specific issue, not the broad tokenization behavior change described in issue 45701." }, { "left": "issue:24643", "right": "issue:29159", "accept": false, - "reason": "Unrelated: DeepSpeed training runtime error vs tokenizer inconsistency." + "reason": "DeepSpeed training weight shape error is unrelated to slow vs fast tokenizer inconsistency." }, { - "left": "issue:44987", - "right": "issue:45907", + "left": "issue:43452", + "right": "issue:43582", "accept": false, - "reason": "Different issues: model loading regression vs list_repo_files revision forwarding." + "reason": "gguf_file loading failures and Apple Silicon caching_allocator_warmup TypeError are separate APIs and code paths." }, { - "left": "issue:41720", - "right": "issue:43540", + "left": "issue:43577", + "right": "issue:46512", "accept": false, - "reason": "Different Qwen3 bugs with different symptoms and code paths." + "reason": "BLIP2 dtype not propagating on load is related only superficially to from_config(dtype=...) leaving config.dtype stale; one is load-time model object state, the other is constructor-time config synchronization." }, { - "left": "issue:42175", - "right": "issue:44488", + "left": "issue:44987", + "right": "issue:45907", "accept": false, - "reason": "Unrelated: package backend selection vs a model loading failure." + "reason": "Loading physical-intelligence/fast and forwarding revision in list_repo_files are unrelated repository/tooling bugs." }, { - "left": "issue:44568", - "right": "issue:45488", + "left": "issue:27957", + "right": "issue:28282", "accept": false, - "reason": "Both tokenizer regressions, but different tokenizer classes and different underlying behavior changes." + "reason": "XLMRoberta Flash Attention 2 support is unrelated to AutoModel failing when PyTorch is absent." }, { - "left": "issue:44792", - "right": "issue:44964", + "left": "issue:41720", + "right": "issue:43540", "accept": false, - "reason": "Different multimodal model failures with no shared concrete fix path." + "reason": "Qwen3 auto device mapping CUDA assert and Qwen3OmniMoe video ValueError are different multimodal/device issues." }, { - "left": "issue:44849", - "right": "issue:45701", + "left": "issue:43540", + "right": "issue:44560", "accept": false, - "reason": "Different issues: Qwen3.5 hidden-state bug vs tokenization version drift." + "reason": "Both concern Qwen3 video handling, but one is a processing ValueError in Qwen3OmniMoe and the other is a StopIteration in Qwen3-vl-embedding; different model families and failure points." }, { - "left": "issue:44857", - "right": "issue:46821", + "left": "issue:44479", + "right": "issue:45381", "accept": false, - "reason": "Unrelated: DETR AMP/CUDA crash vs TimesFM window_size attribute error." + "reason": "The v5 video input regression and wrong vision_position_ids are both Qwen VL-related, but the reported bugs are distinct and target different internal computations." }, { - "left": "issue:44936", - "right": "issue:44964", + "left": "issue:44568", + "right": "issue:45488", "accept": false, - "reason": "Different domains: trainer post-train evaluate bug vs multimodal model loading failure." + "reason": "mdeberta-v3-base add_special_tokens behavior and LlamaTokenizer ByteLevel/Metaspace override are different tokenizer regressions with different tokenization stacks." } ] }, @@ -7163,7 +7247,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 23, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -7176,13 +7260,13 @@ "issue:34567", "issue:36331", "issue:38175", + "issue:42175", "issue:43257", "issue:43452", "issue:43531", "issue:43540", "issue:43577", "issue:43582", - "issue:43697", "issue:43716", "issue:43742", "issue:43756", @@ -7191,26 +7275,25 @@ "issue:44291", "issue:44479", "issue:44488", - "issue:44560", "issue:44568", "issue:44589", "issue:44625", "issue:44737", + "issue:44792", "issue:44821", "issue:44877", "issue:44898", - "issue:44913", + "issue:44964", "issue:45042", "issue:45072", "issue:45081", "issue:45325", - "issue:45381", "issue:45405", "issue:45600", + "issue:45663", "issue:45672", "issue:45674", "issue:45823", - "issue:46121", "issue:46399", "issue:46463", "issue:46468", @@ -7218,241 +7301,227 @@ "issue:46693", "issue:46701", "issue:46821", - "issue:46829" + "issue:46846" ], "soft_pairs": [ - "issue:43540|issue:44560", - "issue:44291|issue:44821", - "issue:38175|issue:44898", - "issue:44479|issue:45381", - "issue:43540|issue:44479", + "issue:44792|issue:44964", "issue:34567|issue:36331", + "issue:42175|issue:44488", + "issue:38175|issue:44898", "issue:43257|issue:43756", + "issue:43540|issue:44479", "issue:44062|issue:45081", + "issue:44291|issue:44821", "issue:45042|issue:45081", - "issue:43742|issue:44589", - "issue:43582|issue:46821", "issue:43756|issue:46399", "issue:43756|issue:45325", + "issue:43582|issue:46821", + "issue:43742|issue:44589", + "issue:43716|issue:46701", "issue:44568|issue:46491", "issue:45405|issue:45600", - "issue:43531|issue:44821", "issue:45672|issue:45674", - "issue:43716|issue:46701", + "issue:43906|issue:46468", + "issue:43531|issue:44821", + "issue:45823|issue:46463", "issue:43577|issue:45072", + "issue:43906|issue:46693", "issue:44737|issue:46463", "issue:24643|issue:29870", - "issue:43906|issue:46468", - "issue:43906|issue:46693", - "issue:45823|issue:46463", "issue:43452|issue:44488", "issue:44625|issue:44877", - "issue:43540|issue:46693", - "issue:43697|issue:46829", - "issue:44913|issue:46121" + "issue:45663|issue:46846" ], - "prepared_review_unit_hash": "7ce86311eda82eb360aa59a78c0acb9cb957e163a2a07f89feb9eaec04d7998b", + "prepared_review_unit_hash": "15937fd2e9b9b52376e6ac398a8249f56a1dac73c0e39845e9819035aa9c1194", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36737, - "estimated_input_tokens": 9185, - "estimated_eval_tokens": 18626 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35534, + "estimated_input_tokens": 8884, + "estimated_eval_tokens": 18024 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "A heterogeneous set of closed/open issue reports with only loose topical overlap. Several pairs share a subsystem or model family, but the concrete failures differ enough that none should be merged as duplicates.", - "confidence": 0.91, - "canonical_issue_reason": "Issue 44479 is the broadest and most representative of the closest family here: a concrete Qwen video-input regression affecting multiple related models in v5.3.0.", + "summary": "This cluster is mostly a loose collection of unrelated transformer bug reports, not a duplicate set. No soft-edge pair is strong enough to merge under strict triage; the closest pair is the Qwen video-input regression reports, but they are still too model/regression-specific to confidently treat as one issue.", + "confidence": 0.89, + "canonical_issue_reason": "issue:44479 is the broadest, most central report in the set: it describes a concrete v5.3.0 regression affecting multiple Qwen multimodal video models, making it the best representative issue if one must be chosen.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44479 is the best single representative issue because it describes a specific, user-visible regression with a clear scope and affected model family.", + "best_issue_reason": "issue:44479 has the clearest cross-model scope and strongest duplicate-triage value among the issues here; it reads like the most actionable umbrella report for a real regression.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43540", - "right": "issue:44560", + "left": "issue:44792", + "right": "issue:44964", "accept": false, - "reason": "Both are Qwen3 multimodal/video failures, but one is a Qwen3OmniMoe ValueError and the other is a Qwen3-vl-embedding StopIteration; different code paths and likely different fixes." + "reason": "Different models and failures: janus test failure vs Phi-4 multimodal load error. No shared code-path bug is evident." }, { - "left": "issue:44291", - "right": "issue:44821", + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "TrainerState token counting bug vs custom compute_loss signature incompatibility. Related to training, but not the same defect." + }, + { + "left": "issue:42175", + "right": "issue:44488", "accept": false, - "reason": "These are unrelated loading problems: init_empty_weights/_is_hf_initialized in model instantiation versus AutoImageProcessor URL loading." + "reason": "Packaging/backend extra issue vs tokenizer/model loading failure for a specific model. Different surfaces and causes." }, { "left": "issue:38175", "right": "issue:44898", "accept": false, - "reason": "Both concern vision models, but one reports unexpected zero probabilities in SigLIP2 and the other is a Perceiver image-classification resize/interpolation failure." + "reason": "Both are vision-model issues, but one is zero-probabilities on SigLIP2 and the other is interpolation failure on Perceiver. Not the same bug." }, { - "left": "issue:44479", - "right": "issue:45381", + "left": "issue:43257", + "right": "issue:43756", "accept": false, - "reason": "Both involve Qwen2.5-VL/Qwen video handling, but one is a broad v5.3.0 regression across several models while the other is specifically wrong vision_position_ids for video input." + "reason": "Different models and failure modes: DeepSpeed/accelerate MoE weight conversion vs SmolLM3 RoPE-layer mismatch." }, { "left": "issue:43540", "right": "issue:44479", "accept": false, - "reason": "Same general Qwen video area, but the symptoms and affected models differ: Qwen3OmniMoe input processing versus a multi-model regression in qwen2_5_vl/qwen3_vl/qwen3_5." + "reason": "Both involve video inputs in Qwen-family models, but one is Qwen3OmniMoe-specific and the other is a broader v5.3.0 regression across several Qwen VL variants. Similar area, not confidently the same underlying bug." }, { - "left": "issue:34567", - "right": "issue:36331", + "left": "issue:44062", + "right": "issue:45081", "accept": false, - "reason": "Training-token accounting in TrainerState is unrelated to the compute_loss signature change causing the unexpected num_items_in_batch TypeError." + "reason": "Tokenizer initialization collision vs Mistral regex patch crash. Both tokenizer-adjacent, but distinct code paths and symptoms." }, { - "left": "issue:43257", - "right": "issue:43756", + "left": "issue:44291", + "right": "issue:44821", "accept": false, - "reason": "Both mention Qwen/MoE-like model families, but one is a DeepSpeed weight-conversion issue and the other is a Smollm3 RoPE-layer count mismatch." + "reason": "init_empty_weights constructor incompatibility vs AutoImageProcessor URL loading. Unrelated failures." }, { - "left": "issue:44062", + "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "These are different tokenizer bugs: AddedToken special-argument duplication versus a Mistral regex patch assuming backend_tokenizer exists." + "reason": "Image processor torchvision dependency regression vs Mistral tokenizer regex crash. Different subsystems." }, { - "left": "issue:45042", - "right": "issue:45081", + "left": "issue:43756", + "right": "issue:46399", "accept": false, - "reason": "One is a PIL image-processor torchvision dependency regression; the other is a tokenizer regex patch AttributeError." + "reason": "Different Qwen/SmolLM issues: RoPE layer omission vs missing decorator for kernel registration. Not mergeable as one bug." }, { - "left": "issue:43742", - "right": "issue:44589", + "left": "issue:43756", + "right": "issue:45325", "accept": false, - "reason": "Loading facebook/MobileLLM-125M key handling is unrelated to the Float8_e4m3fnStorage lookup error." + "reason": "Both mention rotary-position logic, but one is missing RoPE layers in SmolLM3 and the other is a Qwen2.5-VL temporal position scaling bug. Different fixes." }, { "left": "issue:43582", "right": "issue:46821", "accept": false, - "reason": "Apple Silicon caching_allocator_warmup TypeError and TimesFM window_size AttributeError are unrelated runtime API issues." + "reason": "Apple Silicon allocator warmup TypeError vs TimesFM window_size attribute error. No shared underlying defect." }, { - "left": "issue:43756", - "right": "issue:46399", + "left": "issue:43742", + "right": "issue:44589", "accept": false, - "reason": "Both touch RoPE-related areas, but one is a layer-selection/configuration bug in Smollm3 and the other is a missing decorator for apply_rotary_pos_emb registration in Qwen3.5 kernels." + "reason": "MobileLLM load-key issue vs float8 storage lookup failure. Different loading/runtime problems." }, { - "left": "issue:43756", - "right": "issue:45325", + "left": "issue:43716", + "right": "issue:46701", "accept": false, - "reason": "Different bugs in different models: Smollm3 RoPE layer dropping versus Qwen2.5-VL still-image temporal position scaling." + "reason": "Mistral-3 image preprocessor dtype mismatch vs DiffusionGemma pipeline breakage. Different models and code paths." }, { "left": "issue:44568", "right": "issue:46491", "accept": false, - "reason": "Tokenizer regressions, but not the same one: mdeberta special-token addition versus CodeLlama leading-space round-trip loss." + "reason": "Tokenizer regressions, but one is BOS/EOS insertion behavior and the other is encode-decode whitespace preservation. Not the same bug." }, { "left": "issue:45405", "right": "issue:45600", "accept": false, - "reason": "Both are config/versioning related, but one is a PEFT minimum-version bump and the other is stale auto_mappings references to removed Sam3LiteText configs." - }, - { - "left": "issue:43531", - "right": "issue:44821", - "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and AutoImageProcessor URL loading are unrelated." + "reason": "Dependency version bump problem vs auto_mappings referencing removed configs. Different maintenance issues." }, { "left": "issue:45672", "right": "issue:45674", "accept": false, - "reason": "Both affect 4-bit inference, but one is a Gemma4 audio-module torch.finfo(uint8) TypeError and the other is a BitsAndBytesConfig skip-module exclusion regression." + "reason": "Both concern 4-bit inference, but one is a Gemma4 audio-module torch.finfo uint8 issue and the other is BitsAndBytesConfig skip-module handling. Not the same concrete bug." }, { - "left": "issue:43716", - "right": "issue:46701", + "left": "issue:43906", + "right": "issue:46468", "accept": false, - "reason": "Mistral-3 image-preprocessor dtype mismatch is unrelated to the DiffusionGemmaGenerationOutput pipeline failure." + "reason": "An isolated reproduction of an old issue vs a broken Nemotron example. No duplicate signal." }, { - "left": "issue:43577", - "right": "issue:45072", + "left": "issue:43531", + "right": "issue:44821", "accept": false, - "reason": "Both are dtype-related, but one is a Blip2 loading dtype propagation bug and the other is a CI inference mismatch in SwitchTransformers/TimmWrapperModel." + "reason": "Qwen3-MoE sliding-window behavior vs image processor URL loading. Unrelated." }, { - "left": "issue:44737", + "left": "issue:45823", "right": "issue:46463", "accept": false, - "reason": "XLNet CPU allocation in relative positional encoding is unrelated to the T5 decoder-input validation error message issue." + "reason": "Gemma4 device-placement bug vs T5 decoder-input validation message. Different model families and defects." }, { - "left": "issue:24643", - "right": "issue:29870", + "left": "issue:43577", + "right": "issue:45072", "accept": false, - "reason": "DeepSpeed training RuntimeError about weight dimensionality is unrelated to the Bart generate() decoder_attention_mask UnboundLocalError." + "reason": "Both are dtype-related, but one is Blip2 dtype propagation and the other is CI inference mismatches in other models. Not the same issue." }, { "left": "issue:43906", - "right": "issue:46468", + "right": "issue:46693", "accept": false, - "reason": "The isolated reproduction of issue 38071 is not the same bug as the broken official Nemotron example." + "reason": "Reproduction of a prior issue vs FA2 long-context performance regression. Different problems." }, { - "left": "issue:43906", - "right": "issue:46693", + "left": "issue:44737", + "right": "issue:46463", "accept": false, - "reason": "One is a reproduction of a specific issue, while the other is a performance regression in long-context generation with FA2." + "reason": "XLNet CPU placement bug vs T5 error-message issue. Not related." }, { - "left": "issue:45823", - "right": "issue:46463", + "left": "issue:24643", + "right": "issue:29870", "accept": false, - "reason": "Gemma4 device_map='auto' PLE device mismatch and T5 decoder validation are different failures in different stacks." + "reason": "DeepSpeed training runtime error vs Bart generate-time UnboundLocalError. Different failures and subsystems." }, { "left": "issue:43452", "right": "issue:44488", "accept": false, - "reason": "Both concern loading/tokenizer/model-from-pretrained behavior, but gguf_file breakage and inability to load cjvt/sleng-bert are distinct problems." + "reason": "gguf_file loading breakage for tokenizer/model APIs vs a model loading failure for sleng-bert. Distinct root causes." }, { "left": "issue:44625", "right": "issue:44877", "accept": false, - "reason": "Qwen3.5 num_labels propagation and strict config blocking granite_speech loading are separate config-handling bugs." - }, - { - "left": "issue:43540", - "right": "issue:46693", - "accept": false, - "reason": "Both mention Qwen/model generation performance areas only loosely; one is a multimodal input ValueError and the other is a FA2 long-context performance regression." + "reason": "Config propagation bug in Qwen3.5 vs strict config rejection for granite_speech. Different config paths." }, { - "left": "issue:43697", - "right": "issue:46829", - "accept": false, - "reason": "RTDetrV2 output divergence in v5 and VideoPrism hidden_states misassignment are unrelated model-output bugs." - }, - { - "left": "issue:44913", - "right": "issue:46121", + "left": "issue:45663", + "right": "issue:46846", "accept": false, - "reason": "GPTNeoX rotary_pct reload behavior and rope-validation list handling are different configuration serialization problems." + "reason": "Gemma4 FSDP2 attention state bug vs Qwen3.5 TP-plan missing linear_attn entries. Both are parallelization-related, but not the same bug." } ] }, @@ -7468,7 +7537,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 24, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -7484,6 +7553,7 @@ "issue:43452", "issue:43525", "issue:43526", + "issue:43540", "issue:43643", "issue:43653", "issue:43688", @@ -7501,244 +7571,255 @@ "issue:44811", "issue:44843", "issue:44898", + "issue:44913", "issue:44936", "issue:45005", "issue:45200", "issue:45216", + "issue:45356", "issue:45357", "issue:45372", "issue:45375", "issue:45479", - "issue:45542", "issue:45588", - "issue:45663", "issue:45920", "issue:46050", + "issue:46121", "issue:46170", - "issue:46178", "issue:46463", + "issue:46693", "issue:46701", "issue:46829", "issue:46832", - "issue:46846", "issue:46899", "issue:46932" ], "soft_pairs": [ - "issue:45663|issue:46846", + "issue:43697|issue:46829", + "issue:43540|issue:46693", "issue:45357|issue:45920", - "issue:44737|issue:45588", + "issue:44913|issue:46121", "issue:28282|issue:32090", "issue:46463|issue:46899", + "issue:44737|issue:45588", "issue:43697|issue:45216", "issue:43526|issue:45479", "issue:43720|issue:44898", "issue:43643|issue:45375", - "issue:45200|issue:46463", "issue:44466|issue:45005", - "issue:43761|issue:46829", - "issue:44361|issue:44442", + "issue:45200|issue:46463", "issue:44493|issue:46050", + "issue:44361|issue:44442", + "issue:43761|issue:46829", "issue:43653|issue:44361", "issue:41762|issue:44936", "issue:43452|issue:44843", - "issue:38175|issue:43295", "issue:43688|issue:43749", + "issue:38175|issue:43295", "issue:43819|issue:44811", "issue:43994|issue:44898", - "issue:43525|issue:46701", - "issue:46832|issue:46932", "issue:45372|issue:46170", - "issue:45542|issue:46178" + "issue:46832|issue:46932", + "issue:43525|issue:46701", + "issue:43295|issue:45356" ], - "prepared_review_unit_hash": "d295babaa73bfeeb9bd08f44a966b4c713b2605971edee2b60fa568422f7d22b", + "prepared_review_unit_hash": "d4e1bcf03dd5de3a5d72d1f77a3c0e32891a9fc4df732ef12319f6bdc1426886", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 25, - "serialized_chars": 35356, - "estimated_input_tokens": 8839, - "estimated_eval_tokens": 17934 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 36474, + "estimated_input_tokens": 9119, + "estimated_eval_tokens": 18494 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most pairs are clearly different bugs across unrelated models or subsystems. The only plausible duplicate edge is the tied-weights regression in v5: #44466 is a concrete symptom of the broader issue described in #45005.", - "confidence": 0.79, - "canonical_issue_reason": "#45005 is the broadest, most representative issue: it frames the underlying tied-weights regression in v5 at the model-family level, while #44466 reads as a specific device-dependent serialization symptom of that same problem.", + "summary": "These issues are mostly unrelated regression reports across different models/features. I found no soft-edge pairs that are strong enough to merge as duplicates based on title-level evidence alone.", + "confidence": 0.88, + "canonical_issue_reason": "issue:46832 is the most concrete and actionable report here: it names a specific model, a specific failing path (`RTDetrModel.forward`), and a precise trigger (`num_feature_levels > len(backbone outputs)`).", "canonical_pr_reason": null, - "best_issue_reason": "#45005 best captures the shared underlying bug and is the better umbrella issue for deduping #44466.", + "best_issue_reason": "issue:46832 is the best representative issue because it is specific, reproducible, and currently open, making it a cleaner anchor than the broader regression reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:45663", - "right": "issue:46846", + "left": "issue:43697", + "right": "issue:46829", "accept": false, - "reason": "Different models and different failure modes: Gemma4 FSDP2 KeyError vs Qwen3.5 TP plan OOM/shape error." + "reason": "Different models and failures: RTDetrV2 output divergence vs VideoPrism hidden_states misassignment." }, { - "left": "issue:45357", + "left": "issue:43540", + "right": "issue:46693", + "accept": false, + "reason": "Unrelated components: Qwen3OmniMoe video-input validation vs a FlashAttention long-context performance regression." + }, + { + "left": "issue:45356", "right": "issue:45920", "accept": false, - "reason": "Different regressions: Qwen3.5 visual encoder save_pretrained keys vs broad AutoTokenizer token-ID mismatch across models." + "reason": "Both are tokenizer regressions, but they describe different failure modes and likely different code paths." }, { - "left": "issue:44737", - "right": "issue:45588", + "left": "issue:44913", + "right": "issue:46121", "accept": false, - "reason": "Unrelated code paths: XLNet CPU device bug vs flash_attention crashing on sink-less models." + "reason": "GPTNeoX config reload issue is unrelated to rope-validation type handling." }, { "left": "issue:28282", "right": "issue:32090", "accept": false, - "reason": "Completely different errors: missing PyTorch import vs Trainer GPU broadcast TypeError." + "reason": "PyTorch import missing vs Trainer GPU broadcast TypeError are different problems." }, { "left": "issue:46463", "right": "issue:46899", "accept": false, - "reason": "Different models and defects: T5 decoder-input validation message vs Gemma4 4-bit multimodal dtype casting." + "reason": "T5 decoder-input validation message vs Gemma4 quantized multimodal embedder dtype casting are unrelated." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "XLNet CPU tensor allocation bug is unrelated to flash_attention s_aux handling." }, { "left": "issue:43697", "right": "issue:45216", "accept": false, - "reason": "Different areas: RT-DETR output regression vs Qwen3.5 checkpoint save_pretrained corruption." + "reason": "Different symptoms and subsystems: RTDetrV2 output mismatch vs Qwen3.5 save_pretrained regression." }, { "left": "issue:43526", "right": "issue:45479", "accept": false, - "reason": "Different bug classes: BEiT image processor label reduction vs sequence-classification zero-loss behavior." + "reason": "BeitImageProcessorFast label reduction bug is unrelated to sequence-classification loss degeneracy." }, { "left": "issue:43720", "right": "issue:44898", "accept": false, - "reason": "Different models and mechanisms: BitNet packed-weight loading vs Perceiver interpolation failure." + "reason": "BitNet packed-weight loading bug vs Perceiver interpolation failure are different code paths." }, { "left": "issue:43643", "right": "issue:45375", "accept": false, - "reason": "Both are about missing config fields, but the titles point to different concrete mechanisms and config paths; not enough to treat as the same bug." + "reason": "AutoConfig remote-code field loss and Qwen3_5MoeVisionConfig strict-field dropping are similar in theme but not the same concrete bug." }, { - "left": "issue:45200", - "right": "issue:46463", + "left": "issue:44466", + "right": "issue:45005", "accept": false, - "reason": "Different issues: Gemma 4 text-only token-type defaults vs T5 decoder-input error messaging." + "reason": "Both mention tied weights, but one is device-dependent lm_head serialization and the other is translation-model behavior; not enough to merge as the same bug." }, { - "left": "issue:44466", - "right": "issue:45005", - "accept": true, - "reason": "Same tied-weights regression in v5: #44466 is a specific device-dependent serialization symptom, while #45005 describes the broader tied-weights problem on translation models." + "left": "issue:45200", + "right": "issue:46463", + "accept": false, + "reason": "Gemma 4 multimodal token-type IDs are unrelated to T5 decoder-input validation." }, { - "left": "issue:43761", - "right": "issue:46829", + "left": "issue:44493", + "right": "issue:46050", "accept": false, - "reason": "Both mention hidden_states, but the models and output bugs differ: CLIPVision forward vs VideoPrism classification output." + "reason": "Position-id key warnings and sinq save/load quantization breakage are different issues." }, { "left": "issue:44361", "right": "issue:44442", "accept": false, - "reason": "Different tokenizer failures with different models and APIs." + "reason": "Different tokenizer failures affecting different tokenizers and APIs." }, { - "left": "issue:44493", - "right": "issue:46050", + "left": "issue:43761", + "right": "issue:46829", "accept": false, - "reason": "Different serialization/weight-format problems affecting different model families." + "reason": "CLIPVisionModel hidden_states regression is not the same as VideoPrism output field mispopulation." }, { "left": "issue:43653", "right": "issue:44361", "accept": false, - "reason": "BigBird special-token decode bug is unrelated to MLukeTokenizer task AttributeError." + "reason": "BigBird special-token registration bug and MLukeTokenizer task AttributeError are unrelated." }, { "left": "issue:41762", "right": "issue:44936", "accept": false, - "reason": "Different training/evaluation problems: Gemma3 ZeRO-3 load failure vs trainer.evaluate after train." + "reason": "Gemma3 ZeRO-3 loading crash and trainer.evaluate-after-train failure are not the same code-path." }, { "left": "issue:43452", "right": "issue:44843", "accept": false, - "reason": "Both involve from_pretrained, but one is gguf_file handling and the other is offline model_info() access in a regex patch." + "reason": "gguf_file handling and offline model_info() lookup in Mistral regex patch are different tokenizer-loading regressions." }, { - "left": "issue:38175", - "right": "issue:43295", + "left": "issue:43688", + "right": "issue:43749", "accept": false, - "reason": "Unrelated symptoms: Siglip2 zero probabilities vs processor.tokenizer / image-tokenizer regression." + "reason": "Aux-loss normalization bug in MoE models is unrelated to FSDP CPU RAM efficient loading." }, { - "left": "issue:43688", - "right": "issue:43749", + "left": "issue:38175", + "right": "issue:43295", "accept": false, - "reason": "Different subsystems: auxiliary loss normalization vs FSDP CPU RAM efficient loading." + "reason": "Siglip2 zero-probability issue and processor.tokenizer regression are different multimodal bugs." }, { "left": "issue:43819", "right": "issue:44811", "accept": false, - "reason": "DAC latent mismatch is unrelated to Whisper batch_decode skip_special_tokens handling." + "reason": "DAC latent reconstruction mismatch is unrelated to Whisper batch_decode skip_special_tokens behavior." }, { "left": "issue:43994", "right": "issue:44898", "accept": false, - "reason": "Different models and failures: Siglip2 nonsensical outputs vs Perceiver resolution/interpolation bug." + "reason": "Siglip2 nonsensical outputs vs Perceiver interpolation failure are different models and failure modes." }, { - "left": "issue:43525", - "right": "issue:46701", + "left": "issue:45372", + "right": "issue:46170", "accept": false, - "reason": "Different model families and failure modes: Llama4Config missing field vs DiffusionGemma pipeline issue." + "reason": "Gemma 4 processor import breakage is unrelated to Windows ROCm torch.distributed dependency issues." }, { "left": "issue:46832", "right": "issue:46932", "accept": false, - "reason": "Different RT-DETR shape bug vs Gemma4 quantized embedder dtype-casting fix." + "reason": "RT-DETR forward TypeError and Gemma4 quantized embedders are completely different bugs." }, { - "left": "issue:45372", - "right": "issue:46170", + "left": "issue:43525", + "right": "issue:46701", "accept": false, - "reason": "Different import/runtime dependency issues: Gemma4 processor import vs Windows ROCm torch.distributed dependency." + "reason": "Llama4Config pad_token_id and DiffusionGemma generation pipeline failure are unrelated." }, { - "left": "issue:45542", - "right": "issue:46178", + "left": "issue:43295", + "right": "issue:45356", "accept": false, - "reason": "Both are backend-related, but they concern different missing backends and error paths, not the same underlying bug." + "reason": "Processor.tokenizer regression and Kimi tokenizer codec/regex regression are different problems." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The only accepted duplicate edge, #44466 \u2194 #45005, is supported by the titles as a narrow symptom vs. broader tied-weights v5 regression. The other pairwise rejections are justified by clearly different models, subsystems, or failure modes, without overstating evidence." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -7747,7 +7828,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 25, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -7762,7 +7843,6 @@ "issue:36010", "issue:41720", "issue:42898", - "issue:43295", "issue:43452", "issue:43504", "issue:43531", @@ -7790,13 +7870,14 @@ "issue:44843", "issue:44977", "issue:44991", - "issue:45356", "issue:45440", "issue:45491", + "issue:45542", "issue:45701", "issue:45704", "issue:46002", "issue:46050", + "issue:46178", "issue:46402", "issue:46491", "issue:46612", @@ -7804,22 +7885,22 @@ "issue:46821" ], "soft_pairs": [ - "issue:43531|issue:43742", - "issue:43295|issue:45356", "issue:29159|issue:33357", - "issue:42898|issue:46491", + "issue:43531|issue:43742", "issue:43653|issue:44568", + "issue:45542|issue:46178", "issue:43646|issue:43749", + "issue:42898|issue:46491", + "issue:43688|issue:45704", "issue:43756|issue:46402", "issue:44843|issue:44991", - "issue:43688|issue:45704", + "issue:44464|issue:44743", "issue:43653|issue:43927", "issue:43452|issue:43577", - "issue:44464|issue:44743", - "issue:44484|issue:44704", "issue:43688|issue:45701", - "issue:45491|issue:46002", "issue:44479|issue:44625", + "issue:45491|issue:46002", + "issue:44484|issue:44704", "issue:44821|issue:46821", "issue:43688|issue:45440", "issue:33666|issue:36010", @@ -7827,208 +7908,208 @@ "issue:44466|issue:46050", "issue:44734|issue:46612", "issue:44977|issue:46693", - "issue:43504|issue:43720", "issue:43504|issue:43550", "issue:44464|issue:44977", + "issue:43504|issue:43720", "issue:41720|issue:44464", "issue:43577|issue:44554" ], - "prepared_review_unit_hash": "07b7f3dca1defb06d7ba08d2c9bc7ba61e29b5332e7fec7a9d429c0c7c959c01", + "prepared_review_unit_hash": "5097b97dc08548b296d88493c6cf95cccf7187d3d160360dad1601d84261f467", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 35609, - "estimated_input_tokens": 8903, - "estimated_eval_tokens": 18062 + "serialized_chars": 35566, + "estimated_input_tokens": 8892, + "estimated_eval_tokens": 18040 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: it mixes unrelated tokenizer regressions, model-loading failures, generation/cache bugs, and training/runtime issues. None of the soft pairs looks like the same underlying bug or a PR-mergable duplicate.", - "confidence": 0.94, - "canonical_issue_reason": "No single issue cleanly represents the cluster; the items span many unrelated subsystems and failure modes. If a representative had to be chosen, issue:29159 is the broadest tokenizer-related report, but it is not a true canonical for the set.", + "summary": "This cluster is highly heterogeneous: it mixes unrelated tokenizer regressions, model-loading failures, generation bugs, training/dispatch issues, and backend/device-specific crashes. None of the proposed soft pairs look like the same underlying bug or change closely enough to merge.", + "confidence": 0.95, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "No issue here is a strong global best-fit for the whole cluster. Most are too specific to their own model/path, and the tokenizer-related ones still diverge in root cause.", + "best_issue_reason": "Issue 45701 is the broadest tokenizer-regression report and the closest thing to a representative item, but it still only covers one slice of the cluster and does not subsume the many unrelated loading/training/generation bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:29159", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency vs a MacOS CLIP bus error are different failure modes and code paths." + }, { "left": "issue:43531", "right": "issue:43742", "accept": false, - "reason": "Different problems: Qwen3-MoE sliding_window behavior vs a MobileLLM load-time key error." + "reason": "Qwen3-MoE sliding_window behavior and MobileLLM key error are unrelated model-specific loading/runtime issues." }, { - "left": "issue:43295", - "right": "issue:45356", + "left": "issue:43653", + "right": "issue:44568", "accept": false, - "reason": "Both are regressions around tokenizer/processor behavior, but they involve different code paths and symptoms (processor.tokenizer/images vs Kimi-K2.5 codec and warning handling)." + "reason": "BigBirdTokenizer special-token registration and mdeberta add_special_tokens BOS/EOS behavior are different tokenizer bugs." }, { - "left": "issue:29159", - "right": "issue:33357", + "left": "issue:45542", + "right": "issue:46178", "accept": false, - "reason": "Tokenizer slow/fast inconsistency is unrelated to a MacOS bus error loading a CLIP model." + "reason": "TensorBoard/TensorFlow backend mismatch is unrelated to the tensorflow_text backend mapping error." }, { - "left": "issue:42898", - "right": "issue:46491", + "left": "issue:43646", + "right": "issue:43749", "accept": false, - "reason": "Both involve tokenization behavior changes, but one is about clean_up_tokenization_spaces semantics and the other is a CodeLlama round-trip leading-space regression." + "reason": "Custom model init breakage in v5 is not the same as FSDP CPU RAM efficient loading being broken." }, { - "left": "issue:43653", - "right": "issue:44568", + "left": "issue:42898", + "right": "issue:46491", "accept": false, - "reason": "Both are special-token bugs, but they affect different tokenizers and different failure modes." + "reason": "clean_up_tokenization_spaces behavior changes and CodeLlama leading-space round-trip regression are distinct tokenizer semantics bugs." }, { - "left": "issue:43646", - "right": "issue:43749", + "left": "issue:43688", + "right": "issue:45704", "accept": false, - "reason": "Custom model initialization breakage is not the same bug as FSDP CPU RAM efficient loading being broken." + "reason": "Auxiliary-loss normalization in MoE models is unrelated to T5's apex FusedRMSNorm memory leak." }, { "left": "issue:43756", "right": "issue:46402", "accept": false, - "reason": "Different models and failures: RoPE layer dropping vs text encoder weights not loading." + "reason": "Smollm3 RoPE-layer mismatch and Sam3Model text-encoder weight loading are different model implementation defects." }, { "left": "issue:44843", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading regressions, but one is an offline model_info() call issue and the other is a specific tokenizer load failure for EMBEDDIA/est-roberta." + "reason": "Offline model_info() call in AutoTokenizer is unrelated to est-roberta tokenizer loading failure." }, { - "left": "issue:43688", - "right": "issue:45704", + "left": "issue:44464", + "right": "issue:44743", "accept": false, - "reason": "Aux-loss normalization in MoE models is unrelated to a T5 RMSNorm memory leak." + "reason": "Chunked generation with compiled forward and Qwen3.5 recurrent-state reset are different generation/cache bugs." }, { "left": "issue:43653", "right": "issue:43927", "accept": false, - "reason": "Special-token registration in BigBirdTokenizer is a different bug from DiaConfig losing custom token IDs on save/load." + "reason": "BigBird mask-token registration and DiaConfig custom token ID loss are unrelated tokenizer/config serialization issues." }, { "left": "issue:43452", "right": "issue:43577", "accept": false, - "reason": "GGUF loading failure is unrelated to Blip2 dtype propagation." - }, - { - "left": "issue:44464", - "right": "issue:44743", - "accept": false, - "reason": "Both mention generation/cache, but one is compiled forward chunked generation inconsistency and the other is recurrent state reset in modular_qwen3_5." + "reason": "gguf_file loading failure and Blip2 dtype propagation to qformer are different model-loading bugs." }, { - "left": "issue:44484", - "right": "issue:44704", + "left": "issue:43688", + "right": "issue:45701", "accept": false, - "reason": "max_shard_size policy question is unrelated to AutoProcessor not forwarding kwargs to cached_file." + "reason": "MoE auxiliary-loss normalization is not the same as a general tokenization-version regression." }, { - "left": "issue:43688", - "right": "issue:45701", + "left": "issue:44479", + "right": "issue:44625", "accept": false, - "reason": "Both are about model behavior changes, but auxiliary-loss normalization and tokenization changes are unrelated." + "reason": "Video input regression for Qwen VL models and Qwen3.5 num_labels propagation affect different subsystems." }, { "left": "issue:45491", "right": "issue:46002", "accept": false, - "reason": "NaNs in Gemma3 attention embeddings are a different issue from NaNs in classification heads during checkout/init." + "reason": "Gemma3 NaN embeddings from sliding-window attention and classification-head NaNs after checkout/init are distinct numerical issues." }, { - "left": "issue:44479", - "right": "issue:44625", + "left": "issue:44484", + "right": "issue:44704", "accept": false, - "reason": "Video-input regression in Qwen models is unrelated to num_labels propagation in Qwen3.5 config." + "reason": "save_pretrained shard-size behavior and AutoProcessor kwargs forwarding are unrelated API concerns." }, { "left": "issue:44821", "right": "issue:46821", "accept": false, - "reason": "AutoImageProcessor URL loading and TimesFM window_size AttributeError are unrelated." + "reason": "AutoImageProcessor URL loading and TimesFM window_size AttributeError are unrelated loader/config bugs." }, { "left": "issue:43688", "right": "issue:45440", "accept": false, - "reason": "Different MoE/model divergence bugs with no shared concrete failure path." + "reason": "MoE auxiliary-loss normalization does not match DeepseekV3MoE divergence from remote implementation." }, { "left": "issue:33666", "right": "issue:36010", "accept": false, - "reason": "Multi-GPU Qwen2-VL training is unrelated to the GenerationMixin import error." + "reason": "Qwen2-VL multi-GPU training and GenerationMixin import failure are different issues with no shared code path." }, { "left": "issue:32090", "right": "issue:33666", "accept": false, - "reason": "Trainer NoneType broadcast failure is not the same as Qwen2-VL multi-GPU training support/issues." + "reason": "Trainer _gpu_broadcast_one NoneType error and Qwen2-VL multi-GPU training are separate distributed-training failures." }, { "left": "issue:44466", "right": "issue:46050", "accept": false, - "reason": "Serialization of tied lm_head weights is unrelated to quantized sinq save/load breakage." + "reason": "v5 lm_head.weight serialization inconsistency and sinq quantized save/load breakage are different persistence bugs." }, { "left": "issue:44734", "right": "issue:46612", "accept": false, - "reason": "Both involve cache/generation, but the concrete failures differ: KV cache continuation tensor indexing vs beam-search cache reorder being skipped." + "reason": "KV-cache continuation crash in serve and skipped beam-search cache reorder for certain models are not the same generation-path problem." }, { "left": "issue:44977", "right": "issue:46693", "accept": false, - "reason": "A correctness regression in flash-attention generation is different from a performance regression with FA2 long-context generation." - }, - { - "left": "issue:43504", - "right": "issue:43720", - "accept": false, - "reason": "Legacy field load failure in BEiT is unrelated to BitNet packed-weight unpacking during accelerate loading." + "reason": "Qwen3.5 flash-attention output bug and FA2 long-context performance regression differ in symptom and root cause." }, { "left": "issue:43504", "right": "issue:43550", "accept": false, - "reason": "Different load/runtime paths: BEiT legacy-field loading vs Bamba torch.compile+SDPA failure." + "reason": "Beit legacy-field loading failure and Bamba torch.compile SDPA failure are unrelated model-specific bugs." }, { "left": "issue:44464", "right": "issue:44977", "accept": false, - "reason": "Both are generation regressions, but one is compiled-forward chunking and the other is flash-attention behavior on Qwen3.5." + "reason": "Compiled chunked generation inconsistency and Qwen3.5 flash-attention generation failure affect different generation mechanisms." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Legacy field load failure for Beit and BitNet packed-weight unpacking during accelerate loading are different load-time defects." }, { "left": "issue:41720", "right": "issue:44464", "accept": false, - "reason": "Auto device mapping cudaErrorAssert on A800 is not the same bug as compiled-forward chunked generation inconsistency." + "reason": "Qwen3 auto device mapping cuda assert and compiled chunked generation inconsistency are unrelated runtime issues." }, { "left": "issue:43577", "right": "issue:44554", "accept": false, - "reason": "Blip2 dtype loading is unrelated to the MPS attention correctness issue." + "reason": "Blip2 dtype propagation and MPS attention correctness with mismatched value/query heads are different bugs." } ] }, @@ -8044,7 +8125,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 26, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -8071,7 +8152,6 @@ "issue:43643", "issue:43646", "issue:43688", - "issue:43701", "issue:43720", "issue:43782", "issue:43825", @@ -8084,9 +8164,9 @@ "issue:44514", "issue:44554", "issue:44561", + "issue:44568", "issue:44610", "issue:44625", - "issue:44849", "issue:44964", "issue:44977", "issue:45290", @@ -8097,250 +8177,237 @@ "issue:46192", "issue:46209", "issue:46365", + "issue:46489", "issue:46693", "issue:46710", "issue:46846" ], "soft_pairs": [ + "issue:43611|issue:46846", "issue:43643|issue:44561", "issue:43720|issue:43927", - "issue:43611|issue:46846", + "issue:43782|issue:46693", "issue:33666|issue:36331", - "issue:43606|issue:43927", "issue:43278|issue:43688", + "issue:43606|issue:43927", "issue:43504|issue:44610", - "issue:43782|issue:46693", "issue:43262|issue:43295", "issue:43720|issue:44361", - "issue:44514|issue:45290", - "issue:44964|issue:45405", - "issue:43525|issue:43572", "issue:43493|issue:43825", + "issue:44964|issue:45405", + "issue:44514|issue:45290", "issue:44977|issue:46209", "issue:44554|issue:46365", "issue:44625|issue:45584", + "issue:43525|issue:43572", "issue:43452|issue:44451", "issue:28218|issue:33357", "issue:45823|issue:46050", - "issue:42796|issue:44977", "issue:43994|issue:46192", + "issue:42796|issue:44977", "issue:43493|issue:43844", "issue:39692|issue:43295", "issue:43278|issue:43844", + "issue:44568|issue:46489", "issue:43646|issue:44466", - "issue:46209|issue:46710", - "issue:44849|issue:46209", - "issue:33357|issue:33666", - "issue:43278|issue:43701" + "issue:46209|issue:46710" ], - "prepared_review_unit_hash": "e80809bbec7e5671a45618938a8f9e08f961063f13b316407b36a66268d6c165", + "prepared_review_unit_hash": "ef99d0dbf73af3ffb5d295d5069ebb5bb6137310788e27c63df82b6acc80d5f5", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36741, - "estimated_input_tokens": 9186, - "estimated_eval_tokens": 18628 + "soft_pair_count": 28, + "serialized_chars": 36263, + "estimated_input_tokens": 9066, + "estimated_eval_tokens": 18388 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a loose collection of mostly unrelated Transformers issues; the soft pairs share at best broad subsystem overlap, not the same underlying bug or change. No canonical issue/PR stands out from the provided set.", - "confidence": 0.95, - "canonical_issue_reason": null, + "summary": "The cluster is highly heterogeneous: most soft pairs share only broad subsystem labels (tokenizer, config loading, multimodal processors, training) rather than the same concrete bug. I would not merge any of the proposed soft edges; they are too coarse to be true duplicates.", + "confidence": 0.22, + "canonical_issue_reason": "If forced to pick one issue, #28218 is the most generic and discussion-heavy tokenizer bug in the set, but it is only a weak representative because the cluster does not form a single duplicate family.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "#28218 is the best issue candidate by age and generality, but the overall cluster is too mixed for it to be a strong canonical duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43611", + "right": "issue:46846", + "accept": false, + "reason": "Different problems: base_model_prefix model loading vs Qwen3.5 tensor-parallel planning/OOM and shape errors." + }, { "left": "issue:43643", "right": "issue:44561", "accept": false, - "reason": "Both mention trust_remote_code, but one is missing config fields and the other is a v5 util removal regression; different root causes." + "reason": "Both involve trust_remote_code, but the concrete failures differ: missing returned fields vs removal of is_torch_fx_available." }, { "left": "issue:43720", "right": "issue:43927", "accept": false, - "reason": "Packed-weight loading in BitNet vs lost token IDs in DiaConfig are unrelated defects." + "reason": "Unrelated bugs: BitNet packed-weight loading vs DiaConfig losing custom token IDs." }, { - "left": "issue:43611", - "right": "issue:46846", + "left": "issue:43782", + "right": "issue:46693", "accept": false, - "reason": "Model loading via base_model_prefix and Qwen3.5 tensor-parallel plan entries are different code paths and failures." + "reason": "Different code paths and symptoms: weight_only loading error vs FA2 long-context performance regression." }, { "left": "issue:33666", "right": "issue:36331", "accept": false, - "reason": "Multi-GPU training setup vs CustomTrainer signature mismatch are unrelated issues." - }, - { - "left": "issue:43606", - "right": "issue:43927", - "accept": false, - "reason": "CPU offload device mismatch and DiaConfig token ID loss are different bugs." + "reason": "Multi-GPU training Qwen2-VL vs a Trainer.compute_loss signature regression; not the same underlying bug." }, { "left": "issue:43278", "right": "issue:43688", "accept": false, - "reason": "Evaluation dtype drift and auxiliary-loss normalization are separate problems." + "reason": "Different numerical issues: dtype drift between train/eval vs auxiliary-loss normalization." }, { - "left": "issue:43504", - "right": "issue:44610", + "left": "issue:43606", + "right": "issue:43927", "accept": false, - "reason": "Legacy-field model loading and processor/model image-size mismatch are unrelated." + "reason": "Different failures: CPU offload device mismatch vs config token ID persistence bug." }, { - "left": "issue:43782", - "right": "issue:46693", + "left": "issue:43504", + "right": "issue:44610", "accept": false, - "reason": "Weight-only loading error and FA2 long-context performance regression are distinct issues." + "reason": "Different models and bugs: legacy field load failure vs processor image-size mismatch." }, { "left": "issue:43262", "right": "issue:43295", "accept": false, - "reason": "Default sampling-rate handling and processor.tokenizer/images regression are different defects in the processor stack." + "reason": "Both are processor/chat-template regressions, but one is audio sampling-rate defaulting and the other is tokenizer/processor API breakage." }, { "left": "issue:43720", "right": "issue:44361", "accept": false, - "reason": "BitNet accelerate-loading failure and MLukeTokenizer task AttributeError are unrelated." + "reason": "BitNet loading issue and MLukeTokenizer AttributeError are unrelated." }, { - "left": "issue:44514", - "right": "issue:45290", + "left": "issue:43493", + "right": "issue:43825", "accept": false, - "reason": "Both hit apply_chat_template, but batched padding=False and tool-call/no-content crashes appear to be different edge cases." + "reason": "SigLIP2 implementation discrepancy vs pipeline error-message wording; not the same bug." }, { "left": "issue:44964", "right": "issue:45405", "accept": false, - "reason": "Phi-4 multimodal loading and an unreleased peft minimum-version bump are unrelated." - }, - { - "left": "issue:43525", - "right": "issue:43572", - "accept": false, - "reason": "Both are missing pad-token config fields, but on different models and with different attributes; not the same bug." + "reason": "Model loading failure vs a dependency-version bump issue; unrelated." }, { - "left": "issue:43493", - "right": "issue:43825", + "left": "issue:44514", + "right": "issue:45290", "accept": false, - "reason": "SigLIP2 implementation mismatch and an incorrect pipeline error message are unrelated." + "reason": "Both are apply_chat_template crashes, but the triggering inputs and affected logic differ." }, { "left": "issue:44977", "right": "issue:46209", "accept": false, - "reason": "Qwen3.5 flash-attention generation bug and FP8 inference failure on B200 are different failures." + "reason": "Qwen3.5 flash-attention generation bug vs FP8 inference on B200; different runtime issues." }, { "left": "issue:44554", "right": "issue:46365", "accept": false, - "reason": "MPS attention correctness and Trainer import conflicts with datasets are unrelated." + "reason": "MPS attention correctness bug vs Trainer import conflict with datasets; unrelated." }, { "left": "issue:44625", "right": "issue:45584", "accept": false, - "reason": "num_labels propagation and Whisper empty-transcription generation are unrelated." + "reason": "Different regressions: num_labels propagation vs Whisper empty-transcription generation failure." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both are missing token-id fields after v5, but they affect different configs and different missing attributes." }, { "left": "issue:43452", "right": "issue:44451", "accept": false, - "reason": "gguf_file loading regression and ScandiBERT load failure do not look like the same underlying problem." + "reason": "Both concern model loading, but one is gguf_file behavior and the other is a specific model compatibility failure." }, { "left": "issue:28218", "right": "issue:33357", "accept": false, - "reason": "Tokenizer spacing after added tokens is unrelated to a MacOS CLIP bus error." + "reason": "Tokenizer spacing bug and a Mac bus error on a CLIP model are unrelated." }, { "left": "issue:45823", "right": "issue:46050", "accept": false, - "reason": "Device mismatch in Gemma4 PLE forward and quantized save/load breakage are different bugs." + "reason": "Device-map forward mismatch vs quantized save/load breakage; not the same defect." }, { - "left": "issue:42796", - "right": "issue:44977", + "left": "issue:43994", + "right": "issue:46192", "accept": false, - "reason": "Tokenizer BPE marker handling and Qwen3.5 flash-attention generation are unrelated." + "reason": "SigLIP2 wrong outputs vs return_tensors being ignored in text kwargs; unrelated." }, { - "left": "issue:43994", - "right": "issue:46192", + "left": "issue:42796", + "right": "issue:44977", "accept": false, - "reason": "Bad SigLIP2 outputs and return_tensors being ignored are different issues." + "reason": "Tokenizer BPE-marker handling vs Qwen3.5 flash-attention generation failure; different bugs." }, { "left": "issue:43493", "right": "issue:43844", "accept": false, - "reason": "Model-implementation discrepancy and ZeRO-3 gradient growth are unrelated." + "reason": "SigLIP2 implementation discrepancy and DeepSpeed gradient growth are unrelated." }, { "left": "issue:39692", "right": "issue:43295", "accept": false, - "reason": "Documentation example errors and custom processor/tokenizer access regressions are different problems." + "reason": "Docs/example errors for SigLIP2 vs processor.tokenizer / image-passing regression; not the same issue." }, { "left": "issue:43278", "right": "issue:43844", "accept": false, - "reason": "dtype mismatch in eval and ZeRO-3 gradient instability are unrelated." + "reason": "Evaluation dtype change vs gradient explosion under ZeRO-3 are distinct numerical problems." + }, + { + "left": "issue:44568", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer special-token regressions, but they affect different tokenizers and different concrete behaviors." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "Custom model initialization breakage and inconsistent lm_head serialization are different code-path failures." + "reason": "Custom model initialization breakage vs inconsistent lm_head serialization; different code paths." }, { "left": "issue:46209", "right": "issue:46710", "accept": false, - "reason": "FP8 inference failure and accuracy regression after version upgrade are not the same bug." - }, - { - "left": "issue:44849", - "right": "issue:46209", - "accept": false, - "reason": "Qwen3.5 hidden-states bug and FP8 B200 inference failure are unrelated." - }, - { - "left": "issue:33357", - "right": "issue:33666", - "accept": false, - "reason": "MacOS bus error on CLIP and Qwen2-VL multi-GPU training are unrelated." - }, - { - "left": "issue:43278", - "right": "issue:43701", - "accept": false, - "reason": "Training/eval dtype drift and resume_from_checkpoint key mismatch are separate issues." + "reason": "FP8 inference failure on B200 and accuracy regression after upgrade are not the same underlying bug." } ] }, @@ -8356,7 +8423,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 27, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -8364,9 +8431,10 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-502", "nodes": [ - "issue:29942", - "issue:30333", + "issue:33357", + "issue:33666", "issue:43257", + "issue:43278", "issue:43299", "issue:43334", "issue:43504", @@ -8374,6 +8442,7 @@ "issue:43575", "issue:43606", "issue:43653", + "issue:43701", "issue:43824", "issue:43844", "issue:43873", @@ -8390,7 +8459,6 @@ "issue:44792", "issue:44843", "issue:44849", - "issue:44869", "issue:44964", "issue:45072", "issue:45200", @@ -8402,243 +8470,242 @@ "issue:45759", "issue:45812", "issue:46050", + "issue:46209", "issue:46285", "issue:46399", "issue:46424", - "issue:46489", "issue:46701", "issue:46710", - "issue:46752", "issue:46821", "issue:46832", "issue:46846" ], "soft_pairs": [ - "issue:44568|issue:46489", + "issue:44849|issue:46209", + "issue:33357|issue:33666", + "issue:44568|issue:45812", + "issue:43278|issue:43701", "issue:44488|issue:44843", - "issue:43927|issue:44361", "issue:44792|issue:45405", - "issue:44568|issue:45812", - "issue:43824|issue:44589", + "issue:43927|issue:44361", "issue:43575|issue:45405", "issue:44849|issue:45357", "issue:43504|issue:43606", + "issue:43824|issue:44589", "issue:43257|issue:43299", - "issue:43334|issue:43572", + "issue:44493|issue:46285", + "issue:45245|issue:46710", + "issue:46399|issue:46846", "issue:46821|issue:46832", - "issue:44589|issue:45520", "issue:43334|issue:45759", - "issue:45245|issue:46710", "issue:43606|issue:44112", - "issue:44493|issue:46285", - "issue:46399|issue:46846", - "issue:43572|issue:44291", + "issue:43334|issue:43572", "issue:43504|issue:43653", + "issue:44589|issue:45520", "issue:45072|issue:46832", + "issue:43572|issue:44291", "issue:44464|issue:46424", "issue:43937|issue:46701", "issue:43844|issue:44964", "issue:43873|issue:46050", - "issue:29942|issue:30333", - "issue:44869|issue:46752", "issue:45200|issue:45229" ], - "prepared_review_unit_hash": "0d3ce0247a381616012ab4929149b2c553dc0bc3899c7e37d54d3653e78f9b9a", + "prepared_review_unit_hash": "bc315ebd26f3a895244a1bc95e748798be2819ffbe4f70b42a351fb7bbd27ec8", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36278, - "estimated_input_tokens": 9070, - "estimated_eval_tokens": 18396 + "serialized_chars": 36146, + "estimated_input_tokens": 9037, + "estimated_eval_tokens": 18330 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pairs look like true duplicates; they mostly describe different model-specific regressions, config breakages, or unrelated CI failures. I would keep them separate and not merge any PRs by these signals alone.", - "confidence": 0.41, - "canonical_issue_reason": "Issue 45812 is the broadest and most representative bug report in the set: a silent tokenizer regression affecting all Granite models after the v4\u2192v5 change, with clear repro details and multiple references.", + "summary": "The soft pairs are mostly unrelated issues that only share a model family or a broad subsystem (tokenization, multimodal configs, quantization, offload, etc.). I found no pair that clearly describes the same underlying bug/change, so none should be merged as duplicates.", + "confidence": 0.11, + "canonical_issue_reason": "Issue 45812 is the strongest standalone candidate: it describes a broad, concrete regression with clear user impact (wrong token IDs across all Granite models) and a focused failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45812 is the best single issue to anchor triage because it is wide in impact, clearly scoped to one regression, and has stronger community signal than the more model-specific reports.", + "best_issue_reason": "Issue 45812 is the most representative and actionable report in the set because it is broad, specific, and directly tied to a reproducible regression.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44568", - "right": "issue:46489", + "left": "issue:44849", + "right": "issue:46209", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different failure modes (BOS/EOS addition vs incorrect DeepSeek-Coder output); not the same bug." + "reason": "Different failures: Qwen3.5 hidden-state behavior vs FP8 inference on B200. Same general area, not the same bug." }, { - "left": "issue:44488", - "right": "issue:44843", + "left": "issue:33357", + "right": "issue:33666", "accept": false, - "reason": "One is a model load failure for cjvt/sleng-bert; the other is an offline-mode break in _patch_mistral_regex. Different code paths and symptoms." + "reason": "Completely different models and problems: MacOS bus error on CLIP vs Qwen2-VL multi-GPU training." }, { - "left": "issue:43927", - "right": "issue:44361", + "left": "issue:44568", + "right": "issue:45812", "accept": false, - "reason": "DiaConfig save/load token-ID loss causing generation IndexError is unrelated to MLukeTokenizer task AttributeError." + "reason": "Both are tokenizer regressions, but they affect different model families and symptoms; not one concrete code-path bug." }, { - "left": "issue:44792", - "right": "issue:45405", + "left": "issue:43278", + "right": "issue:43701", "accept": false, - "reason": "A failing janus generation test is unrelated to a PEFT minimum-version bump; one is a runtime bug, the other is dependency metadata." + "reason": "BF16/FP32 dtype mismatch in evaluation is unrelated to resume_from_checkpoint key mismatch." }, { - "left": "issue:44568", - "right": "issue:45812", + "left": "issue:44488", + "right": "issue:44843", "accept": false, - "reason": "Both are tokenizer regressions, but they target different model families and behaviors; not one underlying defect." + "reason": "Model loading failure for cjvt/sleng-bert is unrelated to an offline model_info() call in AutoTokenizer." }, { - "left": "issue:43824", - "right": "issue:44589", + "left": "issue:44792", + "right": "issue:45405", "accept": false, - "reason": "Qwen2.5-VL import error and Float8 storage lookup failure are distinct breakages with different causes." + "reason": "Janus test failure and a PEFT version bump are unrelated." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenization/config bugs with different failure modes: custom token IDs loss vs MLukeTokenizer AttributeError." }, { "left": "issue:43575", "right": "issue:45405", "accept": false, - "reason": "TP OOM on Qwen2-57B is unrelated to the unreleased PEFT version bump." + "reason": "OOM during tensor-parallel loading is unrelated to the PEFT minimum-version change." }, { "left": "issue:44849", "right": "issue:45357", "accept": false, - "reason": "Hidden-states behavior in Qwen3.5 and incorrect visual encoder keys on save_pretrained are different bugs." + "reason": "Both mention Qwen3.5, but one is output_hidden_states behavior and the other is incorrect save_pretrained visual keys." }, { "left": "issue:43504", "right": "issue:43606", "accept": false, - "reason": "Beit legacy-field loading and bark-small CPU-offload device mismatch are unrelated." + "reason": "Legacy field loading for BEiT is unrelated to Bark CPU-offload device mismatch." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "ImportError for a missing model class is unrelated to an FP8 storage TypeError." }, { "left": "issue:43257", "right": "issue:43299", "accept": false, - "reason": "Both mention Qwen3 MoE loading, but one is accelerate+deepspeed weight conversion and the other is a dev-version load regression for Qwen3VL; not the same concrete issue." + "reason": "Both concern Qwen MoE loading, but one is accelerate+deepspeed conversion and the other is a transformers dev compatibility break." }, { - "left": "issue:43334", - "right": "issue:43572", + "left": "issue:44493", + "right": "issue:46285", "accept": false, - "reason": "Both are missing config-token-field regressions, but they affect different configs and different missing fields (`pad_token_id` vs `pad_token_idx`)." + "reason": "Unexpected key warnings with position IDs are not the same as the CLIP text-model structure inconsistency." }, { - "left": "issue:46821", - "right": "issue:46832", + "left": "issue:45245", + "right": "issue:46710", "accept": false, - "reason": "TimesFM `window_size` attribute errors and RT-DETR feature-level mismatch are unrelated model-forward bugs." + "reason": "Category cardinality limit error is unrelated to an accuracy regression after upgrade." }, { - "left": "issue:44589", - "right": "issue:45520", + "left": "issue:46399", + "right": "issue:46846", "accept": false, - "reason": "Float8 storage lookup failure and flash_attn import_utils KeyError are separate environment/import issues." + "reason": "Different Qwen3.5 issues: kernel registration crash vs tensor-parallel plan omissions causing OOM/shape errors." }, { - "left": "issue:43334", - "right": "issue:45759", + "left": "issue:46821", + "right": "issue:46832", "accept": false, - "reason": "Both involve Qwen multimodal configs, but one is a missing text-config attribute and the other is `from_config` not unwrapping `text_config`; different failure points." + "reason": "TimesFM window_size AttributeError and RT-DETR forward shape mismatch are unrelated." }, { - "left": "issue:45245", - "right": "issue:46710", + "left": "issue:43334", + "right": "issue:45759", "accept": false, - "reason": "Category-cardinality RuntimeError and an accuracy regression after version upgrade are not the same bug." + "reason": "Both involve Qwen multimodal configs, but the specific bug differs: missing pad_token_id vs not unwrapping text_config." }, { "left": "issue:43606", "right": "issue:44112", "accept": false, - "reason": "CPU offload device mismatch in bark-small is unrelated to a stale GraniteSpeech CI test." + "reason": "CPU offload device mismatch in Bark is unrelated to a stale GraniteSpeech CI test." }, { - "left": "issue:44493", - "right": "issue:46285", + "left": "issue:43334", + "right": "issue:43572", "accept": false, - "reason": "Unexpected key/position-id loading issues are not the same as CLIP text-model structure divergence." + "reason": "Different config fields and models: pad_token_id in Qwen3-VL vs pad_token_idx in StableLmConfig." }, { - "left": "issue:46399", - "right": "issue:46846", + "left": "issue:43504", + "right": "issue:43653", "accept": false, - "reason": "Qwen3.5 kernel registration error and missing linear_attn TP-plan entries produce different breakages and would not be merged as one fix." + "reason": "BEiT legacy-field loading and BigBirdTokenizer special-token registration are unrelated." }, { - "left": "issue:43572", - "right": "issue:44291", + "left": "issue:44589", + "right": "issue:45520", "accept": false, - "reason": "Missing `pad_token_idx` after 5.0 and `_is_hf_initialized` argument errors are distinct config/init incompatibilities." + "reason": "FP8 storage lookup failure is unrelated to flash_attn import_utils behavior on Python 3.13." }, { - "left": "issue:43504", - "right": "issue:43653", + "left": "issue:45072", + "right": "issue:46832", "accept": false, - "reason": "Beit legacy-field loading and BigBirdTokenizer special-token registration are unrelated." + "reason": "dtype mismatches in inference are unrelated to RT-DETR's num_feature_levels/backbone-output error." }, { - "left": "issue:45072", - "right": "issue:46832", + "left": "issue:43572", + "right": "issue:44291", "accept": false, - "reason": "bf16 dtype mismatches in SwitchTransformers/TimmWrapperModel are unrelated to RT-DETR forward shape/type errors." + "reason": "StableLm missing field after 5.0 update is unrelated to the _is_hf_initialized argument error with init_empty_weights." }, { "left": "issue:44464", "right": "issue:46424", "accept": false, - "reason": "Chunked generation inconsistency with compiled forward is unrelated to cache-length handling for quantized cache." + "reason": "Compiled-forward chunked generation inconsistency is unrelated to max_cache_length behavior in quantized cache." }, { "left": "issue:43937", "right": "issue:46701", "accept": false, - "reason": "GLM-5 generation-config validation and DiffusionGemma pipeline failure are different issues." + "reason": "GLM-5 generation config validation is unrelated to the DiffusionGemma image-text-to-text pipeline failure." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "ZeRO-3 gradient growth during training and Phi-4 multimodal load failure are not the same problem." + "reason": "Training instability with ZeRO-3 is unrelated to Phi-4 multimodal loading failure." }, { "left": "issue:43873", "right": "issue:46050", "accept": false, - "reason": "Quantization/offloading behavior and quantized save/load corruption are both quantization-related but not the same concrete code-path." - }, - { - "left": "issue:29942", - "right": "issue:30333", - "accept": false, - "reason": "Flash Attention 2 test failures and MLFlow reporting on failed jobs are unrelated." - }, - { - "left": "issue:44869", - "right": "issue:46752", - "accept": false, - "reason": "Whisper timestamp decode crash and empty-list chat-template IndexError are different edge-case bugs." + "reason": "Offloading/quantization mismatch is not the same as save/load breakage for sinq-quantized models." }, { "left": "issue:45200", "right": "issue:45229", "accept": false, - "reason": "Gemma 4 token-type defaults and multi-GPU CUDA OOM are unrelated issues." + "reason": "Missing mm_token_type_ids defaulting is unrelated to Gemma4 multi-GPU inference OOM." } ] }, @@ -8654,7 +8721,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 28, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -8665,6 +8732,8 @@ "issue:15354", "issue:28218", "issue:29159", + "issue:29942", + "issue:30333", "issue:32090", "issue:34567", "issue:37428", @@ -8673,14 +8742,12 @@ "issue:42490", "issue:43065", "issue:43295", + "issue:43404", "issue:43450", "issue:43493", "issue:43540", - "issue:43550", - "issue:43572", "issue:43696", "issue:43701", - "issue:43844", "issue:43873", "issue:43957", "issue:43994", @@ -8690,6 +8757,7 @@ "issue:44554", "issue:44610", "issue:44734", + "issue:44869", "issue:44977", "issue:45072", "issue:45084", @@ -8708,12 +8776,14 @@ "issue:46369", "issue:46701", "issue:46710", - "issue:46821" + "issue:46752" ], "soft_pairs": [ + "issue:44869|issue:46752", + "issue:45229|issue:45823", "issue:43994|issue:44493", "issue:45588|issue:46050", - "issue:45229|issue:45823", + "issue:29942|issue:30333", "issue:42490|issue:43696", "issue:29159|issue:32090", "issue:28218|issue:34567", @@ -8721,214 +8791,212 @@ "issue:45084|issue:45520", "issue:43065|issue:46369", "issue:44554|issue:45446", - "issue:15354|issue:38175", "issue:44423|issue:44734", + "issue:15354|issue:38175", "issue:43493|issue:43701", "issue:41720|issue:46710", "issue:45406|issue:46121", - "issue:44977|issue:46710", + "issue:43404|issue:46710", "issue:45072|issue:45198", "issue:43295|issue:43450", - "issue:37428|issue:43957", "issue:45588|issue:45759", + "issue:37428|issue:43957", "issue:43540|issue:45200", "issue:43295|issue:43873", "issue:44186|issue:46192", - "issue:43572|issue:46821", - "issue:43295|issue:44610", "issue:45823|issue:46192", - "issue:43550|issue:43844" + "issue:43295|issue:44610" ], - "prepared_review_unit_hash": "d101cfb59d09bbf091d8b8ba9e7224c9993f326c26ca3454c3a7de75f1f5e08a", + "prepared_review_unit_hash": "4d264ab493e3a76e344d7e462ec2606e314320ba4699f92399d89c2b5500390c", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 27, - "serialized_chars": 35247, - "estimated_input_tokens": 8812, - "estimated_eval_tokens": 17880 + "serialized_chars": 35866, + "estimated_input_tokens": 8967, + "estimated_eval_tokens": 18190 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items do not form a single duplicate set; they span unrelated tokenizer, processor, serving, quantization, and model-specific regressions. None of the soft pairs look like the same underlying bug/change, so they should stay separate. If a representative issue is needed, issue 43295 is the broadest recent tokenizer/processor regression anchor.", - "confidence": 0.91, - "canonical_issue_reason": "No exact duplicate anchor stands out; issue 43295 is the broadest and most reusable triage target because it captures a recent regression in processor/tokenizer handling and custom-model integration.", + "summary": "The cluster is a loose collection of unrelated transformer issues rather than a duplicate set. None of the proposed soft pairs look like the same underlying bug/change. The most representative issue is the broad, well-referenced offloading/quantization report.", + "confidence": 0.94, + "canonical_issue_reason": "issue:43873 is the broadest and most centrally referenced report in the set (open, 6 inbound references), so it is the best representative issue for the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43295 is the best representative because it is broad, recent, regression-framed, and centered on a common API break that can absorb nearby processor/tokenizer reports.", + "best_issue_reason": "issue:43873 has the strongest signal for being the cluster anchor: it is general, actively discussed, and more reusable as a canonical placeholder than the model-specific or narrow bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44869", + "right": "issue:46752", + "accept": false, + "reason": "Different code paths and failures: Whisper timestamp decoding vs chat template handling on empty input." + }, + { + "left": "issue:45229", + "right": "issue:45823", + "accept": false, + "reason": "Both mention Gemma4, but one is a CUDA OOM and the other is a forward device-mismatch bug; not the same change." + }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "Different problems: SigLIP2 inference produces bad outputs, while the other is a key/position-id warning during model loading or forward." + "reason": "SigLIP2 bad outputs and unexpected position-id keys are unrelated symptoms with different likely causes." }, { "left": "issue:45588", "right": "issue:46050", "accept": false, - "reason": "One is a flash-attention crash on sink-less models; the other is a quantization save/load breakage in sinq. Different code paths and fixes." + "reason": "Flash-attention sink handling and sinq save/load breakage are distinct bugs in different subsystems." }, { - "left": "issue:45229", - "right": "issue:45823", + "left": "issue:29942", + "right": "issue:30333", "accept": false, - "reason": "Both involve Gemma4, but one is a multi-GPU OOM and the other is a device-map/forward device mismatch. Not the same bug." + "reason": "Flash Attention 2 test failures are unrelated to MLflow job-status reporting on training failure." }, { "left": "issue:42490", "right": "issue:43696", "accept": false, - "reason": "A trailing-slash import warning and a GPT-oss out-of-memory error are unrelated." + "reason": "Import warning for a trailing-slash path is unrelated to GPT-oss CUDA out-of-memory." }, { "left": "issue:29159", "right": "issue:32090", "accept": false, - "reason": "Tokenizer slow/fast inconsistency is unrelated to the Trainer NoneType broadcast error." + "reason": "Tokenizer slow/fast inconsistency does not match the Trainer _gpu_broadcast_one NoneType error." }, { "left": "issue:28218", "right": "issue:34567", "accept": false, - "reason": "Added-token spacing behavior is unrelated to TrainerState token-count tracking." + "reason": "Extra space after added token is a tokenizer formatting issue, not a TrainerState token-count update bug." }, { "left": "issue:44977", "right": "issue:46701", "accept": false, - "reason": "Different failures: Qwen3.5 flash-attention generation versus a DiffusionGemma pipeline output-type issue." + "reason": "Qwen3.5 flash-attention generation failure and DiffusionGemma image-text pipeline breakage are different problems." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "Non-template-node compile failure and flash_attn import-key error are separate issues." + "reason": "Template-node compilation error and flash_attn import KeyError are separate failures." }, { "left": "issue:43065", "right": "issue:46369", "accept": false, - "reason": "Sam3 pixel decoder dummy-layer construction is not the same as a memory overflow when invoking sam3-litetext." + "reason": "A dummy Conv2d in Sam3PixelDecoder is not the same as a memory overflow in sam3-litetext." }, { "left": "issue:44554", "right": "issue:45446", "accept": false, - "reason": "MPS attention correctness is unrelated to a PyTorch version check bug in flex_attention import gating." + "reason": "MPS attention correctness and a PyTorch version-check bug in flex_attention are unrelated." }, { - "left": "issue:15354", - "right": "issue:38175", + "left": "issue:44423", + "right": "issue:44734", "accept": false, - "reason": "TorchScript GeneratorExp export failure and SigLIP2 zero-probability outputs are unrelated." + "reason": "Continuous-batching multimodal crash and KV-cache continuation indexing error are different serve-path bugs." }, { - "left": "issue:44423", - "right": "issue:44734", + "left": "issue:15354", + "right": "issue:38175", "accept": false, - "reason": "Both are serve crashes, but one is a multimodal .to() conversion problem and the other is KV-cache continuation indexing. Different fixes." + "reason": "TorchScript GeneratorExp export failure and SigLIP2 zero-probability outputs are unrelated." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "SigLIP2 numerical discrepancy and resume-from-checkpoint key mismatch are unrelated." + "reason": "SigLIP2 HF-vs-JAX discrepancy and resume_from_checkpoint key mismatch do not share a concrete failure mode." }, { "left": "issue:41720", "right": "issue:46710", "accept": false, - "reason": "CUDA assert from auto device mapping is not the same as an accuracy regression after upgrading Transformers." + "reason": "Qwen3 auto device-mapping CUDA assert and DeepSeek accuracy regression after upgrade are different issues." }, { "left": "issue:45406", "right": "issue:46121", "accept": false, - "reason": "Gemma4Processor missing _tokenizer in serve is unrelated to rope-param validation on JSON-loaded lists." + "reason": "Gemma4Processor missing _tokenizer in serve and rope-param conversion TypeError are unrelated." }, { - "left": "issue:44977", + "left": "issue:43404", "right": "issue:46710", "accept": false, - "reason": "Generation failure with flash-attention and a DeepSeek accuracy regression are different code-path problems." + "reason": "Mistral3 weight tying and DeepSeek accuracy regression are separate model-behavior problems." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "bfloat16 dtype mismatch in CI is unrelated to Wav2Vec2 save_pretrained/tokenization failure." + "reason": "BF16 dtype mismatches in CI and Wav2Vec2 save/tokenization failure are not the same bug." }, { "left": "issue:43295", "right": "issue:43450", "accept": false, - "reason": "Processor/tokenizer API regression and batched video shape handling are different bugs, even though both touch preprocessing." + "reason": "Processor.tokenizer regression and batched video-shape handling are different multimodal processing bugs." }, { - "left": "issue:37428", - "right": "issue:43957", + "left": "issue:45588", + "right": "issue:45759", "accept": false, - "reason": "Flash-attention import failure and meta-device model loading breakage are separate issues." + "reason": "Flash-attention sink-less crash and Qwen multimodal text_config unwrapping are unrelated code paths." }, { - "left": "issue:45588", - "right": "issue:45759", + "left": "issue:37428", + "right": "issue:43957", "accept": false, - "reason": "A flash_attention crash on sink-less models is unrelated to AutoModelForCausalLM.from_config not unwrapping text_config." + "reason": "ImportError for _flash_supports_window_size and meta-device model loading breakage are distinct issues." }, { "left": "issue:43540", "right": "issue:45200", "accept": false, - "reason": "Video-input processing error in Qwen3OmniMoe is not the same as Gemma4 mm_token_type_ids defaulting." + "reason": "Video-input processing failure in Qwen3OmniMoe is unrelated to Gemma4 text-only mm_token_type_ids defaults." }, { "left": "issue:43295", "right": "issue:43873", "accept": false, - "reason": "Processor/tokenizer regression and quantization offloading behavior are unrelated." + "reason": "Processor.tokenizer regression and offloading/quantization behavior are different subsystems and failure modes." }, { "left": "issue:44186", "right": "issue:46192", "accept": false, - "reason": "LayoutLMv2 tokenizer crashes on NER/padding are distinct from return_tensors being ignored when text_kwargs is passed." - }, - { - "left": "issue:43572", - "right": "issue:46821", - "accept": false, - "reason": "StableLmConfig pad_token_idx absence and TimesFM window_size AttributeError are unrelated config issues." - }, - { - "left": "issue:43295", - "right": "issue:44610", - "accept": false, - "reason": "Processor/tokenizer regression is unrelated to OmDet-Turbo's image-size mismatch." + "reason": "LayoutLMv2Tokenizer NER/padding crash and return_tensors being ignored are separate tokenizer bugs." }, { "left": "issue:45823", "right": "issue:46192", "accept": false, - "reason": "Gemma4 device_map forwarding mismatch and return_tensors handling are different failures." + "reason": "Gemma4 device mismatch during forward is unrelated to return_tensors handling." }, { - "left": "issue:43550", - "right": "issue:43844", + "left": "issue:43295", + "right": "issue:44610", "accept": false, - "reason": "torch.compile SDPA failure and DeepSpeed ZeRO-3 gradient inflation are unrelated training/runtime bugs." + "reason": "Processor API regression and OmDet-Turbo image-size mismatch are different model-processing bugs." } ] }, @@ -8944,7 +9012,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 29, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -8964,15 +9032,15 @@ "issue:43262", "issue:43278", "issue:43295", - "issue:43334", "issue:43450", "issue:43452", "issue:43493", - "issue:43525", + "issue:43550", "issue:43572", "issue:43575", "issue:43653", "issue:43824", + "issue:43844", "issue:43856", "issue:43950", "issue:43975", @@ -8997,221 +9065,221 @@ "issue:46209", "issue:46285", "issue:46512", - "issue:46894", - "issue:46932" + "issue:46821", + "issue:46894" ], "soft_pairs": [ + "issue:43572|issue:46821", + "issue:43550|issue:43844", + "issue:45200|issue:45823", "issue:29870|issue:29942", - "issue:45520|issue:46178", "issue:43450|issue:43575", - "issue:45200|issue:45823", + "issue:45520|issue:46178", "issue:43572|issue:44987", - "issue:37428|issue:43824", "issue:44554|issue:45356", "issue:42490|issue:43493", "issue:43232|issue:44849", - "issue:43295|issue:43493", + "issue:37428|issue:43824", "issue:39692|issue:46192", + "issue:43295|issue:43493", "issue:44466|issue:45127", "issue:44521|issue:45290", "issue:43653|issue:44112", "issue:45381|issue:46209", - "issue:28218|issue:35141", "issue:43278|issue:43856", + "issue:28218|issue:35141", "issue:44737|issue:46894", + "issue:43950|issue:46285", "issue:42915|issue:43824", "issue:43824|issue:43975", - "issue:43950|issue:46285", "issue:43452|issue:46512", - "issue:43334|issue:43525", "issue:43262|issue:43450", - "issue:44442|issue:45198", - "issue:45200|issue:46932" + "issue:44442|issue:45198" ], - "prepared_review_unit_hash": "ec5ec89eadf847490aa1f84bd3fe2d978f764424c55104007dfac02c3e6559b0", + "prepared_review_unit_hash": "0a39e980791d03adf183122d31531721f680f06f37012f2f0d203e1a2e95f8c1", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 26, - "serialized_chars": 35304, - "estimated_input_tokens": 8826, - "estimated_eval_tokens": 17908 + "serialized_chars": 35249, + "estimated_input_tokens": 8813, + "estimated_eval_tokens": 17882 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pair looks like the same underlying bug/change, so this cluster should not be collapsed into duplicates.", - "confidence": 0.94, - "canonical_issue_reason": "No single canonical duplicate emerges: the items span unrelated tokenizer, config, generation, import, processor, and test failures.", + "summary": "These items do not form a coherent duplicate cluster: the soft pairs mostly share only broad themes like tokenizers, generation, or import errors, but point to different concrete bugs and code paths. I would not merge any of the proposed soft edges, and I would not choose a single canonical issue for this set.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43950 is the strongest standalone representative because it describes a broad, core 5.x regression in from_pretrained with a clear failure mode.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:29870", - "right": "issue:29942", + "left": "issue:43572", + "right": "issue:46821", "accept": false, - "reason": "Tokenizer spacing bug vs Flash Attention 2 test failures; different code paths and symptoms." + "reason": "Unrelated config bugs: missing pad_token_idx vs window_size AttributeError." }, { - "left": "issue:45520", - "right": "issue:46178", + "left": "issue:43550", + "right": "issue:43844", + "accept": false, + "reason": "Different failures and subsystems: torch.compile+SDPA vs DeepSpeed ZeRO-3 gradient growth." + }, + { + "left": "issue:45200", + "right": "issue:45823", + "accept": false, + "reason": "Same model family, but different bugs: mm_token_type_ids defaulting vs device_map forward mismatch." + }, + { + "left": "issue:29870", + "right": "issue:29942", "accept": false, - "reason": "Flash-attn import_utils KeyError on Python 3.13 vs an unrelated tensorflow_text backend mapping error." + "reason": "Bart generate() decoder_attention_mask error is unrelated to Flash Attention 2 test failures." }, { "left": "issue:43450", "right": "issue:43575", "accept": false, - "reason": "Video processor batching shape bug vs Qwen2-57B tensor-parallel OOM; not the same defect." + "reason": "Video batch shape bug vs tensor-parallel OOM are different problems." }, { - "left": "issue:45200", - "right": "issue:45823", + "left": "issue:45520", + "right": "issue:46178", "accept": false, - "reason": "Default zero mm_token_type_ids vs PLE device mismatch during forward; different Gemma4 issues." + "reason": "Different import/backend issues: flash_attn on Python 3.13 vs tensorflow_text backend mapping." }, { "left": "issue:43572", "right": "issue:44987", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig vs a model-loading failure for physical-intelligence/fast; no shared underlying change." - }, - { - "left": "issue:37428", - "right": "issue:43824", - "accept": false, - "reason": "Different missing imports: flash support helper vs Qwen2_5_VLForConditionalGeneration." + "reason": "Both are loading-related, but one is StableLm config and the other is a model load regression." }, { "left": "issue:44554", "right": "issue:45356", "accept": false, - "reason": "Attention correctness on MPS vs Kimi-K2.5 tokenizer codec regression; unrelated subsystems." + "reason": "MPS attention correctness issue vs Kimi tokenizer regression; no shared code path." }, { "left": "issue:42490", "right": "issue:43493", "accept": false, - "reason": "Trailing-slash local path warning vs SigLIP2 implementation discrepancy; not the same bug." + "reason": "Local-path import warning is unrelated to SigLIP2 HF/JAX discrepancy." }, { "left": "issue:43232", "right": "issue:44849", "accept": false, - "reason": "Generation kwargs update after sync_gpus vs Qwen3.5 output_hidden_states handling; different generation paths." + "reason": "Generation kwargs sync bug vs Qwen3.5 hidden-states bug are distinct." }, { - "left": "issue:43295", - "right": "issue:43493", + "left": "issue:37428", + "right": "issue:43824", "accept": false, - "reason": "Processor/tokenizer API regression vs SigLIP2 model-vs-JAX discrepancy; no shared code-path problem." + "reason": "Different missing symbols and different affected models; both are import errors only at a broad level." }, { "left": "issue:39692", "right": "issue:46192", "accept": false, - "reason": "Broken SigLIP2 docs example vs return_tensors being ignored when text_kwargs is passed; separate issues." + "reason": "Docs/example failure vs silent return_tensors handling bug; not the same issue." + }, + { + "left": "issue:43295", + "right": "issue:43493", + "accept": false, + "reason": "Processor.tokenizer regression is unrelated to SigLIP2 implementation mismatch." }, { "left": "issue:44466", "right": "issue:45127", "accept": false, - "reason": "Device-dependent serialization of lm_head.weight vs LoRA merge collapse with extended vocab; both involve tied weights but not the same failure." + "reason": "lm_head serialization inconsistency and LoRA-collapse after vocab extension are different tied-weight bugs." }, { "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "All-zero assistant masks for multimodal inputs vs crash on tool-call assistant messages; different apply_chat_template bugs." + "reason": "Multimodal assistant masks vs tool-call tokenization crash are different apply_chat_template paths." }, { "left": "issue:43653", "right": "issue:44112", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug vs a stale GraniteSpeech CI test; unrelated." + "reason": "Tokenizer special-token bug is unrelated to a GraniteSpeech CI test failure." }, { "left": "issue:45381", "right": "issue:46209", "accept": false, - "reason": "Qwen2.5-VL video vision_position_ids bug vs FP8 inference failure on B200; different model/runtime problems." + "reason": "Qwen2.5-VL vision_position_ids bug vs FP8 inference failure on B200 are unrelated." }, { - "left": "issue:28218", - "right": "issue:35141", + "left": "issue:43278", + "right": "issue:43856", "accept": false, - "reason": "Added-token spacing issue vs output embedding reinitialization after resizing; related to tokenization, but not the same bug." + "reason": "Embedding dtype regression vs Qwen3 MoE memory inefficiency are different training issues." }, { - "left": "issue:43278", - "right": "issue:43856", + "left": "issue:28218", + "right": "issue:35141", "accept": false, - "reason": "dtype drift between train/eval vs Qwen3 MoE training memory inefficiency; distinct concerns." + "reason": "Tokenizer spacing after added token is unrelated to output embedding reinit on resize." }, { "left": "issue:44737", "right": "issue:46894", "accept": false, - "reason": "CPU tensor creation in XLNet positional encoding vs CohereASR training-loss double-shift bug; no overlap in code path." + "reason": "XLNet CPU device bug is unrelated to CohereASR training-loss shift bug." + }, + { + "left": "issue:43950", + "right": "issue:46285", + "accept": false, + "reason": "Non-persistent buffer corruption in from_pretrained is unrelated to CLIP text-model structure mismatch." }, { "left": "issue:42915", "right": "issue:43824", "accept": false, - "reason": "Qwen3Moe FP8 config failure vs an import error for Qwen2_5_VLForConditionalGeneration; unrelated." + "reason": "Qwen3Moe FP8Config failure is unrelated to Qwen2_5_VL import errors." }, { "left": "issue:43824", "right": "issue:43975", "accept": false, - "reason": "Missing exported model class vs DeepSeek tokenizer detokenization regression; different layers of the stack." - }, - { - "left": "issue:43950", - "right": "issue:46285", - "accept": false, - "reason": "Non-persistent buffer corruption in from_pretrained vs CLIP class structure inconsistency; not the same regression." + "reason": "ImportError for a VL class and DeepSeek detokenization are different bugs entirely." }, { "left": "issue:43452", "right": "issue:46512", "accept": false, - "reason": "gguf_file loading breakage vs stale model.config.dtype after from_config(dtype=...); separate initialization/loading bugs." - }, - { - "left": "issue:43334", - "right": "issue:43525", - "accept": false, - "reason": "Different model configs missing pad_token_id; similar symptom, but no evidence of one shared underlying fix." + "reason": "gguf_file loading breakage is unrelated to from_config dtype/config stale-state behavior." }, { "left": "issue:43262", "right": "issue:43450", "accept": false, - "reason": "Audio chat-template sampling-rate default vs video processor batching shape bug; different processor classes and failures." + "reason": "Audio chat-template sampling-rate bug and video batched-shape bug are separate processor issues." }, { "left": "issue:44442", "right": "issue:45198", "accept": false, - "reason": "AutoTokenizer loading failure for FastSpeech2ConformerTokenizer vs Wav2Vec2 save_pretrained/tokenization failure; not the same defect." - }, - { - "left": "issue:45200", - "right": "issue:46932", - "accept": false, - "reason": "Gemma4 text-only mm_token_type_ids defaulting vs quantized vision/audio embedder dtype casting; same project area, different bug." + "reason": "Different speech model/tokenization failures; not the same underlying defect." } ] }, @@ -9227,7 +9295,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 30, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -9242,25 +9310,24 @@ "issue:37428", "issue:38175", "issue:41720", + "issue:43334", "issue:43335", "issue:43388", "issue:43425", - "issue:43504", + "issue:43525", "issue:43526", "issue:43550", "issue:43606", - "issue:43611", "issue:43653", "issue:43819", "issue:43827", "issue:43901", - "issue:43931", "issue:44112", "issue:44206", + "issue:44479", "issue:44561", "issue:44610", "issue:44792", - "issue:44855", "issue:44857", "issue:44898", "issue:45083", @@ -9281,241 +9348,241 @@ "issue:46512", "issue:46821", "issue:46858", - "issue:46899" + "issue:46899", + "issue:46932" ], "soft_pairs": [ - "issue:44792|issue:45563", + "issue:43334|issue:43525", "issue:45200|issue:46899", - "issue:44898|issue:45412", + "issue:45200|issue:46932", + "issue:44792|issue:45563", "issue:43819|issue:44112", - "issue:43526|issue:46192", + "issue:44898|issue:45412", + "issue:44479|issue:46192", "issue:45200|issue:45325", - "issue:43827|issue:43901", "issue:41720|issue:46209", + "issue:43827|issue:43901", "issue:43335|issue:43550", "issue:44561|issue:46234", - "issue:43388|issue:43526", "issue:43819|issue:45083", + "issue:43388|issue:43526", "issue:44857|issue:45593", "issue:44610|issue:44857", - "issue:46424|issue:46858", "issue:45759|issue:46285", + "issue:46424|issue:46858", + "issue:43606|issue:45083", "issue:46512|issue:46821", "issue:45520|issue:46170", - "issue:43606|issue:45083", - "issue:38175|issue:45412", "issue:44206|issue:45584", + "issue:38175|issue:45412", "issue:43335|issue:43653", - "issue:37428|issue:43425", "issue:29942|issue:32090", "issue:43550|issue:44112", "issue:28282|issue:33357", - "issue:43611|issue:43931", - "issue:43335|issue:43504", - "issue:44855|issue:45083" + "issue:37428|issue:43425" ], - "prepared_review_unit_hash": "21c311c2586d077e82b210586b6baa48548d525836775bc36ca957d7cec32674", + "prepared_review_unit_hash": "099fbe9a3a7f3880d5b1985795187dd3fbc38a6d690c020d4615f8c6c4aeb072", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 29, - "serialized_chars": 36160, - "estimated_input_tokens": 9040, - "estimated_eval_tokens": 18336 + "soft_pair_count": 28, + "serialized_chars": 35853, + "estimated_input_tokens": 8964, + "estimated_eval_tokens": 18184 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: it mixes unrelated model bugs, import/version regressions, doc issues, and hardware-specific failures. None of the soft pairs look like true duplicates or mergeable as the same underlying change, so no single canonical duplicate anchor emerges.", - "confidence": 0.91, - "canonical_issue_reason": "No canonical issue: the items are largely unrelated, so choosing one as the duplicate anchor would be misleading.", + "summary": "Mostly unrelated bug reports; the only clear duplicate pair is the Gemma4 4-bit quantized vision/audio embedder dtype-cast regression (#46899/#46932). The rest share broad themes but not the same concrete bug path.", + "confidence": 0.89, + "canonical_issue_reason": "#46899 is the more complete symptom report for the Gemma4 quantization regression; #46932 is a narrower fix-oriented restatement of the same dtype-casting bug in vision/audio embedders.", "canonical_pr_reason": null, - "best_issue_reason": "No single best issue for the cluster; the reports do not share one concrete bug or code path.", + "best_issue_reason": "#46899 is the clearest representative of the only true duplicate theme in this set and has the most informative problem description.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44792", - "right": "issue:45563", + "left": "issue:43334", + "right": "issue:43525", "accept": false, - "reason": "Different generation bugs: Janus image-generation failure vs a stale paged-generate warning." + "reason": "Same symptom family (missing pad_token_id) but different configs/models and likely separate model-specific regressions." }, { "left": "issue:45200", "right": "issue:46899", "accept": false, - "reason": "Both are Gemma4-related, but one is text-only mm_token_type_ids handling and the other is 4-bit multimodal dtype casting." + "reason": "Both involve Gemma4, but one is about mm_token_type_ids defaults and the other about 4-bit quantized embedder dtype casting; different bugs." }, { - "left": "issue:44898", - "right": "issue:45412", + "left": "issue:45200", + "right": "issue:46932", + "accept": false, + "reason": "Different failure modes: text-only mm_token_type_ids handling vs quantized vision/audio embedder casting." + }, + { + "left": "issue:44792", + "right": "issue:45563", "accept": false, - "reason": "Unrelated failures: Perceiver non-default resolution vs RT-DETR memory cleanup." + "reason": "Different generation paths and symptoms; one is image generation test failure, the other is a stale warning about num_return_sequences." }, { "left": "issue:43819", "right": "issue:44112", "accept": false, - "reason": "Different code paths: DAC latents/forward mismatch vs GraniteSpeech device-override CI test." + "reason": "Unrelated subsystems: DAC latent reconstruction mismatch vs GraniteSpeech CI device override flake." }, { - "left": "issue:43526", + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Perceiver resolution handling and RT-DETR memory release are unrelated issues." + }, + { + "left": "issue:44479", "right": "issue:46192", "accept": false, - "reason": "Separate bugs in different APIs: BEiT image label reduction vs return_tensors being ignored with text_kwargs." + "reason": "Video input regression vs return_tensors being ignored; both are input-processing issues but not the same bug." }, { "left": "issue:45200", "right": "issue:45325", "accept": false, - "reason": "Different multimodal Qwen bugs: defaulting mm_token_type_ids vs still-image rope indexing." + "reason": "Both touch multimodal models, but the bugs differ: missing default token ids vs temporal position scaling in rope index." }, { - "left": "issue:43827", - "right": "issue:43901", + "left": "issue:41720", + "right": "issue:46209", "accept": false, - "reason": "Both are doc issues, but they cover different pipelines and different stale parameters." + "reason": "Different hardware/model-specific inference failures with different root causes and environments." }, { - "left": "issue:41720", - "right": "issue:46209", + "left": "issue:43827", + "right": "issue:43901", "accept": false, - "reason": "Different model/hardware failures: Qwen3 auto device mapping on A800 vs FP8 inference on B200." + "reason": "Both are docs regressions, but they concern different pipeline API deprecations and are not one underlying defect." }, { "left": "issue:43335", "right": "issue:43550", "accept": false, - "reason": "Unrelated: SwitchTransformers sparse-layer config bug vs Bamba torch.compile/SDPA failure." + "reason": "Different models and different code paths: sparse-layer config creation vs torch.compile/SDPA failure." }, { "left": "issue:44561", "right": "issue:46234", "accept": false, - "reason": "Both involve import_utils, but they are different missing-function regressions affecting different model families." + "reason": "Both mention import_utils, but they are missing different symbols with different breakage patterns and no shared concrete bug path." }, { - "left": "issue:43388", - "right": "issue:43526", + "left": "issue:43819", + "right": "issue:45083", "accept": false, - "reason": "Different bug classes: accelerator metric gathering vs BEiT image processor label reduction." + "reason": "DAC latent conversion and qwen3_omni_moe feature-length helper behavior are unrelated." }, { - "left": "issue:43819", - "right": "issue:45083", + "left": "issue:43388", + "right": "issue:43526", "accept": false, - "reason": "Different models and functions: DAC latent reconstruction vs qwen3_omni_moe feature-length helper." + "reason": "Different label-processing bugs in different components; not the same code path." }, { "left": "issue:44857", "right": "issue:45593", "accept": false, - "reason": "Different training bugs: float16 AMP crash in LwDetrImageLoss vs missing auxiliary losses in D-FINE." + "reason": "Different object-detection losses and different failure conditions." }, { "left": "issue:44610", "right": "issue:44857", "accept": false, - "reason": "Different problems in different models: OmDet-Turbo input sizing vs LwDetrImageLoss AMP crash." + "reason": "Processor input-size mismatch vs AMP/CUDA crash; no shared underlying bug." + }, + { + "left": "issue:45759", + "right": "issue:46285", + "accept": false, + "reason": "Composite Qwen text_config unwrapping and CLIP text model structure inconsistency are different API breakages." }, { "left": "issue:46424", "right": "issue:46858", "accept": false, - "reason": "Both are cache-related, but one is max_cache_length semantics and the other is a device-tensor arithmetic bug in _prefill." + "reason": "Both are cache-related, but one is about cache length configuration and the other about device tensors breaking slice arithmetic." }, { - "left": "issue:45759", - "right": "issue:46285", + "left": "issue:43606", + "right": "issue:45083", "accept": false, - "reason": "Unrelated model-structure regressions: Qwen composite config unwrapping vs CLIP text model structure changes." + "reason": "CPU offload device mismatch and qwen3_omni_moe feature-length helper behavior are unrelated." }, { "left": "issue:46512", "right": "issue:46821", "accept": false, - "reason": "Different issues: stale model.config.dtype after from_config vs TimesFM window_size AttributeError." + "reason": "Stale model.config.dtype after from_config(dtype=...) is unrelated to TimesFM window_size AttributeError." }, { "left": "issue:45520", "right": "issue:46170", "accept": false, - "reason": "Both are platform/import failures, but they stem from different dependencies and environments." + "reason": "Different platform/import failures with different missing dependencies and error modes." }, { - "left": "issue:43606", - "right": "issue:45083", + "left": "issue:44206", + "right": "issue:45584", "accept": false, - "reason": "Different subsystems entirely: CPU offload device mismatch vs qwen3_omni_moe feature-length helper." + "reason": "Unsupported feature-arg crash in LasrFeatureExtractor vs Whisper empty-transcription generation failure." }, { "left": "issue:38175", "right": "issue:45412", "accept": false, - "reason": "Unrelated: SigLIP2 zero probabilities vs RT-DETR memory not released." - }, - { - "left": "issue:44206", - "right": "issue:45584", - "accept": false, - "reason": "Different bugs: unsupported center arg in LasrFeatureExtractor vs Whisper empty-transcription generation failure." + "reason": "Siglip2 zero probabilities and RT-DETR memory retention are unrelated." }, { "left": "issue:43335", "right": "issue:43653", "accept": false, - "reason": "Different model-specific regressions: SwitchTransformers sparse-layer config vs BigBirdTokenizer special-token decode issue." - }, - { - "left": "issue:37428", - "right": "issue:43425", - "accept": false, - "reason": "One is a missing flash-attention symbol import, the other is a broad Torch 2.10 compatibility report." + "reason": "SwitchTransformers sparse-layer config bug and BigBirdTokenizer special-token registration are separate issues." }, { "left": "issue:29942", "right": "issue:32090", "accept": false, - "reason": "Unrelated failures: Flash Attention 2 tests vs Trainer _gpu_broadcast_one NoneType error." + "reason": "Failing Flash Attention tests and Trainer gpu_broadcast_one NoneType error are different bugs." }, { "left": "issue:43550", "right": "issue:44112", "accept": false, - "reason": "Different failures: Bamba torch.compile/SDPA vs GraniteSpeech stale device override test." + "reason": "Different code paths and symptoms: compile/SDPA model failure vs stale CI device-override test." }, { "left": "issue:28282", "right": "issue:33357", "accept": false, - "reason": "Different environment issues: missing PyTorch import vs MacOS bus error with a CLIP community model." + "reason": "PyTorch-not-found ImportError and MacOS bus error on a CLIP model are unrelated." }, { - "left": "issue:43611", - "right": "issue:43931", - "accept": false, - "reason": "Different loading bugs: base_model_prefix handling vs Qwen3-VL weight-shape mismatch." - }, - { - "left": "issue:43335", - "right": "issue:43504", + "left": "issue:37428", + "right": "issue:43425", "accept": false, - "reason": "Different BEiT/SwitchTransformers bugs with no shared code path." + "reason": "Missing flash attention helper import and torch 2.10 incompatibility are too broad to be the same concrete bug." }, { - "left": "issue:44855", - "right": "issue:45083", - "accept": false, - "reason": "Different problems: DebertaV2Model import parsing on Python 3.13 vs qwen3_omni_moe feature-length behavior." + "left": "issue:46899", + "right": "issue:46932", + "accept": true, + "reason": "Same Gemma4 quantized vision/audio embedder dtype-casting regression; #46932 is essentially the fix-oriented restatement of #46899." } ] }, @@ -9531,7 +9598,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 31, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -9544,7 +9611,6 @@ "issue:29127", "issue:31501", "issue:32090", - "issue:33357", "issue:34689", "issue:35141", "issue:37428", @@ -9552,16 +9618,17 @@ "issue:42490", "issue:43278", "issue:43335", + "issue:43504", "issue:43526", - "issue:43575", "issue:43606", + "issue:43611", "issue:43701", "issue:43720", "issue:43819", "issue:43825", "issue:43854", + "issue:43931", "issue:44112", - "issue:44206", "issue:44351", "issue:44448", "issue:44493", @@ -9570,6 +9637,7 @@ "issue:44617", "issue:44743", "issue:44779", + "issue:44855", "issue:44857", "issue:45083", "issue:45127", @@ -9580,7 +9648,6 @@ "issue:45478", "issue:45588", "issue:45593", - "issue:45823", "issue:45859", "issue:46234", "issue:46326", @@ -9589,244 +9656,244 @@ "issue:46697" ], "soft_pairs": [ + "issue:43611|issue:43931", "issue:44448|issue:44743", + "issue:43335|issue:43504", + "issue:44855|issue:45083", "issue:43278|issue:43825", - "issue:37428|issue:44351", "issue:43335|issue:45083", - "issue:43335|issue:44112", "issue:43335|issue:45593", "issue:29127|issue:32090", - "issue:44514|issue:44521", + "issue:43335|issue:44112", "issue:43720|issue:44112", - "issue:45290|issue:46326", - "issue:34689|issue:35141", + "issue:44514|issue:44521", "issue:45588|issue:45859", + "issue:45290|issue:46326", "issue:42490|issue:43854", - "issue:46431|issue:46566", + "issue:37428|issue:44351", + "issue:34689|issue:35141", "issue:45127|issue:45245", - "issue:45310|issue:45478", - "issue:44448|issue:44779", + "issue:46431|issue:46566", "issue:43819|issue:44493", "issue:39692|issue:43526", - "issue:43606|issue:44857", + "issue:44448|issue:44779", + "issue:45310|issue:45478", "issue:26413|issue:31501", + "issue:43606|issue:44857", "issue:28218|issue:32090", "issue:45459|issue:46234", - "issue:44617|issue:46697", "issue:43701|issue:45083", "issue:39692|issue:44493", - "issue:44206|issue:45823", - "issue:32090|issue:33357", - "issue:42490|issue:43575" + "issue:44617|issue:46697" ], - "prepared_review_unit_hash": "a78c9e18240c4d357e8942f3e8b62d2a15c71b2eaf97d8c9b4afc474ecedc240", + "prepared_review_unit_hash": "2e5bcc17c51512fd970d872e7bf1827040476f362e04568647b60efc04f1d74d", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36505, - "estimated_input_tokens": 9127, - "estimated_eval_tokens": 18510 + "serialized_chars": 36623, + "estimated_input_tokens": 9156, + "estimated_eval_tokens": 18568 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Only one duplicate pair is clearly grounded in the packet: issue:45310 \u2194 issue:45478, the repeated Qwen3.5 MoE from_pretrained regression. The resume_from_checkpoint items (issue:26413 and issue:31501) are not supported as a true duplicate pair here, and the remaining soft matches are just broad thematic similarity.", - "confidence": 0.97, - "canonical_issue_reason": "issue:45310 is the best duplicate-family representative because issue:45478 repeats the same Qwen3.5 MoE from_pretrained failure with essentially the same title and problem scope.", + "summary": "Only one soft-edge pair is a true duplicate: the two Qwen3.5 MoE from_pretrained reports (45310/45478). The rest are conservative false matches across different bugs or code paths.", + "confidence": 0.89, + "canonical_issue_reason": "issue:45310 is the better canonical issue because it is the earlier report of the same Qwen3.5 MoE from_pretrained failure and can serve as the shared reference for the later near-duplicate issue:45478.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the strongest global issue candidate because it has a direct, well-supported duplicate in issue:45478 and represents a concrete repeated regression.", + "best_issue_reason": "issue:45310 is the best issue overall since it is the earliest and clearest representative of the only confirmed duplicate pair, so it is the most suitable consolidation target.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43611", + "right": "issue:43931", + "accept": false, + "reason": "Both are model-loading failures, but they describe different failure modes and models; not the same bug." + }, { "left": "issue:44448", "right": "issue:44743", "accept": false, - "reason": "Both are version-regression reports, but they hit different models and different code paths." + "reason": "Different problems: Pegasus output mismatch vs Qwen recurrent state reset in cache handling." }, { - "left": "issue:43278", - "right": "issue:43825", + "left": "issue:43335", + "right": "issue:43504", "accept": false, - "reason": "Different bugs: dtype drift during evaluate vs an incorrect pipeline task-support error message." + "reason": "Unrelated model/config bugs in different components; no shared concrete code path." }, { - "left": "issue:37428", - "right": "issue:44351", + "left": "issue:44855", + "right": "issue:45083", "accept": false, - "reason": "Both are import errors, but for different missing symbols in different modules." + "reason": "Python 3.13 import/torchscript parsing error vs qwen3_omni_moe helper length behavior." }, { - "left": "issue:43335", - "right": "issue:45083", + "left": "issue:43278", + "right": "issue:43825", "accept": false, - "reason": "Different defect types: sparse-layer config construction vs feature-length helper behavior." + "reason": "Training/eval dtype drift is unrelated to a pipeline error message regression." }, { "left": "issue:43335", - "right": "issue:44112", + "right": "issue:45083", "accept": false, - "reason": "A config bug and a CI stale-device test failure are not the same underlying issue." + "reason": "Different models and different bugs; only broad subsystem overlap." }, { "left": "issue:43335", "right": "issue:45593", "accept": false, - "reason": "Different model families and failure modes." + "reason": "Sparse-layer config bug and D-FINE auxiliary-loss behavior are unrelated." }, { "left": "issue:29127", "right": "issue:32090", "accept": false, - "reason": "Unrelated: LayoutLMv3 box validation messaging vs Trainer broadcast of NoneType." + "reason": "LayoutLMv3 error clarity and Trainer NoneType broadcast are different issues." }, { - "left": "issue:44514", - "right": "issue:44521", + "left": "issue:43335", + "right": "issue:44112", "accept": false, - "reason": "Both involve apply_chat_template, but one crashes on batched padding=False input and the other miscomputes assistant masks." + "reason": "Different model/config bug vs CI device-override test failure." }, { "left": "issue:43720", "right": "issue:44112", "accept": false, - "reason": "Packed-weight loading in BitNet is unrelated to the GraniteSpeech CI device override issue." + "reason": "Packed-weight loading failure is not the same as a stale CI test." }, { - "left": "issue:45290", - "right": "issue:46326", + "left": "issue:44514", + "right": "issue:44521", "accept": false, - "reason": "Different chat-template/tokenizer bugs with different triggers and symptoms." + "reason": "Both touch apply_chat_template, but one is a batch-padding crash and the other is incorrect assistant masks." }, { - "left": "issue:34689", - "right": "issue:35141", + "left": "issue:45588", + "right": "issue:45859", "accept": false, - "reason": "Different regressions: model loading failure vs post_init reinitializing resized embeddings." + "reason": "Flash attention None aux handling and CPU-offload rotary embedding incompatibility are separate code paths." }, { - "left": "issue:45588", - "right": "issue:45859", + "left": "issue:45290", + "right": "issue:46326", "accept": false, - "reason": "Different flash-attention/embedding failures and different affected code paths." + "reason": "Tool-call/no-content crash is distinct from continue_final_message behavior with mistral_common tokenizers." }, { "left": "issue:42490", "right": "issue:43854", "accept": false, - "reason": "A trailing-slash path warning is unrelated to a model-loading failure in tests." + "reason": "Local-path trailing-slash warning and GLM-4.7-Flash model loading are unrelated." }, { - "left": "issue:46431", - "right": "issue:46566", + "left": "issue:37428", + "right": "issue:44351", "accept": false, - "reason": "Missing chat template support is unrelated to weights staying on meta with device_map=auto." + "reason": "Missing flash-support helper and missing HybridCache export are different import regressions." }, { - "left": "issue:45127", - "right": "issue:45245", + "left": "issue:34689", + "right": "issue:35141", "accept": false, - "reason": "LoRA/vocabulary collapse and a categories-limit RuntimeError are unrelated." + "reason": "Model-loading breakage for Llama 3.2 Vision is unrelated to embedding reinitialization after resize." }, { - "left": "issue:45310", - "right": "issue:45478", - "accept": true, - "reason": "The titles and bug descriptions match the same Qwen3.5 MoE from_pretrained regression." + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA merge collapse and the 2^24 category limit are unrelated bugs." }, { - "left": "issue:44448", - "right": "issue:44779", + "left": "issue:46431", + "right": "issue:46566", "accept": false, - "reason": "Both mention output differences across versions, but they concern different models and mechanisms." + "reason": "Missing chat template on a processor and meta-device weights with device_map='auto' are different failures." }, { "left": "issue:43819", "right": "issue:44493", "accept": false, - "reason": "A model forward/from_latents mismatch is unrelated to unexpected position-id key handling." + "reason": "DAC latent mismatch and unexpected position-id keys are unrelated." }, { "left": "issue:39692", "right": "issue:43526", "accept": false, - "reason": "Documentation/example errors are unrelated to BEiT image-processor label reduction." + "reason": "SigLIP2 doc/example issues and BeitImageProcessorFast label reduction bug are different problems." }, { - "left": "issue:43606", - "right": "issue:44857", + "left": "issue:44448", + "right": "issue:44779", "accept": false, - "reason": "CPU offload device mismatch in Bark is unrelated to an AMP crash in LwDetrImageLoss." + "reason": "Different output on a summarization model vs tokenizer regression in Deepseek; not the same change." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Same title, same reported Qwen3.5 MoE from_pretrained error, and the later issue is a near-duplicate of the same bug." }, { "left": "issue:26413", "right": "issue:31501", "accept": false, - "reason": "resume_from_checkpoint iterator/key failures do not match FineWeb training not starting." + "reason": "Checkpoint resume iterator failure and FineWeb training not starting do not match the same root cause." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU-offload device mismatch in bark-small is distinct from LwDetrImageLoss AMP/CUDA crashing." }, { "left": "issue:28218", "right": "issue:32090", "accept": false, - "reason": "Tokenizer spacing after added tokens is unrelated to a NoneType GPU broadcast error." + "reason": "Tokenizer added-space behavior is unrelated to the Trainer NoneType broadcast error." }, { "left": "issue:45459", "right": "issue:46234", "accept": false, - "reason": "Tokenizer protobuf error handling is unrelated to a missing import_utils availability helper." - }, - { - "left": "issue:44617", - "right": "issue:46697", - "accept": false, - "reason": "CUDA OOM and image over-padding are different classes of problems." + "reason": "Tokenizer protobuf error masking and a missing availability helper are unrelated." }, { "left": "issue:43701", "right": "issue:45083", "accept": false, - "reason": "resume_from_checkpoint key mismatch and qwen3_omni_moe feature-length behavior are unrelated." + "reason": "Resume-from-checkpoint key mismatch is unrelated to qwen3_omni_moe feature-length behavior." }, { "left": "issue:39692", "right": "issue:44493", "accept": false, - "reason": "A docs/model-processor mismatch is not the same as an unexpected position-id key bug." - }, - { - "left": "issue:44206", - "right": "issue:45823", - "accept": false, - "reason": "Unsupported feature-extractor args and a device mismatch are unrelated." - }, - { - "left": "issue:32090", - "right": "issue:33357", - "accept": false, - "reason": "A Trainer broadcast TypeError is unrelated to a MacOS bus error in CLIP loading." + "reason": "Doc/example model-processor mismatch is unrelated to position-id key warnings." }, { - "left": "issue:42490", - "right": "issue:43575", + "left": "issue:44617", + "right": "issue:46697", "accept": false, - "reason": "A path warning is unrelated to Qwen OOM under tensor parallelism." + "reason": "CUDA OOM in Sam3Video and over-padding in Swin2SR are not the same bug." } ] }, "evaluator_result": { "accept": false, - "feedback": "The analysis overstates the duplicate landscape. The packet supports issue:45310 as a clear duplicate-family representative with issue:45478, but it does not support a second true duplicate pair for `resume_from_checkpoint`; the cited 26413/31501 pair is not a duplicate and the analyst itself marks it unrelated. The summary should be more conservative and mention only the one clearly grounded duplicate pair." + "feedback": "Reject: the duplicate call on 45310/45478 is plausible, but the canonical/best-issue rationale overstates the packet. Both issues are shown as closed, so calling 45310 the \"primary open report\" or \"clearest live tracker\" is not grounded. The rest of the soft-edge rejections are generally conservative, but this factual error is enough to fail precision." }, "evaluator_used": true, "retried": true, @@ -9839,7 +9906,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 32, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -9855,6 +9922,7 @@ "issue:29942", "issue:32090", "issue:33290", + "issue:33357", "issue:34689", "issue:35141", "issue:37428", @@ -9867,15 +9935,15 @@ "issue:43381", "issue:43526", "issue:43550", + "issue:43575", "issue:43873", "issue:43927", - "issue:43976", + "issue:44206", "issue:44360", "issue:44448", "issue:44466", "issue:44493", "issue:44561", - "issue:44568", "issue:44779", "issue:44855", "issue:44857", @@ -9885,10 +9953,9 @@ "issue:45229", "issue:45290", "issue:45381", - "issue:45440", - "issue:45491", "issue:45588", "issue:45698", + "issue:45823", "issue:45920", "issue:46192", "issue:46399", @@ -9897,245 +9964,252 @@ "issue:46858" ], "soft_pairs": [ + "issue:44206|issue:45823", + "issue:32090|issue:33357", + "issue:42490|issue:43575", "issue:44857|issue:45072", - "issue:44360|issue:44561", "issue:43550|issue:45083", - "issue:27957|issue:34689", + "issue:44360|issue:44561", "issue:29942|issue:35141", + "issue:27957|issue:34689", "issue:38175|issue:44493", - "issue:29870|issue:32090", "issue:43278|issue:43329", - "issue:44448|issue:44977", + "issue:29870|issue:32090", "issue:45920|issue:46192", + "issue:44448|issue:44977", "issue:45290|issue:45381", "issue:10000|issue:15354", "issue:44977|issue:45229", - "issue:29870|issue:33290", - "issue:43278|issue:43381", - "issue:44855|issue:46399", "issue:43927|issue:45698", - "issue:45290|issue:46655", + "issue:44855|issue:46399", + "issue:43278|issue:43381", + "issue:29870|issue:33290", "issue:40444|issue:43526", - "issue:37428|issue:42175", + "issue:45290|issue:46655", "issue:44857|issue:45588", "issue:32090|issue:33290", + "issue:37428|issue:42175", "issue:46421|issue:46858", "issue:44561|issue:44779", - "issue:44448|issue:44466", "issue:43873|issue:45229", - "issue:25251|issue:32090", "issue:42490|issue:43873", - "issue:43976|issue:45440", - "issue:44568|issue:45491" + "issue:44448|issue:44466", + "issue:25251|issue:32090" ], - "prepared_review_unit_hash": "a6a0fbfe329fd6d7c8d94dc5037e6a326f0255dfc3e6a2eb383532017c88e1cf", + "prepared_review_unit_hash": "89ee10672877a69d9791d6206084616bef6f9f8ca8ea7790402e292882bf6146", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 37034, - "estimated_input_tokens": 9259, - "estimated_eval_tokens": 18774 + "soft_pair_count": 31, + "serialized_chars": 37237, + "estimated_input_tokens": 9310, + "estimated_eval_tokens": 18876 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The soft pairs mostly share broad themes (Flash Attention, cache, v5 regressions, multimodal bugs) but do not look like the same underlying bug/change. I found no strong duplicate pair; the cluster appears heterogeneous.", - "confidence": 0.89, - "canonical_issue_reason": null, + "summary": "This cluster is heterogeneous: the items cover unrelated bugs in generation, tokenization, loading, Flash Attention, dtype/device handling, and training. None of the soft pairs look like the same underlying issue, so no duplicate consolidation is justified.", + "confidence": 0.98, + "canonical_issue_reason": "No single canonical issue fits this cluster; the issues are thematically broad but describe different failures and code paths.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the most active open item in the set, with multiple inbound references and discussion, but it is not a good duplicate canonical for the other issues because the topics are unrelated.", + "best_issue_reason": "No best issue candidate stands out because the cluster is not a duplicate set and lacks one representative underlying bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44857", - "right": "issue:45072", + "left": "issue:44206", + "right": "issue:45823", "accept": false, - "reason": "Both involve dtype-related failures, but they affect different models and different code paths (AMP loss vs bfloat16 inference mismatch)." + "reason": "Both mention model/input handling, but one is a LasrFeatureExtractor crash and the other is a Gemma4 device-mismatch during forward; different bugs and fixes." }, { - "left": "issue:44360", - "right": "issue:44561", + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "NoneType GPU broadcast in Trainer vs MacOS bus error on a community CLIP model; different subsystems and failure modes." + }, + { + "left": "issue:42490", + "right": "issue:43575", + "accept": false, + "reason": "Local-path import warning on trailing slash vs OOM when loading a huge Qwen model with tensor parallelism; not the same bug." + }, + { + "left": "issue:44857", + "right": "issue:45072", "accept": false, - "reason": "One is about a DSA indexer activation detail; the other is a v5 API removal breaking trust_remote_code. No shared underlying bug." + "reason": "AMP/CUDA loss crash in LwDetrImageLoss vs bfloat16 dtype mismatch in SwitchTransformers/TimmWrapperModel; distinct code paths." }, { "left": "issue:43550", "right": "issue:45083", "accept": false, - "reason": "Different subsystems and symptoms: torch.compile/SDPA model failure vs a helper length-calculation bug in qwen3_omni_moe." + "reason": "torch.compile+SDPA failure for Bamba-9B-v2 vs helper-length calculation behavior in qwen3_omni_moe; unrelated issues." }, { - "left": "issue:27957", - "right": "issue:34689", + "left": "issue:44360", + "right": "issue:44561", "accept": false, - "reason": "Both are model-specific regressions, but they concern unrelated models and failures (Flash Attention 2 vs model loading)." + "reason": "DSA indexer activation-function complaint vs removal of is_torch_fx_available breaking trust_remote_code models; no shared underlying defect." }, { "left": "issue:29942", "right": "issue:35141", "accept": false, - "reason": "Failing Flash Attention tests and post_init reinitializing embeddings are unrelated bugs." + "reason": "Flash Attention 2 test failures vs token-embedding resize reinitializing output embeddings; different components and fixes." }, { - "left": "issue:38175", - "right": "issue:44493", + "left": "issue:27957", + "right": "issue:34689", "accept": false, - "reason": "Zero probabilities in one model and unexpected position-id keys across many models do not indicate the same defect." + "reason": "XLMRoberta Flash Attention 2 support vs Llama 3.2 90B Vision model-loading regression; not the same concrete problem." }, { - "left": "issue:29870", - "right": "issue:32090", + "left": "issue:38175", + "right": "issue:44493", "accept": false, - "reason": "Generate-time decoder mask bug vs Trainer GPU broadcast TypeError; different code paths and root causes." + "reason": "Unexpected zero probabilities in SigLIP2 vs unexpected position-id keys in many models; distinct symptoms and likely causes." }, { "left": "issue:43278", "right": "issue:43329", "accept": false, - "reason": "BF16/FP32 dtype drift in evaluate is unrelated to undefined video-token helper variables." + "reason": "BF16-to-FP32 embedding dtype regression during eval vs undefined video-branch variables in multimodal token counting; unrelated." }, { - "left": "issue:44448", - "right": "issue:44977", + "left": "issue:29870", + "right": "issue:32090", "accept": false, - "reason": "Both are generation regressions, but they involve different models and different failure mechanisms." + "reason": "BART generate decoder_attention_mask bug vs Trainer gpu_broadcast NoneType error; different call sites and failure mechanisms." }, { "left": "issue:45920", "right": "issue:46192", "accept": false, - "reason": "Tokenizer ID corruption across models is not the same as return_tensors being ignored when text_kwargs are passed." + "reason": "AutoTokenizer producing wrong token IDs vs return_tensors being ignored when text_kwargs is passed; both tokenizer-related but not the same bug." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Pegasus v4/v5 output difference vs Qwen3.5 flash-attention generation failure; different models and code paths." }, { "left": "issue:45290", "right": "issue:45381", "accept": false, - "reason": "Chat template crash with tool-call messages is separate from the Qwen2.5-VL video position-id bug." + "reason": "apply_chat_template crash with tool-call assistant messages vs Qwen2.5-VL video vision_position_ids bug; both multimodal but distinct failures." }, { "left": "issue:10000", "right": "issue:15354", "accept": false, - "reason": "German DistilBertModel and torch.jit.script GeneratorExp export are unrelated issues." + "reason": "German DistilBertModel issue vs torch.jit.script GeneratorExp unsupported for ViT export; unrelated historical issues." }, { "left": "issue:44977", "right": "issue:45229", "accept": false, - "reason": "Flash-attention generation bug vs multi-GPU CUDA OOM; no shared concrete code-path problem." + "reason": "Flash-attention generation bug in Qwen3.5 vs multi-GPU CUDA OOM for Gemma4; not mergeable into one fix." }, { - "left": "issue:29870", - "right": "issue:33290", - "accept": false, - "reason": "One is a generate() UnboundLocalError; the other is a deepspeed/adafactor OOM. Too different to merge." - }, - { - "left": "issue:43278", - "right": "issue:43381", + "left": "issue:43927", + "right": "issue:45698", "accept": false, - "reason": "Dtype drift in evaluate and gradient checkpointing in eval mode are distinct behaviors." + "reason": "DiaConfig losing custom token IDs on save/load vs wrong custom module loaded after save_pretrained; both serialization-related but different objects and bugs." }, { "left": "issue:44855", "right": "issue:46399", "accept": false, - "reason": "Both mention decorators/parsing, but one is a Python 3.13 import/JIT parsing issue and the other is a missing decorator registration bug." + "reason": "Python 3.13 indentation/JIT parsing failure in DebertaV2Model vs missing decorator causing Qwen3.5 kernel registration error; separate root causes." }, { - "left": "issue:43927", - "right": "issue:45698", + "left": "issue:43278", + "right": "issue:43381", "accept": false, - "reason": "Save/load problems in custom objects are related in spirit, but these are different bugs: lost token IDs vs wrong custom module resolution." + "reason": "Embedding dtype regression during eval vs gradient checkpointing disallowed in eval mode; different behaviors." }, { - "left": "issue:45290", - "right": "issue:46655", + "left": "issue:29870", + "right": "issue:33290", "accept": false, - "reason": "Tool-call chat templating crash and video padding calculation are unrelated preprocessing bugs." + "reason": "generate() decoder_attention_mask bug vs Adafactor+DeepSpeed OOM; one is correctness, the other memory usage." }, { "left": "issue:40444", "right": "issue:43526", "accept": false, - "reason": "Different multimodal preprocessing failures affecting different components and outputs." + "reason": "Qwen2.5-VL iterable dataset with multiple images failing vs BeitImageProcessorFast reduce_labels returning one label; different multimodal/preprocessing issues." }, { - "left": "issue:37428", - "right": "issue:42175", + "left": "issue:45290", + "right": "issue:46655", "accept": false, - "reason": "Flash-attention import error and missing TensorFlow backend in a torch install are unrelated packaging/runtime issues." + "reason": "Chat template crash on tool-call messages vs wrong padding calculation in video processor; different layers of the stack." }, { "left": "issue:44857", "right": "issue:45588", "accept": false, - "reason": "Both touch flash-attention, but one is an AMP dtype crash and the other is an s_aux=None AttributeError in sink-less models." + "reason": "LwDetrImageLoss fp16 crash vs flash_attention.py AttributeError on s_aux=None; separate bugs in different modules." }, { "left": "issue:32090", "right": "issue:33290", "accept": false, - "reason": "Trainer broadcast TypeError and deepspeed Adafactor OOM do not share the same underlying bug." + "reason": "Trainer gpu_broadcast NoneType error vs deepspeed Adafactor OOM; not the same underlying problem." + }, + { + "left": "issue:37428", + "right": "issue:42175", + "accept": false, + "reason": "Missing _flash_supports_window_size import vs TensorFlow not included in torch-only install; unrelated packaging/API issues." }, { "left": "issue:46421", "right": "issue:46858", "accept": false, - "reason": "Both involve cache/prefill/compile, but one is spurious recompilation from lazy init and the other is a device-tensor return breaking slice arithmetic." + "reason": "Cache lazy initialization causing recompilation vs seq_length returning a device tensor and breaking arithmetic; both cache-related but different bugs." }, { "left": "issue:44561", "right": "issue:44779", "accept": false, - "reason": "A removed helper breaking trust_remote_code is not the same as tokenizer output regression in DeepSeek." - }, - { - "left": "issue:44448", - "right": "issue:44466", - "accept": false, - "reason": "Different regressions: output drift for Pegasus vs tied-weight serialization inconsistency." + "reason": "Removal of is_torch_fx_available breaking remote-code models vs DeepSeek tokenizer regression in v5; different functionality and fixes." }, { "left": "issue:43873", "right": "issue:45229", "accept": false, - "reason": "Quantization/offloading behavior and Gemma4 multi-GPU OOM are not the same concrete bug." - }, - { - "left": "issue:25251", - "right": "issue:32090", - "accept": false, - "reason": "Pipeline top_k nesting change is unrelated to the Trainer NoneType broadcast error." + "reason": "Quantization/offloading not working as expected vs Gemma4 multi-GPU CUDA OOM; shared memory theme only, not the same bug." }, { "left": "issue:42490", "right": "issue:43873", "accept": false, - "reason": "Import warning for trailing-slash local paths and quantization offloading are unrelated." + "reason": "Trailing-slash local-model import warning vs quantization offloading behavior; unrelated issues." }, { - "left": "issue:43976", - "right": "issue:45440", + "left": "issue:44448", + "right": "issue:44466", "accept": false, - "reason": "Python-version compatibility and DeepSeekV3MoE divergence are unrelated issues." + "reason": "Pegasus output divergence across versions vs inconsistent lm_head.weight serialization depending on device; different model behaviors and code paths." }, { - "left": "issue:44568", - "right": "issue:45491", + "left": "issue:25251", + "right": "issue:32090", "accept": false, - "reason": "Tokenizer BOS/EOS handling and Gemma3 sliding-window NaNs are different bugs." + "reason": "pipeline top_k changing output shape vs Trainer NoneType GPU broadcast error; no common underlying defect." } ] }, @@ -10151,7 +10225,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 33, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -10171,7 +10245,6 @@ "issue:42915", "issue:42994", "issue:43065", - "issue:43066", "issue:43278", "issue:43381", "issue:43412", @@ -10180,14 +10253,15 @@ "issue:43653", "issue:43701", "issue:43756", + "issue:43976", "issue:43994", "issue:44186", "issue:44206", "issue:44265", "issue:44361", "issue:44442", - "issue:44462", "issue:44479", + "issue:44568", "issue:44743", "issue:44871", "issue:44977", @@ -10209,252 +10283,259 @@ "issue:46693" ], "soft_pairs": [ - "issue:43278|issue:43575", + "issue:43976|issue:45440", + "issue:44568|issue:45491", "issue:44871|issue:44977", - "issue:43381|issue:43701", - "issue:33357|issue:39692", - "issue:29127|issue:34567", "issue:45588|issue:45823", "issue:41720|issue:45229", + "issue:29127|issue:34567", + "issue:43381|issue:43701", + "issue:43278|issue:43575", + "issue:33357|issue:39692", "issue:44977|issue:45563", "issue:45198|issue:45803", - "issue:44361|issue:45198", "issue:44186|issue:45198", - "issue:43653|issue:44743", + "issue:44361|issue:45198", "issue:43611|issue:46121", + "issue:43653|issue:44743", "issue:29870|issue:34689", - "issue:42915|issue:42994", "issue:45372|issue:45588", "issue:43756|issue:45440", - "issue:43653|issue:44442", + "issue:42915|issue:42994", "issue:29942|issue:30064", "issue:44206|issue:44479", - "issue:43412|issue:45588", - "issue:45459|issue:45712", + "issue:43653|issue:44442", "issue:45372|issue:45406", + "issue:45459|issue:45712", "issue:45491|issue:46693", "issue:44265|issue:45198", - "issue:29127|issue:33357", + "issue:43412|issue:45588", "issue:43065|issue:43994", - "issue:43065|issue:45412", - "issue:43575|issue:43701", + "issue:29127|issue:33357", "issue:45245|issue:45563", - "issue:43066|issue:44462" + "issue:43065|issue:45412", + "issue:43575|issue:43701" ], - "prepared_review_unit_hash": "f15ad2d551b40105ae44d07cb06027ebae7aec44af41dca624f1621fdbef5d19", + "prepared_review_unit_hash": "b2c3b36edc6a4c879816c3cbf85dbad4dd6fc883b8e656fd0934e2ffbe9e126b", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 31, - "serialized_chars": 37051, - "estimated_input_tokens": 9263, - "estimated_eval_tokens": 18782 + "soft_pair_count": 32, + "serialized_chars": 37412, + "estimated_input_tokens": 9353, + "estimated_eval_tokens": 18962 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is a loose bundle of Transformers bugs with a few superficial theme overlaps (tokenizers, generation, multimodal models), but the paired items do not describe the same underlying defect. I would not merge any of the soft pairs; they span different models, subsystems, and failure modes.", - "confidence": 0.96, - "canonical_issue_reason": "Issue 44462 is the strongest general-purpose representative because it concerns a core AutoTokenizer loading path and has broader impact than the more model-specific tokenizer bugs. That said, the cluster is not truly a duplicate set.", + "summary": "This cluster is heterogeneous: the items span unrelated bugs in tokenizers, generation, model loading, quantization, device mapping, and documentation. None of the soft pairs look like the same underlying issue, so all are rejected.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "44462 is the best issue to anchor the cluster triage: it is a central, user-facing loading regression with clear reproduction potential and wider applicability than the other, narrower reports.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43278", - "right": "issue:43575", + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Different failure classes: a Python version compatibility break vs a DeepSeek implementation divergence." + }, + { + "left": "issue:44568", + "right": "issue:45491", "accept": false, - "reason": "One is an eval-time dtype drift in embeddings; the other is a tensor-parallel OOM on a specific Qwen2 57B load path. Different bug classes and fixes." + "reason": "Tokenizer special-token behavior is unrelated to Gemma3 NaN embeddings in mixed-length batching." }, { "left": "issue:44871", "right": "issue:44977", "accept": false, - "reason": "EOS token config mismatch in Gemma-3 vs flash-attention generation failure in Qwen3.5. Different models and different code paths." + "reason": "EOS token config mismatch is distinct from a flash-attention generation bug." }, { - "left": "issue:43381", - "right": "issue:43701", + "left": "issue:45588", + "right": "issue:45823", "accept": false, - "reason": "Gradient checkpointing in eval mode is a runtime-mode constraint bug; resume_from_checkpoint key mismatch is a checkpoint-loading naming mismatch. Not the same defect." + "reason": "Different code paths: flash_attention integration crash vs Gemma4 PLE device placement mismatch." }, { - "left": "issue:33357", - "right": "issue:39692", + "left": "issue:41720", + "right": "issue:45229", "accept": false, - "reason": "MacOS bus error with a community CLIP model vs SigLIP2 docs example errors. Same broad vision area, but unrelated failures." + "reason": "Both are inference failures, but one is a CUDA assert during auto device mapping and the other is multi-GPU OOM." }, { "left": "issue:29127", "right": "issue:34567", "accept": false, - "reason": "LayoutLMv3 box/error-message issue vs TrainerState token-count update bug. Completely different components." + "reason": "LayoutLMv3 error messaging is unrelated to TrainerState token-count tracking." }, { - "left": "issue:45588", - "right": "issue:45823", + "left": "issue:43381", + "right": "issue:43701", "accept": false, - "reason": "Flash-attention sink handling crash vs Gemma4 PLE device mismatch. Both are runtime crashes, but they are different code paths and symptoms." + "reason": "Eval-mode gradient checkpointing incompatibility is separate from checkpoint resume key mismatch." }, { - "left": "issue:41720", - "right": "issue:45229", + "left": "issue:43278", + "right": "issue:43575", "accept": false, - "reason": "Qwen3 auto device mapping CUDA assert vs Gemma4 multi-GPU OOM. Different models and different resource-allocation failures." + "reason": "Embedding dtype drift in eval is unrelated to tensor-parallel OOM on a large Qwen model." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "MacOS bus error on CLIP loading is not the same bug as SigLIP2 documentation example errors." }, { "left": "issue:44977", "right": "issue:45563", "accept": false, - "reason": "Qwen3.5 flash-attention generation bug vs a stale warning in paged generate. One is correctness, the other is warning behavior." + "reason": "Generation malfunction with flash-attention is distinct from a stale warning in paged generate()." }, { "left": "issue:45198", "right": "issue:45803", "accept": false, - "reason": "Wav2Vec2 save/tokenization failure vs FuyuBatchFeature swallowing errors. Different subsystems and error mechanisms." - }, - { - "left": "issue:44361", - "right": "issue:45198", - "accept": false, - "reason": "MLuke tokenizer AttributeError vs Wav2Vec2 save_pretrained/tokenization failure. Both tokenizer-related, but not the same model or bug." + "reason": "Wav2Vec2 save/tokenization failures do not match a bare-except bug in Fuyu tensor conversion." }, { "left": "issue:44186", "right": "issue:45198", "accept": false, - "reason": "LayoutLMv2 tokenizer crashes on NER/padding vs Wav2Vec2 save/tokenization failure. Different tokenizer implementations and failure modes." + "reason": "Different models and symptoms: LayoutLMv2 tokenizer crashes vs Wav2Vec2 save/tokenization failures." }, { - "left": "issue:43653", - "right": "issue:44743", + "left": "issue:44361", + "right": "issue:45198", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug vs Qwen3.5 recurrent state reset in modular model code. Unrelated." + "reason": "MLuke task-time AttributeError is unrelated to Wav2Vec2 save/tokenization behavior." }, { "left": "issue:43611", "right": "issue:46121", "accept": false, - "reason": "base_model_prefix loading regression vs rope-validation TypeError from JSON-loaded lists. Different config/loading paths." + "reason": "Base-model-prefix loading breakage is not the same as rope-parameter list handling." }, { - "left": "issue:29870", - "right": "issue:34689", + "left": "issue:43653", + "right": "issue:44743", "accept": false, - "reason": "BART generate decoder_attention_mask initialization bug vs Llama 3.2 Vision model loading breakage. Different models and failure points." + "reason": "BigBirdTokenizer special-token registration and Qwen3.5 recurrent-state reset are unrelated." }, { - "left": "issue:42915", - "right": "issue:42994", + "left": "issue:29870", + "right": "issue:34689", "accept": false, - "reason": "Qwen3Moe fails with FineGrainedFP8Config vs quantized model saving failure. Same quantization domain, but different operations and code paths." + "reason": "BART generate() UnboundLocalError differs from a Llama 3.2 Vision loading regression." }, { "left": "issue:45372", "right": "issue:45588", "accept": false, - "reason": "Gemma 4 processor loading ImportError vs flash_attention sink-less crash. Different error sources and fixes." + "reason": "Gemma4 processor import failure is a different bug than sink-less flash_attention None handling." }, { "left": "issue:43756", "right": "issue:45440", "accept": false, - "reason": "Smollm3 RoPE layer omission vs DeepseekV3MoE divergence from remote implementation. Both architecture discrepancies, but not the same model or bug." + "reason": "Smollm3 RoPE-layer truncation is a different model behavior than DeepSeekV3MoE divergence." }, { - "left": "issue:43653", - "right": "issue:44442", + "left": "issue:42915", + "right": "issue:42994", "accept": false, - "reason": "BigBirdTokenizer special-token registration issue vs AutoTokenizer failing to load a specific tokenizer class. Related to tokenizers, but distinct defects." + "reason": "FineGrainedFP8Config model failure and quantized model saving failure are distinct quantization issues." }, { "left": "issue:29942", "right": "issue:30064", "accept": false, - "reason": "Flash Attention 2 tests failing vs image processor void segmentation map handling. Different subsystems and problem statements." + "reason": "Flash Attention 2 test failures are unrelated to void segmentation-map processing." }, { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Unsupported center arg in LasrFeatureExtractor vs v5.3.0 video input regression for Qwen VL variants. Different APIs and models." + "reason": "Unsupported center arg in LasrFeatureExtractor is not the same as the Qwen video-input regression." }, { - "left": "issue:43412", - "right": "issue:45588", + "left": "issue:43653", + "right": "issue:44442", "accept": false, - "reason": "Gemma3n export guard/erfinv support issue vs flash_attention s_aux=None crash. Different export/runtime paths." + "reason": "BigBirdTokenizer empty decode issue is unrelated to AutoTokenizer loading FastSpeech2ConformerTokenizer." }, { - "left": "issue:45459", - "right": "issue:45712", + "left": "issue:45372", + "right": "issue:45406", "accept": false, - "reason": "Tokenizer error masking when protobuf is absent vs leftover dummy classes leaking into dir(transformers). One is exception handling, the other repo hygiene." + "reason": "Gemma4 processor import error and missing _tokenizer on Gemma4Processor are different failure points." }, { - "left": "issue:45372", - "right": "issue:45406", + "left": "issue:45459", + "right": "issue:45712", "accept": false, - "reason": "Gemma4 processor import failure vs Gemma4Processor missing _tokenizer in serve. Same family, but different failing stages and root causes." + "reason": "Tokenizer error masking with missing protobuf is unrelated to leftover dummy class cleanup." }, { "left": "issue:45491", "right": "issue:46693", "accept": false, - "reason": "Gemma3 NaN embeddings from all-padding windows vs FA2 long-context performance regression. Numerical correctness vs performance regression." + "reason": "NaN embeddings in Gemma3 are a correctness bug, not the same as a performance regression in FA2 long-context generation." }, { "left": "issue:44265", "right": "issue:45198", "accept": false, - "reason": "torch.export compatibility with torch_compilable_check vs Wav2Vec2 tokenization/save failure. Different functionality entirely." + "reason": "torch.export failures from torch_compilable_check are unrelated to Wav2Vec2 save/tokenization bugs." }, { - "left": "issue:29127", - "right": "issue:33357", + "left": "issue:43412", + "right": "issue:45588", "accept": false, - "reason": "LayoutLMv3 box info error message vs CLIP bus error on MacOS. No shared underlying bug." + "reason": "Executorch export guard/erfinv support problems do not match the flash_attention sink-less-model crash." }, { "left": "issue:43065", "right": "issue:43994", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder vs SigLIP2 AutoModel/pipeline bad outputs. Same vision area only at a very high level; not the same defect." + "reason": "A dummy Conv2d in Sam3PixelDecoder is unrelated to SigLIP2 AutoModel/pipeline nonsense outputs." }, { - "left": "issue:43065", - "right": "issue:45412", + "left": "issue:29127", + "right": "issue:33357", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder vs RT-DETR memory not released. Different model internals and symptoms." + "reason": "LayoutLMv3 box-information errors are not the same as the CLIP MacOS bus error." }, { - "left": "issue:43575", - "right": "issue:43701", + "left": "issue:45245", + "right": "issue:45563", "accept": false, - "reason": "Tensor-parallel OOM on Qwen2-57B vs resume_from_checkpoint key mismatch. Resource sizing vs checkpoint key naming." + "reason": "Category-count overflow is unrelated to the stale num_return_sequences warning." }, { - "left": "issue:45245", - "right": "issue:45563", + "left": "issue:43065", + "right": "issue:45412", "accept": false, - "reason": "Category-count limit RuntimeError vs stale generate warning. Unrelated issues in different layers." + "reason": "A dummy convolution placeholder is unrelated to RT-DETR memory not being released." }, { - "left": "issue:43066", - "right": "issue:44462", + "left": "issue:43575", + "right": "issue:43701", "accept": false, - "reason": "Wrong tokenizer decoder type in v5 vs AutoTokenizer ignoring tokenizer.json. Both tokenizer-loading related, but one is decoder-type selection and the other is file precedence/selection." + "reason": "Tensor-parallel OOM and resume_from_checkpoint key mismatch are different problems." } ] }, @@ -10470,7 +10551,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 34, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -10485,6 +10566,7 @@ "issue:35141", "issue:42490", "issue:42915", + "issue:43066", "issue:43334", "issue:43381", "issue:43421", @@ -10502,13 +10584,13 @@ "issue:44279", "issue:44387", "issue:44442", + "issue:44462", "issue:44464", "issue:44514", "issue:45072", "issue:45250", "issue:45372", "issue:45381", - "issue:45406", "issue:45412", "issue:45440", "issue:45459", @@ -10523,242 +10605,241 @@ "issue:46369", "issue:46399", "issue:46640", - "issue:46693", - "issue:46701" + "issue:46693" ], "soft_pairs": [ "issue:45440|issue:45593", "issue:42490|issue:43381", "issue:29127|issue:35141", + "issue:43066|issue:44462", "issue:45372|issue:45459", - "issue:45406|issue:46701", "issue:43421|issue:45491", "issue:43994|issue:45412", - "issue:45412|issue:46369", "issue:43575|issue:44387", + "issue:45412|issue:46369", "issue:44442|issue:45072", "issue:44464|issue:45584", - "issue:43526|issue:45834", "issue:42915|issue:43931", + "issue:43526|issue:45834", "issue:43824|issue:44279", "issue:43526|issue:46693", "issue:27957|issue:29127", - "issue:27957|issue:33290", - "issue:45250|issue:46640", "issue:43994|issue:46369", + "issue:27957|issue:33290", + "issue:43526|issue:44186", "issue:33290|issue:34689", "issue:33290|issue:35141", - "issue:43526|issue:44186", - "issue:43334|issue:43531", "issue:46121|issue:46399", - "issue:45588|issue:45600", + "issue:43334|issue:43531", + "issue:45250|issue:46640", "issue:43381|issue:43873", - "issue:44514|issue:45381", + "issue:45588|issue:45600", + "issue:43761|issue:45859", "issue:43646|issue:44077", - "issue:43761|issue:45859" + "issue:44514|issue:45381" ], - "prepared_review_unit_hash": "6ed3b8a164118e43785dfdbc8eb8da51d2e17a10575c2f4dbe5c9939f2854e26", + "prepared_review_unit_hash": "1262e30054e7228d92dd2a55bc391ea83a9ca6efb203fc57a08b34b4cbe01e49", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 35754, - "estimated_input_tokens": 8939, - "estimated_eval_tokens": 18134 + "serialized_chars": 35674, + "estimated_input_tokens": 8919, + "estimated_eval_tokens": 18094 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the items cover many unrelated bugs/features across model loading, tokenizers, vision, quantization, and flash-attention. None of the soft pairs looks like the same underlying issue, so there is no meaningful duplicate canonical within this set.", - "confidence": 0.97, - "canonical_issue_reason": "No single canonical issue fits: the cluster spans unrelated failure modes and subsystems, so treating any one issue as the representative duplicate would be misleading.", + "summary": "The candidate pairs are largely unrelated and should all be rejected as duplicates. If a representative issue is needed, #43873 is the strongest anchor because it is open, broad, and has the most inbound references/discussion.", + "confidence": 0.42, + "canonical_issue_reason": "#43873 is the most broadly referenced and actively discussed open issue in the set, so it is the best representative anchor if one must be chosen.", "canonical_pr_reason": null, - "best_issue_reason": "No issue is a strong global best representative because the set is not a coherent duplicate cluster; the safest triage is to leave them separate.", + "best_issue_reason": "#43873 is the strongest issue candidate overall: open, high discussion/inbound references, and framed as a broader regression rather than a narrow model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45440", "right": "issue:45593", "accept": false, - "reason": "Different bugs in different models: DeepseekV3MoE divergence vs D-FINE auxiliary-loss behavior." + "reason": "Different model families and different failure modes: DeepseekV3MoE divergence vs D-FINE auxiliary loss behavior." }, { "left": "issue:42490", "right": "issue:43381", "accept": false, - "reason": "Path/import warning on local model loading is unrelated to gradient checkpointing in eval mode." + "reason": "Local-path import warning and gradient-checkpointing eval-mode error are unrelated bugs." }, { "left": "issue:29127", "right": "issue:35141", "accept": false, - "reason": "LayoutLMv3 error messaging and token-embedding reinit are distinct code paths and symptoms." + "reason": "LayoutLMv3 error messaging and embedding reinitialization on resize are separate code paths." }, { - "left": "issue:45372", - "right": "issue:45459", + "left": "issue:43066", + "right": "issue:44462", "accept": false, - "reason": "Gemma 4 processor import breakage vs tokenizer error masking from protobuf handling are different problems." + "reason": "Both touch tokenizer loading, but one is a decoder-type mismatch in v5 and the other is AutoTokenizer ignoring tokenizer.json; not the same bug." }, { - "left": "issue:45406", - "right": "issue:46701", + "left": "issue:45372", + "right": "issue:45459", "accept": false, - "reason": "Gemma4Processor attribute error in serve is unrelated to DiffusionGemmaGenerationOutput pipeline failure." + "reason": "Gemma processor import failure due to mistral_common differs from hidden tokenizer errors when protobuf is missing." }, { "left": "issue:43421", "right": "issue:45491", "accept": false, - "reason": "Runtime special-token post-processor updates and Gemma3 NaN embeddings are different tokenizer/attention issues." + "reason": "Runtime special-token post-processor updates and Gemma3 sliding-window NaNs are unrelated." }, { "left": "issue:43994", "right": "issue:45412", "accept": false, - "reason": "Siglip2 wrong outputs with AutoModel/pipeline is unrelated to RT-DETR memory release on deletion." + "reason": "SigLIP2 bad outputs and RT-DETR memory release issues are different problems." }, { - "left": "issue:45412", - "right": "issue:46369", + "left": "issue:43575", + "right": "issue:44387", "accept": false, - "reason": "RT-DETR GC/memory leak and SAM3-LiteText memory overflow are separate memory issues in different models." + "reason": "Both involve OOMs, but one is tensor-parallel model loading and the other is int4 quantization memory regression." }, { - "left": "issue:43575", - "right": "issue:44387", + "left": "issue:45412", + "right": "issue:46369", "accept": false, - "reason": "Tensor-parallel OOM for Qwen2-57B-A14B-Instruct is not the same as int4 quantization reserved-memory regression." + "reason": "Memory leak on deletion and memory overflow during SAM3 LiteText call are not the same underlying issue." }, { "left": "issue:44442", "right": "issue:45072", "accept": false, - "reason": "Tokenizer loading failure and bfloat16 dtype mismatches are unrelated bugs." + "reason": "Tokenizer loading failure and bf16 dtype mismatch in inference are unrelated." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency is a different path from Whisper empty-transcription failure after special-token alignment." + "reason": "Compiled-forward generation inconsistency and Whisper empty-transcription failure after special-token alignment are different code paths." }, { - "left": "issue:43526", - "right": "issue:45834", + "left": "issue:42915", + "right": "issue:43931", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug and Kosmos2.5 OCR index error do not share the same underlying code defect." + "reason": "Qwen3Moe FP8 config failure and Qwen3-VL shape mismatch are separate model-loading bugs." }, { - "left": "issue:42915", - "right": "issue:43931", + "left": "issue:43526", + "right": "issue:45834", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure and Qwen3-VL weight-shape mismatch are distinct model-loading issues." + "reason": "BeitImageProcessorFast label reduction bug and Kosmos2.5 OCR long-input index error are not the same defect." }, { "left": "issue:43824", "right": "issue:44279", "accept": false, - "reason": "Qwen2_5_VL import error and generic dependency issue are not the same bug." + "reason": "Missing Qwen2_5_VL import and a generic dependency issue are too vague/different to merge." }, { "left": "issue:43526", "right": "issue:46693", "accept": false, - "reason": "BeitImageProcessorFast label reduction and FA2 long-context performance regression are unrelated." + "reason": "Label reduction in an image processor and FA2 long-context performance regression are unrelated." }, { "left": "issue:27957", "right": "issue:29127", "accept": false, - "reason": "XLMRoberta Flash Attention 2 support and LayoutLMv3 error messaging are different features/issues." + "reason": "Flash Attention 2 on XLMRoberta and LayoutLMv3 box-validation messaging are unrelated." }, { - "left": "issue:27957", - "right": "issue:33290", + "left": "issue:43994", + "right": "issue:46369", "accept": false, - "reason": "Flash Attention 2 support for XLMRoberta is unrelated to Adafactor/DeepSpeed OOM." + "reason": "SigLIP2 inference outputs and SAM3 LiteText memory overflow are different issues." }, { - "left": "issue:45250", - "right": "issue:46640", + "left": "issue:27957", + "right": "issue:33290", "accept": false, - "reason": "Both mention Flash Attention 2, but one is a generic topic placeholder and the other is an issue report; not enough evidence of the same concrete bug." + "reason": "Flash Attention 2 support and Adafactor+DeepSpeed OOM are unrelated." }, { - "left": "issue:43994", - "right": "issue:46369", + "left": "issue:43526", + "right": "issue:44186", "accept": false, - "reason": "Siglip2 wrong outputs and SAM3-LiteText memory overflow are unrelated model-specific failures." + "reason": "BeitImageProcessorFast label reduction and LayoutLMv2Tokenizer NER/padding crashes are different tokenizer/processor bugs." }, { "left": "issue:33290", "right": "issue:34689", "accept": false, - "reason": "Adafactor+DeepSpeed OOM is unrelated to Llama 3.2 Vision load breakage." + "reason": "Adafactor OOM in DeepSpeed and Llama 3.2 Vision loading breakage are unrelated." }, { "left": "issue:33290", "right": "issue:35141", "accept": false, - "reason": "OOM in DeepSpeed optimizer use and embedding reinitialization after resize are different defects." + "reason": "OOM during optimizer use and embedding reinitialization after resize are not the same problem." }, { - "left": "issue:43526", - "right": "issue:44186", + "left": "issue:46121", + "right": "issue:46399", "accept": false, - "reason": "BeitImageProcessorFast label handling and LayoutLMv2Tokenizer NER/padding crashes are unrelated tokenization/image-processing bugs." + "reason": "Rope-parameter validation TypeError and missing decorator registration for apply_rotary_pos_emb are different rope-related failures." }, { "left": "issue:43334", "right": "issue:43531", "accept": false, - "reason": "Qwen3-VL missing pad_token_id and Qwen3-MoE sliding_window issues affect different subsystems and failure modes." + "reason": "Qwen3-VL pad_token_id import failure and Qwen3-MoE sliding_window behavior are distinct issues." }, { - "left": "issue:46121", - "right": "issue:46399", + "left": "issue:45250", + "right": "issue:46640", "accept": false, - "reason": "RoPE validation TypeError with JSON-loaded lists is unrelated to missing decorator registration for apply_rotary_pos_emb." + "reason": "Both mention Flash Attention 2, but the titles are too generic and provide no evidence of the same concrete bug." }, { - "left": "issue:45588", - "right": "issue:45600", + "left": "issue:43381", + "right": "issue:43873", "accept": false, - "reason": "Flash-attention sink handling crash and auto_mappings stale config references are unrelated." + "reason": "Gradient checkpointing in eval mode and offloading with quantization are separate training/runtime behaviors." }, { - "left": "issue:43381", - "right": "issue:43873", + "left": "issue:45588", + "right": "issue:45600", "accept": false, - "reason": "Gradient checkpointing in eval mode and quantization offloading behavior are different problems." + "reason": "Flash-attention sinkless-model crash and auto_mappings CI breakage are unrelated." }, { - "left": "issue:44514", - "right": "issue:45381", + "left": "issue:43761", + "right": "issue:45859", "accept": false, - "reason": "Batched apply_chat_template padding crash and Qwen2.5-VL video vision_position_ids bug are distinct Qwen2.5-VL issues." + "reason": "CLIP hidden_states regression and Qwen3_5Moe CPU-offload incompatibility are different model-specific bugs." }, { "left": "issue:43646", "right": "issue:44077", "accept": false, - "reason": "Custom model initialization breakage in Transformers 5.0.0 is not the same as patchtsmixer post_init policy." + "reason": "Custom model initialization breakage in Transformers 5.0 and patchtsmixer's post_init policy are not the same underlying issue." }, { - "left": "issue:43761", - "right": "issue:45859", + "left": "issue:44514", + "right": "issue:45381", "accept": false, - "reason": "CLIPVisionModel hidden_states regression and Qwen3_5Moe CPU offload incompatibility are unrelated." + "reason": "Qwen2.5-VL batched chat-template crash and video position-id bug are related only by model family, not the same code-path problem." } ] }, @@ -10774,7 +10855,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 35, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -10840,216 +10921,216 @@ "issue:45003|issue:45440", "issue:43720|issue:44442", "issue:42175|issue:46234", - "issue:27957|issue:29870", - "issue:43316|issue:43525", "issue:43493|issue:43827", - "issue:44062|issue:45907", "issue:44077|issue:44479", - "issue:43873|issue:44871", + "issue:27957|issue:29870", + "issue:44062|issue:45907", + "issue:43316|issue:43525", "issue:45325|issue:46697", + "issue:43873|issue:44871", "issue:44855|issue:45137", - "issue:43643|issue:44913", "issue:43749|issue:43931", + "issue:43643|issue:44913", "issue:43316|issue:45759", - "issue:44987|issue:46178", - "issue:46829|issue:46932", "issue:44190|issue:44496", "issue:44442|issue:45803", "issue:43761|issue:44188", "issue:43761|issue:44079", + "issue:44987|issue:46178", "issue:43931|issue:45216", - "issue:43232|issue:43388" + "issue:43232|issue:43388", + "issue:46829|issue:46932" ], - "prepared_review_unit_hash": "ad98071974817fa3a58575f8ad6cd23f819a62c152472d40c9ee3669f2dee0eb", + "prepared_review_unit_hash": "8345dc218082e15e323ddae14fbb7b7bfbd615b30a2321055f4d0f5c675f2588", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 27, - "serialized_chars": 36180, - "estimated_input_tokens": 9045, - "estimated_eval_tokens": 18346 + "serialized_chars": 36183, + "estimated_input_tokens": 9046, + "estimated_eval_tokens": 18348 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The candidate pairs are all superficial similarities at best; they describe different models, APIs, code paths, or failure modes. I would not merge any of them as duplicates.", - "confidence": 0.97, - "canonical_issue_reason": "No single canonical issue: this set does not represent one underlying bug, but many unrelated reports across generation, config, loading, audio, vision, quantization, and docs.", + "summary": "The cluster is a collection of largely unrelated issue reports across many different models and subsystems; none of the soft pairs look like the same underlying bug or change.", + "confidence": 0.99, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "No best issue for duplicate triage here; none of the issues is a credible umbrella duplicate target for the others.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44977", "right": "issue:45245", "accept": false, - "reason": "Qwen3.5 flash-attention generation bug vs a categories-cardinality runtime error; unrelated failures." + "reason": "Different failures in different areas: Qwen3.5 flash-attn generation vs a categories cardinality runtime error." }, { "left": "issue:44743", "right": "issue:44811", "accept": false, - "reason": "Different components: recurrent-state cache handling vs Whisper batch_decode skip_special_tokens." + "reason": "Unrelated subsystems: Qwen recurrent cache state reset vs Whisper batch_decode skip_special_tokens handling." }, { "left": "issue:44442", "right": "issue:44898", "accept": false, - "reason": "Tokenizer loading failure vs Perceiver image-classification resizing bug; different subsystems and symptoms." + "reason": "Tokenizer loading failure for FastSpeech2Conformer is unrelated to Perceiver image-classification interpolation/padding behavior." }, { "left": "issue:43262", "right": "issue:44206", "accept": false, - "reason": "Audio chat template sampling-rate default vs feature extractor center-arg crash; not the same bug." + "reason": "Audio chat-template sampling-rate defaulting is not the same bug as LastFeatureExtractor passing an unsupported center arg." }, { "left": "issue:42490", "right": "issue:44743", "accept": false, - "reason": "Import warning on trailing-slash local model paths vs Qwen recurrent-state reset; unrelated." + "reason": "Local-path import warning and Qwen recurrent-state reset are different code paths and symptoms." }, { "left": "issue:45003", "right": "issue:45440", "accept": false, - "reason": "Unsafe sys.modules access in modeling_utils vs DeepSeekV3MoE divergence from remote implementation; different code paths." + "reason": "Both mention modeling code, but one is unsafe sys.modules access and the other is DeepseekV3MoE behavior divergence." }, { "left": "issue:43720", "right": "issue:44442", "accept": false, - "reason": "BitNet packed-weight loading bug vs FastSpeech2ConformerTokenizer loading failure; unrelated." + "reason": "BitNet accelerate-loading packed-weight bug is unrelated to tokenizer loading for FastSpeech2Conformer." }, { "left": "issue:42175", "right": "issue:46234", "accept": false, - "reason": "Missing tensorflow in torch extra vs missing import_utils helper; packaging/import issues but not the same defect." + "reason": "Package/backend availability issues, but for different backends and different missing functionality." }, { - "left": "issue:27957", - "right": "issue:29870", + "left": "issue:43493", + "right": "issue:43827", "accept": false, - "reason": "Flash Attention 2 support for XLMRoberta vs BART generate decoder_attention_mask UnboundLocalError; different models and bugs." + "reason": "SigLIP2 implementation mismatch is unrelated to docs still referencing removed pipeline() APIs." }, { - "left": "issue:43316", - "right": "issue:43525", + "left": "issue:44077", + "right": "issue:44479", "accept": false, - "reason": "Two config-field discrepancies, but for different models and different missing fields; not one underlying issue." + "reason": "Different models and bugs: PatchTSMixer post_init policy vs Qwen video-input regression." }, { - "left": "issue:43493", - "right": "issue:43827", + "left": "issue:27957", + "right": "issue:29870", "accept": false, - "reason": "SigLIP2 implementation mismatch vs docs still referencing removed pipeline(); unrelated model vs documentation issue." + "reason": "XLMRoberta flash-attention support and BART decoder_attention_mask generation error are unrelated." }, { "left": "issue:44062", "right": "issue:45907", "accept": false, - "reason": "AddedToken special-argument duplication vs list_repo_files revision forwarding; different APIs and failures." - }, - { - "left": "issue:44077", - "right": "issue:44479", - "accept": false, - "reason": "patchtsmixer post_init validation vs Qwen video-input regression; no shared code path." + "reason": "AddedToken special-argument conflict is unrelated to list_repo_files failing to forward revision." }, { - "left": "issue:43873", - "right": "issue:44871", + "left": "issue:43316", + "right": "issue:43525", "accept": false, - "reason": "Quantization/offloading behavior vs Gemma-3 eos_token_id inconsistency; unrelated." + "reason": "Both are config-related, but they concern different models and different missing/incorrect fields." }, { "left": "issue:45325", "right": "issue:46697", "accept": false, - "reason": "Qwen2.5-VL rope-index temporal scaling bug vs Swin2SR over-padding images; different preprocessing/code paths." + "reason": "Qwen2.5-VL rope-index temporal scaling and Swin2SR over-padding are distinct image/video preprocessing bugs." }, { - "left": "issue:44855", - "right": "issue:45137", + "left": "issue:43873", + "right": "issue:44871", "accept": false, - "reason": "Python 3.13 TorchScript parsing error vs DeepSpeed ZeRO3 deque underflow; unrelated runtime failures." + "reason": "Quantization/offloading behavior is unrelated to Gemma-3 eos_token_id configuration mismatch." }, { - "left": "issue:43643", - "right": "issue:44913", + "left": "issue:44855", + "right": "issue:45137", "accept": false, - "reason": "trust_remote_code missing fields on load vs rotary_pct reset on reload; both config-related but distinct bugs." + "reason": "Python 3.13 decorator parsing issue is unrelated to DeepSpeed ZeRO3 deque underflow." }, { "left": "issue:43749", "right": "issue:43931", "accept": false, - "reason": "FSDP CPU RAM efficient loading broken vs Qwen3-VL weight-shape mismatch; different loading problems." + "reason": "FSDP CPU RAM efficient loading and Qwen3-VL weight-shape mismatch are different loading problems." }, { - "left": "issue:43316", - "right": "issue:45759", - "accept": false, - "reason": "Gemma3TextConfig API mismatch vs AutoModelForCausalLM not unwrapping text_config for composite Qwen configs; different issues." - }, - { - "left": "issue:44987", - "right": "issue:46178", + "left": "issue:43643", + "right": "issue:44913", "accept": false, - "reason": "Model loading regression for physical-intelligence/fast vs backend mapping error for tensorflow_text; unrelated." + "reason": "Remote-code config field loss and GPTNeoX rotary_pct reload reset are separate config serialization issues." }, { - "left": "issue:46829", - "right": "issue:46932", + "left": "issue:43316", + "right": "issue:45759", "accept": false, - "reason": "VideoPrism output field assignment bug vs Gemma4 quantized embedder dtype casting; different models and symptoms." + "reason": "Different composite-config problems: Gemma3TextConfig API discrepancy vs AutoModelForCausalLM not unwrapping text_config for Qwen 3.5/3.6." }, { "left": "issue:44190", "right": "issue:44496", "accept": false, - "reason": "Local dataset loading in training script vs unrecognized model/config.json error; unrelated." + "reason": "Dataset loading in example script is unrelated to AutoConfig failing on a model without model_type." }, { "left": "issue:44442", "right": "issue:45803", "accept": false, - "reason": "Tokenizer loading failure vs bare-except swallowing errors in FuyuBatchFeature; different components." + "reason": "Tokenizer loading failure is unrelated to FuyuBatchFeature swallowing exceptions in tensor conversion." }, { "left": "issue:43761", "right": "issue:44188", "accept": false, - "reason": "CLIPVisionModel hidden_states regression vs attention kernel divergence under torch.compile; unrelated." + "reason": "CLIPVisionModel hidden_states regression is unrelated to attention-kernel divergence under torch.compile." }, { "left": "issue:43761", "right": "issue:44079", "accept": false, - "reason": "Hidden_states regression in CLIPVisionModel vs ModelOutput key assignment bug; different code paths." + "reason": "Hidden_states being None and ModelOutput key assignment with prior None are different output-handling bugs." + }, + { + "left": "issue:44987", + "right": "issue:46178", + "accept": false, + "reason": "Loading a specific model with transformers>=5.1.0 is unrelated to a missing BACKENDS_MAPPING entry for tensorflow_text." }, { "left": "issue:43931", "right": "issue:45216", "accept": false, - "reason": "Qwen3-VL load-time shape mismatch vs Qwen3.5 save_pretrained checkpoint regression; related family, but different concrete bugs." + "reason": "Both involve Qwen models, but one is a load-time shape mismatch and the other is a save_pretrained checkpoint regression." }, { "left": "issue:43232", "right": "issue:43388", "accept": false, - "reason": "Generation kwargs update after sync_gpus vs gather_for_metrics dropping tuple labels; different distributed-training bugs." + "reason": "Generation kwargs update after sync_gpus and gather_for_metrics label truncation are unrelated distributed-training bugs." + }, + { + "left": "issue:46829", + "right": "issue:46932", + "accept": false, + "reason": "VideoPrism output field mixup and Gemma4 quantized embedders dtype casting are unrelated model issues." } ] }, @@ -11065,7 +11146,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 36, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -11079,7 +11160,6 @@ "issue:27957", "issue:28282", "issue:32090", - "issue:42175", "issue:42617", "issue:42907", "issue:43295", @@ -11120,15 +11200,16 @@ "issue:46421", "issue:46424", "issue:46506", + "issue:46821", "issue:46894" ], "soft_pairs": [ "issue:43701|issue:46506", "issue:42907|issue:45698", "issue:42617|issue:43575", - "issue:27957|issue:32090", "issue:43550|issue:43701", "issue:43575|issue:43742", + "issue:27957|issue:32090", "issue:43606|issue:43701", "issue:25251|issue:28282", "issue:46234|issue:46291", @@ -11137,217 +11218,217 @@ "issue:43295|issue:44871", "issue:44871|issue:45584", "issue:44464|issue:46421", - "issue:43316|issue:43716", - "issue:42175|issue:43316", "issue:44077|issue:44534", + "issue:43316|issue:43716", + "issue:43316|issue:46821", "issue:44871|issue:46050", "issue:43646|issue:44913", - "issue:45803|issue:46365", "issue:43452|issue:44496", + "issue:45803|issue:46365", "issue:44162|issue:44792", - "issue:44242|issue:46894", "issue:43404|issue:43441", "issue:45632|issue:46424", + "issue:44242|issue:46894", "issue:45803|issue:46170", - "issue:43650|issue:46202", - "issue:44610|issue:46369" + "issue:44610|issue:46369", + "issue:43650|issue:46202" ], - "prepared_review_unit_hash": "9e23f7273151ed00f58db88cb47378d2504b932a9c9a87177dd3fc788fc95d83", + "prepared_review_unit_hash": "3d771bf70d637dd17ff9b8f344b658d6b5db8cec16d037e7d140220d30600d5f", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 35736, - "estimated_input_tokens": 8934, - "estimated_eval_tokens": 18124 + "serialized_chars": 35721, + "estimated_input_tokens": 8931, + "estimated_eval_tokens": 18118 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pair looks like the same underlying bug or change. The items span unrelated model-loading, cache, quantization, generation, and dependency issues, so this cluster should not be merged into a single duplicate set.", - "confidence": 0.97, - "canonical_issue_reason": "No single issue is a clear canonical duplicate target; the candidate pairs are thematically related only at a very broad Transformers level and are mostly distinct bugs.", + "summary": "This cluster is mostly a grab bag of unrelated Transformers issues; the soft-similarity pairs share broad themes (loading, cache, generation, model configs) but not the same concrete bug or change. No soft pair is a safe duplicate merge candidate.", + "confidence": 0.71, + "canonical_issue_reason": "issue:45698 is the broadest and most active report here: it is open, has the most discussion/inbound references, and describes a concrete `from_pretrained`/`save_pretrained` custom-module loading failure that could serve as the cluster\u2019s representative loading bug.", "canonical_pr_reason": null, - "best_issue_reason": "None of the issues is a good global representative for the cluster because the set is heterogeneous and lacks a common concrete code-path failure.", + "best_issue_reason": "issue:45698 is the strongest standalone issue for follow-up because it is specific, reproducible, and has the clearest user impact among the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43701", "right": "issue:46506", "accept": false, - "reason": "Key mismatch on checkpoint resume is unrelated to CPU/CUDA inference divergence." + "reason": "Different failures: checkpoint key mismatch vs CPU/CUDA result divergence." }, { "left": "issue:42907", "right": "issue:45698", "accept": false, - "reason": "Both involve save/load flows, but one is about dequantized model saving and the other about loading the wrong custom module after save_pretrained." + "reason": "Both involve saving/loading, but one is dequantized model save failure and the other is wrong custom module loaded after save/load." }, { "left": "issue:42617", "right": "issue:43575", "accept": false, - "reason": "3d_parallel.py execution failure is not the same bug as tp-related OOM while loading Qwen2." - }, - { - "left": "issue:27957", - "right": "issue:32090", - "accept": false, - "reason": "FlashAttention support for XLM-Roberta is unrelated to the Trainer _gpu_broadcast_one NoneType error." + "reason": "Parallel script startup failure and tensor-parallel OOM are distinct problems." }, { "left": "issue:43550", "right": "issue:43701", "accept": false, - "reason": "torch.compile with SDPA on Bamba is unrelated to resume_from_checkpoint key mismatches." + "reason": "torch.compile + SDPA bug is unrelated to resume-from-checkpoint key mismatch." }, { "left": "issue:43575", "right": "issue:43742", "accept": false, - "reason": "OOM on a large Qwen2 model is not the same as a key error loading MobileLLM-125M." + "reason": "OOM during tensor-parallel load is not the same as a key error while loading a model." + }, + { + "left": "issue:27957", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention support issue and Trainer broadcast TypeError are unrelated." }, { "left": "issue:43606", "right": "issue:43701", "accept": false, - "reason": "CPU offload device mismatch is unrelated to checkpoint key mismatch." + "reason": "CPU offload device mismatch is not the same as checkpoint key mismatch." }, { "left": "issue:25251", "right": "issue:28282", "accept": false, - "reason": "Pipeline top_k output shape change is unrelated to a missing PyTorch dependency." + "reason": "Pipeline output-shape regression and missing PyTorch import are unrelated." }, { "left": "issue:46234", "right": "issue:46291", "accept": false, - "reason": "Missing import_utils helper and pipeline import failure with kernels v0.15 are different dependency/import bugs." + "reason": "Both are import-time dependency issues, but one is a missing utility symbol and the other is a kernels version/revision requirement." }, { "left": "issue:43295", "right": "issue:43825", "accept": false, - "reason": "Custom processor.tokenizer regression is unrelated to the pipeline translation error message bug." + "reason": "Processor/tokenizer regression and misleading pipeline error messaging are different bugs." }, { "left": "issue:16998", "right": "issue:22355", "accept": false, - "reason": "A model_max_length question is unrelated to the transformers.onnx import error." + "reason": "model_max_length behavior and missing `transformers.onnx` module are unrelated." }, { "left": "issue:43295", "right": "issue:44871", "accept": false, - "reason": "Processor/tokenizer regression is unrelated to Gemma eos_token_id configuration inconsistency." + "reason": "Processor API regression and eos_token_id config mismatch are not the same underlying issue." }, { "left": "issue:44871", "right": "issue:45584", "accept": false, - "reason": "Gemma eos_token_id mismatch and Whisper empty-transcription failure are distinct special-token issues with different code paths." + "reason": "Both involve special tokens, but one is a config inconsistency and the other is a Whisper empty-transcription generation failure." }, { "left": "issue:44464", "right": "issue:46421", "accept": false, - "reason": "Compiled-forward output inconsistency is not the same bug as spurious recompilation from chunked prefill cache lazy initialization." + "reason": "Chunked generation inconsistency and cache lazy-init recompilation are related to compilation, but not the same bug." }, { - "left": "issue:43316", - "right": "issue:43716", + "left": "issue:44077", + "right": "issue:44534", "accept": false, - "reason": "Gemma3TextConfig API discrepancy is unrelated to image preprocessor/model dtype mismatch." + "reason": "These are separate v5 issues: post_init policy vs non-persistent buffer initialization." }, { - "left": "issue:42175", - "right": "issue:43316", + "left": "issue:43316", + "right": "issue:43716", "accept": false, - "reason": "Missing TensorFlow from the torch extra is unrelated to a config API discrepancy." + "reason": "Config API discrepancy and image-preprocessor dtype mismatch are unrelated." }, { - "left": "issue:44077", - "right": "issue:44534", + "left": "issue:43316", + "right": "issue:46821", "accept": false, - "reason": "Optional post_init restriction in patchtsmixer is unrelated to non-persistent buffers being filled with junk." + "reason": "Gemma3TextConfig API mismatch and TimesFM window_size AttributeError are different model-specific issues." }, { "left": "issue:44871", "right": "issue:46050", "accept": false, - "reason": "Gemma eos_token_id inconsistency is unrelated to quantized model save/load roundtrip failure." + "reason": "eos_token_id mismatch is not the same as quantized model save/load breakage." }, { "left": "issue:43646", "right": "issue:44913", "accept": false, - "reason": "Generic Transformers 5 init breakage is not the same as GPTNeoX rotary_pct not persisting on reload." - }, - { - "left": "issue:45803", - "right": "issue:46365", - "accept": false, - "reason": "Bare except in Fuyu tensor conversion is unrelated to importing Trainer only after uninstalling datasets." + "reason": "Custom model initialization breakage and GPTNeoX rotary_pct reload regression are distinct configuration problems." }, { "left": "issue:43452", "right": "issue:44496", "accept": false, - "reason": "gguf_file handling in from_pretrained is unrelated to a missing model_type key in config.json." + "reason": "From-pretrained loader issues, but one is gguf handling and the other is missing model_type metadata." }, { - "left": "issue:44162", - "right": "issue:44792", + "left": "issue:45803", + "right": "issue:46365", "accept": false, - "reason": "ESM2 breakage is unrelated to the janus image-generation test failure." + "reason": "Bare except swallowing errors in Fuyu preprocessing is unrelated to importing Trainer with datasets installed." }, { - "left": "issue:44242", - "right": "issue:46894", + "left": "issue:44162", + "right": "issue:44792", "accept": false, - "reason": "Missing MoE load-balancing loss and ASR training-loss double-shift are different loss-calculation bugs." + "reason": "ESM2 workflow breakage and Janus image-generation test failure are unrelated." }, { "left": "issue:43404", "right": "issue:43441", "accept": false, - "reason": "lm_head weight tying in Mistral3 is unrelated to FlashAttention failures in Ministral-3." + "reason": "Weight tying bug and FlashAttention incompatibility in the Mistral/Ministral family are different code paths." }, { "left": "issue:45632", "right": "issue:46424", "accept": false, - "reason": "Remote-code cache path collisions are unrelated to max_cache_length being honored only for Quantized Cache." + "reason": "Cache-path collision for trust_remote_code is unrelated to max_cache_length only applying to quantized cache." }, { - "left": "issue:45803", - "right": "issue:46170", + "left": "issue:44242", + "right": "issue:46894", "accept": false, - "reason": "Fuyu tensor conversion error handling is unrelated to the Windows ROCm torch.distributed dependency issue." + "reason": "Load-balancing loss omission and training-loss double-shift are not the same bug." }, { - "left": "issue:43650", - "right": "issue:46202", + "left": "issue:45803", + "right": "issue:46170", "accept": false, - "reason": "A placeholder 'ADD THE DATA' issue is unrelated to Gemma pipeline roundtrip failures." + "reason": "Tensor conversion exception handling and Windows/ROCm distributed import failure are unrelated." }, { "left": "issue:44610", "right": "issue:46369", "accept": false, - "reason": "OmDet-Turbo image-size mismatch is unrelated to sam3-litetext memory overflow." + "reason": "Processor image-size mismatch and SAM memory overflow are different failures." + }, + { + "left": "issue:43650", + "right": "issue:46202", + "accept": false, + "reason": "One issue contains no technical bug content; the other is a Gemma pipeline roundtrip failure." } ] }, @@ -11363,7 +11444,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 37, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -11389,6 +11470,7 @@ "issue:43827", "issue:43994", "issue:44190", + "issue:44479", "issue:44514", "issue:44534", "issue:44617", @@ -11400,7 +11482,6 @@ "issue:44913", "issue:44964", "issue:44977", - "issue:45072", "issue:45127", "issue:45200", "issue:45245", @@ -11417,214 +11498,206 @@ "issue:46640", "issue:46697", "issue:46832", - "issue:46858", - "issue:46932" + "issue:46858" ], "soft_pairs": [ - "issue:43761|issue:46932", - "issue:45542|issue:46234", + "issue:43761|issue:44479", "issue:43504|issue:45698", "issue:43540|issue:44514", - "issue:43572|issue:46178", - "issue:43352|issue:46640", - "issue:43065|issue:44617", - "issue:44792|issue:45600", + "issue:45542|issue:46234", + "issue:43388|issue:43688", "issue:44792|issue:44977", + "issue:44792|issue:45600", + "issue:43065|issue:44617", + "issue:43572|issue:46178", "issue:45127|issue:45200", - "issue:43388|issue:43688", "issue:43673|issue:44743", - "issue:41950|issue:43475", + "issue:43352|issue:46640", "issue:44190|issue:44964", - "issue:43525|issue:43531", - "issue:35141|issue:42490", + "issue:41950|issue:43475", "issue:44534|issue:44913", + "issue:43525|issue:43531", "issue:44734|issue:45406", + "issue:35141|issue:42490", "issue:46082|issue:46697", "issue:45632|issue:46617", - "issue:45072|issue:46932", - "issue:44855|issue:45588", "issue:45245|issue:46858", - "issue:44869|issue:46832", - "issue:43827|issue:43994" + "issue:43827|issue:43994", + "issue:44855|issue:45588", + "issue:44869|issue:46832" ], - "prepared_review_unit_hash": "d55a0cdba482b15ae2253fbb7d6b0724bd1d0e2d1ed66356e5690198db71daef", + "prepared_review_unit_hash": "ea7c6cf7dd2a6e93731466514538c94ca96a56e36ed57e83754f4f826b61970a", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 25, - "serialized_chars": 35404, - "estimated_input_tokens": 8851, - "estimated_eval_tokens": 17958 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 24, + "serialized_chars": 34531, + "estimated_input_tokens": 8633, + "estimated_eval_tokens": 17522 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose semantic grouping of unrelated Transformers issues. The only plausible common thread is model-loading/cache/trust_remote_code regressions, but the individual reports still point to different code paths and failure modes.", - "confidence": 0.39, - "canonical_issue_reason": "issue:45698 is the broadest, most actively discussed open loading/regression report in the set, so it is the least bad representative for this loose cluster.", + "summary": "This cluster is very heterogeneous: the issues span unrelated model families, config/loading regressions, processor/pipeline bugs, cache behavior, and training/runtime edge cases. I don\u2019t see any soft pair that clearly describes the same underlying bug or a PR that would merge into a single fix path.", + "confidence": 0.95, + "canonical_issue_reason": "No clear canonical issue emerges; the cluster appears to be a noisy grouping of unrelated bugs. If a representative must be chosen, issue 45698 is the broadest/openest loading bug, but it does not subsume the others.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45698 has the strongest activity signal and is broad enough to stand in for the cluster, even though the set is not a true duplicate group.", + "best_issue_reason": "No issue in this set is a strong global duplicate hub. Issue 45698 is the most broadly scoped and best-engaged open issue, but only as a weak representative, not as a true canonical duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43761", - "right": "issue:46932", - "accept": false, - "reason": "Different regressions: CLIPVision hidden-state return bug vs Gemma4 quantized embedder dtype casting." - }, - { - "left": "issue:45542", - "right": "issue:46234", + "right": "issue:44479", "accept": false, - "reason": "Unrelated subsystems: backend detection for tensorflow/tensorboard vs missing import_utils symbol for Falcon Mamba." + "reason": "Both are video/vision-regression themed, but they affect different model stacks and failure modes; not the same concrete bug." }, { "left": "issue:43504", "right": "issue:45698", "accept": false, - "reason": "Both concern loading, but one is a legacy preset field on Beit segmentation and the other is wrong custom-module resolution after save/load." + "reason": "Legacy pretrained preset loading for BEiT semantic segmentation is unrelated to custom-module loading after save_pretrained." }, { "left": "issue:43540", "right": "issue:44514", "accept": false, - "reason": "Both are multimodal processor issues, but they affect different models and different API paths." + "reason": "Qwen3Omni video-input processing and Qwen2_5_VL chat-template batching are distinct code paths and symptoms." }, { - "left": "issue:43572", - "right": "issue:46178", + "left": "issue:45542", + "right": "issue:46234", "accept": false, - "reason": "Different config/backend problems: StableLm pad token migration vs tensorflow_text backend mapping error." + "reason": "These are different backend/import-utils issues: tensorboard/TensorFlow detection vs a missing falcon-mamba availability helper." }, { - "left": "issue:43352", - "right": "issue:46640", + "left": "issue:43388", + "right": "issue:43688", "accept": false, - "reason": "Same broad Flash Attention theme, but one is an explicit model-support guard and the other is a separate FA2 failure report." + "reason": "One is gather_for_metrics dropping tuple labels; the other is auxiliary-loss normalization. Different subsystems and fixes." }, { - "left": "issue:43065", - "right": "issue:44617", + "left": "issue:44792", + "right": "issue:44977", "accept": false, - "reason": "Different SAM3 failures: a dummy Conv2d implementation issue vs CUDA OOM in video inference." + "reason": "Janus image-generation test failure and Qwen3.5 flash-attention generation regression are unrelated model-specific problems." }, { "left": "issue:44792", "right": "issue:45600", "accept": false, - "reason": "Unrelated: janus image-generation test failure vs stale auto_mappings references after config removal." + "reason": "A failed janus test does not match an auto_mappings CI break from removed Sam3LiteText configs." }, { - "left": "issue:44792", - "right": "issue:44977", + "left": "issue:43065", + "right": "issue:44617", "accept": false, - "reason": "Different models and symptoms; one is a janus test failure, the other is Qwen3.5 flash-attention generation breakage." + "reason": "Dummy Conv2d in Sam3PixelDecoder and SAM3 video OOM are different issues, not one fixable code-path bug." }, { - "left": "issue:45127", - "right": "issue:45200", + "left": "issue:43572", + "right": "issue:46178", "accept": false, - "reason": "Different bugs: LoRA merge with extended vocab and tied embeddings vs Gemma4 text-only mm_token_type_ids defaulting." + "reason": "StableLmConfig pad_token_idx regression and a backend-mapping error for tensorflow_text are unrelated configuration/import issues." }, { - "left": "issue:43388", - "right": "issue:43688", + "left": "issue:45127", + "right": "issue:45200", "accept": false, - "reason": "Different training/eval bugs: last-batch metric gathering vs auxiliary-loss normalization." + "reason": "LoRA merge collapse with tie_word_embeddings=True is not the same as Gemma 4 mm_token_type_ids defaulting behavior." }, { "left": "issue:43673", "right": "issue:44743", "accept": false, - "reason": "Both involve cache/state handling, but one is chunked-prefill GenerationMixin cache absence and the other is recurrent-state reset with seq_len>1." + "reason": "Chunked prefill cache handling and Qwen3.5 recurrent-state reset are both cache-related but on different mechanisms and code paths." }, { - "left": "issue:41950", - "right": "issue:43475", + "left": "issue:43352", + "right": "issue:46640", "accept": false, - "reason": "Different pipeline/model issues: image processor lookup in video classification vs missing fpn_position_embeddings in Sam3 vision output." + "reason": "NemotronH Flash Attention 2 unsupported error and a generic Flash Attention 2 issue are not concrete duplicates." }, { "left": "issue:44190", "right": "issue:44964", "accept": false, - "reason": "Different loading failures in different example/model contexts; no shared underlying bug is evident." + "reason": "Local dataset loading in an example script is unrelated to loading the Phi-4 multimodal model." }, { - "left": "issue:43525", - "right": "issue:43531", + "left": "issue:41950", + "right": "issue:43475", "accept": false, - "reason": "Separate config bugs: missing pad_token_id in Llama4Config vs sliding_window behavior in Qwen3-MoE." + "reason": "Video-classification processor lookup and a missing fpn_position_embeddings attribute in SAM3 video are different bugs." }, { - "left": "issue:35141", - "right": "issue:42490", + "left": "issue:44534", + "right": "issue:44913", "accept": false, - "reason": "Different issues: embedding resize/post_init reinitialization vs local-path trailing-slash import warning." + "reason": "Non-persistent buffer corruption and GPTNeoX rotary_pct reload behavior are separate configuration/runtime problems." }, { - "left": "issue:44534", - "right": "issue:44913", + "left": "issue:43525", + "right": "issue:43531", "accept": false, - "reason": "Both are serialization/config persistence regressions, but they affect different mechanisms and code paths." + "reason": "Llama4 pad_token_id missing and Qwen3-MoE sliding_window behavior are unrelated config/model issues." }, { "left": "issue:44734", "right": "issue:45406", "accept": false, - "reason": "Different failures: server KV-cache continuation indexing vs Gemma4Processor missing _tokenizer." + "reason": "Serve KV-cache continuation indexing and Gemma4Processor missing _tokenizer are distinct failures in different components." + }, + { + "left": "issue:35141", + "right": "issue:42490", + "accept": false, + "reason": "Embedding resize/post_init reinitialization and path-with-trailing-slash import warnings do not share a bug mechanism." }, { "left": "issue:46082", "right": "issue:46697", "accept": false, - "reason": "Completely different areas: config validation for head_dim vs Swin2SR image padding." + "reason": "LlamaConfig head_dim validation and Swin2SR over-padding are unrelated model-configuration vs image-processing bugs." }, { "left": "issue:45632", "right": "issue:46617", "accept": false, - "reason": "Both touch trust_remote_code/cache, but one is a leaf-directory cache collision and the other is a FileNotFoundError when loading from cache; not the same concrete bug." + "reason": "Trust-remote-code cache collision by leaf directory name is not the same as FileNotFoundError when loading from cache." }, { - "left": "issue:45072", - "right": "issue:46932", + "left": "issue:45245", + "right": "issue:46858", "accept": false, - "reason": "Different dtype issues in different models/components; not a single fixable code-path problem." + "reason": "Category-count limit and device-tensor seq_length arithmetic are completely different runtime failures." }, { - "left": "issue:44855", - "right": "issue:45588", + "left": "issue:43827", + "right": "issue:43994", "accept": false, - "reason": "Unrelated: Python 3.13 decorator parsing/IndentationError vs flash_attention crash on s_aux=None." + "reason": "Docs still referencing pipeline() and SigLIP2 producing nonsensical AutoModel/pipeline results are related only by pipelines broadly, not by the same bug." }, { - "left": "issue:45245", - "right": "issue:46858", + "left": "issue:44855", + "right": "issue:45588", "accept": false, - "reason": "Different low-level runtime failures: category-count limit vs static-cache slice arithmetic on non-CUDA backends." + "reason": "Python 3.13 indentation/parsing import failure and flash_attention.py crashing on s_aux=None are unrelated code paths." }, { "left": "issue:44869", "right": "issue:46832", "accept": false, - "reason": "Different model bugs: Whisper timestamp decoding edge case vs RT-DETR feature-level mismatch." - }, - { - "left": "issue:43827", - "right": "issue:43994", - "accept": false, - "reason": "Docs cleanup for removed pipeline APIs vs a model/pipeline correctness issue for SigLIP2." + "reason": "Whisper timestamp decoding and RT-DETR feature-level indexing are different model-specific bugs." } ] }, @@ -11640,7 +11713,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 38, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -11657,7 +11730,6 @@ "issue:43065", "issue:43295", "issue:43335", - "issue:43526", "issue:43550", "issue:43644", "issue:43673", @@ -11678,6 +11750,7 @@ "issue:44877", "issue:44913", "issue:44933", + "issue:45072", "issue:45127", "issue:45245", "issue:45276", @@ -11693,253 +11766,243 @@ "issue:46421", "issue:46489", "issue:46655", - "issue:46697", "issue:46752", - "issue:46897" + "issue:46897", + "issue:46932" ], "soft_pairs": [ + "issue:45072|issue:46932", "issue:43673|issue:46421", "issue:29870|issue:35141", - "issue:43931|issue:44060", "issue:44242|issue:46897", + "issue:43931|issue:44060", "issue:44617|issue:45584", "issue:43644|issue:44913", - "issue:25251|issue:27957", - "issue:44933|issue:45362", "issue:43335|issue:44855", + "issue:25251|issue:27957", "issue:45563|issue:45584", + "issue:44933|issue:45362", "issue:43550|issue:44190", "issue:44077|issue:44877", - "issue:45381|issue:46369", "issue:41720|issue:45245", + "issue:45381|issue:46369", "issue:44479|issue:44734", "issue:43335|issue:44190", "issue:43065|issue:44610", - "issue:44464|issue:45563", "issue:39692|issue:44190", + "issue:44464|issue:45563", "issue:45127|issue:45357", "issue:45381|issue:46655", - "issue:44617|issue:46369", + "issue:44779|issue:46489", "issue:44112|issue:45600", "issue:43295|issue:45276", - "issue:44779|issue:46489", - "issue:44462|issue:45907", + "issue:44617|issue:46369", "issue:45290|issue:46752", - "issue:43526|issue:46697", - "issue:45276|issue:45584" + "issue:44462|issue:45907" ], - "prepared_review_unit_hash": "28d36aedf18145191c1b564b434227910eadeb1e8b1fc0aa009cc685af3c99bb", + "prepared_review_unit_hash": "1cb8da91e37cc83c813ffd99477578ac99541e38b331fc7e0e712b8d4692c434", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36450, - "estimated_input_tokens": 9113, - "estimated_eval_tokens": 18482 + "soft_pair_count": 28, + "serialized_chars": 36129, + "estimated_input_tokens": 9033, + "estimated_eval_tokens": 18322 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous. The only strong duplicate-like pair is the DeepSeek tokenizer regression in v5; the other soft links look related only at a broad subsystem level, not the same concrete bug.", - "confidence": 0.74, - "canonical_issue_reason": "issue:44779 is the best canonical issue: it\u2019s the broadest description of the DeepSeek tokenizer v5 regression, and issue:46489 reads like a specific DeepSeek-Coder v1 manifestation/gap in the same fix.", + "summary": "The set is mostly a bundle of unrelated bug reports that only match at a very broad subsystem level. None of the soft-edge pairs look like the same underlying bug/change, so all should be rejected.", + "confidence": 0.93, + "canonical_issue_reason": "Issue 45276 is the clearest self-contained bug report in the set: it describes a concrete Gemma4 embedding-resize failure with a specific expected/actual mismatch and a narrow fix surface.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44779 is the strongest representative issue because it describes the underlying tokenizer regression at the right scope and already has multiple inbound references.", + "best_issue_reason": "Issue 45276 is the best anchor candidate because it is specific, reproducible, and actionable; the rest are either broader regressions, unrelated model issues, or follow-on reports rather than true duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:45072", + "right": "issue:46932", + "accept": false, + "reason": "Both mention dtype/casting problems, but they affect different model families and different code paths (generic bfloat16 inference vs quantized Gemma4 embedders)." + }, { "left": "issue:43673", "right": "issue:46421", "accept": false, - "reason": "Both involve cache/chunked prefill, but one is a missing-cache bug and the other is a recompilation issue; different concrete failure modes." + "reason": "Both involve chunked prefill/cache behavior, but one is about missing cache state in v5.0.0 while the other is about spurious recompilation from lazy cache init." }, { "left": "issue:29870", "right": "issue:35141", "accept": false, - "reason": "Unrelated bugs: BART generate() decoder mask initialization vs token embedding resize/post_init reinitialization." + "reason": "These are unrelated failures: BART generate() mask initialization vs token-embedding resize reinitialization." }, { - "left": "issue:43931", - "right": "issue:44060", + "left": "issue:44242", + "right": "issue:46897", "accept": false, - "reason": "Different problems: Qwen3-VL load shape mismatch vs Qwen3-Next tied-weights warning on the wrong parameter." + "reason": "One is an MoE load-balancing loss bug, the other is a Florence2 training-loss shift bug; they are not the same defect." }, { - "left": "issue:44242", - "right": "issue:46897", + "left": "issue:43931", + "right": "issue:44060", "accept": false, - "reason": "Both mention training loss behavior, but they are different models and different mechanisms; not the same bug." + "reason": "Different models and failures: Qwen3-VL weight-shape loading mismatch vs Qwen3-Next tied-weights warning misbinding." }, { "left": "issue:44617", "right": "issue:45584", "accept": false, - "reason": "CUDA OOM in Sam3Video vs Whisper empty-transcription failure; no shared code-path bug." + "reason": "CUDA OOM in Sam3Video is not the same as Whisper failing on empty transcription after token alignment." }, { "left": "issue:43644", "right": "issue:44913", "accept": false, - "reason": "Global buffer initialization bug vs GPTNeoX config reload defaulting; distinct issues." + "reason": "Non-persistent buffer junk in v5 is unrelated to GPTNeoXConfig rotary_pct reverting on reload." }, { - "left": "issue:25251", - "right": "issue:27957", + "left": "issue:43335", + "right": "issue:44855", "accept": false, - "reason": "Pipeline top_k nesting bug is unrelated to XLMRoberta Flash Attention 2 support." + "reason": "Sparse-layer construction in SwitchTransformers and a Python 3.13 import indentation failure are different issues." }, { - "left": "issue:44933", - "right": "issue:45362", + "left": "issue:25251", + "right": "issue:27957", "accept": false, - "reason": "Import/export issue from image_utils vs Qwen3.5 chat crash; different surfaces and causes." + "reason": "Pipeline top_k nesting behavior and XLMRoberta Flash Attention 2 are unrelated." }, { - "left": "issue:43335", - "right": "issue:44855", + "left": "issue:45563", + "right": "issue:45584", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug vs Python 3.13 TorchScript parsing failure; unrelated." + "reason": "A stale warning for num_return_sequences is not the same as a Whisper crash on empty transcription." }, { - "left": "issue:45563", - "right": "issue:45584", + "left": "issue:44933", + "right": "issue:45362", "accept": false, - "reason": "A stale warning in paged generate() is not the same bug as Whisper failing on empty transcription." + "reason": "A missing import from image_utils and a Qwen3.5 chat crash are unrelated." }, { "left": "issue:43550", "right": "issue:44190", "accept": false, - "reason": "torch.compile/SDPA model failure vs local dataset loading in an example script; unrelated." + "reason": "Torch.compile/SDPA failure in Bamba is unrelated to the local dataset loading bug in the example script." }, { "left": "issue:44077", "right": "issue:44877", "accept": false, - "reason": "Optional post_init policy change vs strict config loading for granite_speech; different fixes." + "reason": "Both touch config validation, but one is about disallowing optional post_init and the other is about a strict config blocking granite_speech loading." }, { - "left": "issue:45381", - "right": "issue:46369", + "left": "issue:41720", + "right": "issue:45245", "accept": false, - "reason": "Qwen2.5-VL video position IDs regression vs Sam3-LiteText memory overflow; unrelated." + "reason": "Qwen3 auto device mapping cudaErrorAssert and the 2^24 category limit error are different failures." }, { - "left": "issue:41720", - "right": "issue:45245", + "left": "issue:45381", + "right": "issue:46369", "accept": false, - "reason": "Auto device mapping cuda assert vs category-count limit error; different model/runtime issues." + "reason": "Qwen2.5-VL video-position indexing and Sam3LiteText memory overflow are not the same bug." }, { "left": "issue:44479", "right": "issue:44734", "accept": false, - "reason": "Video input regression in Qwen VL models vs serve KV-cache continuation indexing crash; not the same bug." + "reason": "Both are generation regressions, but one is a video-input regression in Qwen models and the other is a serve-side KV-cache indexing crash." }, { "left": "issue:43335", "right": "issue:44190", "accept": false, - "reason": "Sparse layer creation bug and dataset-loading failure are unrelated." + "reason": "SwitchTransformers sparse-layer config failure and dataset loading in an example script are unrelated." }, { "left": "issue:43065", "right": "issue:44610", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder vs processor output-size mismatch in OmDet-Turbo; different code paths." + "reason": "A dummy Conv2d in Sam3PixelDecoder is not the same as OmDet-Turbo producing the wrong image resolution." }, { - "left": "issue:44464", - "right": "issue:45563", + "left": "issue:39692", + "right": "issue:44190", "accept": false, - "reason": "Inconsistent compiled chunked generation outputs vs a stale warning about num_return_sequences; not duplicate symptoms." + "reason": "Documentation example errors for SigLIP2 and a local dataset loading failure do not describe the same bug." }, { - "left": "issue:39692", - "right": "issue:44190", + "left": "issue:44464", + "right": "issue:45563", "accept": false, - "reason": "SigLIP2 documentation errors vs local dataset loading in image classification script; unrelated." + "reason": "Compiled-forward chunked generation inconsistencies are different from a stale warning about num_return_sequences." }, { "left": "issue:45127", "right": "issue:45357", "accept": false, - "reason": "LoRA merge with extended vocab causing collapse vs incorrect visual-encoder keys in save_pretrained; different bugs." + "reason": "LoRA merge collapse with extended vocab and incorrect visual-encoder keys in save_pretrained are unrelated." }, { "left": "issue:45381", "right": "issue:46655", "accept": false, - "reason": "Both are video-processing issues, but one is a vision-position-id regression and the other is padding calculation in SmolVLM." + "reason": "Both are video-related, but they describe different model families and different processor bugs (vision_position_ids vs padding calculation)." }, { - "left": "issue:44617", - "right": "issue:46369", + "left": "issue:44779", + "right": "issue:46489", "accept": false, - "reason": "OOM in Sam3Video vs memory overflow in Sam3-LiteText; different models and failure causes." + "reason": "These are related by theme, but the reports are not identical enough: generic DeepSeek tokenizer regression vs a DeepSeek-Coder v1 tokenizer gap tied to a specific prior fix." }, { "left": "issue:44112", "right": "issue:45600", "accept": false, - "reason": "CI failure from stale GraniteSpeech test vs broken auto_mappings references; both are CI-related, but not the same defect." + "reason": "A stale GraniteSpeech test and broken auto_mappings references to removed configs are separate CI problems." }, { "left": "issue:43295", "right": "issue:45276", "accept": false, - "reason": "Processor/tokenizer API regression vs Gemma4 embedding-resize behavior; unrelated bugs." - }, - { - "left": "issue:44779", - "right": "issue:46489", - "accept": true, - "reason": "Both describe the same DeepSeek tokenizer regression in v5 producing incorrect output; 46489 is a specific DeepSeek-Coder v1 case/gap in the same fix." + "reason": "Processor.tokenizer regression and Gemma4 embedding-resize behavior affect different APIs and code paths." }, { - "left": "issue:44462", - "right": "issue:45907", + "left": "issue:44617", + "right": "issue:46369", "accept": false, - "reason": "AutoTokenizer ignoring tokenizer.json vs list_repo_files not forwarding revision; unrelated repository-loading issues." + "reason": "Both mention Sam3 models, but one is OOM in Sam3Video while the other is memory overflow in sam3-litetext; not the same concrete bug." }, { "left": "issue:45290", "right": "issue:46752", "accept": false, - "reason": "apply_chat_template crashes in two different edge cases: empty list vs assistant tool-call messages without content." - }, - { - "left": "issue:43526", - "right": "issue:46697", - "accept": false, - "reason": "BeitImageProcessorFast reduce_labels bug vs Swin2SR over-padding bug; different processors and logic." + "reason": "Both crash in apply_chat_template, but one is specifically about assistant tool-call messages with no content and the other is about an empty list input." }, { - "left": "issue:45276", - "right": "issue:45584", + "left": "issue:44462", + "right": "issue:45907", "accept": false, - "reason": "Gemma4 resize_token_embeddings/output-embedding bug is unrelated to Whisper empty-transcription generation failure." + "reason": "AutoTokenizer ignoring tokenizer.json and list_repo_files not forwarding revision are unrelated repository/loading issues." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays mostly conservative. The canonical issue choice is supported by the titles and inbound references, and the only accepted soft edge is plausibly the same DeepSeek tokenizer regression/fix gap. I don\u2019t see overclaiming beyond what the packet supports." - }, - "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 }, @@ -11948,7 +12011,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 39, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -11967,11 +12030,10 @@ "issue:41950", "issue:42175", "issue:42796", - "issue:43295", "issue:43404", + "issue:43526", "issue:43531", "issue:43540", - "issue:43582", "issue:43643", "issue:43650", "issue:43673", @@ -11990,9 +12052,10 @@ "issue:44964", "issue:45003", "issue:45216", - "issue:45335", + "issue:45276", "issue:45381", "issue:45542", + "issue:45584", "issue:45600", "issue:45632", "issue:45803", @@ -12001,228 +12064,228 @@ "issue:46234", "issue:46424", "issue:46506", - "issue:46858", - "issue:46932" + "issue:46697", + "issue:46858" ], "soft_pairs": [ - "issue:44938|issue:45542", - "issue:43540|issue:44734", + "issue:43526|issue:46697", "issue:44737|issue:45803", - "issue:25251|issue:29870", + "issue:45276|issue:45584", + "issue:43540|issue:44734", + "issue:44938|issue:45542", "issue:44521|issue:45381", + "issue:25251|issue:29870", "issue:43673|issue:46858", "issue:43854|issue:44964", - "issue:44246|issue:45003", "issue:45632|issue:46234", - "issue:44964|issue:45600", "issue:43643|issue:44496", + "issue:44246|issue:45003", + "issue:44964|issue:45600", "issue:44112|issue:44964", "issue:27957|issue:35141", "issue:42175|issue:44279", "issue:46082|issue:46121", - "issue:36246|issue:41628", "issue:42796|issue:44521", - "issue:36246|issue:38617", + "issue:36246|issue:41628", "issue:32090|issue:35141", "issue:43404|issue:45216", - "issue:44295|issue:46506", - "issue:41950|issue:44016", + "issue:36246|issue:38617", "issue:43531|issue:45600", - "issue:43650|issue:44220", + "issue:41950|issue:44016", + "issue:44295|issue:46506", "issue:43673|issue:46424", - "issue:43582|issue:46932", - "issue:43295|issue:45335" + "issue:43650|issue:44220" ], - "prepared_review_unit_hash": "992bf8fe1e93826d700647009e355fa070e82b92f767853049d2009666cf31f6", + "prepared_review_unit_hash": "5e17d18e0e56ef97e2ef5309b612752a1889764456aa2520b7a6ce571b196170", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 27, - "serialized_chars": 35382, - "estimated_input_tokens": 8846, - "estimated_eval_tokens": 17948 + "serialized_chars": 35411, + "estimated_input_tokens": 8853, + "estimated_eval_tokens": 17962 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a grab bag of unrelated Transformers issues across model loading, multimodal processing, cache behavior, import regressions, and performance. None of the suggested soft pairs look like the same underlying bug closely enough to merge, so this is not a strong duplicate cluster.", - "confidence": 0.38, - "canonical_issue_reason": "issue:44521 is the clearest actively relevant bug report here: it is open, has inbound references, and describes a concrete reproducible failure in a user-facing multimodal API.", + "summary": "This cluster is very noisy and mostly contains unrelated issues that only share superficial transformer/modeling vocabulary. I would not merge any of the suggested soft edges; they are not the same underlying bug or change.", + "confidence": 0.23, + "canonical_issue_reason": "issue:44521 is the most discussion-heavy open issue with inbound references, so it is the best available anchor, but the cluster is heterogeneous and it is not a strong canonical representative.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44521 is the best representative issue because it is current, well-scoped, and tied to a specific broken code path rather than a vague symptom.", + "best_issue_reason": "issue:44521 is the strongest issue candidate by activity and references, though only as a weak representative because the surrounding items do not form a true duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44938", - "right": "issue:45542", + "left": "issue:43526", + "right": "issue:46697", "accept": false, - "reason": "Different failures: Python 3.14 import startup vs TensorFlow backend detection." + "reason": "Both are image-processor bugs, but they affect different processors and different failure modes (label reduction vs over-padding)." }, { - "left": "issue:43540", - "right": "issue:44734", + "left": "issue:44737", + "right": "issue:45803", "accept": false, - "reason": "Both involve multimodal/generation, but the bugs are different: video input validation vs KV-cache continuation indexing." + "reason": "Unrelated bugs: one is XLNet device placement in positional encoding, the other is a broad exception-handling issue in Fuyu batching." }, { - "left": "issue:44737", - "right": "issue:45803", + "left": "issue:45276", + "right": "issue:45584", "accept": false, - "reason": "Unrelated code paths: XLNet device placement vs Fuyu tensor conversion exception handling." + "reason": "Different subsystems and symptoms: token-embedding resizing/tied weights versus Whisper empty-transcription generation." }, { - "left": "issue:25251", - "right": "issue:29870", + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Both touch video/generation-related code, but the concrete problems are unrelated: Qwen3OmniMoe video input validation versus serve KV-cache indexing." + }, + { + "left": "issue:44938", + "right": "issue:45542", "accept": false, - "reason": "Different symptoms and subsystems: pipeline output nesting vs BART generate() decoder_attention_mask initialization." + "reason": "Different causes of import/backend failure: Python 3.14 loading issue versus missing TensorFlow in a torch-only install." }, { "left": "issue:44521", "right": "issue:45381", "accept": false, - "reason": "Both are multimodal, but one is broken assistant_masks and the other is wrong vision_position_ids; not the same bug." + "reason": "Both involve multimodal handling, but one is assistant mask computation and the other is vision position IDs for Qwen2.5-VL video input." + }, + { + "left": "issue:25251", + "right": "issue:29870", + "accept": false, + "reason": "No shared bug: pipeline top_k output nesting versus Bart generate decoder_attention_mask initialization." }, { "left": "issue:43673", "right": "issue:46858", "accept": false, - "reason": "Both mention cache, but one is missing GenerationMixin cache in v5.0.0 and the other is device-tensor arithmetic in static prefill." + "reason": "Both are cache-related, but one is about missing cache during chunked prefill and the other is a device-tensor return breaking slice arithmetic on non-CUDA backends." }, { "left": "issue:43854", "right": "issue:44964", "accept": false, - "reason": "Different model-loading failures for different models and likely different root causes." + "reason": "Both are model-loading failures, but they concern different models and likely different loader paths." }, { - "left": "issue:44246", - "right": "issue:45003", + "left": "issue:45632", + "right": "issue:46234", "accept": false, - "reason": "Import slowness is not the same as unsafe sys.modules access; related area, different issue." + "reason": "Different import/runtime issues: trust_remote_code cache-path collision versus a missing import_utils helper symbol." }, { - "left": "issue:45632", - "right": "issue:46234", + "left": "issue:43643", + "right": "issue:44496", "accept": false, - "reason": "Trust-remote-code cache collision vs missing import_utils symbol are unrelated." + "reason": "Both are config-loading problems, but one drops fields with trust_remote_code and the other rejects a model missing model_type." }, { - "left": "issue:44964", - "right": "issue:45600", + "left": "issue:44246", + "right": "issue:45003", "accept": false, - "reason": "Different problems: model loading regression vs stale auto_mappings reference breaking CI." + "reason": "Import performance and sys.modules access are related only broadly; they are not the same concrete defect." }, { - "left": "issue:43643", - "right": "issue:44496", + "left": "issue:44964", + "right": "issue:45600", "accept": false, - "reason": "Both are config-loading issues, but one is trust_remote_code field loss and the other is unrecognized model/config missing model_type." + "reason": "Different failures: loading a multimodal model versus CI breakage from stale auto_mappings references." }, { "left": "issue:44112", "right": "issue:44964", "accept": false, - "reason": "CI test flake in GraniteSpeech is unrelated to Phi-4 multimodal loading." + "reason": "CI test instability in GraniteSpeech is unrelated to multimodal model loading." }, { "left": "issue:27957", "right": "issue:35141", "accept": false, - "reason": "Flash Attention support and embedding reinitialization are different bugs." + "reason": "Different model features and failures: Flash Attention 2 support versus embedding reinitialization after resize." }, { "left": "issue:42175", "right": "issue:44279", "accept": false, - "reason": "A packaging/backend extras issue is not the same as a general dependency issue." + "reason": "One is a packaging/backend install bug; the other is a vague dependency issue with no same code path." }, { "left": "issue:46082", "right": "issue:46121", "accept": false, - "reason": "Distinct config-validation bugs: head_dim rejection vs rope-parameter list type handling." - }, - { - "left": "issue:36246", - "right": "issue:41628", - "accept": false, - "reason": "Both are import errors, but for different symbols and likely different package-export regressions." + "reason": "Distinct config validation bugs: head_dim validation versus rope-parameter list handling." }, { "left": "issue:42796", "right": "issue:44521", "accept": false, - "reason": "Tokenizer BPE marker handling is unrelated to multimodal chat-template assistant masks." + "reason": "Tokenizer BPE-marker handling is unrelated to multimodal chat template assistant masks." }, { "left": "issue:36246", - "right": "issue:38617", + "right": "issue:41628", "accept": false, - "reason": "Different missing imports from different modules; no shared underlying defect." + "reason": "Both are import errors, but they involve different missing symbols and different export surfaces." }, { "left": "issue:32090", "right": "issue:35141", "accept": false, - "reason": "Trainer broadcast TypeError is unrelated to embedding weight reinitialization." + "reason": "Trainer GPU broadcast NoneType error is unrelated to output embedding reinitialization." }, { "left": "issue:43404", "right": "issue:45216", "accept": false, - "reason": "Model weight tying and save_pretrained checkpoint regression are different code paths." + "reason": "Both involve model weights/checkpoints, but one is weight tying and the other is save_pretrained regression." }, { - "left": "issue:44295", - "right": "issue:46506", - "accept": false, - "reason": "Position_ids buffer access and CPU/CUDA output divergence are not the same bug." - }, - { - "left": "issue:41950", - "right": "issue:44016", + "left": "issue:36246", + "right": "issue:38617", "accept": false, - "reason": "Pipeline processor lookup and a notebook syntax error are completely unrelated." + "reason": "Different missing imports in different modules; no shared underlying bug." }, { "left": "issue:43531", "right": "issue:45600", "accept": false, - "reason": "Qwen3-MoE sliding-window behavior and auto_mappings cleanup are unrelated." + "reason": "Qwen3-MoE sliding-window behavior and stale auto_mappings references are unrelated." }, { - "left": "issue:43650", - "right": "issue:44220", + "left": "issue:41950", + "right": "issue:44016", "accept": false, - "reason": "No meaningful overlap beyond being generic errors; separate problem reports." + "reason": "Pipeline image-processor lookup bug versus a notebook syntax error are clearly unrelated." }, { - "left": "issue:43673", - "right": "issue:46424", + "left": "issue:44295", + "right": "issue:46506", "accept": false, - "reason": "Both mention cache configuration, but one is v5 cache availability and the other is max_cache_length behavior for quantized cache only." + "reason": "Position_ids buffer access bug is unrelated to CPU/CUDA result divergence in PPDocLayoutV3." }, { - "left": "issue:43582", - "right": "issue:46932", + "left": "issue:43673", + "right": "issue:46424", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError and Gemma4 dtype casting are unrelated." + "reason": "Both mention cache, but the defects differ: missing cache in one path versus max_cache_length being honored only for quantized cache." }, { - "left": "issue:43295", - "right": "issue:45335", + "left": "issue:43650", + "right": "issue:44220", "accept": false, - "reason": "Processor.tokenizer regression and t5gemma decoder embedding resize are different bugs." + "reason": "The titles are too generic; no evidence these are the same code-path or bug." } ] }, @@ -12238,7 +12301,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 40, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -12254,19 +12317,17 @@ "issue:43441", "issue:43479", "issue:43526", - "issue:43577", + "issue:43582", "issue:43673", "issue:43824", "issue:43901", "issue:43976", "issue:44016", "issue:44062", - "issue:44075", "issue:44220", "issue:44279", "issue:44327", "issue:44336", - "issue:44368", "issue:44496", "issue:44521", "issue:44683", @@ -12293,254 +12354,249 @@ "issue:46506", "issue:46515", "issue:46693", - "issue:46752" + "issue:46752", + "issue:46932" ], "soft_pairs": [ "issue:43976|issue:44987", - "issue:44016|issue:44220", "issue:45083|issue:46506", - "issue:42907|issue:43441", - "issue:44683|issue:45446", + "issue:43582|issue:46932", + "issue:43295|issue:45335", + "issue:44016|issue:44220", + "issue:44496|issue:45042", "issue:43526|issue:43901", - "issue:43976|issue:44279", "issue:43381|issue:45593", - "issue:43479|issue:45375", - "issue:44496|issue:45042", - "issue:43673|issue:44336", + "issue:44683|issue:45446", + "issue:42907|issue:43441", "issue:43901|issue:44496", + "issue:43673|issue:44336", + "issue:43479|issue:45375", + "issue:43976|issue:44279", + "issue:44496|issue:45759", "issue:44792|issue:46424", "issue:44016|issue:44062", - "issue:44496|issue:45759", - "issue:43824|issue:44933", "issue:43441|issue:45588", + "issue:43824|issue:44933", "issue:46123|issue:46369", - "issue:44327|issue:46752", - "issue:44279|issue:46515", "issue:44521|issue:46693", + "issue:44279|issue:46515", + "issue:44327|issue:46752", "issue:45335|issue:45584", "issue:43299|issue:45216", - "issue:44279|issue:45405", "issue:46082|issue:46399", + "issue:44279|issue:45405", "issue:43295|issue:44779", - "issue:44075|issue:44368", "issue:39692|issue:43901", - "issue:43976|issue:45405", - "issue:43577|issue:45216" + "issue:43976|issue:45405" ], - "prepared_review_unit_hash": "5dd98e12c3aecae48fea9b40c4617edabcf11887334f764d86a937a9fdd108be", + "prepared_review_unit_hash": "d4aadf20de926eb81e55b1f56a090441c0f40c9bc29accce83ae46bdfdb0336e", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36740, - "estimated_input_tokens": 9185, - "estimated_eval_tokens": 18626 + "serialized_chars": 36078, + "estimated_input_tokens": 9020, + "estimated_eval_tokens": 18296 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the soft-similarity pairs are mostly broad topical matches (model loading, multimodal configs, tokenizer/processor behavior, flex_attention, etc.) rather than the same concrete bug or change. I would not merge any of the proposed pairs as duplicates.", - "confidence": 0.96, - "canonical_issue_reason": null, + "summary": "The cluster is heterogeneous: most pairs share only a broad subsystem or release context, not the same underlying bug. No soft edge looks safe to merge as a duplicate.", + "confidence": 0.88, + "canonical_issue_reason": "If a single issue must represent the set, #43976 is the cleanest broad regression report, but the cluster is too mixed to have a strong true canonical issue.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "#43976 is the most representative general version-compatibility regression among these disparate reports, though it is not a duplicate umbrella for the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43976", "right": "issue:44987", "accept": false, - "reason": "Both are version/regression reports, but one is a Python compatibility problem and the other is a model-loading failure for a specific repo; different root causes." + "reason": "Both mention 5.1.0, but one is Python-version incompatibility and the other is a model-loading failure; different root causes." }, { - "left": "issue:44016", - "right": "issue:44220", + "left": "issue:45083", + "right": "issue:46506", "accept": false, - "reason": "A notebook syntax error and an audio feature extraction bug are unrelated code paths." + "reason": "Different models and failure modes: helper length calculation vs CPU/CUDA output mismatch." }, { - "left": "issue:45083", - "right": "issue:46506", + "left": "issue:43582", + "right": "issue:46932", "accept": false, - "reason": "Both mention differing behavior, but one is a helper length calculation bug and the other is CPU/CUDA output mismatch for a layout model." + "reason": "Apple Silicon TypeError in allocator warmup is unrelated to Gemma4 quantized embedder dtype casting." }, { - "left": "issue:42907", - "right": "issue:43441", + "left": "issue:43295", + "right": "issue:45335", "accept": false, - "reason": "Saving a dequantized model and FlashAttention failures for Ministral are different failures in different subsystems." + "reason": "One is a processor/tokenizer API regression; the other is embedding resize not propagating to decoder weights." }, { - "left": "issue:44683", - "right": "issue:45446", + "left": "issue:44016", + "right": "issue:44220", "accept": false, - "reason": "Both involve flex_attention, but one is a compile/runtime failure on torch>=2.9 and the other is a bad version check for AuxRequest import; not the same bug." + "reason": "Notebook syntax error and fbank feature extraction bug are unrelated." }, { - "left": "issue:43526", - "right": "issue:43901", + "left": "issue:44496", + "right": "issue:45042", "accept": false, - "reason": "A Beit image processor label-reduction bug and a docs mismatch for text classification are unrelated." + "reason": "Unrecognized model config and torchvision requirement in PIL image processors are different issues." }, { - "left": "issue:43976", - "right": "issue:44279", + "left": "issue:43526", + "right": "issue:43901", "accept": false, - "reason": "Python-version incompatibility and a generic dependency issue are not the same underlying defect." + "reason": "One is a BeitImageProcessorFast label-reduction bug; the other is a docs/behavior mismatch for a pipeline flag." }, { "left": "issue:43381", "right": "issue:45593", "accept": false, - "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary loss behavior are separate training-time issues." + "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary-loss behavior are not the same bug." }, { - "left": "issue:43479", - "right": "issue:45375", + "left": "issue:44683", + "right": "issue:45446", "accept": false, - "reason": "Both are config-related, but they affect different multimodal model families and different missing/default fields." + "reason": "Both touch flex_attention, but one is a compiled torch>=2.9 failure and the other is a version-check/import guard bug." }, { - "left": "issue:44496", - "right": "issue:45042", + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Failing to save dequantized Ministrals is distinct from a FlashAttention runtime failure in Ministral-3." + }, + { + "left": "issue:43901", + "right": "issue:44496", "accept": false, - "reason": "One is an unrecognized model/config problem; the other is an image-processor torchvision dependency regression." + "reason": "A docs mismatch about return_all_scores is unrelated to model auto-loading/config recognition." }, { "left": "issue:43673", "right": "issue:44336", "accept": false, - "reason": "Generation cache handling and ANSI-color terminal output are unrelated." + "reason": "Generation cache regression and ANSI-color logging output are unrelated." }, { - "left": "issue:43901", - "right": "issue:44496", + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Both are config issues, but for different models and different missing/default fields." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "Python-version breakage and a generic dependency issue are not the same concrete bug." + }, + { + "left": "issue:44496", + "right": "issue:45759", "accept": false, - "reason": "A pipeline docs/behavior mismatch is not the same as a model config loading error." + "reason": "Both involve model/config loading, but one is an unrecognized model type and the other is failure to unwrap text_config for composite Qwen multimodal configs." }, { "left": "issue:44792", "right": "issue:46424", "accept": false, - "reason": "A test failure around image generation and a cache max length setting issue do not share the same code-path problem." + "reason": "Janus image-generation test failure and cache-length semantics are unrelated." }, { "left": "issue:44016", "right": "issue:44062", "accept": false, - "reason": "Notebook syntax breakage and AddedToken keyword collisions are unrelated." + "reason": "Notebook syntax error versus AddedToken keyword collision; no shared code path." }, { - "left": "issue:44496", - "right": "issue:45759", + "left": "issue:43441", + "right": "issue:45588", "accept": false, - "reason": "Both touch model/config loading, but one is an unrecognized model and the other is missing text_config unwrapping for composite multimodal configs." + "reason": "Both mention FlashAttention, but they target different model classes and different null/attribute handling paths." }, { "left": "issue:43824", "right": "issue:44933", "accept": false, - "reason": "An import error for a Qwen VL class and a non-existent image_utils import are different missing-symbol issues." - }, - { - "left": "issue:43441", - "right": "issue:45588", - "accept": false, - "reason": "FlashAttention failure for Ministral-3 and an AttributeError on s_aux=None in flash_attention integration are distinct bugs." + "reason": "Missing Qwen2.5-VL import and nonexistent image_utils import are separate API/export problems." }, { "left": "issue:46123", "right": "issue:46369", "accept": false, - "reason": "Partial-batch mask generation dropping results and a memory overflow in sam3-litetext are different symptoms and code paths." + "reason": "Partial-batch result dropping and memory overflow in sam3-litetext are different failures." }, { - "left": "issue:44327", - "right": "issue:46752", + "left": "issue:44521", + "right": "issue:46693", "accept": false, - "reason": "A QA decode_spans top-k bounds error and an empty-list apply_chat_template crash are unrelated." + "reason": "All-zero assistant masks in chat templating is unrelated to a long-context FA2 performance regression." }, { "left": "issue:44279", "right": "issue:46515", "accept": false, - "reason": "A dependency issue and a torch-version incompatibility are not the same defect." + "reason": "General dependency trouble and torch 2.4.1 incompatibility are too broad to be the same bug." }, { - "left": "issue:44521", - "right": "issue:46693", + "left": "issue:44327", + "right": "issue:46752", "accept": false, - "reason": "assistant_mask generation for multimodal chat templates and FA2 long-context performance regression are different behaviors." + "reason": "QA decode_spans kth-out-of-bounds and empty-list chat_template IndexError are unrelated edge cases." }, { "left": "issue:45335", "right": "issue:45584", "accept": false, - "reason": "resize_token_embeddings not updating decoder embeddings and Whisper empty-transcription generation failure are unrelated." + "reason": "Embedding resize propagation and Whisper empty-transcription generation failure are different code paths." }, { "left": "issue:43299", "right": "issue:45216", "accept": false, - "reason": "Qwen3VL MoE loading regression and Qwen3.5 checkpoint save corruption are different model/version failures." + "reason": "Qwen3VL model-loading regression and Qwen3.5 save_pretrained checkpoint regression are distinct." }, { - "left": "issue:45375", - "right": "issue:45584", + "left": "issue:46082", + "right": "issue:46399", "accept": false, - "reason": "A Qwen multimodal config field omission and Whisper special-token alignment behavior are unrelated." + "reason": "LlamaConfig head_dim validation and Qwen3.5 rotary-pos-emb registration are unrelated." }, { "left": "issue:44279", "right": "issue:45405", "accept": false, - "reason": "A generic dependency issue and an unreleased PEFT minimum-version bump are not duplicates." - }, - { - "left": "issue:46082", - "right": "issue:46399", - "accept": false, - "reason": "A LlamaConfig head_dim validation issue and a missing decorator for apply_rotary_pos_emb registration are different problems." + "reason": "A generic dependency issue is not the same as a PEFT minimum-version bump problem." }, { "left": "issue:43295", "right": "issue:44779", "accept": false, - "reason": "Both mention tokenizer/processor regressions, but one is a processor.tokenizer API break and the other is incorrect Deepseek tokenization output." - }, - { - "left": "issue:44075", - "right": "issue:44368", - "accept": false, - "reason": "Optimizer SGD arguments not being used and a tie_word_embeddings warning are unrelated." + "reason": "Processor/tokenizer API regression and Deepseek tokenizer output corruption do not share a concrete fix." }, { "left": "issue:39692", "right": "issue:43901", "accept": false, - "reason": "A SigLIP2 example bug and a text-classification docs mismatch are not the same issue." + "reason": "SigLIP2 docs example errors and return_all_scores docs mismatch are separate documentation issues." }, { "left": "issue:43976", "right": "issue:45405", "accept": false, - "reason": "Python support incompatibility and an unreleased dependency version bump are distinct release issues." - }, - { - "left": "issue:43577", - "right": "issue:45216", - "accept": false, - "reason": "Blip2 dtype propagation and Qwen3.5 save_pretrained checkpoint correctness are separate model-specific bugs." + "reason": "Python compatibility regression and unreleased PEFT minimum-version bump are unrelated." } ] }, @@ -12556,7 +12612,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 41, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -12573,18 +12629,18 @@ "issue:43441", "issue:43479", "issue:43525", + "issue:43577", "issue:43650", "issue:43653", "issue:43854", "issue:44016", + "issue:44075", "issue:44242", - "issue:44261", - "issue:44263", "issue:44291", - "issue:44297", "issue:44336", "issue:44351", "issue:44361", + "issue:44368", "issue:44462", "issue:44466", "issue:44623", @@ -12595,6 +12651,7 @@ "issue:45081", "issue:45103", "issue:45200", + "issue:45216", "issue:45276", "issue:45278", "issue:45357", @@ -12613,217 +12670,224 @@ "issue:46851" ], "soft_pairs": [ - "issue:44351|issue:45278", - "issue:42796|issue:43650", + "issue:44075|issue:44368", + "issue:43577|issue:45216", "issue:46082|issue:46512", + "issue:44351|issue:45278", "issue:44779|issue:45812", + "issue:42796|issue:43650", "issue:41950|issue:45020", "issue:43299|issue:45357", "issue:44466|issue:45276", "issue:43299|issue:43854", "issue:43653|issue:45593", - "issue:41950|issue:45823", "issue:44242|issue:45200", + "issue:41950|issue:45823", "issue:43650|issue:46209", - "issue:44297|issue:44462", + "issue:44462|issue:45081", "issue:43479|issue:43525", - "issue:44683|issue:45542", "issue:44016|issue:44291", + "issue:44683|issue:45542", "issue:44361|issue:45593", "issue:45081|issue:45907", - "issue:33290|issue:45656", "issue:43441|issue:45834", + "issue:33290|issue:45656", + "issue:45103|issue:45372", + "issue:44336|issue:46851", "issue:42994|issue:43317", "issue:44623|issue:44749", - "issue:44336|issue:46851", - "issue:44261|issue:44263", - "issue:45103|issue:45372", "issue:45859|issue:46851" ], - "prepared_review_unit_hash": "ca741d3bfb59e12e44a944357aff3e93349a00f361f17deac9e184e0ea4633f9", + "prepared_review_unit_hash": "2df0bab1e9ac84a467e65059473a1af99e7d8a80cfe217d0dc538ee72ce22f12", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 26, - "serialized_chars": 34967, - "estimated_input_tokens": 8742, - "estimated_eval_tokens": 17740 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35931, + "estimated_input_tokens": 8983, + "estimated_eval_tokens": 18222 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a set of unrelated bugs across tokenizers, model loading, saving, offload, and config regressions. None of the soft pairs look like true duplicates of the same underlying bug, so no soft edge is accepted.", - "confidence": 0.9, - "canonical_issue_reason": "If one issue must anchor the set, issue 45020 is the broadest and most representative regression report, with substantial detail and discussion. It still does not actually subsume the other issues.", + "summary": "The items are mostly distinct Transformers regression reports, not true duplicates. Several are superficially similar around v5-era loading/saving/tokenizer issues, but they hit different models and code paths, so none of the soft pairs should be merged.", + "confidence": 0.93, + "canonical_issue_reason": "No clear canonical issue: these reports span separate bugs across tokenizers, configs, save/load, offload, and optimizer behavior.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45020 is the best single issue to represent the cluster only because it is broad and well-documented; however, the items are not duplicates and the issue should not be treated as a real canonical for them.", + "best_issue_reason": "Issue 45216 is the best single representative if one is needed: it is a clear, versioned save_pretrained regression with a specific model and failure mode.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44351", - "right": "issue:45278", + "left": "issue:44075", + "right": "issue:44368", "accept": false, - "reason": "Both mention import problems, but one is a missing HybridCache import and the other is many import errors after an upgrade; not the same concrete failure." + "reason": "SGD optimizer args being ignored is unrelated to a Qwen3.5 tie_word_embeddings warning." }, { - "left": "issue:42796", - "right": "issue:43650", + "left": "issue:43577", + "right": "issue:45216", "accept": false, - "reason": "Tokenizer BPE marker handling is unrelated to the unrelated placeholder/troll issue title 'ADD THE DATA'." + "reason": "Both involve model state/dtype concerns, but one is Blip2 dtype propagation and the other is Qwen3.5 checkpoint serialization." }, { "left": "issue:46082", "right": "issue:46512", "accept": false, - "reason": "One is LlamaConfig head_dim validation; the other is stale config.dtype after from_config(dtype=...). Different code paths and bugs." + "reason": "Explicit head_dim validation and stale config.dtype after from_config are different config bugs." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "HybridCache import failure is a specific symbol breakage, while 45278 is a broader upgrade-induced import error report." }, { "left": "issue:44779", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and likely different tokenizer logic." + "reason": "Deepseek token IDs and Granite token IDs are separate tokenizer regressions with different model families and likely different root causes." + }, + { + "left": "issue:42796", + "right": "issue:43650", + "accept": false, + "reason": "Tokenizer BPE-marker handling is unrelated to the nonspecific 'ADD THE DATA' issue." }, { "left": "issue:41950", "right": "issue:45020", "accept": false, - "reason": "Video-classification processor lookup and remote_code model breakage are different subsystems and failure modes." + "reason": "A video-classification processor lookup bug is unrelated to remote_code model breakages." }, { "left": "issue:43299", "right": "issue:45357", "accept": false, - "reason": "Qwen3VL MoE loading and Qwen3.5 save_pretrained visual-key serialization are distinct bugs." + "reason": "Qwen3VL MoE loading and Qwen3.5 save_pretrained visual key serialization are different failure paths." }, { "left": "issue:44466", "right": "issue:45276", "accept": false, - "reason": "lm_head.weight serialization inconsistency is unrelated to resize_token_embeddings not updating tied/output embeddings." + "reason": "Inconsistent lm_head serialization and Gemma4 resize_token_embeddings behavior are distinct save vs resize bugs." }, { "left": "issue:43299", "right": "issue:43854", "accept": false, - "reason": "Different models and symptoms: Qwen3VL MoE loading vs GLM-4.7 unit-test loading." + "reason": "Different models and different loading failures; not the same underlying bug." }, { "left": "issue:43653", "right": "issue:45593", "accept": false, - "reason": "BigBirdTokenizer special-token registration is unrelated to D-FINE auxiliary-loss behavior." + "reason": "BigBirdTokenizer special-token decoding and D-FINE auxiliary-loss behavior are unrelated." }, { - "left": "issue:41950", - "right": "issue:45823", + "left": "issue:44242", + "right": "issue:45200", "accept": false, - "reason": "Video pipeline processor lookup and Gemma4 PLE device mismatch are different bugs." + "reason": "Missing load-balancing loss and mm_token_type_ids default handling affect different parts of model logic." }, { - "left": "issue:44242", - "right": "issue:45200", + "left": "issue:41950", + "right": "issue:45823", "accept": false, - "reason": "Missing load-balancing loss and Gemma 4 mm_token_type_ids defaulting are separate model/training issues." + "reason": "Pipeline processor selection and Gemma4 device-mismatch forward errors are unrelated." }, { "left": "issue:43650", "right": "issue:46209", "accept": false, - "reason": "No meaningful overlap; the titles and contexts are unrelated." + "reason": "The first item is not a concrete bug report, and the second is an FP8 inference-on-B200 issue." }, { - "left": "issue:44297", - "right": "issue:44462", + "left": "issue:44462", + "right": "issue:45081", "accept": false, - "reason": "Tokenizer save_pretrained metadata mismatch and AutoTokenizer ignoring tokenizer.json are related to serialization/loading, but not the same defect." + "reason": "Ignoring tokenizer.json and crashing in fix_mistral_regex are separate tokenizer-loading problems." }, { "left": "issue:43479", "right": "issue:43525", "accept": false, - "reason": "Phi4MultimodalConfig default-init behavior and missing pad_token_id on Llama4Config are separate config bugs." + "reason": "Different config classes and different missing-default fields; not the same bug." }, { - "left": "issue:44683", - "right": "issue:45542", + "left": "issue:44016", + "right": "issue:44291", "accept": false, - "reason": "Compiled flex_attention on torch 2.9 and undefined tf backend from tensorboard-only install are unrelated." + "reason": "Notebook syntax error and init_empty_weights argument mismatch are unrelated." }, { - "left": "issue:44016", - "right": "issue:44291", + "left": "issue:44683", + "right": "issue:45542", "accept": false, - "reason": "Notebook syntax error and init_empty_weights unexpected argument are unrelated." + "reason": "Compiled flex_attention on new torch and a TensorBoard-without-TF backend error are unrelated." }, { "left": "issue:44361", "right": "issue:45593", "accept": false, - "reason": "MLukeTokenizer AttributeError and D-FINE auxiliary-loss omission are unrelated." + "reason": "MLukeTokenizer task AttributeError and D-FINE loss handling are unrelated." }, { "left": "issue:45081", "right": "issue:45907", "accept": false, - "reason": "Mistral regex patch crash and list_repo_files revision forwarding are different API paths and bugs." - }, - { - "left": "issue:33290", - "right": "issue:45656", - "accept": false, - "reason": "AdaFactor OOM in DeepSpeed and optimizer.step being called twice are not the same underlying issue." + "reason": "Mistral regex patching and list_repo_files revision forwarding are completely different code paths." }, { "left": "issue:43441", "right": "issue:45834", "accept": false, - "reason": "FlashAttention failure for Ministral-3 and Kosmos2.5 OCR index error are unrelated model-specific bugs." + "reason": "Ministral FlashAttention failure and Kosmos2.5 long-OCR index error are different model-specific bugs." }, { - "left": "issue:42994", - "right": "issue:43317", + "left": "issue:33290", + "right": "issue:45656", "accept": false, - "reason": "Quantized model saving failure and dequantized model offload loading failure are different code paths." + "reason": "Adafactor OOM under deepspeed and optimizer step called twice under deepspeed are not the same defect." }, { - "left": "issue:44623", - "right": "issue:44749", + "left": "issue:45103", + "right": "issue:45372", "accept": false, - "reason": "processor.save_pretrained missing files and a Chinese report about slowdown after upgrading are unrelated." + "reason": "Auto-docstring annotation parsing and Gemma4 processor import failure are unrelated." }, { "left": "issue:44336", "right": "issue:46851", "accept": false, - "reason": "ANSI codes in loading_report and Qwen3.5 variable-length position_ids handling are unrelated." + "reason": "ANSI escape leakage in loading_report and Qwen3.5 variable-length position_ids are unrelated." }, { - "left": "issue:44261", - "right": "issue:44263", + "left": "issue:42994", + "right": "issue:43317", "accept": false, - "reason": "MLA q_a_layernorm epsilon precision issue and GlmMoeDsaIndexer torch.split return values are different bugs." + "reason": "Quantized-model saving and dequantized-model loading with device_map=auto are different failure modes." }, { - "left": "issue:45103", - "right": "issue:45372", + "left": "issue:44623", + "right": "issue:44749", "accept": false, - "reason": "auto_docstring crash with future annotations and Gemma 4 processor ImportError are unrelated." + "reason": "Missing processor.save_pretrained files and a reported slowdown after upgrade are not the same concrete bug." }, { "left": "issue:45859", "right": "issue:46851", "accept": false, - "reason": "CPU offload incompatibility in Qwen3_5MoeTextRotaryEmbedding and variable-length position_ids in Qwen3.5 are different defects." + "reason": "Both mention Qwen3.5, but CPU offload incompatibility and variable-length position_ids are different code-path problems." } ] }, @@ -12839,7 +12903,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 42, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -12849,11 +12913,9 @@ "nodes": [ "issue:41628", "issue:42175", - "issue:42796", "issue:43441", "issue:43525", "issue:43526", - "issue:43540", "issue:43643", "issue:43673", "issue:43723", @@ -12863,10 +12925,13 @@ "issue:43881", "issue:43901", "issue:43976", + "issue:43994", "issue:44016", "issue:44162", "issue:44188", "issue:44246", + "issue:44261", + "issue:44263", "issue:44297", "issue:44336", "issue:44373", @@ -12874,6 +12939,7 @@ "issue:44589", "issue:44623", "issue:44671", + "issue:44683", "issue:44779", "issue:45003", "issue:45083", @@ -12892,242 +12958,248 @@ "issue:46399", "issue:46655", "issue:46726", - "issue:46752", "issue:46851" ], "soft_pairs": [ + "issue:44261|issue:44263", "issue:45600|issue:45712", "issue:42175|issue:43976", - "issue:44623|issue:45397", "issue:44671|issue:44779", + "issue:44623|issue:45397", + "issue:43673|issue:43761", + "issue:43901|issue:43994", "issue:43723|issue:44162", "issue:43901|issue:44246", - "issue:43673|issue:43761", - "issue:45542|issue:45800", "issue:43824|issue:43976", + "issue:43441|issue:44683", "issue:43525|issue:43881", - "issue:42796|issue:43441", - "issue:44188|issue:44336", "issue:46291|issue:46399", - "issue:44016|issue:45823", - "issue:44336|issue:44373", - "issue:44589|issue:45362", + "issue:44188|issue:44336", + "issue:45542|issue:45800", + "issue:44016|issue:45823", "issue:43525|issue:43643", - "issue:46655|issue:46726", + "issue:44589|issue:45362", + "issue:44336|issue:44373", "issue:45325|issue:46851", + "issue:46655|issue:46726", "issue:41628|issue:43976", - "issue:44297|issue:45397", - "issue:43784|issue:45542", - "issue:44246|issue:44496", "issue:45083|issue:46851", + "issue:44246|issue:44496", + "issue:43784|issue:45542", + "issue:44297|issue:45397", "issue:46399|issue:46851", "issue:45561|issue:45563", "issue:45103|issue:46399", - "issue:43526|issue:45003", - "issue:43540|issue:46752" + "issue:43526|issue:45003" ], - "prepared_review_unit_hash": "4de9e3c886c0324f6b06821a78079054063b9ab200cf6694523368872f9a00b1", + "prepared_review_unit_hash": "2435a7998784a022af76eb97ac80dd26fbdcb40ceae9394e54f74792e594afd4", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 29, - "serialized_chars": 35737, - "estimated_input_tokens": 8935, - "estimated_eval_tokens": 18126 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36637, + "estimated_input_tokens": 9160, + "estimated_eval_tokens": 18576 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues are a broad mix of unrelated regressions across tokenizers, processors, import paths, configs, kernels, docs, and packaging. No soft pair looks like the same underlying bug/change, so none should be merged as duplicates.", - "confidence": 0.96, - "canonical_issue_reason": "issue:43824 is the strongest single representative: it has the clearest concrete import regression, strong discussion activity, and inbound references, making it a good triage anchor for this mixed cluster.", + "summary": "No true duplicate cluster emerges here: the items cover unrelated bugs, packaging issues, docs, and model-specific regressions. The soft edges are mostly superficial title similarity or shared subsystem vocabulary, so all should be rejected.", + "confidence": 0.95, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43824 is the best overall issue candidate because it is specific, reproducible, and appears to have the most external impact/attention among the set.", + "best_issue_reason": "issue:45561 is the strongest standalone bug report here: it is open, well-scoped, has discussion and inbound references, and describes a concrete race in testing utils. It is not, however, representative of the rest of the heterogeneous set.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "Different bugs: one is a MLA layernorm precision/config issue, the other is a torch.split return-value handling bug in GlmMoeDsaIndexer." + }, { "left": "issue:45600", "right": "issue:45712", "accept": false, - "reason": "Different repo-maintenance failures: one is a stale auto_mappings reference, the other is leftover dummy classes leaking into checks/dir()." + "reason": "Both are repo-maintenance problems, but one is auto_mappings CI breakage while the other is leftover dummy classes leaking into dir(transformers); not the same underlying change." }, { "left": "issue:42175", "right": "issue:43976", "accept": false, - "reason": "Packaging/backend inclusion bug vs Python version compatibility regression; unrelated symptoms and code paths." + "reason": "Packaging/backend dependency issue versus Python version compatibility regression; different failure modes and fixes." }, { - "left": "issue:44623", - "right": "issue:45397", + "left": "issue:44671", + "right": "issue:44779", "accept": false, - "reason": "Missing processor save files and a Zero3 from_pretrained issue are separate processor vs distributed-loading problems." + "reason": "Different model regressions: CamemBERT masked LM outputs vs Deepseek tokenizer behavior." }, { - "left": "issue:44671", - "right": "issue:44779", + "left": "issue:44623", + "right": "issue:45397", "accept": false, - "reason": "CamemBERT MLM prediction regression and Deepseek tokenizer/BPE handling are different model/tokenizer bugs." + "reason": "Processor save_pretrained file omission is unrelated to gemma-4 zero3 from_pretrained behavior." }, { - "left": "issue:43723", - "right": "issue:44162", + "left": "issue:43673", + "right": "issue:43761", "accept": false, - "reason": "AutoTokenizer loading failure and ESM2 model breakage are distinct failures in different model families." + "reason": "Generation cache/chunked_prefill bug versus CLIPVisionModel hidden_states regression; separate code paths." }, { "left": "issue:43901", - "right": "issue:44246", + "right": "issue:43994", "accept": false, - "reason": "Docs mismatch for return_all_scores is unrelated to intermittent slow imports." + "reason": "Docs mismatch for text classification pipeline is not the same as SIGLIP2 model inference producing nonsense results." }, { - "left": "issue:43673", - "right": "issue:43761", + "left": "issue:43723", + "right": "issue:44162", "accept": false, - "reason": "Chunked-prefill cache regression in generation is unrelated to CLIPVisionModel hidden_states output behavior." + "reason": "Tokenizer loading regression and ESM2 model breakage are different layers of the stack." }, { - "left": "issue:45542", - "right": "issue:45800", + "left": "issue:43901", + "right": "issue:44246", "accept": false, - "reason": "TensorFlow backend installation issue and torch/transformers version incompatibility are different dependency problems." + "reason": "Pipeline docs/behavior mismatch does not match intermittent import-time slowness." }, { "left": "issue:43824", "right": "issue:43976", "accept": false, - "reason": "Importing a specific Qwen2.5-VL class is unrelated to a Python 3.9/3.10 compatibility report." + "reason": "Qwen2.5-VL missing importable class and Python version support are unrelated issues." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "Ministral FlashAttention failure and compiled flex_attention failure are different model/kernel regressions." }, { "left": "issue:43525", "right": "issue:43881", "accept": false, - "reason": "Missing pad_token_id on Llama4Config and glm-4v loading failure are different model/config loading bugs." + "reason": "Llama4Config missing attribute and glm-4v-9b loading failure are separate model configuration problems." }, { - "left": "issue:42796", - "right": "issue:43441", + "left": "issue:46291", + "right": "issue:46399", "accept": false, - "reason": "Tokenizer BPE marker handling and FlashAttention failure are separate Ministral issues affecting different parts of the stack." + "reason": "Importing pipelines due to kernels package requirements is not the same as missing decorator registration for apply_rotary_pos_emb." }, { "left": "issue:44188", "right": "issue:44336", "accept": false, - "reason": "torch.compile attention-kernel divergence is unrelated to ANSI codes emitted by loading_report." + "reason": "torch.compile attention-kernel divergence is unrelated to ANSI code leakage in loading_report." }, { - "left": "issue:46291", - "right": "issue:46399", + "left": "issue:45542", + "right": "issue:45800", "accept": false, - "reason": "Kernel dependency/version import failure and a Qwen3.5 rotary-pos-emb registration crash are not the same code-path problem." + "reason": "Undefined tf backend from optional installs is not the same as a torch/transformers version incompatibility." }, { "left": "issue:44016", "right": "issue:45823", "accept": false, - "reason": "Notebook syntax error and Gemma4 device mismatch are unrelated documentation vs runtime issues." + "reason": "Notebook syntax error and Gemma4 PLE device mismatch are unrelated." }, { - "left": "issue:44336", - "right": "issue:44373", + "left": "issue:43525", + "right": "issue:43643", "accept": false, - "reason": "Terminal ANSI escape leakage and a wrong docstring are unrelated documentation/formatting issues." + "reason": "Missing pad_token_id on Llama4Config is not the same as trust_remote_code causing missing fields in AutoConfig." }, { "left": "issue:44589", "right": "issue:45362", "accept": false, - "reason": "Float8 storage lookup failure and Qwen3.5 chat crash have different causes and subsystems." + "reason": "Storage backend/type error and Qwen3.5 chat crash are different problems." }, { - "left": "issue:43525", - "right": "issue:43643", + "left": "issue:44336", + "right": "issue:44373", "accept": false, - "reason": "A missing config attribute on Llama4Config is not the same as AutoConfig dropping fields under trust_remote_code." + "reason": "Terminal ANSI escaping bug versus incorrect docstring text; no shared code defect." }, { - "left": "issue:46655", - "right": "issue:46726", + "left": "issue:45325", + "right": "issue:46851", "accept": false, - "reason": "Video padding calculation bug and missing-vision-backend import error are different processor failure modes." + "reason": "Both involve Qwen3.5/Qwen2.5-VL positional handling, but these are distinct model-specific bugs with different symptoms and fixes." }, { - "left": "issue:45325", - "right": "issue:46851", + "left": "issue:46655", + "right": "issue:46726", "accept": false, - "reason": "Qwen2.5-VL still-image temporal scaling and Qwen3.5 variable-length position_ids are separate position-id bugs." + "reason": "Incorrect video padding calculation in SmolVLM is not the same as AutoProcessor erroring when vision backend is missing." }, { "left": "issue:41628", "right": "issue:43976", "accept": false, - "reason": "AutoImageProcessor import failure is unrelated to Python version compatibility with transformers 5.1.0." + "reason": "AutoImageProcessor import failure and Python version compatibility are unrelated." }, { - "left": "issue:44297", - "right": "issue:45397", + "left": "issue:45083", + "right": "issue:46851", "accept": false, - "reason": "Tokenizer save_pretrained metadata mismatch and Gemma4 Zero3 from_pretrained are unrelated issues." + "reason": "Helper output-length computation bug in qwen3_omni_moe is different from variable-length position_ids handling in Qwen3.5." }, { - "left": "issue:43784", - "right": "issue:45542", + "left": "issue:44246", + "right": "issue:44496", "accept": false, - "reason": "sentence-transformers import NameError and undefined tf backend error are separate backend/import failures." + "reason": "Import slowness versus unrecognized model/config loading failure are different issues." }, { - "left": "issue:44246", - "right": "issue:44496", + "left": "issue:43784", + "right": "issue:45542", "accept": false, - "reason": "Intermittent slow imports and unrecognized model/config loading are different issue classes." + "reason": "Sentence-transformers import NameError and missing TensorFlow backend are unrelated dependency issues." }, { - "left": "issue:45083", - "right": "issue:46851", + "left": "issue:44297", + "right": "issue:45397", "accept": false, - "reason": "Feature-extraction length helper behavior and Qwen3.5 variable-length position_ids are unrelated model-specific bugs." + "reason": "Tokenizer save_pretrained metadata mismatch is not the same as gemma-4 zero3 loading failure." }, { "left": "issue:46399", "right": "issue:46851", "accept": false, - "reason": "Kernel registration/decorator crash and variable-length position_ids are different Qwen3.5 failure modes." + "reason": "Missing decorator for a rotary-pos-emb registration error is separate from variable-length input position_ids bugs." }, { "left": "issue:45561", "right": "issue:45563", "accept": false, - "reason": "pytest-xdist file race and stale generation warning are unrelated testing vs generation issues." + "reason": "Race on captured_info.txt in testing utils and a stale num_return_sequences warning in paged generate() are different behaviors." }, { "left": "issue:45103", "right": "issue:46399", "accept": false, - "reason": "auto_docstring future-annotations crash is unrelated to Qwen3.5 kernel registration." + "reason": "auto_docstring annotation handling crash is unrelated to kernel registration/decorator errors." }, { "left": "issue:43526", "right": "issue:45003", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug and unsafe sys.modules access are separate implementation issues." - }, - { - "left": "issue:43540", - "right": "issue:46752", - "accept": false, - "reason": "Qwen3OmniMoe video processing ValueError and empty-list chat template IndexError are unrelated input-validation bugs." + "reason": "BeitImageProcessorFast label reduction bug and unsafe sys.modules access in modeling_utils are not the same defect." } ] }, @@ -13143,7 +13215,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 43, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -13174,7 +13246,6 @@ "issue:44479", "issue:44485", "issue:44496", - "issue:44749", "issue:44779", "issue:44871", "issue:45003", @@ -13189,7 +13260,9 @@ "issue:45588", "issue:45593", "issue:45759", + "issue:45859", "issue:46132", + "issue:46399", "issue:46506", "issue:46614", "issue:46655", @@ -13200,15 +13273,16 @@ "issue:46851" ], "soft_pairs": [ + "issue:43540|issue:46752", "issue:45072|issue:46829", "issue:46614|issue:46665", - "issue:44295|issue:45397", - "issue:43531|issue:45362", "issue:42491|issue:43299", + "issue:43531|issue:45362", "issue:44246|issue:44871", - "issue:44336|issue:44485", - "issue:44242|issue:45593", "issue:45588|issue:46851", + "issue:44295|issue:45397", + "issue:44242|issue:45593", + "issue:44336|issue:44485", "issue:45003|issue:46821", "issue:42886|issue:46132", "issue:44451|issue:46752", @@ -13216,100 +13290,106 @@ "issue:44779|issue:45292", "issue:36246|issue:42757", "issue:36246|issue:44351", - "issue:41950|issue:44220", "issue:43901|issue:45003", "issue:43257|issue:45137", + "issue:41950|issue:44220", "issue:45003|issue:45561", "issue:44496|issue:45003", "issue:43540|issue:46655", - "issue:44162|issue:45200", "issue:45375|issue:45759", + "issue:44162|issue:45200", "issue:43262|issue:44246", - "issue:44295|issue:44749", - "issue:44479|issue:46655" + "issue:44479|issue:46655", + "issue:45859|issue:46399" ], - "prepared_review_unit_hash": "f8b23237c12613b182d28fd5b463858d905d235ea260df33658e0dcc7c24ecac", + "prepared_review_unit_hash": "b7e2d938e808d3b420a1eb3429323678dc390a3c69bc70b5f788a915d1500868", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 27, - "serialized_chars": 35210, - "estimated_input_tokens": 8803, - "estimated_eval_tokens": 17862 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36046, + "estimated_input_tokens": 9012, + "estimated_eval_tokens": 18280 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is heterogeneous and does not contain any strong duplicate pairs. Most items are distinct model- or subsystem-specific bugs; the soft similarities are mostly superficial (shared model family, similar failure mode, or same broad subsystem).", - "confidence": 0.82, - "canonical_issue_reason": "Issue 44479 is the most representative broad regression report: it covers multiple Qwen VLM variants, has a concrete video-input failure, and reads like a central upstream bug rather than a narrow edge case.", + "summary": "This cluster is mostly a loose similarity set of unrelated Transformers issues, with only broad thematic overlap around model loading, video/audio processors, and Qwen variants. None of the soft pairs look like the same concrete bug/change, so all are rejected.", + "confidence": 0.93, + "canonical_issue_reason": "Best representative is issue 45003 because it is a core internal bug in modeling_utils with broad potential impact across many loading paths, unlike the rest which are mostly model-specific regressions.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44479 is the best single issue to anchor this cluster because it is the clearest high-impact, actionable regression among the set. It is still not a duplicate of the others.", + "best_issue_reason": "Issue 45003 is the most cross-cutting and infrastructure-level item in the set; it is more suitable as a cluster representative than the many narrowly scoped model bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43540", + "right": "issue:46752", + "accept": false, + "reason": "Different code paths: Qwen3OmniMoe video input processing vs empty-chat-template handling. Same broad area, not the same bug." + }, { "left": "issue:45072", "right": "issue:46829", "accept": false, - "reason": "Different models and different failure classes: bfloat16 dtype mismatch vs incorrect output field mapping." + "reason": "Unrelated failures: bfloat16 dtype mismatch in two models vs hidden_states field wiring in video classification output." }, { "left": "issue:46614", "right": "issue:46665", "accept": false, - "reason": "Unrelated bugs: stale phonemizer backend state vs from_pretrained behavior inside Python classes." + "reason": "Stale phonemizer backend state is unrelated to from_pretrained behavior inside Python classes." }, { - "left": "issue:44295", - "right": "issue:45397", + "left": "issue:42491", + "right": "issue:43299", "accept": false, - "reason": "One is a position_ids buffer access bug; the other is a DeepSpeed ZeRO3 load issue for Gemma-4." + "reason": "Both mention Qwen3/MoE loading, but one is a LoRA compatibility problem and the other is a model-loading breakage in v5; not the same concrete defect." }, { "left": "issue:43531", "right": "issue:45362", "accept": false, - "reason": "Both involve Qwen, but one is sliding_window handling and the other is chat crashes for Qwen3.5-35B." + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to Qwen3.5 chat crashes." }, { - "left": "issue:42491", - "right": "issue:43299", + "left": "issue:44246", + "right": "issue:44871", "accept": false, - "reason": "Different loading regressions with different scopes: LoRA compatibility vs Qwen3VL MoE loading on dev branch." + "reason": "Import-time performance and eos_token_id config mismatch are unrelated." }, { - "left": "issue:44246", - "right": "issue:44871", + "left": "issue:45588", + "right": "issue:46851", "accept": false, - "reason": "No shared bug path: training loss omission vs eos_token_id config mismatch." + "reason": "Flash-attention s_aux=None crash is unrelated to variable-length position_ids handling." }, { - "left": "issue:44336", - "right": "issue:44485", + "left": "issue:44295", + "right": "issue:45397", "accept": false, - "reason": "Completely different areas: terminal ANSI output vs GLM-5 RoPE implementation." + "reason": "Buffer access after registering position_ids is unrelated to Gemma-4 ZeRO3 from_pretrained loading." }, { "left": "issue:44242", "right": "issue:45593", "accept": false, - "reason": "Both mention missing losses, but for different models and different mechanisms; not the same defect." + "reason": "Both concern losses not being added under a condition, but for different models and loss mechanisms; not mergeable as one bug." }, { - "left": "issue:45588", - "right": "issue:46851", + "left": "issue:44336", + "right": "issue:44485", "accept": false, - "reason": "Different code paths: flash_attention sink handling vs variable-length position_ids handling." + "reason": "ANSI code output in loading_report is unrelated to GLM-5 RoPE implementation." }, { "left": "issue:45003", @@ -13321,103 +13401,103 @@ "left": "issue:42886", "right": "issue:46132", "accept": false, - "reason": "Offline cache loading and AttentionInterface registration behavior are unrelated." + "reason": "Offline tokenizer cache loading is unrelated to AttentionInterface.register behavior changes." }, { "left": "issue:44451", "right": "issue:46752", "accept": false, - "reason": "Model loading failure vs empty-chat-template IndexError; different bugs." + "reason": "ScandiBERT loading failure is unrelated to empty-list apply_chat_template crashes." }, { "left": "issue:43550", "right": "issue:46506", "accept": false, - "reason": "Different subsystems and symptoms: torch.compile/SDPA vs CPU/CUDA output divergence." + "reason": "torch.compile+SDPA failure in Bamba is unrelated to CPU/CUDA result divergence in PPDocLayoutV3." }, { "left": "issue:44779", "right": "issue:45292", "accept": false, - "reason": "Tokenizer correctness regression is not the same as output_embeddings not resizing." + "reason": "DeepSeek tokenizer regression is unrelated to resize_token_embeddings not updating output embeddings." }, { "left": "issue:36246", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but for different symbols and unrelated packages/paths." + "reason": "Both are import errors, but for different symbols and different packages; not the same underlying change." }, { "left": "issue:36246", "right": "issue:44351", "accept": false, - "reason": "Different missing exports from transformers; likely version mismatch, not one shared bug." - }, - { - "left": "issue:41950", - "right": "issue:44220", - "accept": false, - "reason": "Video-classification image-processor lookup is unrelated to _torch_extract_fbank_features()." + "reason": "Missing Qwen2_5_VLImageProcessor export and missing HybridCache export are distinct API breakages." }, { "left": "issue:43901", "right": "issue:45003", "accept": false, - "reason": "Docs mismatch for pipeline args does not match sys.modules import safety bug." + "reason": "Documentation mismatch for return_all_scores is unrelated to sys.modules access in modeling_utils." }, { "left": "issue:43257", "right": "issue:45137", "accept": false, - "reason": "DeepSpeed-related, but one is Qwen3 MoE conversion and the other is ZeRO3 deque underflow." + "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed and Zero3 deque underflow are separate DeepSpeed-related failures." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification image-processor lookup and _torch_extract_fbank_features are different subsystems." }, { "left": "issue:45003", "right": "issue:45561", "accept": false, - "reason": "Unsafe module lookup and pytest-xdist file race are unrelated." + "reason": "Core loading bug is unrelated to pytest-xdist file race in testing utils." }, { "left": "issue:44496", "right": "issue:45003", "accept": false, - "reason": "Unrecognized model config vs unsafe sys.modules access are different failure modes." + "reason": "Unrecognized model_type in config loading is a model registry issue, not the sys.modules access bug." }, { "left": "issue:43540", "right": "issue:46655", "accept": false, - "reason": "Both are video-processing bugs, but in different models and likely different root causes." + "reason": "Both are video-related, but Qwen3OmniMoe video-input ValueError and SmolVLM padding calculation are model-specific and not evidently the same bug." }, { - "left": "issue:44162", - "right": "issue:45200", + "left": "issue:45375", + "right": "issue:45759", "accept": false, - "reason": "ESM2 breakage and Gemma-4 mm_token_type_ids defaults are unrelated." + "reason": "Both involve Qwen 3.5 multimodal config handling, but one is a missing vision config field and the other is from_config text_config unwrapping; different fixes." }, { - "left": "issue:45375", - "right": "issue:45759", + "left": "issue:44162", + "right": "issue:45200", "accept": false, - "reason": "Both concern composite Qwen multimodal configs, but one drops a field under strict parsing while the other mishandles text_config unwrapping." + "reason": "ESM2 breakage and Gemma-4 mm_token_type_ids defaulting are unrelated model-specific issues." }, { "left": "issue:43262", "right": "issue:44246", "accept": false, - "reason": "Audio chat-template sampling-rate default is unrelated to load-balancing loss behavior." + "reason": "Audio chat-template sampling-rate default and import performance are unrelated." }, { - "left": "issue:44295", - "right": "issue:44749", + "left": "issue:44479", + "right": "issue:46655", "accept": false, - "reason": "Position_ids buffer error and slow filtering after upgrade are not the same bug." + "reason": "Both concern video inputs, but the reported regressions target different model families and different failure modes." }, { - "left": "issue:44479", - "right": "issue:46655", + "left": "issue:45859", + "right": "issue:46399", "accept": false, - "reason": "Both involve video inputs, but the affected models and described failure mechanisms differ." + "reason": "Qwen3.5 CPU offload rotary embedding incompatibility and missing decorator on apply_rotary_pos_emb registration are distinct rotary/kernel issues." } ] }, @@ -13433,7 +13513,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 44, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -13446,12 +13526,11 @@ "issue:36246", "issue:42199", "issue:42886", - "issue:43065", "issue:43295", "issue:43352", - "issue:43408", "issue:43475", "issue:43526", + "issue:43540", "issue:43596", "issue:43653", "issue:44038", @@ -13459,10 +13538,12 @@ "issue:44077", "issue:44162", "issue:44242", + "issue:44295", "issue:44368", "issue:44466", "issue:44479", "issue:44623", + "issue:44749", "issue:44857", "issue:44877", "issue:44977", @@ -13479,222 +13560,235 @@ "issue:45584", "issue:45600", "issue:45656", - "issue:45859", "issue:46123", "issue:46170", "issue:46369", - "issue:46399", "issue:46506", "issue:46726", + "issue:46832", "issue:46851", "issue:46894", "issue:46897" ], "soft_pairs": [ - "issue:45859|issue:46399", "issue:43475|issue:46369", + "issue:44295|issue:44749", "issue:45137|issue:45600", - "issue:36246|issue:43596", - "issue:44877|issue:45375", - "issue:43295|issue:45292", "issue:45292|issue:45584", + "issue:43295|issue:45292", + "issue:44877|issue:45375", "issue:44466|issue:45335", - "issue:44623|issue:46726", + "issue:36246|issue:43596", "issue:44242|issue:44857", + "issue:44623|issue:46726", "issue:43526|issue:46897", - "issue:44038|issue:46506", "issue:44987|issue:45459", + "issue:44038|issue:46506", "issue:29870|issue:30333", - "issue:43352|issue:45250", "issue:42199|issue:46851", + "issue:43352|issue:45250", "issue:43526|issue:46894", - "issue:44060|issue:44077", - "issue:44368|issue:44977", - "issue:43653|issue:44162", "issue:45412|issue:46123", + "issue:43653|issue:44162", "issue:44479|issue:46170", - "issue:45397|issue:45656", + "issue:44060|issue:44077", + "issue:44368|issue:44977", "issue:42886|issue:45229", - "issue:43065|issue:43408" + "issue:45397|issue:45656", + "issue:44877|issue:46170", + "issue:43475|issue:43540", + "issue:46170|issue:46832" ], - "prepared_review_unit_hash": "637e8a121434b25fffbbcbb98be5f553fdc380eddfc9f3e88947c27a5a68196e", + "prepared_review_unit_hash": "72de5f4c9e64dbed4636a6426c0775e2c0000b5e7e726d9f77f875c932b49390", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 25, - "serialized_chars": 35310, - "estimated_input_tokens": 8828, - "estimated_eval_tokens": 17912 + "soft_pair_count": 27, + "serialized_chars": 35969, + "estimated_input_tokens": 8993, + "estimated_eval_tokens": 18242 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are unrelated bug reports across different models/subsystems, so there is no true duplicate set here. All soft edges should be rejected.", - "confidence": 0.89, - "canonical_issue_reason": "Issue 45229 is the strongest representative only in the sense of being the most substantial single report in the set, with the highest discussion activity and a clear, high-impact failure mode; however, the cluster as a whole is not a duplicate group.", - "canonical_pr_reason": null, - "best_issue_reason": "Issue 45229 is the best overall representative because it is detailed, concrete, and drew the most discussion, making it the most useful anchor if one issue must be chosen.", - "best_pr_reason": null, + "summary": "This set is highly heterogeneous; the soft pairs are topical matches at best, not true duplicates. I would not merge any of the proposed pairs.", + "confidence": 0.96, + "canonical_issue_reason": "No clear canonical issue: the items span unrelated bugs across models, tokenizers, configs, deepspeed, and pipelines.", + "canonical_pr_reason": "No PRs are present in this cluster.", + "best_issue_reason": "No single issue clearly subsumes the rest, so there is no defensible global best issue for duplicate triage.", + "best_pr_reason": "No PRs are present in this cluster.", "soft_edge_verdicts": [ - { - "left": "issue:45859", - "right": "issue:46399", - "accept": false, - "reason": "Both are Qwen3.5-related, but one is a CPU-offload rotary embedding incompatibility and the other is a kernel registration/decorator crash; different bugs and code paths." - }, { "left": "issue:43475", "right": "issue:46369", "accept": false, - "reason": "One is a missing attribute on Sam3VisionEncoderOutput, the other is a memory overflow in sam3-litetext; not the same underlying issue." + "reason": "Both mention SAM 3, but one is an attribute error in vision encoder output and the other is a memory overflow report; different failures and code paths." }, { - "left": "issue:45137", - "right": "issue:45600", + "left": "issue:44295", + "right": "issue:44749", "accept": false, - "reason": "ZeRO3 deque IndexError and stale auto_mappings references are unrelated failures; no shared concrete bug." + "reason": "Unrelated symptoms: position_ids buffer access error vs a performance regression after upgrading Transformers." }, { - "left": "issue:36246", - "right": "issue:43596", + "left": "issue:45137", + "right": "issue:45600", "accept": false, - "reason": "Missing Qwen2_5_VLImageProcessor import and a DeepSpeed/BertModel index error are different problems." + "reason": "DeepSpeed ZeRO3 deque IndexError and broken auto_mappings CI are unrelated issues." }, { - "left": "issue:44877", - "right": "issue:45375", + "left": "issue:45292", + "right": "issue:45584", "accept": false, - "reason": "Both involve strict config handling, but they affect different model families and different missing fields, so they are not duplicates." + "reason": "resize_token_embeddings behavior is unrelated to Whisper empty-transcription generation after align_special_tokens." }, { "left": "issue:43295", "right": "issue:45292", "accept": false, - "reason": "Processor.tokenizer/image-input regression and resize_token_embeddings not updating output embeddings are separate API bugs." + "reason": "Processor/tokenizer regression in v4.57.5 is not the same underlying bug as output_embeddings not updating on resize." }, { - "left": "issue:45292", - "right": "issue:45584", + "left": "issue:44877", + "right": "issue:45375", "accept": false, - "reason": "Embedding resize inconsistency and Whisper empty-transcription failure after align_special_tokens are different code paths." + "reason": "Both involve strict config handling, but they affect different configs and different missing fields/load failures." }, { "left": "issue:44466", "right": "issue:45335", "accept": false, - "reason": "Inconsistent lm_head.weight serialization and t5gemma resize_token_embeddings behavior are not the same bug, despite both touching tied embeddings." + "reason": "Serialization of tied weights on device vs t5gemma resize_token_embeddings not updating decoder.embed_tokens are different code paths." }, { - "left": "issue:44623", - "right": "issue:46726", + "left": "issue:36246", + "right": "issue:43596", "accept": false, - "reason": "processor.save_pretrained missing files and AutoProcessor backend-missing errors are different failure modes." + "reason": "Qwen2.5-VL import error and BertModel zero3 IndexError are unrelated." }, { "left": "issue:44242", "right": "issue:44857", "accept": false, - "reason": "Missing load-balancing loss when output_router_logits=False and an AMP/CUDA loss crash are unrelated." + "reason": "Load-balancing loss omission in MoE routing is unrelated to an AMP/CUDA crash in LwDetrImageLoss." }, { - "left": "issue:43526", - "right": "issue:46897", + "left": "issue:44623", + "right": "issue:46726", "accept": false, - "reason": "BEiT reduce_labels returning one label and Florence2 double-shift training loss are distinct model-specific bugs." + "reason": "Missing files in processor.save_pretrained is different from AutoProcessor failing when a vision backend is absent." }, { - "left": "issue:44038", - "right": "issue:46506", + "left": "issue:43526", + "right": "issue:46897", "accept": false, - "reason": "Qwen3-VL-Moe/transformers 5.0 compatibility and PPDocLayoutV3 CPU-vs-CUDA differences are unrelated." + "reason": "BeitImageProcessorFast reducing labels incorrectly is not the same bug as Florence2 training-loss double shifting." }, { "left": "issue:44987", "right": "issue:45459", "accept": false, - "reason": "A loading failure for physical-intelligence/fast and protobuf-missing tokenizer error masking are different problems." + "reason": "Loading failure in transformers>=5.1.0 and protobuf-related tokenizer error masking are different failures." }, { - "left": "issue:29870", - "right": "issue:30333", + "left": "issue:44038", + "right": "issue:46506", "accept": false, - "reason": "BART generate() UnboundLocalError and MLflow job-status reporting are completely unrelated." + "reason": "Qwen3-VL-Moe compatibility issue and CPU/CUDA output mismatch in PPDocLayoutV3 are unrelated." }, { - "left": "issue:43352", - "right": "issue:45250", + "left": "issue:29870", + "right": "issue:30333", "accept": false, - "reason": "Nemotron flash-attn unsupported-model gating and a generic Flash Attention 2.0 issue are not the same concrete bug." + "reason": "Bart generate() UnboundLocalError and MLFlow job-state reporting are unrelated." }, { "left": "issue:42199", "right": "issue:46851", "accept": false, - "reason": "DETR cardinality/background-class logic and Qwen3.5 variable-length position_ids are unrelated." + "reason": "DETR cardinality error and Qwen3.5 variable-length position_ids issue affect different models and logic." }, { - "left": "issue:43526", - "right": "issue:46894", - "accept": false, - "reason": "BEiT label reduction bug and CohereASR training-loss double-shift bug are different issues." + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "A specific model not supporting Flash Attention 2.0 yet is not the same as a generic Flash Attention 2.0 issue." }, { - "left": "issue:44060", - "right": "issue:44077", + "left": "issue:43526", + "right": "issue:46894", "accept": false, - "reason": "Incorrect tied-weights warning and patchtsmixer post_init policy are not the same underlying change." + "reason": "Label reduction in BeitImageProcessorFast is unrelated to the ASR training-loss double-shift bug." }, { - "left": "issue:44368", - "right": "issue:44977", + "left": "issue:45412", + "right": "issue:46123", "accept": false, - "reason": "One is a tied-weights warning during LoRA fine-tuning, the other is a flash-attention generation failure; different code paths." + "reason": "RT-DETR memory leak and MaskGenerationPipeline partial-batch dropping are different bugs." }, { "left": "issue:43653", "right": "issue:44162", "accept": false, - "reason": "BigBird tokenizer special-token registration and ESM2 breakage are unrelated." + "reason": "BigBirdTokenizer special-token decode bug and ESM2 breakage are not the same underlying defect." }, { - "left": "issue:45412", - "right": "issue:46123", + "left": "issue:44479", + "right": "issue:46170", "accept": false, - "reason": "RT-DETR memory not released on deletion and MaskGenerationPipeline final-batch handling are different bugs." + "reason": "Qwen video-input regression and Windows ROCm/distributed import failure are unrelated." }, { - "left": "issue:44479", - "right": "issue:46170", + "left": "issue:44060", + "right": "issue:44077", "accept": false, - "reason": "A multimodal video regression and a Windows/ROCm torch.distributed import issue are unrelated." + "reason": "Tied-weights warning in Qwen3-Next and patchtsmixer post_init policy are different model/config issues." }, { - "left": "issue:45397", - "right": "issue:45656", + "left": "issue:44368", + "right": "issue:44977", "accept": false, - "reason": "Gemma-4 Zero3 from_pretrained and double optimizer.step() under deepspeed are different failures." + "reason": "A warning about tie_word_embeddings and a flash-attention generation failure are not the same bug." }, { "left": "issue:42886", "right": "issue:45229", "accept": false, - "reason": "Offline tokenizer cache loading and Gemma4 multi-GPU CUDA OOM are unrelated." + "reason": "Offline tokenizer cache loading and Gemma4 multi-GPU OOM are unrelated." }, { - "left": "issue:43065", - "right": "issue:43408", + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Gemma-4 ZeRO3 from_pretrained failure and deepspeed optimizer double-step are different code paths." + }, + { + "left": "issue:44877", + "right": "issue:46170", "accept": false, - "reason": "Both are SAM3-related, but one is a dummy Conv2d implementation detail and the other is a model-type mismatch warning; not the same bug." + "reason": "Strict config loading and Windows ROCm torch.distributed dependency errors are unrelated." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "SAM 3 video encoder attribute error and Qwen3OmniMoe video ValueError are different model families and failures." + }, + { + "left": "issue:46170", + "right": "issue:46832", + "accept": false, + "reason": "Windows ROCm import/type error and RT-DETR feature-level TypeError are unrelated bugs." } ] }, @@ -13710,7 +13804,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 45, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -13718,13 +13812,15 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-502", "nodes": [ + "issue:28282", + "issue:30990", "issue:42994", "issue:43065", "issue:43232", + "issue:43408", "issue:43441", "issue:43475", "issue:43482", - "issue:43540", "issue:43643", "issue:43646", "issue:43673", @@ -13742,7 +13838,6 @@ "issue:44623", "issue:44749", "issue:44871", - "issue:44877", "issue:44995", "issue:45103", "issue:45200", @@ -13756,257 +13851,256 @@ "issue:46082", "issue:46123", "issue:46132", - "issue:46170", "issue:46202", "issue:46421", "issue:46431", "issue:46710", "issue:46726", "issue:46752", - "issue:46832", + "issue:46821", "issue:46851", "issue:46858" ], "soft_pairs": [ - "issue:43475|issue:43540", - "issue:44877|issue:46170", - "issue:46170|issue:46832", + "issue:43065|issue:43408", "issue:43723|issue:46710", "issue:44368|issue:45563", "issue:44995|issue:45200", "issue:43931|issue:44368", + "issue:45987|issue:46082", "issue:44188|issue:46202", "issue:44483|issue:46202", "issue:43906|issue:46202", - "issue:45987|issue:46082", - "issue:44995|issue:45588", "issue:44188|issue:46421", + "issue:44995|issue:45588", "issue:45685|issue:46431", "issue:45103|issue:46858", - "issue:44297|issue:44623", "issue:43065|issue:46123", "issue:44361|issue:44749", + "issue:44297|issue:44623", "issue:43994|issue:46726", "issue:44749|issue:45656", "issue:43701|issue:44749", "issue:44871|issue:46132", - "issue:44623|issue:45656", "issue:43646|issue:44162", + "issue:44623|issue:45656", "issue:43441|issue:46752", - "issue:45375|issue:45478", + "issue:43475|issue:43643", "issue:42994|issue:43482", + "issue:28282|issue:30990", + "issue:45375|issue:45478", "issue:44188|issue:46851", - "issue:43475|issue:43643", + "issue:43475|issue:46821", "issue:43232|issue:43673" ], - "prepared_review_unit_hash": "3fff90d8ec0b5fe93e8157f2dba9544747a918584fcab52c27a738c9e81a2d98", + "prepared_review_unit_hash": "21e0208a65cc23330133d4b9ce07645c9d371230ccfbdb8a164e668e0c5723bf", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36822, - "estimated_input_tokens": 9206, - "estimated_eval_tokens": 18668 + "serialized_chars": 36783, + "estimated_input_tokens": 9196, + "estimated_eval_tokens": 18648 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This batch is mostly a set of unrelated closed issue reports with some loose topical similarity around cache/config/save_pretrained regressions. None of the provided soft pairs are strong enough to treat as duplicates; the closest ones are still different code paths or different failure modes.", - "confidence": 0.64, - "canonical_issue_reason": "Issue 46858 is the most specific, actionable report: it pinpoints `_prefill`, `cache.get_seq_length()`, the bad tensor return type, and the non-CUDA/torch.compile slice-arithmetic failure. It best represents the concrete bug class among the issues shown.", + "summary": "Mostly a loose collection of unrelated Transformers issues spanning model loading, generation/cache, serialization, and model-specific regressions. No pull requests are present. I found no soft pair that is clearly the same underlying bug with enough specificity to merge, so I kept all soft edges rejected.", + "confidence": 0.91, + "canonical_issue_reason": "If forced to pick one representative issue, 46858 is the clearest concrete core-runtime regression: it names a specific failing path (_prefill/cache.get_seq_length on static cache with torch.compile on non-CUDA backends) rather than a broad symptom.", "canonical_pr_reason": null, - "best_issue_reason": "46858 is the best overall issue candidate because it has the clearest root-cause description and the most directly fixable code path. It is more precise than the broader cache-related reports.", + "best_issue_reason": "46858 is the best global representative because it is precise, actionable, and tied to a core cache/generation bug that is easy to understand and likely to have high impact.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43475", - "right": "issue:43540", - "accept": false, - "reason": "Both are model-specific runtime failures, but one is a missing attribute in SAM 3 video and the other is a video-input ValueError in Qwen3OmniMoe; different subsystems and failure mechanisms." - }, - { - "left": "issue:44877", - "right": "issue:46170", - "accept": false, - "reason": "One is strict config loading for granite_speech, the other is a Windows/ROCm torch.distributed import dependency issue; unrelated code paths." - }, - { - "left": "issue:46170", - "right": "issue:46832", + "left": "issue:43065", + "right": "issue:43408", "accept": false, - "reason": "Import/dependency breakage on Windows is not the same as RT-DETR forward shape handling when feature levels exceed backbone outputs." + "reason": "Both are SAM3-related, but one is a dummy Conv2d implementation choice and the other is a config/model-type mismatch warning; different bug classes." }, { "left": "issue:43723", "right": "issue:46710", "accept": false, - "reason": "Tokenizer loading in v5 and an accuracy regression after a version upgrade are different bug classes; no shared concrete failure path." + "reason": "Tokenizer loading failure and an accuracy regression after upgrade are unrelated failure modes." }, { "left": "issue:44368", "right": "issue:45563", "accept": false, - "reason": "A warning about `tie_word_embeddings` during LoRA fine-tuning is unrelated to a stale `num_return_sequences` warning in paged generate()." + "reason": "Both mention warnings, but one is about tie_word_embeddings and the other is a stale num_return_sequences warning; not the same issue." }, { "left": "issue:44995", "right": "issue:45200", "accept": false, - "reason": "GLM MoE stale indexer-cache crash on second forward is not the same as Gemma 4 requiring `mm_token_type_ids` for text-only fine-tuning." + "reason": "Stale indexer cache on second forward pass and missing mm_token_type_ids defaults are different problems." }, { "left": "issue:43931", "right": "issue:44368", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch and a Qwen3.5 embedding warning are different model-loading/fine-tuning issues." + "reason": "Weight-shape mismatch for Qwen3-VL is unrelated to a tie_word_embeddings warning." + }, + { + "left": "issue:45987", + "right": "issue:46082", + "accept": false, + "reason": "StaticCache seq_length return-type bug and LlamaConfig head_dim validation are separate code paths." }, { "left": "issue:44188", "right": "issue:46202", "accept": false, - "reason": "Diverging attention kernels under torch.compile is unrelated to Gemma pipeline roundtrip tokenization behavior." + "reason": "torch.compile attention-kernel branching and Gemma pipeline roundtrip behavior are unrelated." }, { "left": "issue:44483", "right": "issue:46202", "accept": false, - "reason": "API request rejection on `/v1/chat/completions` is not the same as Gemma tokenizer roundtrip failure for the underscore-like token." + "reason": "Chat-completions API acceptance failure is not the same as a Gemma pipeline token roundtrip bug." }, { "left": "issue:43906", "right": "issue:46202", "accept": false, - "reason": "An isolated reproduction of an earlier issue is not the same as a Gemma pipeline tokenization roundtrip bug." + "reason": "An isolated reproduction of another issue is not the same as a Gemma pipeline failure." }, { - "left": "issue:45987", - "right": "issue:46082", + "left": "issue:44188", + "right": "issue:46421", "accept": false, - "reason": "StaticCache returning a shape-(1,) tensor is a cache API contract bug; LlamaConfig rejecting explicit head_dim is a separate config validation issue." + "reason": "Both involve cache/compile themes, but one is divergent attention kernels from branching and the other is spurious recompilation from lazy cache init; not the same bug." }, { "left": "issue:44995", "right": "issue:45588", "accept": false, - "reason": "Both mention runtime failures, but one is stale indexer cache on second forward and the other is flash-attention crashing on `s_aux=None`; different implementations." - }, - { - "left": "issue:44188", - "right": "issue:46421", - "accept": false, - "reason": "Both are cache/compile related, but one is attention-kernel divergence from branching and the other is spurious recompilation from lazy cache initialization; not the same bug." + "reason": "Second-forward stale cache and flash_attention s_aux=None crash are different failures." }, { "left": "issue:45685", "right": "issue:46431", "accept": false, - "reason": "An MPS histogram dtype error in MoE code is unrelated to missing chat templates in processors." + "reason": "MPS histogram dtype support and missing chat template on a processor are unrelated." }, { "left": "issue:45103", "right": "issue:46858", "accept": false, - "reason": "Auto-docstring annotation handling and cache sequence-length tensor arithmetic are entirely different code paths." - }, - { - "left": "issue:44297", - "right": "issue:44623", - "accept": false, - "reason": "Both involve save_pretrained, but tokenizer_class mismatch and missing processor files are different serialization failures in different components." + "reason": "Docstring helper crash under future annotations and static-cache seq_length/device-tensor arithmetic are unrelated." }, { "left": "issue:43065", "right": "issue:46123", "accept": false, - "reason": "A dummy Conv2d in SAM3 pixel decoding is unrelated to `is_last` being false on the final partial batch in MaskGenerationPipeline." + "reason": "Dummy Conv2d in SAM3 and MaskGenerationPipeline final-batch handling are different issues." }, { "left": "issue:44361", "right": "issue:44749", "accept": false, - "reason": "MLukeTokenizer task AttributeError and a data-processing slowdown after upgrade do not share the same underlying bug." + "reason": "Tokenizer AttributeError and a performance regression during filtering are unrelated." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "Tokenizer config mismatch on save and missing processor files on save are both serialization-related, but they are different APIs and symptoms; not enough to call them the same bug." }, { "left": "issue:43994", "right": "issue:46726", "accept": false, - "reason": "SigLIP2 nonsensical outputs through AutoModel/pipeline are not the same as AutoProcessor reporting an unrecognized image processor when a backend is missing." + "reason": "Nonsensical SigLIP2 model outputs and missing vision-backend image processor handling are different problems." }, { "left": "issue:44749", "right": "issue:45656", "accept": false, - "reason": "A performance regression after upgrading Transformers is unrelated to DeepSpeed calling optimizer.step twice." + "reason": "Data filtering slowdown and duplicated optimizer steps under deepspeed are unrelated." }, { "left": "issue:43701", "right": "issue:44749", "accept": false, - "reason": "Checkpoint key mismatch on resume is a distinct training-state bug from a filtering slowdown regression." + "reason": "Checkpoint key mismatch and a performance regression do not describe the same failure." }, { "left": "issue:44871", "right": "issue:46132", "accept": false, - "reason": "Gemma eos_token_id config inconsistency is unrelated to AttentionInterface.register changing registered-function behavior." + "reason": "Gemma eos_token_id inconsistency and AttentionInterface.register behavior change are unrelated." }, { - "left": "issue:44623", - "right": "issue:45656", + "left": "issue:43646", + "right": "issue:44162", "accept": false, - "reason": "Processor save_pretrained missing files and DeepSpeed double optimizer.step are unrelated issues." + "reason": "A general v5 custom-model-init breakage and ESM2-specific breakage are not the same underlying bug." }, { - "left": "issue:43646", - "right": "issue:44162", + "left": "issue:44623", + "right": "issue:45656", "accept": false, - "reason": "Custom model initialization breakage in Transformers v5 is not the same as ESM2 being broken; different model families and failure modes." + "reason": "Missing processor save files and doubled deepspeed optimizer steps are different subsystems and symptoms." }, { "left": "issue:43441", "right": "issue:46752", "accept": false, - "reason": "Ministral FlashAttention failure and empty-list apply_chat_template IndexError are distinct bugs." + "reason": "FlashAttention failure for Ministral-3 and empty-list apply_chat_template IndexError are unrelated." }, { - "left": "issue:45375", - "right": "issue:45478", + "left": "issue:43475", + "right": "issue:43643", "accept": false, - "reason": "Both concern Qwen3.5 MoE config/loading, but one is a strict-field drop in a config object and the other is a generic from_pretrained error; not enough evidence they are the same concrete bug." + "reason": "Missing fpn_position_embeddings on Sam3 output and missing config fields with trust_remote_code are different issues." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are opposite directions of the model lifecycle and different code paths." + "reason": "Quantized model saving failure and GGUF loading failure are opposite directions and distinct bugs." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "ImportError for missing PyTorch and Sentence Transformers hanging on load are unrelated." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "These are related in model family and loading path, but one is a specific config field being dropped while the other is a broad from_pretrained error; not enough evidence they are the same concrete bug." }, { "left": "issue:44188", "right": "issue:46851", "accept": false, - "reason": "Torch.compile attention-kernel divergence is unrelated to Qwen3.5 variable-length position_ids handling." + "reason": "Attention-kernel branching under torch.compile and Qwen3.5 variable-length position_ids handling are different failures." }, { "left": "issue:43475", - "right": "issue:43643", + "right": "issue:46821", "accept": false, - "reason": "A missing `fpn_position_embeddings` attribute in SAM 3 video output is not the same as missing fields from `trust_remote_code=True` config loading." + "reason": "Sam3 vision-encoder missing field and TimesFM window_size AttributeError are unrelated." }, { "left": "issue:43232", "right": "issue:43673", "accept": false, - "reason": "`_update_model_kwargs_for_generation` after sync_gpus and missing GenerationMixin cache during chunked_prefill are related to generation, but they describe different breakages and code paths." + "reason": "A generation kwargs update bug after sync_gpus and a missing cache during chunked_prefill are both generation-related but clearly different code paths." } ] }, @@ -14022,7 +14116,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 46, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -14030,16 +14124,15 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-502", "nodes": [ - "issue:28282", - "issue:30990", "issue:36010", "issue:39692", "issue:41628", "issue:42371", "issue:42673", + "issue:42831", "issue:42994", "issue:43278", - "issue:43475", + "issue:43504", "issue:43525", "issue:43704", "issue:43723", @@ -14067,6 +14160,7 @@ "issue:45375", "issue:45397", "issue:45406", + "issue:45561", "issue:45593", "issue:45753", "issue:46202", @@ -14075,259 +14169,252 @@ "issue:46463", "issue:46701", "issue:46726", - "issue:46821", "issue:46851", "issue:46877" ], "soft_pairs": [ - "issue:45230|issue:45397", - "issue:28282|issue:30990", - "issue:41628|issue:44933", "issue:44964|issue:45137", - "issue:45310|issue:45375", - "issue:43475|issue:46821", - "issue:43986|issue:46726", "issue:44869|issue:45356", - "issue:44038|issue:44623", + "issue:43986|issue:46726", + "issue:45310|issue:45375", + "issue:41628|issue:44933", + "issue:46463|issue:46877", "issue:39692|issue:46726", + "issue:45230|issue:45397", + "issue:44038|issue:44623", "issue:44373|issue:46851", - "issue:46463|issue:46877", - "issue:42994|issue:43525", - "issue:42371|issue:45092", "issue:43723|issue:46326", + "issue:42371|issue:45092", + "issue:42994|issue:43525", "issue:43873|issue:44945", - "issue:43784|issue:46399", "issue:45406|issue:46202", - "issue:41628|issue:44351", + "issue:43784|issue:46399", "issue:44945|issue:45753", "issue:44857|issue:46877", "issue:45593|issue:46877", - "issue:42673|issue:43704", - "issue:44038|issue:45310", - "issue:44222|issue:45310", + "issue:41628|issue:44351", "issue:36010|issue:41628", "issue:43278|issue:45137", - "issue:42994|issue:43824", + "issue:45081|issue:45103", "issue:46202|issue:46701", - "issue:45081|issue:45103" + "issue:44038|issue:45310", + "issue:42673|issue:43704", + "issue:44222|issue:45310", + "issue:42994|issue:43824", + "issue:42831|issue:44222", + "issue:43504|issue:45561" ], - "prepared_review_unit_hash": "b62d181327f14df2b5269847a8ede6f4930fd0e5fba2ac67181919646c9622e3", + "prepared_review_unit_hash": "81b1ab48c68a08b796c4a8e08b4cef0ffc0063c2a26bc3152ce3fa35570df57b", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36412, - "estimated_input_tokens": 9103, - "estimated_eval_tokens": 18462 + "serialized_chars": 35841, + "estimated_input_tokens": 8961, + "estimated_eval_tokens": 18178 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly unrelated import/loading/tokenizer reports. The strongest duplicate signal is the Qwen3ForCausalLM VRAM-leak pair, which shares the exact title and the same failure mode, but the packet does not explicitly confirm linkage beyond that match.", - "confidence": 0.84, - "canonical_issue_reason": "issue:42673 is the best-supported canonical within the likely duplicate pair: it is earlier and has fuller discussion, while issue:43704 repeats the exact title and same VRAM-leak scenario.", + "summary": "This cluster is mostly heterogeneous and does not form a broad duplicate set. The only clear duplicate is the repeated Qwen3ForCausalLM VRAM leak report (issue:42673 and issue:43704); the rest are separate bugs spanning imports, tokenizers, quantization, processors, and model-specific regressions.", + "confidence": 0.92, + "canonical_issue_reason": "issue:42673 is the strongest anchor because it has a true duplicate in issue:43704 with the exact same title and bug description, while also having much richer discussion context.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42673 is the strongest issue to anchor the cluster because it is the earliest, most detailed report and has the clearest duplicate-style counterpart with the same title and failure mode.", + "best_issue_reason": "issue:42673 is the best representative issue for the only confirmed duplicate pair in this cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:45230", - "right": "issue:45397", + "left": "issue:44964", + "right": "issue:45137", "accept": false, - "reason": "Generic bug report vs a specific gemma-4 ZeRO3 from_pretrained failure; not enough evidence they are the same bug." + "reason": "Different failures: Phi-4 multimodal model loading vs DeepSpeed ZeRO3 deque underflow." }, { - "left": "issue:28282", - "right": "issue:30990", + "left": "issue:44869", + "right": "issue:45356", "accept": false, - "reason": "Different problems: missing PyTorch dependency import error vs Sentence Transformers hanging on load." + "reason": "Both involve tokenization, but one is a Whisper timestamp decode crash and the other is a Kimi tokenizer regression with mistral regex handling." }, { - "left": "issue:41628", - "right": "issue:44933", + "left": "issue:43986", + "right": "issue:46726", "accept": false, - "reason": "Both are import-related, but they concern different symbols and failure paths, so they do not look like the same concrete bug." + "reason": "Both mention processor/backend availability, but they are different code paths and errors (torchvision-missing video model vs unrecognized image processor)." }, { - "left": "issue:44964", - "right": "issue:45137", + "left": "issue:45310", + "right": "issue:45375", "accept": false, - "reason": "Completely different failures: Phi-4 multimodal loading vs a DeepSpeed ZeRO3 deque IndexError." + "reason": "Same model family, but one is a from_pretrained error and the other is a missing config field silently dropped; not enough to treat as the same bug." }, { - "left": "issue:45310", - "right": "issue:45375", + "left": "issue:41628", + "right": "issue:44933", "accept": false, - "reason": "Same model family, but one is a from_pretrained error and the other is a missing config-field/serialization issue; likely related but not clearly the same bug." + "reason": "Both are import errors, but they concern different symbols and modules; no evidence of one underlying fix." }, { - "left": "issue:43475", - "right": "issue:46821", + "left": "issue:46463", + "right": "issue:46877", "accept": false, - "reason": "Both are AttributeError reports, but they hit unrelated models and attributes." + "reason": "Unrelated: T5 decoder-input validation message vs auto_docstring import-time diagnostics output." }, { - "left": "issue:43986", + "left": "issue:39692", "right": "issue:46726", "accept": false, - "reason": "Both involve AutoProcessor and vision backends, but they describe different code paths and different failures." + "reason": "SigLIP2 documentation/example errors are not the same bug as AutoProcessor backend-missing runtime handling." }, { - "left": "issue:44869", - "right": "issue:45356", + "left": "issue:45230", + "right": "issue:45397", "accept": false, - "reason": "Tokenizer/codec-related, but these are distinct bugs affecting different models and workflows." + "reason": "Generic bug report vs Gemma-4 ZeRO3 from_pretrained failure; no shared underlying code-path." }, { "left": "issue:44038", "right": "issue:44623", "accept": false, - "reason": "Different areas: Qwen3-VL-Moe / Transformers 5 compatibility vs processor.save_pretrained missing files." - }, - { - "left": "issue:39692", - "right": "issue:46726", - "accept": false, - "reason": "A docs example with model/processor mismatch and quantization failure is not the same as a backend-missing AutoProcessor error." + "reason": "Qwen3-VL-Moe load incompatibility and processor.save_pretrained missing files are separate issues." }, { "left": "issue:44373", "right": "issue:46851", "accept": false, - "reason": "One is a wrong docstring, the other is a functional position_ids bug; not duplicates." + "reason": "Docstring mismatch for position_ids is unrelated to a runtime variable-length input bug." }, { - "left": "issue:46463", - "right": "issue:46877", - "accept": false, - "reason": "Different subsystems: T5 decoder input validation messaging vs auto_docstring import-time diagnostics." - }, - { - "left": "issue:42994", - "right": "issue:43525", + "left": "issue:43723", + "right": "issue:46326", "accept": false, - "reason": "Quantized model save failure vs missing Llama4Config attribute; unrelated bugs." + "reason": "Tokenizer loading in v5 and continue_final_message for mistral_common are different bugs despite both touching tokenization." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF32 API guidance vs InternVL2 checkpoint/meta-init incompatibility; no shared underlying defect." + "reason": "TF32 API settings guidance and InternVL2 meta-initialization incompatibility are unrelated." }, { - "left": "issue:43723", - "right": "issue:46326", + "left": "issue:42994", + "right": "issue:43525", "accept": false, - "reason": "AutoTokenizer v5 loading issue vs mistral_common continue_final_message bug; different tokenizer paths." + "reason": "Quantized model saving failure and missing pad_token_id on Llama4Config are different defects." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Quantization/offloading behavior vs pipeline parallelism output correctness; not the same code-path problem." - }, - { - "left": "issue:43784", - "right": "issue:46399", - "accept": false, - "reason": "Importing sentence-transformers with latest transformers vs a Qwen3.5 kernel registration error; unrelated." + "reason": "Offloading/quantization behavior and pipeline-parallel incorrect output are separate issues." }, { "left": "issue:45406", "right": "issue:46202", "accept": false, - "reason": "Gemma4Processor missing _tokenizer in transformers serve vs Gemma pipeline roundtrip on \u2581; different bugs." + "reason": "Gemma processor missing _tokenizer and Gemma roundtrip on \u2581 are different failure modes." }, { - "left": "issue:41628", - "right": "issue:44351", + "left": "issue:43784", + "right": "issue:46399", "accept": false, - "reason": "Both are missing exports from transformers, but for different symbols and likely different regressions." + "reason": "Sentence-transformers import NameError and Qwen3.5 kernel registration crash are unrelated." }, { "left": "issue:44945", "right": "issue:45753", "accept": false, - "reason": "Pipeline parallelism correctness vs a Qwen3.5 image-triggered infinite loop; not the same defect." + "reason": "Pipeline parallelism output corruption and Qwen3.5 infinite-loop-on-image are different bugs." }, { "left": "issue:44857", "right": "issue:46877", "accept": false, - "reason": "CUDA float16 AMP crash in LwDetrImageLoss is unrelated to auto_docstring stdout diagnostics." + "reason": "LwDetrImageLoss AMP crash and auto_docstring stdout diagnostics are unrelated." }, { "left": "issue:45593", "right": "issue:46877", "accept": false, - "reason": "D-FINE auxiliary-loss behavior has nothing to do with auto_docstring import-time logging." - }, - { - "left": "issue:42673", - "right": "issue:43704", - "accept": true, - "reason": "Exact same title and matching VRAM-leak scenario in Qwen3ForCausalLM with multiple dataloader threads; this is the best-supported duplicate pair." - }, - { - "left": "issue:44038", - "right": "issue:45310", - "accept": false, - "reason": "Both mention Qwen/Moe and loading, but they are different models and failure modes; not safe to merge." + "reason": "D-FINE auxiliary-loss behavior is unrelated to auto_docstring import-time logging." }, { - "left": "issue:44222", - "right": "issue:45310", + "left": "issue:41628", + "right": "issue:44351", "accept": false, - "reason": "FP8 save_pretrained failure and from_pretrained load failure are different operations and bugs." + "reason": "Both are import/export regressions, but they involve different missing symbols and different modules." }, { "left": "issue:36010", "right": "issue:41628", "accept": false, - "reason": "Different missing imports from transformers; same general theme, not the same concrete issue." + "reason": "GenerationMixin import failure and AutoImageProcessor import failure are different missing-export regressions." }, { "left": "issue:43278", "right": "issue:45137", "accept": false, - "reason": "Embedding dtype mismatch during evaluate is unrelated to DeepSpeed ZeRO3 deque underflow." + "reason": "Embedding dtype drift in evaluation is unrelated to a ZeRO3 deque error." }, { - "left": "issue:42994", - "right": "issue:43824", + "left": "issue:45081", + "right": "issue:45103", "accept": false, - "reason": "Quantized saving vs missing Qwen2_5_VLForConditionalGeneration export are unrelated." + "reason": "Tokenizer regex patch crash and auto_docstring future-annotations crash are distinct code paths." }, { "left": "issue:46202", "right": "issue:46701", "accept": false, - "reason": "Different Gemma-related failures: tokenizer roundtrip on \u2581 vs image-text-to-text pipeline breakage." + "reason": "Both involve Gemma, but the pipeline roundtrip bug and the image-text-to-text pipeline failure are not the same underlying issue." }, { - "left": "issue:45081", - "right": "issue:45103", - "accept": false, - "reason": "Different helper crashes in different components: mistral regex patching vs auto_docstring parameter processing." - } - ] - }, - "evaluator_result": { - "accept": false, - "feedback": "The duplicate pair choice is plausible, but the writeup overstates the evidence: the packet does not explicitly confirm any duplicate linkage, so phrases like \"only confirmed duplicate target\" are not grounded. Please keep the verdict conservative (e.g. \"best-supported duplicate pair\" based on identical title and matching failure mode) rather than claiming confirmation." + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Exact same VRAM leak report for Qwen3ForCausalLM in multiple dataloader threads; clearly the same bug." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "FP8 save_pretrained moe and Qwen3.5 from_pretrained error are different operations and failures." + }, + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Quantized model saving failure is unrelated to Qwen2_5_VL import failure." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "Both mention FP8, but one is an accuracy problem and the other is a saving bug; not the same fix." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Legacy-field model loading bug and pytest-xdist file race are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: the only clearly duplicate pair is issue:42673 and issue:43704, which share the exact same title, and the discussion-count contrast supports the canonical anchor choice. The other soft-edge rejections are appropriately conservative and do not overclaim equivalence from weak title similarity alone." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -14337,7 +14424,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 47, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -14355,11 +14442,10 @@ "issue:42175", "issue:42222", "issue:42673", - "issue:42831", "issue:42898", "issue:43232", + "issue:43295", "issue:43479", - "issue:43504", "issue:43723", "issue:43784", "issue:43824", @@ -14367,7 +14453,6 @@ "issue:43976", "issue:44016", "issue:44079", - "issue:44222", "issue:44496", "issue:44568", "issue:44623", @@ -14381,6 +14466,7 @@ "issue:45103", "issue:45200", "issue:45310", + "issue:45356", "issue:45405", "issue:45561", "issue:45600", @@ -14395,15 +14481,13 @@ "issue:46909" ], "soft_pairs": [ - "issue:42831|issue:44222", "issue:43232|issue:44792", - "issue:42175|issue:44933", - "issue:43504|issue:45561", "issue:45561|issue:45600", - "issue:43976|issue:44933", + "issue:42175|issue:44933", "issue:44496|issue:45561", - "issue:30333|issue:33290", "issue:45405|issue:46877", + "issue:43976|issue:44933", + "issue:30333|issue:33290", "issue:30990|issue:35141", "issue:42898|issue:44568", "issue:42175|issue:45712", @@ -14412,228 +14496,230 @@ "issue:44623|issue:45310", "issue:42673|issue:46693", "issue:43479|issue:44869", - "issue:43723|issue:44568", "issue:42898|issue:45812", + "issue:43723|issue:44568", "issue:44016|issue:46231", "issue:44625|issue:46877", - "issue:42898|issue:44779", "issue:39692|issue:46909", + "issue:42898|issue:44779", "issue:43479|issue:44079", - "issue:43874|issue:45092", "issue:43784|issue:43824", - "issue:27957|issue:30333", "issue:44734|issue:45103", + "issue:43874|issue:45092", + "issue:27957|issue:30333", + "issue:42898|issue:43295", "issue:30333|issue:34689", - "issue:42222|issue:45200" + "issue:42222|issue:45200", + "issue:42898|issue:45356" ], - "prepared_review_unit_hash": "cdaf5c95085b7318b63e9233c935f920b3b62a1d4c8128c176e9994e15b1a2e6", + "prepared_review_unit_hash": "fc90859cc6d22c238cdeda93e0f9c6dabb80094b3d5c808020df59e56683eb80", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36794, - "estimated_input_tokens": 9199, - "estimated_eval_tokens": 18654 + "serialized_chars": 36369, + "estimated_input_tokens": 9093, + "estimated_eval_tokens": 18442 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the items span unrelated model bugs, tokenizer regressions, docs issues, packaging problems, and infra/test failures. None of the soft-edge pairs look like the same underlying bug or change, so all should be rejected.", - "confidence": 0.97, - "canonical_issue_reason": "No canonical issue: there is no single underlying bug here; the issues cover distinct code paths and subsystems.", + "summary": "This cluster is heterogeneous; none of the soft pairs look like the same underlying bug/change. I\u2019d keep issue 45812 as the best representative standalone issue because it describes a broad, high-impact silent tokenizer regression.", + "confidence": 0.81, + "canonical_issue_reason": "Issue 45812 is the strongest canonical issue here: it reports a broad, concrete regression affecting all Granite models, with clear user impact and a precise failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "No single issue is a good global representative because the cluster is not a duplicate set; it is a mix of unrelated reports.", + "best_issue_reason": "Issue 45812 is also the best standalone issue in the cluster: it is actionable, high-impact, and more central than the narrower infra/docs/config reports around it.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:42831", - "right": "issue:44222", + "left": "issue:43232", + "right": "issue:44792", "accept": false, - "reason": "Both mention FP8, but one is an accuracy issue and the other is a save_pretrained/moe serialization bug; different failures." + "reason": "Different problems: generation kwargs/sync_gpus handling vs janus image-generation test failure." }, { - "left": "issue:43232", - "right": "issue:44792", + "left": "issue:45561", + "right": "issue:45600", "accept": false, - "reason": "Generation kwargs after sync_gpus is unrelated to a failed janus image-generation test case." + "reason": "Unrelated: xdist race on captured_info.txt vs broken auto_mappings references to removed configs." }, { "left": "issue:42175", "right": "issue:44933", "accept": false, - "reason": "Packaging/backends install behavior is unrelated to a missing image_utils import." + "reason": "Different code paths: missing TensorFlow backend in install extras vs nonexistent image_utils import." }, { - "left": "issue:43504", + "left": "issue:44496", "right": "issue:45561", "accept": false, - "reason": "A legacy config field load bug and a pytest-xdist file race are unrelated." + "reason": "Unrelated model loading/config error vs testing-utils file race." }, { - "left": "issue:45561", - "right": "issue:45600", + "left": "issue:45405", + "right": "issue:46877", "accept": false, - "reason": "A captured_info.txt race and stale auto_mappings references are different CI issues with different causes." + "reason": "Package version bump issue vs auto_docstring stdout noise; not the same bug." }, { "left": "issue:43976", "right": "issue:44933", "accept": false, - "reason": "Python-version compatibility and an import export bug are not the same code-path problem." - }, - { - "left": "issue:44496", - "right": "issue:45561", - "accept": false, - "reason": "Model config recognition failure is unrelated to a testing-utils file race." + "reason": "Python version compatibility problem vs image_utils import error." }, { "left": "issue:30333", "right": "issue:33290", "accept": false, - "reason": "MLflow reporting on failed jobs and a Deepspeed OOM are different training failures." - }, - { - "left": "issue:45405", - "right": "issue:46877", - "accept": false, - "reason": "A dependency version bump and import-time docstring logging are unrelated." + "reason": "MLflow reporting on training failure vs deepspeed adafactor OOM; different symptoms and fixes." }, { "left": "issue:30990", "right": "issue:35141", "accept": false, - "reason": "A loading hang and embedding reinitialization after resize are different bugs." + "reason": "Loading hang in Sentence Transformers vs embedding reinitialization after resize; not the same failure." }, { "left": "issue:42898", "right": "issue:44568", "accept": false, - "reason": "Both are tokenizer regressions, but one is clean_up_tokenization_spaces behavior and the other is add_special_tokens/BOS-EOS handling for a specific tokenizer." + "reason": "Both are tokenizer regressions in v5, but they are different behaviors and model-specific bugs." }, { "left": "issue:42175", "right": "issue:45712", "accept": false, - "reason": "Backend-install packaging and dummy class cleanup are unrelated." + "reason": "Install/backend dependency issue vs leftover dummy classes leaking into namespace/check_repo." }, { "left": "issue:43479", "right": "issue:46829", "accept": false, - "reason": "Multimodal config defaults and VideoPrism output-field wiring are different code paths." + "reason": "Multimodal config initialization bug vs classifier output field misassignment." }, { "left": "issue:46077", "right": "issue:46123", "accept": false, - "reason": "Attention-mask construction and pipeline batch-finalization logic are unrelated." + "reason": "Causal mask construction bug vs batch-finalization bug in MaskGenerationPipeline; unrelated." }, { "left": "issue:44623", "right": "issue:45310", "accept": false, - "reason": "Missing processor save files and Qwen3.5 from_pretrained errors are different load/save failures." + "reason": "Saving processor artifacts vs Qwen3.5 MoE from_pretrained load failure; different paths." }, { "left": "issue:42673", "right": "issue:46693", "accept": false, - "reason": "A thread-related VRAM leak in one model is not the same as a broader FA2 long-context performance regression." + "reason": "VRAM leak with multiple dataloader threads vs FA2 long-context performance regression; not the same defect." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Config initialization with None and Whisper timestamp decoding crash are unrelated." + "reason": "Different multimodal/model bugs: config defaults vs Whisper word-timestamp decode crash." }, { - "left": "issue:43723", - "right": "issue:44568", + "left": "issue:42898", + "right": "issue:45812", "accept": false, - "reason": "Generic tokenizer loading failure and a specific add_special_tokens regression are not the same bug." + "reason": "Both concern v5 tokenizer behavior, but one is a generic cleanup-space change and the other is a Granite-specific token-id regression." }, { - "left": "issue:42898", - "right": "issue:45812", + "left": "issue:43723", + "right": "issue:44568", "accept": false, - "reason": "Token cleanup behavior and wrong Granite token IDs are different tokenizer regressions." + "reason": "AutoTokenizer loading failure vs missing BOS/EOS insertion for one tokenizer model." }, { "left": "issue:44016", "right": "issue:46231", "accept": false, - "reason": "A notebook syntax error and an API doc parameter-name typo are unrelated documentation issues." + "reason": "Notebook syntax error vs docs parameter-name mismatch." }, { "left": "issue:44625", "right": "issue:46877", "accept": false, - "reason": "Qwen3.5 num_labels propagation and docstring lint output are unrelated." + "reason": "Qwen3.5 num_labels propagation bug vs docstring import-time stdout diagnostics." }, { - "left": "issue:42898", - "right": "issue:44779", + "left": "issue:39692", + "right": "issue:46909", "accept": false, - "reason": "These are both tokenizer-related v5 regressions, but they affect different models and behaviors; not the same concrete bug." + "reason": "SigLIP2 example errors vs DIA documentation example breakage; different docs and failures." }, { - "left": "issue:39692", - "right": "issue:46909", + "left": "issue:42898", + "right": "issue:44779", "accept": false, - "reason": "Broken documentation examples, but for different models/examples and different errors." + "reason": "Tokenizer behavior change in v5 vs Deepseek tokenizer producing incorrect results; related area but not the same bug." }, { "left": "issue:43479", "right": "issue:44079", "accept": false, - "reason": "A multimodal config defaulting bug is unrelated to ModelOutput key assignment." + "reason": "Multimodal config defaults vs ModelOutput key assignment logic; unrelated." }, { - "left": "issue:43874", - "right": "issue:45092", + "left": "issue:43784", + "right": "issue:43824", "accept": false, - "reason": "Missing image-processor method and remote-code meta-init incompatibility are different multimodal issues." + "reason": "Different import/loading issues: sentence-transformers NameError vs missing Qwen2_5_VL export." }, { - "left": "issue:43784", - "right": "issue:43824", + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Generation/KV-cache indexing bug vs auto_docstring crash with future annotations; unrelated." + }, + { + "left": "issue:43874", + "right": "issue:45092", "accept": false, - "reason": "An import-time NameError in sentence-transformers is unrelated to a missing Qwen2.5-VL import." + "reason": "Image processor API missing method vs old InternVL2 remote-code checkpoint/meta-init incompatibility." }, { "left": "issue:27957", "right": "issue:30333", "accept": false, - "reason": "Flash Attention support for XLM-R and MLflow reporting on failed jobs are unrelated." + "reason": "Flash Attention 2 model support vs MLflow job-reporting bug; no shared underlying change." }, { - "left": "issue:44734", - "right": "issue:45103", + "left": "issue:42898", + "right": "issue:43295", "accept": false, - "reason": "Serve KV-cache indexing crash and auto_docstring future-annotations crash are different subsystems and failures." + "reason": "Tokenizer cleanup behavior in v5 vs regression in custom code accessing processor.tokenizer/passing images; different breakages." }, { "left": "issue:30333", "right": "issue:34689", "accept": false, - "reason": "Training-job MLflow reporting and Llama 3.2 model loading are unrelated." + "reason": "MLflow reporting bug vs Llama 3.2 Vision model-loading regression; unrelated." }, { "left": "issue:42222", "right": "issue:45200", "accept": false, - "reason": "VitPose model breakage and Gemma 4 token-type defaults are different model bugs." + "reason": "Broken VitPose model files vs Gemma 4 multimodal token-type-id defaulting." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Generic v5 tokenizer cleanup change vs Kimi-K2.5 codec/regex regression; same subsystem but different concrete bug." } ] }, @@ -14649,7 +14735,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 48, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -14660,15 +14746,14 @@ "issue:22355", "issue:28282", "issue:30333", - "issue:35141", + "issue:36331", "issue:41628", "issue:42757", "issue:42796", "issue:42831", - "issue:42898", "issue:43116", "issue:43122", - "issue:43295", + "issue:43232", "issue:43479", "issue:43644", "issue:43723", @@ -14692,259 +14777,260 @@ "issue:45125", "issue:45245", "issue:45310", - "issue:45356", "issue:45405", "issue:45656", "issue:45712", + "issue:45834", "issue:46082", "issue:46132", "issue:46209", "issue:46231", + "issue:46424", "issue:46614", "issue:46710", "issue:46762", "issue:46932" ], "soft_pairs": [ - "issue:42898|issue:43295", - "issue:42898|issue:45356", "issue:44295|issue:45656", "issue:43116|issue:44190", - "issue:30333|issue:35141", - "issue:44222|issue:44297", + "issue:30333|issue:36331", "issue:45405|issue:45712", "issue:22355|issue:28282", + "issue:44222|issue:44297", "issue:44060|issue:45125", "issue:44075|issue:46231", + "issue:44448|issue:46132", "issue:44933|issue:45712", - "issue:44360|issue:46762", "issue:42831|issue:44623", - "issue:44297|issue:45656", - "issue:44222|issue:44623", + "issue:44360|issue:46762", "issue:42796|issue:43723", - "issue:44448|issue:46132", + "issue:44297|issue:45656", "issue:45245|issue:46132", + "issue:44222|issue:44623", "issue:43479|issue:43644", + "issue:46132|issue:46710", "issue:41628|issue:42757", "issue:44373|issue:44485", "issue:44297|issue:45310", - "issue:46132|issue:46710", "issue:45071|issue:46082", - "issue:44222|issue:44484", - "issue:46614|issue:46932", "issue:43723|issue:44977", + "issue:46614|issue:46932", + "issue:44222|issue:44484", "issue:43122|issue:46209", + "issue:44190|issue:46082", "issue:44263|issue:45092", - "issue:44190|issue:46082" + "issue:43232|issue:46424", + "issue:43232|issue:45834" ], - "prepared_review_unit_hash": "d9a5efa4f0eea3ddacc168749bbfb96afcde614ad0004d2a73da8ca98bdd3c10", + "prepared_review_unit_hash": "f6410a1fa95f9236a1de11a57a86ed1b41f7942735395b3bd9e91575a671e691", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 35896, - "estimated_input_tokens": 8974, - "estimated_eval_tokens": 18204 + "serialized_chars": 35760, + "estimated_input_tokens": 8940, + "estimated_eval_tokens": 18136 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is a mixed bag of unrelated Transformers issues (imports, tokenizers, model configs, FP8, examples, and training). None of the soft pairs look like the same underlying bug or change, so I rejected all of them.", + "summary": "This cluster is heterogeneous: most items are unrelated import, training, tokenizer, config, and serialization bugs. None of the soft pairs look like the same underlying change or defect, so all soft edges are rejected.", "confidence": 0.96, - "canonical_issue_reason": "Issue #43723 is the most representative, concrete user-facing regression in the set: a broad tokenizer-loading failure in v5 with clear API impact.", + "canonical_issue_reason": "No single canonical issue stands out; the issues span multiple unrelated subsystems and failure modes, so there is no clear duplicate representative.", "canonical_pr_reason": null, - "best_issue_reason": "#43723 is the best canonical issue if one must be chosen, because it describes a specific, high-impact regression with a clean failure mode and broad relevance.", + "best_issue_reason": "No issue is a good global representative because the cluster is not a duplicate set; the closest-looking items still differ in concrete code path and symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:42898", - "right": "issue:43295", - "accept": false, - "reason": "Both involve v5 regressions around tokenization/processor behavior, but they are different bugs: one is cleanup-space behavior, the other is processor.tokenizer/image passthrough." - }, - { - "left": "issue:42898", - "right": "issue:45356", - "accept": false, - "reason": "Different tokenizer regressions for different models and symptoms; no shared code-path bug is evident." - }, { "left": "issue:44295", "right": "issue:45656", "accept": false, - "reason": "Unrelated subsystems: position_ids buffer access vs optimizer stepping under DeepSpeed." + "reason": "Position_ids buffer read error is unrelated to DeepSpeed calling optimizer.step twice." }, { "left": "issue:43116", "right": "issue:44190", "accept": false, - "reason": "Both are example-script related, but they affect different scripts and different failures." + "reason": "Empty multi-label results in an example script is not the same bug as failing to load a local dataset." }, { "left": "issue:30333", - "right": "issue:35141", - "accept": false, - "reason": "MLFlow reporting on job failure is unrelated to embedding resizing and post_init reinitialization." - }, - { - "left": "issue:44222", - "right": "issue:44297", + "right": "issue:36331", "accept": false, - "reason": "Both mention save_pretrained, but one is FP8 MoE model serialization and the other is tokenizer config metadata; different code paths." + "reason": "MLflow reporting on failed training and CustomTrainer.compute_loss signature mismatch are different problems." }, { "left": "issue:45405", "right": "issue:45712", "accept": false, - "reason": "A dependency-version bump issue is not the same as leftover dummy classes leaking into the API surface." + "reason": "A PEFT version bump and leftover dummy classes failing repo checks are unrelated." }, { "left": "issue:22355", "right": "issue:28282", "accept": false, - "reason": "Distinct import errors from different causes: missing transformers.onnx module vs missing PyTorch dependency." + "reason": "Missing transformers.onnx and missing PyTorch for AutoModel are distinct import errors." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "Both involve save_pretrained, but one is FP8 MoE saving and the other is tokenizer_class metadata mismatch." }, { "left": "issue:44060", "right": "issue:45125", "accept": false, - "reason": "Both are Qwen3-related, but one is a tied-weights warning and the other is missing tensor-parallel plan support." + "reason": "Qwen3-Next tied-weights warning and missing _tp_plan for Qwen3_5MoE are different model bugs." }, { "left": "issue:44075", "right": "issue:46231", "accept": false, - "reason": "One is a runtime optimizer-argument bug; the other is a documentation parameter-name mismatch." + "reason": "Unused SGD args and a docstring parameter name mismatch are not the same defect." }, { - "left": "issue:44933", - "right": "issue:45712", + "left": "issue:44448", + "right": "issue:46132", "accept": false, - "reason": "Different API-surface cleanup issues with no common failing behavior." + "reason": "Pegasus v4/v5 output regression and AttentionInterface.register behavior change are separate code paths." }, { - "left": "issue:44360", - "right": "issue:46762", + "left": "issue:44933", + "right": "issue:45712", "accept": false, - "reason": "DSA indexer implementation details are unrelated to Minimax attention numerical differences." + "reason": "A missing import from image_utils is unrelated to dummy classes leaking into dir(transformers)." }, { "left": "issue:42831", "right": "issue:44623", "accept": false, - "reason": "FP8 accuracy behavior and processor.save_pretrained missing files are separate problems." - }, - { - "left": "issue:44297", - "right": "issue:45656", - "accept": false, - "reason": "Tokenizer save metadata bug is unrelated to DeepSpeed optimizer stepping." + "reason": "FineGrainedFP8 accuracy issues are not the same as processor.save_pretrained missing files." }, { - "left": "issue:44222", - "right": "issue:44623", + "left": "issue:44360", + "right": "issue:46762", "accept": false, - "reason": "Both touch save_pretrained, but one concerns model FP8/MoE save logic and the other processor file emission." + "reason": "A missing ReLU in the DSA indexer is not the same as Minimax MSA numerical drift." }, { "left": "issue:42796", "right": "issue:43723", "accept": false, - "reason": "Both involve tokenizers, but one is a specific BPE-marker handling bug and the other is a v5 tokenizer loading/import issue." + "reason": "Tokenizer BPE marker handling and AutoTokenizer.from_pretrained loading failures are different issues." }, { - "left": "issue:44448", - "right": "issue:46132", + "left": "issue:44297", + "right": "issue:45656", "accept": false, - "reason": "Different behaviors: output regression for a summarization model vs AttentionInterface.register changing registered functions." + "reason": "Tokenizer save_pretrained metadata mismatch is unrelated to DeepSpeed optimizer stepping." }, { "left": "issue:45245", "right": "issue:46132", "accept": false, - "reason": "Category-cardinality runtime error is unrelated to attention registration behavior." + "reason": "Category-count limits and AttentionInterface.register behavior are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "FP8 MoE save_pretrained and processor.save_pretrained missing files are different serialization bugs." }, { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Config None-handling for multimodal defaults is not the same as non-persistent buffers being filled with junk." + "reason": "Phi4MultimodalConfig None-handling and non-persistent buffer junk are unrelated initialization problems." + }, + { + "left": "issue:46132", + "right": "issue:46710", + "accept": false, + "reason": "AttentionInterface.register behavior change does not match the DeepSeek accuracy regression." }, { "left": "issue:41628", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but they concern different symbols and different dependency layers." + "reason": "Missing AutoImageProcessor from transformers and missing is_offline_mode from huggingface_hub are distinct import issues." }, { "left": "issue:44373", "right": "issue:44485", "accept": false, - "reason": "A wrong docstring is not the same as a RoPE implementation bug." + "reason": "A wrong docstring for position_ids is not the same as a GLM-5 RoPE implementation bug." }, { "left": "issue:44297", "right": "issue:45310", "accept": false, - "reason": "Tokenizer save-pretrained metadata mismatch is not the same as Qwen3.5 MoE from_pretrained failure." - }, - { - "left": "issue:46132", - "right": "issue:46710", - "accept": false, - "reason": "AttentionInterface.register semantics and DeepSeek accuracy regression are different issues." + "reason": "Tokenizer save_pretrained metadata mismatch and Qwen3.5 MoE from_pretrained error are unrelated." }, { "left": "issue:45071", "right": "issue:46082", "accept": false, - "reason": "Both are config-related, but one is type-checking breakage and the other is explicit head_dim validation in LlamaConfig." + "reason": "PretrainedConfig type checking breakage and LlamaConfig head_dim validation are different config bugs." }, { - "left": "issue:44263", - "right": "issue:45092", + "left": "issue:43723", + "right": "issue:44977", "accept": false, - "reason": "torch.split return-value handling in one model component is unrelated to remote-code checkpoint meta-init incompatibility." + "reason": "Tokenizer loading failure and Qwen3.5 flash-attention generation regression are not the same bug." }, { "left": "issue:46614", "right": "issue:46932", "accept": false, - "reason": "Stale phonemizer backend state and Gemma4 quantized embedder dtype casting are unrelated." + "reason": "Stale phonemizer backend state and Gemma4 quantized vision/audio dtype casting are unrelated." }, { - "left": "issue:43723", - "right": "issue:44977", + "left": "issue:44222", + "right": "issue:44484", "accept": false, - "reason": "Tokenizer loading failure and flash-attention generation regression are different code-path problems." + "reason": "Both mention save_pretrained, but one is FP8 MoE saving and the other is the max_shard_size default." }, { "left": "issue:43122", "right": "issue:46209", "accept": false, - "reason": "Tokenization drift across versions is unrelated to FP8 inference failure on B200." + "reason": "Tokenization changes across versions and FP8 inference failure on B200 are unrelated." + }, + { + "left": "issue:44190", + "right": "issue:46082", + "accept": false, + "reason": "Local dataset loading failure is not the same as LlamaConfig head_dim validation." }, { "left": "issue:44263", "right": "issue:45092", "accept": false, - "reason": "Different failures and different layers: indexer split semantics vs checkpoint loading/meta initialization." + "reason": "GlmMoeDsaIndexer torch.split output handling and remote-code meta init incompatibility are different issues." }, { - "left": "issue:44190", - "right": "issue:46082", + "left": "issue:43232", + "right": "issue:46424", + "accept": false, + "reason": "Generation kwargs update after sync_gpus and cache max length handling are unrelated cache/generation behaviors." + }, + { + "left": "issue:43232", + "right": "issue:45834", "accept": false, - "reason": "Local dataset loading in an example script is unrelated to LlamaConfig head_dim validation." + "reason": "Generation kwargs update and Kosmos2.5 OCR index errors are not the same bug." } ] }, @@ -14960,7 +15046,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 49, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -14987,6 +15073,7 @@ "issue:43644", "issue:43824", "issue:44038", + "issue:44188", "issue:44261", "issue:44263", "issue:44279", @@ -15003,8 +15090,8 @@ "issue:45357", "issue:45381", "issue:45406", + "issue:45593", "issue:45753", - "issue:45834", "issue:46002", "issue:46050", "issue:46209", @@ -15012,244 +15099,257 @@ "issue:46328", "issue:46369", "issue:46402", - "issue:46424", "issue:46431", "issue:46655", "issue:46909" ], "soft_pairs": [ - "issue:43232|issue:46424", - "issue:44295|issue:44623", - "issue:43232|issue:45834", - "issue:46209|issue:46431", "issue:43475|issue:46305", - "issue:44263|issue:44360", - "issue:44263|issue:44373", + "issue:46209|issue:46431", + "issue:44295|issue:44623", "issue:46305|issue:46655", "issue:30990|issue:38175", - "issue:42831|issue:45310", + "issue:44263|issue:44360", + "issue:44263|issue:44373", "issue:45092|issue:45357", - "issue:46369|issue:46909", "issue:42907|issue:46050", + "issue:42831|issue:45310", "issue:44279|issue:46328", + "issue:46369|issue:46909", "issue:45381|issue:46305", - "issue:44261|issue:45092", - "issue:22355|issue:27957", "issue:44483|issue:45406", - "issue:42831|issue:44038", - "issue:44829|issue:46002", + "issue:22355|issue:27957", "issue:43644|issue:46305", + "issue:44261|issue:45092", + "issue:44829|issue:46002", + "issue:42831|issue:44038", "issue:41628|issue:43824", "issue:43408|issue:46402", "issue:42831|issue:45753", "issue:22355|issue:28218", "issue:39401|issue:44568", "issue:42886|issue:43066", + "issue:27957|issue:30990", + "issue:44188|issue:44373", "issue:42886|issue:42994", - "issue:42757|issue:45278" + "issue:42757|issue:45278", + "issue:43408|issue:46369", + "issue:43232|issue:45593" ], - "prepared_review_unit_hash": "0385128b49302f39b059149e5801edc882cc60e81d66a7d640ed625d96b17f66", + "prepared_review_unit_hash": "40b4395056d22cefa8ec528b1e92fd03f3731fb13b83be3a6a56ec9c579d887c", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 35863, - "estimated_input_tokens": 8966, - "estimated_eval_tokens": 18188 + "soft_pair_count": 31, + "serialized_chars": 36516, + "estimated_input_tokens": 9129, + "estimated_eval_tokens": 18514 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of distinct bug reports across unrelated areas (imports, tokenizers, multimodal processors, quantization, generation, and model-specific regressions). None of the soft pairs look like the same underlying bug/change closely enough to merge.", - "confidence": 0.81, - "canonical_issue_reason": "issue:42886 is the most concrete, narrowly scoped, and reproducible report: it names the exact failure mode, environment flag, and affected version, making it a strong canonical representative for a bug cluster.", + "summary": "No soft pair looks like the same underlying bug/change closely enough to merge. The set is mostly unrelated issues within Transformers, often sharing only broad themes like tokenizers, imports, model loading, or serialization.", + "confidence": 0.94, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:42886 is the best standalone issue because it is specific, actionable, and tied to a clear regression with a precise repro condition.", + "best_issue_reason": "issue:46305 is the most specific, concrete, and actionable bug report in the set (a precise off-by-one in sliding_window_mask_function with clear expected vs actual behavior).", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43232", - "right": "issue:46424", + "left": "issue:43475", + "right": "issue:46305", "accept": false, - "reason": "Both touch generation/cache behavior, but they describe different failure modes and code paths: sync_gpus kwargs handling vs max_cache_length being ignored." + "reason": "Different SAM3 memory/attribute failure vs MoonshineStreaming mask off-by-one; unrelated code paths." }, { - "left": "issue:44295", - "right": "issue:44623", + "left": "issue:46209", + "right": "issue:46431", "accept": false, - "reason": "A buffer/position_ids read error is unrelated to processor.save_pretrained missing files." + "reason": "FP8 inference on B200 vs missing chat template error; no shared bug mechanism." }, { - "left": "issue:43232", - "right": "issue:45834", + "left": "issue:44295", + "right": "issue:44623", "accept": false, - "reason": "Generation kwargs handling and Kosmos2.5 OCR index errors are unrelated bugs." + "reason": "Buffer registration/readback issue vs processor save_pretrained file omission; both serialization-adjacent but not the same defect." }, { - "left": "issue:46209", - "right": "issue:46431", + "left": "issue:46305", + "right": "issue:46655", "accept": false, - "reason": "FP8 inference on B200 and missing chat template in a processor are different problems in different subsystems." + "reason": "Different streaming/video processor logic and different off-by-one/padding behavior." }, { - "left": "issue:43475", - "right": "issue:46305", + "left": "issue:30990", + "right": "issue:38175", "accept": false, - "reason": "Sam3 vision encoder missing attribute and Moonshine streaming mask off-by-one are distinct model-specific bugs." + "reason": "Loading stall vs zero probabilities; different runtime symptoms and model paths." }, { "left": "issue:44263", "right": "issue:44360", "accept": false, - "reason": "Both mention DSA/indexer code, but one is about torch.split return values and the other about a missing ReLU; not the same defect." + "reason": "DSA indexer split behavior vs lack of ReLU in a different indexer; not the same bug." }, { "left": "issue:44263", "right": "issue:44373", "accept": false, - "reason": "A DSA indexer tensor-shape issue is not the same as a wrong docstring for position_ids." + "reason": "Indexer torch.split values vs docstring typo for position_ids; one is functional, one documentation-related." }, { - "left": "issue:46305", - "right": "issue:46655", + "left": "issue:45092", + "right": "issue:45357", "accept": false, - "reason": "Both are video-related, but one is a streaming attention mask off-by-one and the other is padding calculation in a video processor." + "reason": "Meta initialization incompatibility vs incorrect visual encoder keys on save_pretrained; distinct save/load issues." }, { - "left": "issue:30990", - "right": "issue:38175", + "left": "issue:42907", + "right": "issue:46050", "accept": false, - "reason": "Sentence Transformers loading and SigLIP2 zero probabilities are unrelated model/runtime issues." + "reason": "Saving dequantized Ministral/Devstral models vs sinq quantized model round-trip breakage; different quantization pipelines." }, { "left": "issue:42831", "right": "issue:45310", "accept": false, - "reason": "FineGrainedFP8 accuracy regression and Qwen3.5 MoE from_pretrained failure are different concrete bugs." + "reason": "FineGrainedFP8 accuracy regression vs Qwen3.5 from_pretrained error; different failure modes." }, { - "left": "issue:45092", - "right": "issue:45357", + "left": "issue:44279", + "right": "issue:46328", "accept": false, - "reason": "Meta-initialization incompatibility for old InternVL2 checkpoints is different from incorrect visual encoder keys being saved." + "reason": "Generic dependency issue vs pad performance inefficiency; unrelated." }, { "left": "issue:46369", "right": "issue:46909", "accept": false, - "reason": "Sam3-litetext memory overflow is unrelated to a broken documentation example for DIA." - }, - { - "left": "issue:42907", - "right": "issue:46050", - "accept": false, - "reason": "Both involve quantized models and saving/loading, but they are different quantization paths and failure modes." - }, - { - "left": "issue:44279", - "right": "issue:46328", - "accept": false, - "reason": "A generic dependency issue is not the same as a SequenceFeatureExtractor.pad performance bug." + "reason": "SAM3-LiteText memory overflow vs documentation example bug; unrelated." }, { "left": "issue:45381", "right": "issue:46305", "accept": false, - "reason": "Wrong qwen2.5-vl vision_position_ids and Moonshine sliding-window masking are unrelated model-specific logic errors." + "reason": "Qwen2.5-VL video position_ids bug vs MoonshineStreaming mask off-by-one; different models and code paths." }, { - "left": "issue:44261", - "right": "issue:45092", + "left": "issue:44483", + "right": "issue:45406", "accept": false, - "reason": "Missing rms_norm_eps in MLA q_a_layernorm is unrelated to InternVL2 meta-init incompatibility." + "reason": "Chat-completions request rejection vs Gemma4Processor missing _tokenizer; unrelated API failures." }, { "left": "issue:22355", "right": "issue:27957", "accept": false, - "reason": "An import failure for transformers.onnx and XLMRoberta Flash Attention 2 are not the same bug." + "reason": "onnx import failure vs XLMRoberta Flash Attention 2 behavior; unrelated." }, { - "left": "issue:44483", - "right": "issue:45406", + "left": "issue:43644", + "right": "issue:46305", "accept": false, - "reason": "Chat/completions request rejection and Gemma4Processor missing _tokenizer are different API failures." + "reason": "Non-persistent buffer junk in v5.0 vs streaming mask off-by-one; different subsystems." }, { - "left": "issue:42831", - "right": "issue:44038", + "left": "issue:44261", + "right": "issue:45092", "accept": false, - "reason": "An FP8 accuracy issue and a Qwen3-VL-Moe from_pretrained issue do not share the same concrete code-path problem." + "reason": "MLA layernorm epsilon precision issue vs remote-code/meta-init incompatibility; not the same defect." }, { "left": "issue:44829", "right": "issue:46002", "accept": false, - "reason": "Degenerate training with flash_attention_3 and NaNs at checkout/init are different symptoms and likely different causes." + "reason": "Flash Attention 3 degenerate training vs NaNs after checkout+init; different training-path problems." }, { - "left": "issue:43644", - "right": "issue:46305", + "left": "issue:42831", + "right": "issue:44038", "accept": false, - "reason": "Non-persistent buffer junk in v5.0.0 is unrelated to Moonshine streaming window off-by-one." + "reason": "FP8 accuracy issue vs Qwen3-VL-Moe from_pretrained error; not the same bug." }, { "left": "issue:41628", "right": "issue:43824", "accept": false, - "reason": "Both are import errors, but they affect different symbols and likely different lazy-import/export paths; too broad to merge." + "reason": "Two different import errors for different symbols; same broad area but not one underlying fix." }, { "left": "issue:43408", "right": "issue:46402", "accept": false, - "reason": "A sam3_video vs sam3_tracker type warning is not the same issue as Sam3Model text encoder weights not loading." + "reason": "sam3_tracker class mismatch warning vs sam3 model text encoder weights not loaded; different SAM3 issues." }, { "left": "issue:42831", "right": "issue:45753", "accept": false, - "reason": "FineGrainedFP8 accuracy regression and Qwen3_5 infinite-loop-on-image are different bugs." + "reason": "FP8 accuracy regression vs infinite loop on a specific image; unrelated symptoms and causes." }, { "left": "issue:22355", "right": "issue:28218", "accept": false, - "reason": "A missing module import and tokenizer added-space behavior are unrelated." + "reason": "onnx import failure vs tokenizer added-space behavior; unrelated tokenizer/import topics." }, { "left": "issue:39401", "right": "issue:44568", "accept": false, - "reason": "Wrong offset_mapping and missing BOS/EOS on add_special_tokens are both tokenizer bugs, but they are different tokenization behaviors." + "reason": "offset_mapping bug vs add_special_tokens not adding BOS/EOS; both tokenizer issues but different mechanics." }, { "left": "issue:42886", "right": "issue:43066", "accept": false, - "reason": "Offline cache-loading failure and wrong tokenizer decoder type are different tokenizer regressions." + "reason": "Offline cache load failure vs wrong tokenizer decoder type in v5; different tokenizer regressions." + }, + { + "left": "issue:27957", + "right": "issue:30990", + "accept": false, + "reason": "Flash Attention 2 model support vs Sentence Transformers loading stall; unrelated." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "torch.compile attention-kernel divergence vs docstring error; not the same kind of bug." }, { "left": "issue:42886", "right": "issue:42994", "accept": false, - "reason": "Tokenizer offline cache loading and quantized model saving failure are unrelated." + "reason": "Offline tokenizer cache loading vs quantized model saving failure; unrelated save/load paths." }, { "left": "issue:42757", "right": "issue:45278", "accept": false, - "reason": "An import from huggingface_hub missing is_offline_mode and broad import errors after upgrade are related only at a very high level, not the same concrete bug." + "reason": "Single missing huggingface_hub symbol vs many import errors after upgrade; same ecosystem, not same underlying issue." + }, + { + "left": "issue:43408", + "right": "issue:46369", + "accept": false, + "reason": "SAM3 model-type warning vs SAM3-LiteText memory overflow; different failures." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation kwargs update after sync_gpus vs D-FINE auxiliary loss logic; unrelated." } ] }, @@ -15265,7 +15365,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 50, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -15273,7 +15373,8 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-502", "nodes": [ - "issue:27957", + "issue:22355", + "issue:29159", "issue:30990", "issue:33357", "issue:39401", @@ -15282,7 +15383,6 @@ "issue:42796", "issue:42907", "issue:43122", - "issue:43232", "issue:43408", "issue:43452", "issue:43475", @@ -15296,12 +15396,11 @@ "issue:44038", "issue:44060", "issue:44117", - "issue:44188", + "issue:44222", "issue:44263", "issue:44295", "issue:44297", "issue:44360", - "issue:44373", "issue:44485", "issue:44488", "issue:44521", @@ -15310,266 +15409,273 @@ "issue:44829", "issue:44945", "issue:44964", + "issue:45216", "issue:45278", "issue:45290", "issue:45381", - "issue:45593", + "issue:45478", "issue:45600", "issue:46002", "issue:46172", "issue:46326", - "issue:46369", + "issue:46614", "issue:46762", "issue:46909" ], "soft_pairs": [ - "issue:44188|issue:44373", - "issue:27957|issue:30990", - "issue:43408|issue:46369", - "issue:43232|issue:45593", "issue:42796|issue:43122", "issue:43646|issue:46002", "issue:45381|issue:46909", "issue:43540|issue:46909", - "issue:42907|issue:44945", - "issue:44295|issue:44297", "issue:30990|issue:33357", - "issue:43576|issue:43784", - "issue:42371|issue:44263", + "issue:42907|issue:44945", "issue:42907|issue:44964", "issue:39401|issue:44488", + "issue:44295|issue:44297", "issue:43976|issue:44945", - "issue:46172|issue:46762", + "issue:43576|issue:43784", + "issue:42371|issue:44263", "issue:43756|issue:44060", - "issue:41628|issue:45278", + "issue:46172|issue:46762", "issue:42907|issue:43937", - "issue:44117|issue:44704", + "issue:41628|issue:45278", "issue:43784|issue:45600", + "issue:44117|issue:44704", "issue:39401|issue:44521", - "issue:44704|issue:45290", "issue:44625|issue:44829", - "issue:44360|issue:44485", + "issue:44704|issue:45290", "issue:43408|issue:43475", "issue:43122|issue:44521", - "issue:44038|issue:44297", + "issue:44360|issue:44485", "issue:43122|issue:46326", "issue:42907|issue:43452", - "issue:44038|issue:44295" + "issue:44038|issue:44297", + "issue:22355|issue:29159", + "issue:44038|issue:44295", + "issue:45216|issue:45478", + "issue:39401|issue:46614", + "issue:44038|issue:44222" ], - "prepared_review_unit_hash": "04d23c134f727c8417955c92283ce5bd8a43855d5c6523017eb0108420f84556", + "prepared_review_unit_hash": "7fdd2f55d0218527b06b30d09e858ff09e2c0530c2e3d29cb331a2a1d5582c0d", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 36775, - "estimated_input_tokens": 9194, - "estimated_eval_tokens": 18644 + "serialized_chars": 36833, + "estimated_input_tokens": 9209, + "estimated_eval_tokens": 18674 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pair looks like a true duplicate or the same fix. Most pairs are only loosely related by subsystem or version regression theme, but they target different concrete bugs. The broadest issue here is the v5 import/regression report.", - "confidence": 0.9, - "canonical_issue_reason": "Issue 45278 is the broadest, most umbrella-like regression report in the set: it summarizes widespread import breakage after upgrading to 5.5.0, making it the most plausible canonical issue among these mostly specific bug reports.", + "summary": "This cluster is heterogeneous: it mixes unrelated tokenizer, model-loading, multimodal, and API/regression bugs. The soft pairs mostly share only broad Transformers similarity, not the same concrete underlying defect, so none should be merged as duplicates.", + "confidence": 0.34, + "canonical_issue_reason": "issue:45278 is the broadest regression report in the set and aligns with several import/breakage complaints, making it the least specific anchor.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45278 is the best general representative because it describes a wide upgrade regression rather than a narrow model-specific or API-specific failure.", + "best_issue_reason": "Best representative of the cluster\u2019s shared theme of Transformers v5 breakages; broad user impact and generic wording make it the most usable single issue anchor.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44188", - "right": "issue:44373", - "accept": false, - "reason": "Attention-kernel branching under torch.compile and a wrong docstring for position_ids are unrelated." - }, - { - "left": "issue:27957", - "right": "issue:30990", - "accept": false, - "reason": "Both involve loading/model usage, but one is an XLM-RoBERTa Flash Attention 2 issue and the other is Sentence Transformers hanging; different bugs." - }, - { - "left": "issue:43408", - "right": "issue:46369", - "accept": false, - "reason": "Both mention SAM3, but one is an incorrect model-type warning and the other is a memory overflow in a different model path." - }, - { - "left": "issue:43232", - "right": "issue:45593", - "accept": false, - "reason": "Generation kwargs handling after sync_gpus is unrelated to D-FINE auxiliary-loss behavior." - }, { "left": "issue:42796", "right": "issue:43122", "accept": false, - "reason": "Both are tokenizer-related, but one is a specific Ministral BPE-marker bug and the other is a version-wide tokenization regression." + "reason": "Both are tokenizer-related, but one is a BPE-marker handling bug and the other is an offset_mapping regression; different failure modes." }, { "left": "issue:43646", "right": "issue:46002", "accept": false, - "reason": "Custom model initialization breakage is not the same underlying bug as NaNs in classification heads after checkout/init." + "reason": "Custom model initialization breakage and NaNs in classification heads are unrelated bugs." }, { "left": "issue:45381", "right": "issue:46909", "accept": false, - "reason": "Qwen2.5-VL video position IDs and a broken DIA docs example are different issues." + "reason": "Video position-id regression vs a broken documentation example; not the same issue." }, { "left": "issue:43540", "right": "issue:46909", "accept": false, - "reason": "Qwen3OmniMoe video-processing errors are unrelated to a documentation example failure." + "reason": "Qwen3Omni video-input ValueError and a docs-example issue are unrelated." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Loading hang for Sentence Transformers is not the same as a MacOS bus error on a CLIP model." }, { "left": "issue:42907", "right": "issue:44945", "accept": false, - "reason": "Dequantized model saving and incorrect output under pipeline parallelism are different code paths." + "reason": "Dequantized save failure and incorrect pipeline-parallel LLM output are different code paths." }, { - "left": "issue:44295", - "right": "issue:44297", + "left": "issue:42907", + "right": "issue:44964", "accept": false, - "reason": "Reading position_ids after buffer registration is unrelated to tokenizer_class mismatches on save_pretrained." + "reason": "Saving dequantized Ministrals/Devstrals and loading Phi-4 multimodal are separate failures." }, { - "left": "issue:30990", - "right": "issue:33357", + "left": "issue:39401", + "right": "issue:44488", "accept": false, - "reason": "Sentence Transformers hanging on load and a MacOS bus error with a CLIP model are distinct failures." + "reason": "Qwen3 tokenizer offset mapping and a model-load failure for sleng-bert are unrelated." }, { - "left": "issue:43576", - "right": "issue:43784", + "left": "issue:44295", + "right": "issue:44297", "accept": false, - "reason": "A broken env CLI command and a NameError during sentence-transformers import are not the same bug." + "reason": "Buffer access on position_ids and tokenizer_class mismatch on save_pretrained are different bugs." }, { - "left": "issue:42371", - "right": "issue:44263", + "left": "issue:43976", + "right": "issue:44945", "accept": false, - "reason": "TF32 API guidance and a torch.split return-value issue in GlmMoeDsaIndexer are unrelated." + "reason": "Python version compatibility and pipeline parallelism output corruption are not the same defect." }, { - "left": "issue:42907", - "right": "issue:44964", + "left": "issue:43576", + "right": "issue:43784", "accept": false, - "reason": "Saving dequantized Ministral/Devstral models is unrelated to loading Phi-4 multimodal instruct." + "reason": "Broken env command and a NameError during sentence-transformers import are distinct import-time issues." }, { - "left": "issue:39401", - "right": "issue:44488", + "left": "issue:42371", + "right": "issue:44263", "accept": false, - "reason": "Wrong offset_mapping in Qwen3 tokenizer and inability to load a specific model are different issues." + "reason": "TF32 settings guidance and a torch.split return-value bug have no concrete overlap." }, { - "left": "issue:43976", - "right": "issue:44945", + "left": "issue:43756", + "right": "issue:44060", "accept": false, - "reason": "Python version compatibility and pipeline-parallelism output corruption are unrelated." + "reason": "Smollm3 RoPE-layer mismatch and Qwen3-Next tied-weights warning are unrelated model-specific problems." }, { "left": "issue:46172", "right": "issue:46762", "accept": false, - "reason": "Both are numeric/attention correctness issues, but they concern different models and different kernels." + "reason": "DeepSeek-v4 transpose logic and Minimax MSA numerical differences are different implementations and symptoms." }, { - "left": "issue:43756", - "right": "issue:44060", + "left": "issue:42907", + "right": "issue:43937", "accept": false, - "reason": "Smollm3 RoPE-layer placement and Qwen3-Next tied-weights warnings are separate model-definition bugs." + "reason": "Saving dequantized models and invalid GenerationConfig handling are separate issues." }, { "left": "issue:41628", "right": "issue:45278", "accept": false, - "reason": "Both are import-related regressions, but 41628 is a specific AutoImageProcessor export failure while 45278 is a broader collection of import errors after upgrade." + "reason": "Both involve imports, but one is a specific AutoImageProcessor import error and the other is a broad post-upgrade import regression." }, { - "left": "issue:42907", - "right": "issue:43937", + "left": "issue:43784", + "right": "issue:45600", "accept": false, - "reason": "Model-saving failures and invalid GenerationConfig errors are different problems." + "reason": "Sentence-transformers import NameError and broken auto_mappings references are unrelated import problems." }, { "left": "issue:44117", "right": "issue:44704", "accept": false, - "reason": "Tokenizer mapping returning None and AutoProcessor not forwarding kwargs to cached_file are distinct API bugs." + "reason": "Tokenizer mapping None handling and AutoProcessor kwarg forwarding are different loader bugs." }, { - "left": "issue:43784", - "right": "issue:45600", + "left": "issue:39401", + "right": "issue:44521", "accept": false, - "reason": "Importing sentence-transformers with latest transformers and stale auto_mappings references to removed Sam3LiteText configs are not the same bug." + "reason": "Tokenizer offset_mapping and multimodal assistant_masks are not the same underlying bug." }, { - "left": "issue:39401", - "right": "issue:44521", + "left": "issue:44625", + "right": "issue:44829", "accept": false, - "reason": "Qwen3 tokenizer offset mapping and multimodal assistant_masks in apply_chat_template are different tokenization issues." + "reason": "Qwen3.5 num_labels propagation and FA3-induced degenerate training are different failures." }, { "left": "issue:44704", "right": "issue:45290", "accept": false, - "reason": "AutoProcessor kwarg forwarding and apply_chat_template crashing on tool-call assistant messages are unrelated code paths." + "reason": "cached_file kwarg forwarding and chat-template crashing on tool-call messages are unrelated." }, { - "left": "issue:44625", - "right": "issue:44829", + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Model-type warning and missing fpn_position_embeddings are separate SAM 3 issues." + }, + { + "left": "issue:43122", + "right": "issue:44521", "accept": false, - "reason": "num_labels propagation in Qwen3.5 config and degenerate training with flash_attention_3 are different failures." + "reason": "Tokenization differences across versions and multimodal assistant_masks zeroing are different bugs." }, { "left": "issue:44360", "right": "issue:44485", "accept": false, - "reason": "A missing ReLU in the DSA indexer and GLM-5 RoPE implementation concerns are unrelated." + "reason": "DSA indexer missing ReLU and GLM-5 RoPE implementation are not the same defect." }, { - "left": "issue:44038", - "right": "issue:44297", + "left": "issue:43122", + "right": "issue:46326", "accept": false, - "reason": "Qwen3-VL-Moe breakage and tokenizer_class mismatch on save are not the same issue." + "reason": "General tokenization changes and mistral_common continue_final_message behavior are different code paths." }, { - "left": "issue:43122", - "right": "issue:44521", + "left": "issue:42907", + "right": "issue:43452", "accept": false, - "reason": "A version-wide tokenization difference and all-zero multimodal assistant masks are related only at a high level, not as the same concrete bug." + "reason": "Dequantized save failure and gguf_file loading breakage are unrelated model I/O bugs." }, { - "left": "issue:43122", - "right": "issue:46326", + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe model bug and tokenizer_class mismatch on save_pretrained are different issues." + }, + { + "left": "issue:22355", + "right": "issue:29159", "accept": false, - "reason": "General tokenization differences across versions and continue_final_message handling for mistral_common tokenizers are different defects." + "reason": "Missing transformers.onnx module and slow/fast tokenizer inconsistency are unrelated." }, { "left": "issue:44038", "right": "issue:44295", "accept": false, - "reason": "Qwen3-VL-Moe issues do not match the position_ids buffer-reading bug." + "reason": "Qwen3-VL-Moe behavior and position_ids buffer access are different bugs." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Both touch Qwen3.5 save/load flows, but one is a save_pretrained regression and the other is a from_pretrained error; not the same concrete bug." + }, + { + "left": "issue:39401", + "right": "issue:46614", + "accept": false, + "reason": "Qwen3 tokenizer offset mapping and stale phonemizer language state are unrelated." }, { "left": "issue:44038", - "right": "issue:44295", + "right": "issue:44222", "accept": false, - "reason": "Qwen3-VL-Moe issues do not match the position_ids buffer-reading bug." + "reason": "Qwen3-VL-Moe loading/processing and FP8 save_pretrained for moe are different failure modes." } ] }, @@ -15585,7 +15691,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 51, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -15595,10 +15701,8 @@ "nodes": [ "issue:22355", "issue:29127", - "issue:29159", "issue:36296", "issue:38617", - "issue:39401", "issue:42175", "issue:42371", "issue:42673", @@ -15610,25 +15714,27 @@ "issue:43278", "issue:43493", "issue:43611", + "issue:43618", "issue:43824", + "issue:43825", "issue:43874", "issue:43937", "issue:43976", "issue:44038", + "issue:44075", "issue:44188", - "issue:44222", "issue:44351", "issue:44355", "issue:44485", "issue:44610", "issue:44625", "issue:44779", + "issue:44936", "issue:44938", "issue:44945", "issue:45072", "issue:45083", "issue:45200", - "issue:45216", "issue:45278", "issue:45357", "issue:45446", @@ -15638,249 +15744,243 @@ "issue:46231", "issue:46328", "issue:46614", + "issue:46829", "issue:46877", "issue:46909" ], "soft_pairs": [ - "issue:45216|issue:45478", - "issue:22355|issue:29159", - "issue:44038|issue:44222", - "issue:39401|issue:46614", "issue:43824|issue:44351", "issue:43116|issue:45200", "issue:43122|issue:44779", - "issue:42757|issue:44351", - "issue:36296|issue:43116", "issue:43278|issue:44945", - "issue:42175|issue:44938", + "issue:36296|issue:43116", + "issue:42757|issue:44351", "issue:22355|issue:29127", + "issue:42175|issue:44938", "issue:43493|issue:45753", - "issue:46231|issue:46877", "issue:44188|issue:46328", + "issue:46231|issue:46877", "issue:43937|issue:44485", - "issue:38617|issue:45278", "issue:42898|issue:46614", + "issue:38617|issue:45278", "issue:43066|issue:44779", - "issue:44038|issue:45478", "issue:43611|issue:46002", "issue:43976|issue:45446", - "issue:44355|issue:46231", - "issue:42673|issue:45072", + "issue:44038|issue:45478", "issue:43116|issue:43493", + "issue:42673|issue:45072", "issue:45357|issue:45478", + "issue:44355|issue:46231", "issue:44625|issue:46002", "issue:42371|issue:43874", "issue:45083|issue:45446", - "issue:44610|issue:46909" + "issue:44610|issue:46909", + "issue:44075|issue:44936", + "issue:43618|issue:46829", + "issue:43116|issue:43825" ], - "prepared_review_unit_hash": "4a3f4943041aa5cb33a101380fb3df339e2a4872f47ffd6e051a299cb49f98ff", + "prepared_review_unit_hash": "ef23e0b9ed7bcba275751fe665d3856a538fd429956baad05054371a3b7889e9", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 30, - "serialized_chars": 35711, - "estimated_input_tokens": 8928, - "estimated_eval_tokens": 18112 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 35979, + "estimated_input_tokens": 8995, + "estimated_eval_tokens": 18246 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly noise: the soft pairs share broad keywords like tokenizers, imports, or Qwen/GLM model names, but they describe different concrete bugs and change requests. No soft edge looks like a true duplicate pair.", - "confidence": 0.96, - "canonical_issue_reason": "issue:45478 is the most central node in the only somewhat coherent mini-group (Qwen3.5/Qwen3.5-Moe save/load regressions), so it is the best representative issue in this noisy cluster.", + "summary": "This set is mostly a loose similarity cluster around Transformers v5 regressions, import/API breakages, tokenizer/model-specific bugs, and assorted unrelated issues. I did not find any soft pair that clearly describes the same underlying bug/change.", + "confidence": 0.83, + "canonical_issue_reason": "issue:45278 is the broadest representative of the cluster because it explicitly summarizes widespread import breakage after upgrading to 5.5.0; most other items are narrower, model-specific, or unrelated.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45478 is the strongest representative of the densest related subgroup, but the overall cluster is too heterogeneous for it to serve as a true canonical for all items.", + "best_issue_reason": "issue:45278 is the best single issue to anchor the cluster since it captures the general upgrade/regression theme better than the more specific one-off bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45216", - "right": "issue:45478", - "accept": false, - "reason": "Related Qwen3.5 save/load regressions, but one is incorrect checkpoint saving and the other is a from_pretrained load error; not the same bug." - }, - { - "left": "issue:22355", - "right": "issue:29159", - "accept": false, - "reason": "Different subsystems: missing transformers.onnx module vs tokenizer slow/fast inconsistency." - }, - { - "left": "issue:44038", - "right": "issue:44222", - "accept": false, - "reason": "Both mention MoE/Qwen-related behavior, but one is a model bug in transformers 5.0 and the other is FP8 save_pretrained behavior; not the same change." - }, - { - "left": "issue:39401", - "right": "issue:46614", - "accept": false, - "reason": "Different tokenizers and failure modes: Qwen3 offset mapping vs Wav2Vec2 phonemizer language state." - }, { "left": "issue:43824", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but they concern different missing symbols in different APIs/packages." + "reason": "Both are import errors in Transformers, but they involve different missing symbols (`Qwen2_5_VLForConditionalGeneration` vs `HybridCache`) and likely different breakages." }, { "left": "issue:43116", "right": "issue:45200", "accept": false, - "reason": "Different functionality: example script multi-label outputs vs Gemma 4 multimodal token type ids in fine-tuning." + "reason": "One is a multi-label example-script bug; the other is a Gemma 4 multimodal config default issue. Different code paths and symptoms." }, { "left": "issue:43122", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different tokenization outputs." + "reason": "Both are tokenizer regressions in v5, but for different models and outputs. Too broad to treat as the same bug." }, { - "left": "issue:42757", - "right": "issue:44351", + "left": "issue:43278", + "right": "issue:44945", "accept": false, - "reason": "Different missing imports from different packages; no shared underlying bug is evident." + "reason": "Different failures: dtype mismatch between train/eval versus incorrect outputs under pipeline parallelism." }, { "left": "issue:36296", "right": "issue:43116", "accept": false, - "reason": "Tensor parallel training bug and example-script classification output issue are unrelated." - }, - { - "left": "issue:43278", - "right": "issue:44945", - "accept": false, - "reason": "BF16/FP32 dtype drift in evaluation is unrelated to incorrect output under pipeline parallelism." + "reason": "Tensor parallel training and a classification example script issue are unrelated." }, { - "left": "issue:42175", - "right": "issue:44938", + "left": "issue:42757", + "right": "issue:44351", "accept": false, - "reason": "Packaging/backend dependency issue vs Python 3.14 load failure; different root causes." + "reason": "Both are import-related, but they fail on different missing exports from different dependency surfaces; not the same underlying defect." }, { "left": "issue:22355", "right": "issue:29127", "accept": false, - "reason": "Missing onnx submodule import and LayoutLMv3 error-message clarity are unrelated." + "reason": "A missing module import and an error-message clarity issue for LayoutLMv3 are unrelated." }, { - "left": "issue:43493", - "right": "issue:45753", + "left": "issue:42175", + "right": "issue:44938", "accept": false, - "reason": "Different models and symptoms: SigLIP2 HF/JAX mismatch vs Qwen3_5 infinite loop on one image." + "reason": "Different environment/version problems: backend extras not including TensorFlow versus Python 3.14 load failure." }, { - "left": "issue:46231", - "right": "issue:46877", + "left": "issue:43493", + "right": "issue:45753", "accept": false, - "reason": "Docs parameter-name mismatch vs import-time docstring diagnostics; not the same issue." + "reason": "Both involve Qwen models, but one is a SigLIP2 implementation discrepancy and the other is an infinite-loop image-specific generation bug." }, { "left": "issue:44188", "right": "issue:46328", "accept": false, - "reason": "Attention-kernel divergence under torch.compile and numpy padding inefficiency are unrelated." + "reason": "One is a torch.compile attention-kernel divergence; the other is a performance issue in feature padding. Not the same bug." }, { - "left": "issue:43937", - "right": "issue:44485", + "left": "issue:46231", + "right": "issue:46877", "accept": false, - "reason": "Both are GLM-5 related, but one is GenerationConfig validation and the other is RoPE implementation; distinct bugs." + "reason": "Documentation parameter mismatch versus import-time doc-lint stdout noise; both are doc-related, but not the same change or defect." }, { - "left": "issue:38617", - "right": "issue:45278", + "left": "issue:43937", + "right": "issue:44485", "accept": false, - "reason": "A specific missing import and a broad post-upgrade import breakage report are not clearly the same underlying defect." + "reason": "Both concern GLM-5, but one is invalid GenerationConfig handling and the other is RoPE implementation discussion. Different concrete issues." }, { "left": "issue:42898", "right": "issue:46614", "accept": false, - "reason": "Tokenization-space behavior change and stale phonemizer backend state are different bugs." + "reason": "Tokenizer cleanup behavior in v5 is unrelated to stale phonemizer language state in Wav2Vec2 phonemization." }, { - "left": "issue:43066", - "right": "issue:44779", + "left": "issue:38617", + "right": "issue:45278", "accept": false, - "reason": "Both are v5 tokenizer regressions, but they target different models and different failure modes." + "reason": "The latter is a broad import-error umbrella, but these are not clearly the same defect; one specific missing import is not enough to merge them." }, { - "left": "issue:44038", - "right": "issue:45478", + "left": "issue:43066", + "right": "issue:44779", "accept": false, - "reason": "Same model family, but one is a Qwen3-VL-Moe bug and the other is a Qwen3.5-Moe loading error; not mergeable as one fix." + "reason": "Both are tokenizer regressions in v5, but they describe different model/tokenization problems and likely different root causes." }, { "left": "issue:43611", "right": "issue:46002", "accept": false, - "reason": "Base_model_prefix loading regression and NaNs in classification head initialization are unrelated." + "reason": "Model loading via `base_model_prefix` and NaNs in classification heads are unrelated." }, { "left": "issue:43976", "right": "issue:45446", "accept": false, - "reason": "Both involve version checks, but one is Python support gating and the other is a PyTorch import-version check; different code paths." + "reason": "Python-version compatibility and a PyTorch version check for AuxRequest are separate version-guard bugs." }, { - "left": "issue:44355", - "right": "issue:46231", + "left": "issue:44038", + "right": "issue:45478", "accept": false, - "reason": "Compiled Python file errors and a documentation parameter mismatch are unrelated." + "reason": "Both mention Qwen3.5/Moe, but one is a 5.0 bug and the other is a 5.4+ from_pretrained error; insufficient evidence they are the same underlying issue." }, { - "left": "issue:42673", - "right": "issue:45072", + "left": "issue:43116", + "right": "issue:43493", "accept": false, - "reason": "VRAM leak in multithreaded dataloader use and dtype mismatches in bfloat16 inference are unrelated." + "reason": "Classification example-script output bug versus SigLIP2 implementation discrepancy are unrelated." }, { - "left": "issue:43116", - "right": "issue:43493", + "left": "issue:42673", + "right": "issue:45072", "accept": false, - "reason": "Example-script classification issue and SigLIP2 implementation mismatch are unrelated." + "reason": "VRAM leak in multi-threaded loading and dtype mismatches in bfloat16 inference are different code paths." }, { "left": "issue:45357", "right": "issue:45478", "accept": false, - "reason": "Both are Qwen3.5 regressions, but one is incorrect save_pretrained visual keys and the other is a from_pretrained error; different concrete bugs." + "reason": "Both are Qwen3.5 issues, but one is save_pretrained key serialization and the other is a from_pretrained loading error; not the same bug." + }, + { + "left": "issue:44355", + "right": "issue:46231", + "accept": false, + "reason": "Compiled Python-file loading errors and a documentation parameter-name mismatch are unrelated." }, { "left": "issue:44625", "right": "issue:46002", "accept": false, - "reason": "Config propagation bug and NaNs during init are different issues." + "reason": "A missing num_labels propagation bug and NaNs in classification heads are different defects." }, { "left": "issue:42371", "right": "issue:43874", "accept": false, - "reason": "TF32 API guidance and GLM46V image-patch helper failure are unrelated." + "reason": "TF32 behavior settings and a missing image-patch-count method are unrelated." }, { "left": "issue:45083", "right": "issue:45446", "accept": false, - "reason": "Different code paths: qwen3_omni_moe feature-length helper vs flex_attention version check." + "reason": "Unexpected helper output lengths in qwen3_omni_moe and an AuxRequest import version check are not the same issue." }, { "left": "issue:44610", "right": "issue:46909", "accept": false, - "reason": "Processor/model input-size mismatch and a broken documentation example are unrelated." + "reason": "Processor/model input-size mismatch versus a broken documentation example are unrelated." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "SGD argument handling and trainer evaluate-after-train failure are different training issues." + }, + { + "left": "issue:43618", + "right": "issue:46829", + "accept": false, + "reason": "Both touch output dataclasses, but they concern different models and different missing/wrong fields." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "A multi-label example script bug and an incorrect pipeline translation-task error message are unrelated." } ] }, @@ -15896,7 +15996,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 52, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -15912,17 +16012,13 @@ "issue:42796", "issue:43064", "issue:43066", - "issue:43116", "issue:43441", "issue:43450", - "issue:43618", "issue:43673", - "issue:43825", "issue:43827", "issue:43937", "issue:43986", "issue:43992", - "issue:44075", "issue:44117", "issue:44263", "issue:44291", @@ -15931,254 +16027,272 @@ "issue:44466", "issue:44704", "issue:44779", - "issue:44936", + "issue:44829", "issue:44977", "issue:44995", "issue:45071", + "issue:45081", "issue:45092", + "issue:45216", "issue:45310", + "issue:45341", "issue:45405", "issue:45446", + "issue:45561", "issue:45563", "issue:45588", "issue:46002", "issue:46077", "issue:46328", + "issue:46512", "issue:46614", "issue:46640", "issue:46655", "issue:46697", "issue:46710", "issue:46752", - "issue:46762", - "issue:46829" + "issue:46762" ], "soft_pairs": [ "issue:43992|issue:44704", - "issue:44075|issue:44936", - "issue:43618|issue:46829", - "issue:43116|issue:43825", - "issue:28348|issue:46640", "issue:43066|issue:46710", "issue:46614|issue:46752", - "issue:44263|issue:46762", + "issue:28348|issue:46640", "issue:44995|issue:46697", + "issue:44263|issue:46762", "issue:45092|issue:46762", - "issue:42222|issue:43441", "issue:43827|issue:45446", + "issue:42222|issue:43441", "issue:33453|issue:39401", "issue:43064|issue:43673", "issue:42673|issue:42796", - "issue:44380|issue:46077", "issue:43937|issue:45563", + "issue:44380|issue:46077", "issue:43673|issue:46328", + "issue:44291|issue:45071", "issue:42673|issue:46655", "issue:42222|issue:46752", "issue:45405|issue:45446", - "issue:44291|issue:45071", "issue:44117|issue:45588", - "issue:39401|issue:44779", "issue:44466|issue:46002", - "issue:44410|issue:45310", + "issue:39401|issue:44779", "issue:43937|issue:44977", - "issue:43450|issue:43986" + "issue:44410|issue:45310", + "issue:43450|issue:43986", + "issue:45341|issue:45561", + "issue:45071|issue:46512", + "issue:45341|issue:46328", + "issue:44829|issue:45216", + "issue:44117|issue:45081" ], - "prepared_review_unit_hash": "c144afe6561b030fbf4eb0d15a5ad399d1951683a2207bd797cd642a22be02d4", + "prepared_review_unit_hash": "06ebe3068c4a0c9edbc583b5f8ab298bd46158aa64d29b902bf0715da3ed6a2a", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 28, - "serialized_chars": 35845, - "estimated_input_tokens": 8962, - "estimated_eval_tokens": 18180 + "soft_pair_count": 30, + "serialized_chars": 36673, + "estimated_input_tokens": 9169, + "estimated_eval_tokens": 18594 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: most items are unrelated Transformers bugs spanning tokenizers, generation, vision/video processing, flash-attention, training, and config/init regressions. None of the soft pairs look like the same underlying bug or change, so I would not merge any of them.", - "confidence": 0.93, - "canonical_issue_reason": "issue:44779 is the clearest self-contained bug report in the cluster: it names a specific model, a version regression, and a concrete tokenizer misbehavior. The rest are mostly unrelated and do not form a single duplicate set.", + "summary": "The set is a heterogeneous mix of Transformers v5 tokenizer, loading, attention, generation, and model-specific regressions. None of the suggested soft pairs look like the same underlying bug or change; they mostly share only broad subsystem keywords.", + "confidence": 0.89, + "canonical_issue_reason": "issue:33453 is the broadest and most umbrella-like report in the set, centered on tokenizer loading regressions rather than a narrow model-specific symptom.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44779 is the best representative issue because it is specific, reproducible, and clearly scoped, even though the overall cluster is too mixed to treat as one duplicate family.", + "best_issue_reason": "issue:33453 has the clearest general loading-regression framing and the highest discussion activity, making it the best representative issue for this heterogeneous cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43992", "right": "issue:44704", "accept": false, - "reason": "Different subsystems and failures: model weight loading vs processor kwarg forwarding." - }, - { - "left": "issue:44075", - "right": "issue:44936", - "accept": false, - "reason": "Unrelated training/evaluation behavior versus optimizer argument handling." - }, - { - "left": "issue:43618", - "right": "issue:46829", - "accept": false, - "reason": "Both are output-structure bugs, but they affect different models and fields with different symptoms." - }, - { - "left": "issue:43116", - "right": "issue:43825", - "accept": false, - "reason": "One is example-script multi-label inference, the other is a pipeline error-message/docs regression." - }, - { - "left": "issue:28348", - "right": "issue:46640", - "accept": false, - "reason": "Both mention FlashAttention, but one is GPT-2 support and the other is an unspecified issue with no clear shared code path." + "reason": "Different code paths: missing model weight handling vs AutoProcessor kwarg propagation." }, { "left": "issue:43066", "right": "issue:46710", "accept": false, - "reason": "Tokenizer decoder-type regression is unrelated to the DeepSeek accuracy regression." + "reason": "Tokenizer decoder typing issue vs a DeepSeek accuracy regression; not the same bug." }, { "left": "issue:46614", "right": "issue:46752", "accept": false, - "reason": "Tokenizer phonemizer state bug versus chat-template empty-input crash." + "reason": "Stale phonemizer backend state vs empty chat template crash; unrelated symptoms." }, { - "left": "issue:44263", - "right": "issue:46762", + "left": "issue:28348", + "right": "issue:46640", "accept": false, - "reason": "Different attention/indexing problems in different models; no evidence they share the same bug." + "reason": "Feature request for GPT-2 FlashAttention support vs a vague Flash Attention 2 issue; not clearly one defect." }, { "left": "issue:44995", "right": "issue:46697", "accept": false, - "reason": "Stale indexer cache crash is unrelated to image over-padding." + "reason": "GLM-MoE stale indexer cache crash vs Swin2SR image over-padding; unrelated." }, { - "left": "issue:45092", + "left": "issue:44263", "right": "issue:46762", "accept": false, - "reason": "Meta-init incompatibility for remote-code checkpoints is not the same as Minimax attention numerical drift." + "reason": "torch.split return semantics in GlmMoeDsaIndexer vs numerical mismatch in Minimax attention; different problems." }, { - "left": "issue:42222", - "right": "issue:43441", + "left": "issue:45092", + "right": "issue:46762", "accept": false, - "reason": "Vitpose loading breakage and Ministral FlashAttention failure are separate model-specific issues." + "reason": "Meta-init incompatibility for InternVL2 vs Minimax attention numeric drift; no shared failure mode." }, { "left": "issue:43827", "right": "issue:45446", "accept": false, - "reason": "Docs still referencing pipeline() is unrelated to a flex_attention import version check." + "reason": "Docs referencing removed pipeline() vs PyTorch version check bug in flex_attention import." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "VitPose model breakage vs Ministral FlashAttention failure; only broad model-loading similarity." }, { "left": "issue:33453", "right": "issue:39401", "accept": false, - "reason": "Generic tokenizer-loading regression and Qwen3 offset_mapping bug are different failures." + "reason": "Generic tokenizer-loading regression vs Qwen3 offset_mapping bug; different concrete defects." }, { "left": "issue:43064", "right": "issue:43673", "accept": false, - "reason": "Distributed optimizer state corruption is unrelated to chunked_prefill cache handling." + "reason": "FSDP2/PEFT optimizer-state corruption vs missing generation cache during chunked prefill." }, { "left": "issue:42673", "right": "issue:42796", "accept": false, - "reason": "VRAM leak in Qwen3ForCausalLM is unrelated to Ministral tokenizer BPE-marker handling." + "reason": "VRAM leak in multi-threaded Qwen3 inference vs tokenizer BPE marker handling." }, { - "left": "issue:44380", - "right": "issue:46077", + "left": "issue:43937", + "right": "issue:45563", "accept": false, - "reason": "Both involve attention behavior, but the concrete bugs are different: scaling ignored vs bidirectional masking wrong." + "reason": "Invalid GenerationConfig vs stale warning in paged generate(); unrelated." }, { - "left": "issue:43937", - "right": "issue:45563", + "left": "issue:44380", + "right": "issue:46077", "accept": false, - "reason": "GenerationConfig validation error and a stale warning for num_return_sequences are different code paths." + "reason": "Ignored GPT-2 attention scaling under SDPA/FlashAttention vs Gemma4 mask construction bug." }, { "left": "issue:43673", "right": "issue:46328", "accept": false, - "reason": "Generation cache handling and SequenceFeatureExtractor padding performance are unrelated." + "reason": "Generation cache missing vs SequenceFeatureExtractor padding inefficiency; unrelated." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "init_empty_weights unexpected argument vs PretrainedConfig type checking regression; not the same issue." }, { "left": "issue:42673", "right": "issue:46655", "accept": false, - "reason": "VRAM leak in causal LM threading is unrelated to video processor padding." + "reason": "VRAM leak in Qwen3 threading vs SmolVLM video padding miscalculation." }, { "left": "issue:42222", "right": "issue:46752", "accept": false, - "reason": "Vitpose model breakage and empty-chat-template IndexError are not the same bug." + "reason": "VitPose model breakage vs empty-list chat template crash; unrelated." }, { "left": "issue:45405", "right": "issue:45446", "accept": false, - "reason": "A released-version dependency bump and an incorrect PyTorch version check are unrelated maintenance issues." + "reason": "Dependency version bump problem vs flex_attention import version check bug." }, { - "left": "issue:44291", - "right": "issue:45071", + "left": "issue:44117", + "right": "issue:45588", "accept": false, - "reason": "Meta-init constructor argument mismatch and PretrainedConfig type checking are different compatibility regressions." + "reason": "Tokenizer mapping fallback bug vs flash_attention.py crashing on s_aux=None; different layers." }, { - "left": "issue:44117", - "right": "issue:45588", + "left": "issue:44466", + "right": "issue:46002", "accept": false, - "reason": "Tokenizer mapping None handling is unrelated to flash_attention auxiliary-state crashes." + "reason": "Serialization of tied lm_head weights vs NaNs in classification heads; not the same defect." }, { "left": "issue:39401", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different observable outputs." + "reason": "Qwen3 offset_mapping bug vs DeepSeek tokenizer regression; similar area, different model and behavior." }, { - "left": "issue:44466", - "right": "issue:46002", + "left": "issue:43937", + "right": "issue:44977", "accept": false, - "reason": "State-dependent lm_head serialization differs from NaNs in classification heads; no shared code path is evident." + "reason": "GenerationConfig validation failure vs FlashAttention generation failure in Qwen3.5." }, { "left": "issue:44410", "right": "issue:45310", "accept": false, - "reason": "Missing QKV/gate projections in one model is not the same as a Qwen3.5 MoE from_pretrained failure." + "reason": "Missing qkv/gate projections in qwen3next vs Qwen3.5 MoE from_pretrained error." }, { - "left": "issue:43937", - "right": "issue:44977", + "left": "issue:43450", + "right": "issue:43986", "accept": false, - "reason": "Invalid GenerationConfig and FlashAttention generation failure are distinct issues." + "reason": "Video processor output shape bug vs confusing missing-torchvision crash through AutoProcessor." }, { - "left": "issue:43450", - "right": "issue:43986", + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "A testing_utils bug vs pytest-xdist race on captured_info.txt; not enough to treat as duplicate." + }, + { + "left": "issue:45071", + "right": "issue:46512", "accept": false, - "reason": "Batched video processor shape bug is unrelated to AutoProcessor crashing without torchvision." + "reason": "Type checking regression in PretrainedConfig vs stale config.dtype after from_config()." + }, + { + "left": "issue:45341", + "right": "issue:46328", + "accept": false, + "reason": "testing_utils issue vs feature extractor padding performance regression; unrelated." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "FlashAttention-3 causes degenerate training vs Qwen3.5 checkpoint save regression." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer mapping None-handling vs Mistral regex patch crash; same broad subsystem, different concrete bug." } ] }, @@ -16194,7 +16308,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 53, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -16202,6 +16316,7 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-502", "nodes": [ + "issue:27957", "issue:38617", "issue:39401", "issue:41628", @@ -16212,292 +16327,291 @@ "issue:43097", "issue:43122", "issue:43208", + "issue:43352", "issue:43408", "issue:43673", "issue:43742", + "issue:43874", "issue:43931", "issue:43937", "issue:44075", - "issue:44117", "issue:44220", "issue:44261", "issue:44263", "issue:44410", + "issue:44483", "issue:44484", "issue:44485", "issue:44493", "issue:44496", - "issue:44829", "issue:44877", "issue:44933", "issue:45071", - "issue:45081", - "issue:45216", "issue:45278", "issue:45310", "issue:45341", "issue:45362", + "issue:45464", "issue:45478", "issue:45542", - "issue:45561", "issue:45656", + "issue:46172", "issue:46178", "issue:46209", "issue:46231", "issue:46298", "issue:46328", - "issue:46512", "issue:46617", "issue:46762", "issue:46894" ], "soft_pairs": [ - "issue:45341|issue:46328", - "issue:44829|issue:45216", - "issue:45341|issue:45561", - "issue:45071|issue:46512", - "issue:44117|issue:45081", "issue:43408|issue:43742", - "issue:38617|issue:41628", - "issue:43122|issue:46298", "issue:42617|issue:45341", + "issue:43122|issue:46298", + "issue:38617|issue:41628", "issue:45362|issue:46328", "issue:42222|issue:46178", + "issue:44877|issue:46231", "issue:44484|issue:45310", + "issue:43937|issue:44496", "issue:42371|issue:44261", - "issue:44877|issue:46231", "issue:39401|issue:46209", - "issue:43937|issue:44496", + "issue:44261|issue:46762", "issue:42371|issue:46617", "issue:44410|issue:45478", - "issue:44261|issue:46762", "issue:44933|issue:45278", - "issue:43208|issue:46894", "issue:38617|issue:44220", "issue:44075|issue:45656", + "issue:43208|issue:46894", "issue:43066|issue:43122", "issue:44263|issue:44485", + "issue:44483|issue:45464", "issue:43097|issue:44493", "issue:43673|issue:45341", - "issue:42222|issue:45542", "issue:43931|issue:44410", + "issue:42222|issue:45542", "issue:41628|issue:45071", - "issue:45071|issue:45542" - ], - "prepared_review_unit_hash": "214edae498bf5222586fbab8820807304139197969cb71003d349f651112ee2b", - "component_budget": { + "issue:45071|issue:45542", + "issue:44484|issue:45478", + "issue:43874|issue:44261", + "issue:44263|issue:46172", + "issue:27957|issue:43352" + ], + "prepared_review_unit_hash": "9579dd932e16b800e1873277670db8e529f06b52e72a39402d6d0ae86f72edc0", + "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36564, - "estimated_input_tokens": 9141, - "estimated_eval_tokens": 18538 + "serialized_chars": 36277, + "estimated_input_tokens": 9070, + "estimated_eval_tokens": 18396 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of unrelated Transformers bugs (imports, tokenization, model loading, config, training, and testing). None of the soft pairs look like the same underlying bug or change, so they are all rejected. The strongest canonical issue is the exact Qwen3.5 MoE from_pretrained duplicate pair.", - "confidence": 0.82, - "canonical_issue_reason": "Issue 45310 is the best canonical issue: it is the earliest of the exact duplicate Qwen3.5 MoE from_pretrained reports (paired with 45478) and describes a concrete regression.", - "canonical_pr_reason": null, - "best_issue_reason": "45310 is the strongest global issue representative because it is a clear, concrete bug report and has an exact same-title duplicate nearby.", - "best_pr_reason": null, + "summary": "This cluster is not a duplicate set: it mixes unrelated import errors, tokenizer regressions, model-loading/config issues, API behavior changes, and performance/logging bugs. The soft pairs mostly share only broad subsystem names, not the same underlying defect.", + "confidence": 0.97, + "canonical_issue_reason": "No canonical issue: the items are heterogeneous and do not appear to describe one shared bug or change.", + "canonical_pr_reason": "No canonical PR candidate is present.", + "best_issue_reason": "No single issue is a strong representative because there is no clear duplicate cluster here.", + "best_pr_reason": "No PR candidate is present.", "soft_edge_verdicts": [ - { - "left": "issue:45341", - "right": "issue:46328", - "accept": false, - "reason": "Testing-utils bug vs sequence feature padding performance issue; unrelated symptoms and code paths." - }, - { - "left": "issue:44829", - "right": "issue:45216", - "accept": false, - "reason": "Flash-attention training regression and Qwen3.5 save_pretrained checkpoint regression are different bugs." - }, - { - "left": "issue:45341", - "right": "issue:45561", - "accept": false, - "reason": "Both touch testing helpers, but one is a doc/small bug and the other is a pytest-xdist file race; not the same issue." - }, - { - "left": "issue:45071", - "right": "issue:46512", - "accept": false, - "reason": "Type-checking regression in PreTrainedConfig vs stale dtype state after from_config; different failure modes." - }, - { - "left": "issue:44117", - "right": "issue:45081", - "accept": false, - "reason": "Tokenizer mapping None handling and Mistral regex patch crash are distinct code paths." - }, { "left": "issue:43408", "right": "issue:43742", "accept": false, - "reason": "Different model-loading failures for different models; no shared concrete bug." + "reason": "Different model-loading problems: SAM3 tracker/type warning vs MobileLLM key error." }, { - "left": "issue:38617", - "right": "issue:41628", + "left": "issue:42617", + "right": "issue:45341", "accept": false, - "reason": "Both are import errors, but for different missing symbols and likely different regressions." + "reason": "Unrelated: a runtime failure in 3d_parallel.py vs a small testing_utils.py bug." }, { "left": "issue:43122", "right": "issue:46298", "accept": false, - "reason": "Tokenizer behavior regression vs a noisy logging message; unrelated." + "reason": "Tokenizer regression vs noisy logging; no shared underlying defect." }, { - "left": "issue:42617", - "right": "issue:45341", + "left": "issue:38617", + "right": "issue:41628", "accept": false, - "reason": "A script runtime failure and a testing-utils bug are not the same underlying defect." + "reason": "Both are import errors, but for different symbols and different root causes." }, { "left": "issue:45362", "right": "issue:46328", "accept": false, - "reason": "Qwen3.5 chat crash and sequence padding performance issue are unrelated." + "reason": "Chat model crash vs SequenceFeatureExtractor padding inefficiency; unrelated." }, { "left": "issue:42222", "right": "issue:46178", "accept": false, - "reason": "Vitpose model breakage and tensorflow_text backend mapping error are different bugs." + "reason": "VitPose model breakage vs tensorflow_text backend mapping error; different bugs." + }, + { + "left": "issue:44877", + "right": "issue:46231", + "accept": false, + "reason": "Strict config loading issue vs documentation parameter mismatch; not the same bug." }, { "left": "issue:44484", "right": "issue:45310", "accept": false, - "reason": "Shard-size question and Qwen3.5 MoE loading error are unrelated." + "reason": "Different topics: save_pretrained shard size behavior vs Qwen3.5 MoE from_pretrained failure." }, { - "left": "issue:42371", - "right": "issue:44261", + "left": "issue:43937", + "right": "issue:44496", "accept": false, - "reason": "TF32 API guidance issue and MLA rms_norm_eps precision bug do not share the same underlying fix." + "reason": "GenerationConfig validation issue vs unrecognized model/config loading error." }, { - "left": "issue:44877", - "right": "issue:46231", + "left": "issue:42371", + "right": "issue:44261", "accept": false, - "reason": "Strict config loading bug and a documentation parameter-name mismatch are not duplicates." + "reason": "TF32 API guidance vs MLA layernorm precision/config bug; unrelated." }, { "left": "issue:39401", "right": "issue:46209", "accept": false, - "reason": "Tokenizer offset mapping bug and FP8 inference on B200 are unrelated." + "reason": "Tokenizer offset_mapping bug vs FP8 inference failure on B200; no match." }, { - "left": "issue:43937", - "right": "issue:44496", + "left": "issue:44261", + "right": "issue:46762", "accept": false, - "reason": "Invalid GenerationConfig for GLM-5 and unrecognized model config are different loading/validation failures." + "reason": "Both mention numerical differences, but one is a missing config epsilon and the other is an attention implementation mismatch." }, { "left": "issue:42371", "right": "issue:46617", "accept": false, - "reason": "TF32 control API issue and trust_remote_code cache FileNotFoundError are unrelated." + "reason": "TF32 settings guidance vs cache FileNotFoundError for trust_remote_code; unrelated." }, { "left": "issue:44410", "right": "issue:45478", "accept": false, - "reason": "Missing qkv/gate projections in qwen3next and Qwen3.5 MoE from_pretrained error are different model bugs." - }, - { - "left": "issue:44261", - "right": "issue:46762", - "accept": false, - "reason": "Both mention numerical differences, but they concern different models and different implementations." + "reason": "Qwen3Next missing projections vs Qwen3.5 MoE from_pretrained error; different code paths." }, { "left": "issue:44933", "right": "issue:45278", "accept": false, - "reason": "Specific missing image_utils import vs broad import errors after upgrade; not the same bug." - }, - { - "left": "issue:43208", - "right": "issue:46894", - "accept": false, - "reason": "xLSTM training blockers and CohereASR double-shift bug are unrelated." + "reason": "A missing image_utils import vs broad import errors after upgrade; too general to be the same bug." }, { "left": "issue:38617", "right": "issue:44220", "accept": false, - "reason": "Different import/runtime issues in unrelated modules." + "reason": "ImportError for layer_type_validation vs audio feature extraction issue; unrelated." }, { "left": "issue:44075", "right": "issue:45656", "accept": false, - "reason": "SGD argument handling and optimizer step being called twice under deepspeed are different optimizer problems." + "reason": "SGD args not used vs optimizer step called twice with deepspeed; different optimizer bugs." + }, + { + "left": "issue:43208", + "right": "issue:46894", + "accept": false, + "reason": "xLSTM training blockers vs CohereASR double-shift loss bug; no shared defect." }, { "left": "issue:43066", "right": "issue:43122", "accept": false, - "reason": "Tokenizer decoder-type regression and a broader tokenization difference are not the same concrete bug." + "reason": "Both are tokenizer-related regressions, but one is decoder type mismatch and the other is changed tokenization output; not the same bug." }, { "left": "issue:44263", "right": "issue:44485", "accept": false, - "reason": "torch.split return handling in an indexer and GLM-5 RoPE implementation are unrelated." + "reason": "torch.split return issue in GLM-MoE indexer vs GLM-5 RoPE implementation; distinct code paths." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Both hit chat/completions, but one is request acceptance failure and the other is streaming inference failure on a specific model." }, { "left": "issue:43097", "right": "issue:44493", "accept": false, - "reason": "Removed tie_embeddings flag and unexpected key warnings with position ids are different regression surfaces." + "reason": "Removed tie_embeddings flag vs unexpected position-id keys; unrelated model/config regressions." }, { "left": "issue:43673", "right": "issue:45341", "accept": false, - "reason": "Generation cache missing in chunked_prefill and a testing-utils bug are unrelated." + "reason": "Cache missing in chunked_prefill vs a testing utility bug; no overlap." }, { - "left": "issue:42222", - "right": "issue:45542", + "left": "issue:43931", + "right": "issue:44410", "accept": false, - "reason": "Vitpose model breakage and TensorFlow backend detection error are different issues." + "reason": "Qwen3-VL weight shape mismatch vs Qwen3Next missing projections; different model-loading problems." }, { - "left": "issue:43931", - "right": "issue:44410", + "left": "issue:42222", + "right": "issue:45542", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch and qwen3next missing projections are different model-loading bugs." + "reason": "VitPose model breakage vs TensorFlow backend detection error; unrelated." }, { "left": "issue:41628", "right": "issue:45071", "accept": false, - "reason": "Missing AutoImageProcessor import and PretrainedConfig type-checking regression are unrelated." + "reason": "ImportError for AutoImageProcessor vs PreTrainedConfig type checking regression; different issues." }, { "left": "issue:45071", "right": "issue:45542", "accept": false, - "reason": "Type-checking regression and TensorFlow backend error are different problems." + "reason": "Config type-checking regression vs undefined tf backend error; unrelated." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "save_pretrained shard size behavior vs Qwen3.5 MoE loading error; different bugs." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Missing image patch helper vs missing rms_norm_eps precision bug; different multimodal/model issues." + }, + { + "left": "issue:44263", + "right": "issue:46172", + "accept": false, + "reason": "GLM-MoE indexing bug vs deepseek-v4 tensor transpose issue; unrelated." + }, + { + "left": "issue:27957", + "right": "issue:43352", + "accept": false, + "reason": "Both mention Flash Attention 2 support, but for different models and likely separate model-specific support gaps." } ] }, @@ -16513,7 +16627,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 54, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -16521,7 +16635,6 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-502", "nodes": [ - "issue:27957", "issue:38617", "issue:42757", "issue:42890", @@ -16529,16 +16642,17 @@ "issue:43064", "issue:43097", "issue:43208", - "issue:43352", "issue:43441", "issue:43452", "issue:43531", "issue:43577", "issue:43673", + "issue:43873", "issue:43874", "issue:43931", "issue:43986", "issue:44038", + "issue:44062", "issue:44117", "issue:44261", "issue:44263", @@ -16546,7 +16660,6 @@ "issue:44351", "issue:44360", "issue:44410", - "issue:44483", "issue:44484", "issue:44485", "issue:44829", @@ -16556,8 +16669,6 @@ "issue:45372", "issue:45375", "issue:45440", - "issue:45464", - "issue:45478", "issue:45588", "issue:45753", "issue:46050", @@ -16567,263 +16678,266 @@ "issue:46369", "issue:46655", "issue:46664", + "issue:46701", "issue:46710", + "issue:46752", + "issue:46821", "issue:46897" ], "soft_pairs": [ - "issue:44483|issue:45464", - "issue:44484|issue:45478", - "issue:43874|issue:44261", - "issue:44263|issue:46172", "issue:44360|issue:46172", - "issue:27957|issue:43352", "issue:43931|issue:44038", - "issue:43012|issue:43577", "issue:43441|issue:44829", + "issue:43012|issue:43577", "issue:43097|issue:43452", "issue:45003|issue:45341", "issue:43986|issue:45341", "issue:44261|issue:44485", "issue:45753|issue:46369", - "issue:43874|issue:44263", "issue:44117|issue:44987", - "issue:45440|issue:46172", "issue:43531|issue:45341", + "issue:43874|issue:44263", + "issue:45440|issue:46172", "issue:43097|issue:45588", "issue:43673|issue:46123", - "issue:38617|issue:42757", "issue:44410|issue:45375", + "issue:43986|issue:45372", "issue:44038|issue:44410", + "issue:38617|issue:42757", + "issue:43097|issue:46050", "issue:43208|issue:46897", "issue:43064|issue:44315", - "issue:43986|issue:45372", - "issue:43097|issue:46050", "issue:43986|issue:44484", "issue:46655|issue:46664", + "issue:46298|issue:46752", "issue:42890|issue:46298", + "issue:46172|issue:46710", + "issue:44062|issue:44117", "issue:38617|issue:44351", - "issue:46172|issue:46710" + "issue:43097|issue:43873", + "issue:43012|issue:46821", + "issue:44117|issue:46701" ], - "prepared_review_unit_hash": "07e444fc648a09febf294b2b192da8a6bad55e42ead3bf66f763be9183c94c16", + "prepared_review_unit_hash": "0cd39a55a124aaec4a5ba26bb1d8b7212055ab271ccf538ee7126c938dabf2b2", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37118, - "estimated_input_tokens": 9280, - "estimated_eval_tokens": 18816 + "serialized_chars": 37183, + "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 very heterogeneous: most items are unrelated model-specific regressions, import errors, or test issues that only share broad Transformers subsystems. I found no soft pair that is clearly the same underlying bug/change, so all candidate duplicate links should be rejected.", - "confidence": 0.41, - "canonical_issue_reason": "issue:43097 is the most general and representative regression-style report in the set: it concerns a core v5 API removal with explicit downstream impact and inbound references, making it a better anchor than the more model-specific reports.", + "summary": "These items are mostly unrelated Transformers bug reports. The soft-similarity links are largely false positives across different models, APIs, or failure modes; none look like the same underlying bug or change.", + "confidence": 0.78, + "canonical_issue_reason": "issue:43873 is the broadest and most triage-relevant item here (open, high discussion, inbound refs), so it is the closest thing to an umbrella anchor\u2014but the cluster is still too heterogeneous to treat it as a true duplicate canonical.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43097 is the best single issue to represent this cluster because it is broad, clearly scoped, and more central than the many highly specific model bugs; the rest do not cohere around one duplicate root cause.", + "best_issue_reason": "issue:43873 is the best representative issue because it has the strongest evidence of active impact and ongoing discussion, even though it does not subsume the other reports as duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44483", - "right": "issue:45464", - "accept": false, - "reason": "Both mention chat/completions, but one is a general v5 request-validation failure and the other is a Qwen3.5 streaming-specific failure; different bugs." - }, - { - "left": "issue:44484", - "right": "issue:45478", - "accept": false, - "reason": "Unrelated problems: one is about save_pretrained shard size defaults, the other is a Qwen3.5-MoE from_pretrained error." - }, - { - "left": "issue:43874", - "right": "issue:44261", - "accept": false, - "reason": "Different model paths and different failures: missing image-patch API vs missing rms_norm_eps in config." - }, - { - "left": "issue:44263", - "right": "issue:46172", - "accept": false, - "reason": "Both are GLM/DeepSeek-related but the concrete code-paths differ completely: torch.split handling vs dtype/transpose behavior." - }, { "left": "issue:44360", "right": "issue:46172", "accept": false, - "reason": "Different bugs in different components: missing ReLU in a DSA indexer vs DeepSeek-V4 tensor conversion/transpose issue." - }, - { - "left": "issue:27957", - "right": "issue:43352", - "accept": false, - "reason": "Both involve Flash Attention 2 support, but they are separate model-support issues for different architectures." + "reason": "Different models and different bugs: DSA indexer math vs DeepSeek-v4 dtype/transpose behavior." }, { "left": "issue:43931", "right": "issue:44038", "accept": false, - "reason": "Both are Qwen3-VL-related, but the reported breakages are different model-loading/configuration failures." + "reason": "Both are Qwen3-VL-related, but one is a weight-shape mismatch and the other is a Qwen3-VL-Moe bug; not the same failure." }, { - "left": "issue:43012", - "right": "issue:43577", + "left": "issue:43441", + "right": "issue:44829", "accept": false, - "reason": "A compilation precision warning is not the same as a Blip2 dtype propagation bug." + "reason": "Both mention flash attention, but one is a Ministral-3 load failure and the other is a training regression in sequence classification." }, { - "left": "issue:43441", - "right": "issue:44829", + "left": "issue:43012", + "right": "issue:43577", "accept": false, - "reason": "Both mention Flash Attention, but one is a model-loading failure and the other is a training-quality regression in sequence classification." + "reason": "Both involve dtype/precision, but one is a PyTorch warning during compile and the other is incorrect dtype propagation when loading Blip2." }, { "left": "issue:43097", "right": "issue:43452", "accept": false, - "reason": "A v5 API removal is unrelated to a gguf_file / from_pretrained loading breakage." + "reason": "A v5 API removal issue and a gguf tokenizer/model loading breakage are different code paths and symptoms." }, { "left": "issue:45003", "right": "issue:45341", "accept": false, - "reason": "modeling_utils sys.modules access and a testing_utils bug are unrelated." + "reason": "sys.modules access in modeling_utils is unrelated to a testing_utils bug." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "AutoProcessor video-loading failure without torchvision is unrelated to a testing utility bug." + "reason": "Video-model processor import crash without torchvision is unrelated to a testing utility issue." }, { "left": "issue:44261", "right": "issue:44485", "accept": false, - "reason": "Different GLM issues: one is a missing config epsilon, the other is a RoPE implementation discussion." + "reason": "Different model internals: MLA q_a_layernorm eps handling vs GLM-5 RoPE implementation." }, { "left": "issue:45753", "right": "issue:46369", "accept": false, - "reason": "Different symptoms and subsystems: Qwen3.5 infinite-loop behavior vs Sam3-LiteText memory overflow." + "reason": "Different symptoms and subsystems: Qwen3_5 infinite loop on an image vs SAM memory overflow." }, { - "left": "issue:43874", - "right": "issue:44263", + "left": "issue:44117", + "right": "issue:44987", "accept": false, - "reason": "Separate GLM46V bugs: missing image patch counting API vs incorrect torch.split return handling." + "reason": "Tokenizer mapping None handling is a different bug from loading physical-intelligence/fast." }, { - "left": "issue:44117", - "right": "issue:44987", + "left": "issue:43531", + "right": "issue:45341", "accept": false, - "reason": "Different root causes: tokenizer mapping returning None vs a specific model-loading failure on transformers>=5.1.0." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to a testing_utils bug." }, { - "left": "issue:45440", - "right": "issue:46172", + "left": "issue:43874", + "right": "issue:44263", "accept": false, - "reason": "DeepSeek implementation divergence is not the same as the DeepSeek-V4 comb.to(dtype).transpose bug." + "reason": "Image patch-count helper missing in GLM46V is unrelated to torch.split output handling in GlmMoeDsaIndexer." }, { - "left": "issue:43531", - "right": "issue:45341", + "left": "issue:45440", + "right": "issue:46172", "accept": false, - "reason": "A Qwen3-MoE sliding_window bug and a testing utility bug are unrelated." + "reason": "DeepSeekV3MoE divergence and DeepSeek-v4 tensor conversion/transposition are different bugs." }, { "left": "issue:43097", "right": "issue:45588", "accept": false, - "reason": "API removal in model config handling is unrelated to a flash_attention.py AttributeError on s_aux=None." + "reason": "One is a removed config attribute; the other is a flash_attention crash on s_aux=None." }, { "left": "issue:43673", "right": "issue:46123", "accept": false, - "reason": "Generation cache handling during chunked_prefill and MaskGenerationPipeline batch-finalization are different bugs." + "reason": "Generation cache missing during chunked prefill is unrelated to MaskGenerationPipeline batch-finalization logic." }, { - "left": "issue:38617", - "right": "issue:42757", + "left": "issue:44410", + "right": "issue:45375", "accept": false, - "reason": "Both are ImportError reports, but they are for different missing symbols and different dependency surfaces." + "reason": "Missing attn projections in qwen3next is unrelated to a missing config field being dropped by strict loading." }, { - "left": "issue:44410", - "right": "issue:45375", + "left": "issue:43986", + "right": "issue:45372", "accept": false, - "reason": "One is a missing projection layer in qwen3next, the other a config field being dropped by @strict." + "reason": "Different import failures in different processor stacks: missing torchvision vs missing ReasoningEffort." }, { "left": "issue:44038", "right": "issue:44410", "accept": false, - "reason": "These are different Qwen3-family regressions with different missing components and failure modes." + "reason": "Qwen3-VL-Moe bug and qwen3next projection omission are distinct model families and failures." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they are different missing symbols from different modules and likely different fixes." + }, + { + "left": "issue:43097", + "right": "issue:46050", + "accept": false, + "reason": "Config API removal is unrelated to save/load breakage for sinq-quantized models." }, { "left": "issue:43208", "right": "issue:46897", "accept": false, - "reason": "xLSTM training blockers and Florence2 training-loss double-shift are unrelated." + "reason": "xLSTM training blockers and Florence2 loss-shift bug are different model-specific training issues." }, { "left": "issue:43064", "right": "issue:44315", "accept": false, - "reason": "Wrong optimizer states under FSDP2/PEFT and Liger Kernel not applied with model_init are separate training-stack issues." + "reason": "Wrong optimizer states under FSDP2+PEFT is a different code path from Liger Kernel not being applied with model_init." }, { "left": "issue:43986", - "right": "issue:45372", + "right": "issue:44484", "accept": false, - "reason": "A missing torchvision dependency path and a mistral_common ReasoningEffort import error are different crashes." + "reason": "A missing torchvision dependency crash is unrelated to max_shard_size behavior." }, { - "left": "issue:43097", - "right": "issue:46050", + "left": "issue:46655", + "right": "issue:46664", "accept": false, - "reason": "A v5 API removal is unrelated to quantized save/load corruption after sinq serialization." + "reason": "Both are SmolVLM video-processor issues, but one is padding calculation and the other is the meaning of num_frames." }, { - "left": "issue:46655", - "right": "issue:46664", + "left": "issue:46298", + "right": "issue:46752", "accept": false, - "reason": "Both are SmolVLM video-processor reports, but one is padding math and the other is num_frames semantics." + "reason": "Noisy logging and empty-list chat-template IndexError are unrelated." }, { "left": "issue:42890", "right": "issue:46298", "accept": false, - "reason": "Test flakiness from missing set_seed and noisy logging are not the same bug." + "reason": "Missing set_seed in tests is unrelated to a logging-message issue." + }, + { + "left": "issue:46172", + "right": "issue:46710", + "accept": false, + "reason": "DeepSeek-v4 tensor conversion bug is unrelated to an accuracy regression after upgrading Transformers." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "AddedToken special-argument conflict is unrelated to TOKENIZER_MAPPING_NAMES returning None." }, { "left": "issue:38617", "right": "issue:44351", "accept": false, - "reason": "Both are import-name failures, but they concern different missing exports and different root causes." + "reason": "Different missing imports from different surfaces; same general class, but not the same bug." }, { - "left": "issue:43986", - "right": "issue:44484", + "left": "issue:43097", + "right": "issue:43873", "accept": false, - "reason": "Video AutoProcessor crash without torchvision has nothing to do with max_shard_size defaults." + "reason": "A removed config field and a quantization offloading problem are different issues." }, { - "left": "issue:46172", - "right": "issue:46710", + "left": "issue:43012", + "right": "issue:46821", + "accept": false, + "reason": "Precision-warning behavior is unrelated to a window_size AttributeError in TimesFM 2.5." + }, + { + "left": "issue:44117", + "right": "issue:46701", "accept": false, - "reason": "A tensor conversion/transpose bug is not the same as an accuracy regression after a Transformers upgrade." + "reason": "Tokenizer mapping None handling is unrelated to the DiffusionGemma generation-output pipeline failure." } ] }, @@ -16839,7 +16953,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 55, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -16848,6 +16962,7 @@ "cluster_id": "cluster-10000-502", "nodes": [ "issue:36683", + "issue:38617", "issue:42175", "issue:42490", "issue:42757", @@ -16859,23 +16974,25 @@ "issue:43208", "issue:43352", "issue:43502", + "issue:43526", "issue:43550", "issue:43582", + "issue:43618", "issue:43824", "issue:43825", - "issue:43873", "issue:43906", "issue:43976", "issue:43986", "issue:43992", - "issue:44062", + "issue:44016", "issue:44077", - "issue:44117", "issue:44162", "issue:44222", "issue:44368", "issue:44380", + "issue:44483", "issue:44484", + "issue:44514", "issue:44655", "issue:44704", "issue:44936", @@ -16891,279 +17008,287 @@ "issue:46178", "issue:46298", "issue:46365", - "issue:46421", - "issue:46701", - "issue:46752", - "issue:46821" + "issue:46421" ], "soft_pairs": [ - "issue:46298|issue:46752", - "issue:44062|issue:44117", - "issue:43097|issue:43873", - "issue:43012|issue:46821", - "issue:44117|issue:46701", "issue:43502|issue:43992", "issue:44936|issue:46365", "issue:44704|issue:45103", "issue:44380|issue:46421", "issue:44938|issue:46178", - "issue:43986|issue:44117", - "issue:42907|issue:44655", + "issue:43986|issue:44016", "issue:42907|issue:45753", + "issue:42907|issue:44655", "issue:43825|issue:44655", "issue:43012|issue:43550", "issue:45712|issue:46298", "issue:43208|issue:43550", "issue:43012|issue:44368", "issue:43992|issue:44484", - "issue:43352|issue:44380", "issue:43352|issue:43976", + "issue:43352|issue:44380", "issue:43066|issue:43906", - "issue:44162|issue:46172", "issue:43097|issue:46298", - "issue:44380|issue:46132", + "issue:44162|issue:46172", "issue:45584|issue:46365", - "issue:44222|issue:45478", + "issue:44380|issue:46132", "issue:44077|issue:46365", + "issue:44222|issue:45478", "issue:36683|issue:43582", "issue:44380|issue:45479", "issue:42175|issue:46365", "issue:42757|issue:43824", "issue:43064|issue:44380", - "issue:42490|issue:43012" + "issue:42490|issue:43012", + "issue:44483|issue:44514", + "issue:43618|issue:44380", + "issue:43208|issue:43526", + "issue:38617|issue:43502" ], - "prepared_review_unit_hash": "b7e94ce5ef693886bc6a151c19ec0390e376507a5279db74b0955194efd84c77", + "prepared_review_unit_hash": "469f5493a9ce934df1cf2b8dad579db0bd2fe57e3a62eb456b208888c51f1af4", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 34, - "serialized_chars": 37694, - "estimated_input_tokens": 9424, - "estimated_eval_tokens": 19104 + "soft_pair_count": 33, + "serialized_chars": 37438, + "estimated_input_tokens": 9360, + "estimated_eval_tokens": 18976 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are mostly unrelated bug reports across loading, tokenization, quantization, training, serialization, and backend/dependency issues. None of the soft pairs look like the same underlying bug or change, so all soft edges should be rejected. If one issue must serve as an anchor, 43873 is the most active and best-documented, but it is not a duplicate target for the rest.", - "confidence": 0.91, - "canonical_issue_reason": "No strong canonical duplicate exists in this heterogeneous cluster; if forced to pick an anchor, issue 43873 is the most active and best-documented report.", + "summary": "The cluster is heterogeneous; none of the soft pairs look like the same underlying bug or change, so all are rejected. No PRs are present, and there is no clear canonical duplicate center.", + "confidence": 0.92, + "canonical_issue_reason": "No strong duplicate center exists in this set; if one issue must anchor triage, #46365 is the clearest user-facing regression with broad impact and active discussion.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43873 is the best overall anchor because it is open, has substantial discussion/inbound references, and includes a concrete repro for a clearly scoped bug.", + "best_issue_reason": "Issue #46365 is the best standalone representative because it is a straightforward import regression, clearly actionable, and likely to matter to many users.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:46298", - "right": "issue:46752", - "accept": false, - "reason": "Noisy logging vs empty-list chat-template IndexError; unrelated symptoms and code paths." - }, - { - "left": "issue:44062", - "right": "issue:44117", - "accept": false, - "reason": "AddedToken keyword collision in tokenizer conversion vs TOKENIZER_MAPPING_NAMES returning None; different layers." - }, - { - "left": "issue:43097", - "right": "issue:43873", - "accept": false, - "reason": "API removal warning about tie_embeddings vs offloading/quantization behavior; not the same bug." - }, - { - "left": "issue:43012", - "right": "issue:46821", - "accept": false, - "reason": "bfloat16 compile warning vs TimesFM window_size AttributeError; different subsystems." - }, - { - "left": "issue:44117", - "right": "issue:46701", - "accept": false, - "reason": "Tokenizer mapping fallback bug vs DiffusionGemma pipeline failure; no shared code path." - }, { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "local_files_only leaking API requests vs missing embed_tokens.weight on UMT5Encoder load/save; unrelated." + "reason": "Different failures: unexpected network access with local_files_only vs missing encoder weights during from_pretrained." }, { "left": "issue:44936", "right": "issue:46365", "accept": false, - "reason": "evaluate() failing after train() vs importing Trainer blocked by datasets; runtime state vs import resolution." + "reason": "Trainer evaluation/runtime behavior is unrelated to the separate import error involving datasets." }, { "left": "issue:44704", "right": "issue:45103", "accept": false, - "reason": "AutoProcessor kwargs forwarding vs docstring parser crash with postponed annotations; unrelated." + "reason": "Kwargs propagation in AutoProcessor is unrelated to the docstring/annotations AttributeError." }, { "left": "issue:44380", "right": "issue:46421", "accept": false, - "reason": "GPT2 SDPA/FlashAttention scaling bug vs cache lazy-init recompilation; distinct performance paths." + "reason": "GPT2 attention scaling backend bug is unrelated to cache lazy initialization and recompilation." }, { "left": "issue:44938", "right": "issue:46178", "accept": false, - "reason": "Python 3.14 import failure vs invalid tensorflow_text backend mapping; different failure modes." + "reason": "Python 3.14 import/load failure is not the same as an invalid BACKENDS_MAPPING backend name." }, { "left": "issue:43986", - "right": "issue:44117", + "right": "issue:44016", "accept": false, - "reason": "AutoProcessor crash without torchvision vs tokenizer mapping None; unrelated." + "reason": "AutoProcessor/torchvision crash and notebook syntax error are completely different issues." }, { "left": "issue:42907", - "right": "issue:44655", + "right": "issue:45753", "accept": false, - "reason": "dequantized model save failure vs saving Pipeline objects; different serialization targets." + "reason": "Saving dequantized models is unrelated to a Qwen3_5 image generation loop." }, { "left": "issue:42907", - "right": "issue:45753", + "right": "issue:44655", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals vs Qwen3_5 infinite image loop; unrelated." + "reason": "Pipeline save_pretrained failure is a different save-path than dequantized model saving." }, { "left": "issue:43825", "right": "issue:44655", "accept": false, - "reason": "Translation-task warning in pipeline() vs save_pretrained on Pipeline objects; no overlap." + "reason": "Error-message wording for translation tasks is unrelated to Pipeline serialization." }, { "left": "issue:43012", "right": "issue:43550", "accept": false, - "reason": "Precision warning on bfloat16 compile vs actual torch.compile/SDPA failure; different root causes." + "reason": "A PyTorch precision warning is unrelated to a torch.compile SDPA model failure." }, { "left": "issue:45712", "right": "issue:46298", "accept": false, - "reason": "Dummy-class cleanup vs noisy logging message; unrelated maintenance issues." + "reason": "Dummy class cleanup and noisy logging are unrelated repository maintenance problems." }, { "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "xLSTM training blockers vs Bamba torch.compile/SDPA failure; different model-specific bugs." + "reason": "xLSTM training blockers are unrelated to Bamba SDPA compile failures." }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "Compile precision warning vs tie_word_embeddings warning in Qwen3.5 LoRA; different triggers and fixes." + "reason": "Float32 warning on bfloat16 compile is unrelated to tie_word_embeddings warnings." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "Missing embed_tokens.weight during UMT5Encoder load vs shard-size default question; serialization mismatch vs API default." + "reason": "Missing embedding weights on load is unrelated to save_pretrained shard-size defaults." }, { "left": "issue:43352", - "right": "issue:44380", + "right": "issue:43976", "accept": false, - "reason": "Flash Attention 2 unsupported error vs GPT2 attention scaling ignored under SDPA/FA; different model/backend issues." + "reason": "Flash Attention support in a model class is unrelated to the Python version compatibility issue." }, { "left": "issue:43352", - "right": "issue:43976", + "right": "issue:44380", "accept": false, - "reason": "Model-specific FlashAttention support error vs Python-version incompatibility; unrelated." + "reason": "Different model/backends and different symptoms; these are not the same code-path bug." }, { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "Wrong tokenizer decoder type regression vs isolated reproduction of another issue; no shared bug described." - }, - { - "left": "issue:44162", - "right": "issue:46172", - "accept": false, - "reason": "ESM2 broken model issue vs deepseek comb transpose bug; unrelated architectures and failure points." + "reason": "Tokenizer decoder type regression is not the same as a reproduction of another issue." }, { "left": "issue:43097", "right": "issue:46298", "accept": false, - "reason": "tie_embeddings removal regression vs noisy logging; no common code path." + "reason": "A removed config field is unrelated to a noisy logging message." }, { - "left": "issue:44380", - "right": "issue:46132", + "left": "issue:44162", + "right": "issue:46172", "accept": false, - "reason": "GPT2 SDPA/FlashAttention scaling bug vs AttentionInterface.register behavior change; different APIs." + "reason": "ESM2 breakage and a DeepSeek tensor transpose bug are unrelated." }, { "left": "issue:45584", "right": "issue:46365", "accept": false, - "reason": "Whisper empty transcription failure vs Trainer import conflict with datasets; unrelated." + "reason": "Whisper empty-transcription failure is unrelated to importing Trainer with datasets installed." }, { - "left": "issue:44222", - "right": "issue:45478", + "left": "issue:44077", + "right": "issue:46365", "accept": false, - "reason": "Both are MoE-related, but one is FP8 save_pretrained and the other is from_pretrained loading; different concrete problems." + "reason": "PatchTSMixer post_init validation is unrelated to the Trainer import dependency issue." }, { - "left": "issue:44077", - "right": "issue:46365", + "left": "issue:44222", + "right": "issue:45478", "accept": false, - "reason": "patchtsmixer post_init policy vs Trainer import conflict; unrelated." + "reason": "FP8 save_pretrained and Qwen3.5 MoE from_pretrained errors involve different directions and different failures." }, { "left": "issue:36683", "right": "issue:43582", "accept": false, - "reason": "Gemma3Config missing vocab_size vs Apple Silicon caching_allocator_warmup TypeError; different stacks." + "reason": "Gemma3Config vocab_size and AppleSilicon allocator warmup are unrelated." }, { "left": "issue:44380", "right": "issue:45479", "accept": false, - "reason": "Attention scaling ignored vs degenerate zero-loss classification bug; no relation." + "reason": "Attention scaling backend behavior is unrelated to degenerate zero-loss classification." }, { - "left": "issue:42175", - "right": "issue:46365", + "left": "issue:43618", + "right": "issue:44380", "accept": false, - "reason": "Packaging/backend selection issue vs importing Trainer with datasets installed; unrelated dependency problems." + "reason": "CLIPOutput attentions assignment and GPT2 attention scaling are separate model bugs." }, { "left": "issue:42757", "right": "issue:43824", "accept": false, - "reason": "Missing is_offline_mode import vs missing Qwen2_5_VLForConditionalGeneration export; both import errors but different packages and causes." + "reason": "Two different import errors for different symbols; not the same missing-export bug." + }, + { + "left": "issue:42175", + "right": "issue:46365", + "accept": false, + "reason": "Backend packaging for TensorFlow is unrelated to importing Trainer with datasets present." }, { "left": "issue:43064", "right": "issue:44380", "accept": false, - "reason": "FSDP2/PEFT optimizer-state corruption vs GPT2 attention scaling backend bug; unrelated." + "reason": "FSDP2/PEFT optimizer-state corruption is a different training bug than GPT2 attention scaling." }, { "left": "issue:42490", "right": "issue:43012", "accept": false, - "reason": "Local-path trailing-slash import warning vs bfloat16 compile warning; different loading vs compile paths." + "reason": "Path-trailing-slash import warning is unrelated to a bfloat16 precision warning." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "Chat-completions API request rejection is unrelated to Qwen2_5_VLProcessor batching crash." + }, + { + "left": "issue:44655", + "right": "issue:45103", + "accept": false, + "reason": "Pipeline serialization and docstring annotation parsing are unrelated." + }, + { + "left": "issue:45478", + "right": "issue:45479", + "accept": false, + "reason": "MoE from_pretrained regression and zero-loss classification are different model-path problems." + }, + { + "left": "issue:45584", + "right": "issue:46365", + "accept": false, + "reason": "Whisper generation edge case is unrelated to Trainer import dependency behavior." + }, + { + "left": "issue:45712", + "right": "issue:46298", + "accept": false, + "reason": "Leftover dummy classes and noisy logging do not describe the same fix." + }, + { + "left": "issue:46132", + "right": "issue:46172", + "accept": false, + "reason": "AttentionInterface registration semantics and a tensor transpose bug are not the same issue." + }, + { + "left": "issue:46178", + "right": "issue:44938", + "accept": false, + "reason": "Backend mapping validation and Python 3.14 loading failure are different import/runtime problems." } ] }, @@ -17179,7 +17304,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 56, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -17188,10 +17313,10 @@ "cluster_id": "cluster-10000-502", "nodes": [ "issue:29870", + "issue:33453", "issue:36010", "issue:36296", "issue:36683", - "issue:38617", "issue:42757", "issue:42994", "issue:43010", @@ -17200,15 +17325,12 @@ "issue:43299", "issue:43352", "issue:43475", - "issue:43502", "issue:43525", - "issue:43526", "issue:43611", "issue:43618", "issue:43756", "issue:43761", "issue:43867", - "issue:43901", "issue:43937", "issue:43976", "issue:43992", @@ -17217,279 +17339,282 @@ "issue:44077", "issue:44230", "issue:44355", - "issue:44380", "issue:44483", "issue:44485", "issue:44492", - "issue:44514", "issue:44625", "issue:44655", + "issue:44855", + "issue:44908", "issue:44938", + "issue:45071", "issue:45216", "issue:45276", "issue:45278", - "issue:45290", "issue:45310", + "issue:46050", "issue:46077", "issue:46082", "issue:46365", + "issue:46369", "issue:46829", + "issue:46851", "issue:46877" ], "soft_pairs": [ - "issue:44483|issue:44514", - "issue:43208|issue:43526", - "issue:43618|issue:44380", - "issue:38617|issue:43502", - "issue:44038|issue:44230", "issue:43756|issue:44485", - "issue:42994|issue:44655", "issue:43618|issue:43761", + "issue:44038|issue:44230", + "issue:42994|issue:44655", "issue:43299|issue:43867", "issue:36010|issue:42757", "issue:44625|issue:46365", "issue:45276|issue:46077", + "issue:43611|issue:44655", "issue:43352|issue:43937", - "issue:44483|issue:45290", + "issue:44483|issue:46369", "issue:44355|issue:46877", "issue:36683|issue:43475", - "issue:43611|issue:44655", - "issue:43901|issue:44655", - "issue:43867|issue:45216", - "issue:43976|issue:44938", + "issue:44655|issue:46050", "issue:36683|issue:44077", + "issue:43976|issue:44938", "issue:44625|issue:46077", + "issue:43867|issue:45216", "issue:46077|issue:46082", "issue:43618|issue:45216", "issue:43208|issue:46829", "issue:36683|issue:43525", "issue:36683|issue:44625", "issue:29870|issue:36296", + "issue:43010|issue:45278", "issue:43992|issue:45310", "issue:43064|issue:44492", - "issue:43010|issue:45278", - "issue:44075|issue:46365" + "issue:44075|issue:46365", + "issue:44908|issue:46851", + "issue:29870|issue:33453", + "issue:44938|issue:45071", + "issue:44355|issue:44855" ], - "prepared_review_unit_hash": "120feb8d20884d94dbfea053d9e98c37072c8b4f6d705433dc9a0634e37d4c9f", + "prepared_review_unit_hash": "123236c6e6bd81a591baa1c4477f7f0bd7bff8be9b942b250fb9978e38a82c1f", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 36970, - "estimated_input_tokens": 9243, - "estimated_eval_tokens": 18742 + "serialized_chars": 36860, + "estimated_input_tokens": 9215, + "estimated_eval_tokens": 18686 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The items are a loose collection of distinct Transformers issues rather than a true duplicate cluster. None of the soft pairs look like the same underlying bug/change with enough specificity to merge.", - "confidence": 0.97, - "canonical_issue_reason": "No clear canonical issue: the set spans unrelated regressions, import errors, model-specific config bugs, and API behavior changes.", + "summary": "This set is a loose grab bag of unrelated Transformers issues (imports, model config regressions, save/load bugs, training bugs). No PRs are present, and none of the soft-edge pairs are strong enough to merge as duplicates or the same fix.", + "confidence": 0.94, + "canonical_issue_reason": "issue:45216 is the cleanest representative: it is a concrete, versioned regression (`save_pretrained` on Qwen3.5 since 5.4.0) with a specific failure mode and likely actionable reproduction.", "canonical_pr_reason": null, - "best_issue_reason": "No single best issue for duplicate triage; the cluster does not cohere around one underlying defect.", + "best_issue_reason": "issue:45216 has the strongest issue quality overall: clear regression boundary, specific model family, and a focused save-path corruption problem.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44483", - "right": "issue:44514", - "accept": false, - "reason": "One is an OpenAI-style chat/completions API request validation problem; the other is a chat template crash on batched input with padding disabled." - }, - { - "left": "issue:43208", - "right": "issue:43526", + "left": "issue:43756", + "right": "issue:44485", "accept": false, - "reason": "xLSTM training bugs and BeitImageProcessorFast label reduction are unrelated subsystems and failure modes." + "reason": "Both mention RoPE, but they are different models and different problems (Smollm3 layer dropping vs GLM-5 RoPE implementation)." }, { "left": "issue:43618", - "right": "issue:44380", - "accept": false, - "reason": "CLIP attentions missing and GPT2 attention scaling ignored under SDPA/FlashAttention are different code paths and symptoms." - }, - { - "left": "issue:38617", - "right": "issue:43502", + "right": "issue:43761", "accept": false, - "reason": "An import error for layer_type_validation is unrelated to local_files_only still allowing network calls." + "reason": "Both are CLIP v5 regressions, but one is missing attentions and the other is missing hidden_states; likely separate bugs in the same area, not clearly one fix." }, { "left": "issue:44038", "right": "issue:44230", "accept": false, - "reason": "Both mention Qwen3-VL/Qwen3.5 MoE, but one is a loading regression and the other is fp8 inference support." - }, - { - "left": "issue:43756", - "right": "issue:44485", - "accept": false, - "reason": "Both mention RoPE, but they concern different models and different implementation issues." + "reason": "Different Qwen3-VL-Moe topics: general breakage/loading regression vs fp8 inference support." }, { "left": "issue:42994", "right": "issue:44655", "accept": false, - "reason": "Quantized model saving and Pipeline.save_pretrained are distinct save-path problems." - }, - { - "left": "issue:43618", - "right": "issue:43761", - "accept": false, - "reason": "Missing attentions in CLIPOutput and missing hidden_states in CLIPVisionModel are related but not the same concrete bug." + "reason": "Both involve saving, but one is quantized model persistence and the other is Pipeline.save_pretrained; different code paths and artifacts." }, { "left": "issue:43299", "right": "issue:43867", "accept": false, - "reason": "Qwen3VL-Moe loading failure and state_dict sorting causing load errors are different issues." + "reason": "Both are loading failures, but the symptoms and likely root causes differ (Qwen3.5 model loading vs sorted state_dict ordering)." }, { "left": "issue:36010", "right": "issue:42757", "accept": false, - "reason": "GenerationMixin import failure and is_offline_mode import failure are separate dependency/import regressions." + "reason": "Both are import errors, but for different symbols and different dependency breakages." }, { "left": "issue:44625", "right": "issue:46365", "accept": false, - "reason": "Qwen3.5 num_labels propagation and Trainer import failing because of datasets are unrelated." + "reason": "Unrelated: Qwen3.5 num_labels propagation vs Trainer import affected by datasets installation." }, { "left": "issue:45276", "right": "issue:46077", "accept": false, - "reason": "Embedding resize propagation and bidirectional attention mask construction are different Gemma4 bugs." + "reason": "Same model family, but one is token embedding resize propagation and the other is attention mask construction; different bugs." + }, + { + "left": "issue:43611", + "right": "issue:44655", + "accept": false, + "reason": "Different problem areas: base_model_prefix loading regression vs pipeline serialization." }, { "left": "issue:43352", "right": "issue:43937", "accept": false, - "reason": "Flash Attention support gating and invalid GenerationConfig validation are unrelated failures." + "reason": "Different failures: Flash Attention unsupported for a model vs invalid GenerationConfig for GLM-5." }, { "left": "issue:44483", - "right": "issue:45290", + "right": "issue:46369", "accept": false, - "reason": "Chat/completions request acceptance and assistant tool-call chat template crashes are different API surfaces." + "reason": "Completely different symptoms and subsystems: chat completions request rejection vs sam3-litetext memory overflow." }, { "left": "issue:44355", "right": "issue:46877", "accept": false, - "reason": "Compiled Python file loading errors and auto_docstring import-time stderr/stdout diagnostics are unrelated." + "reason": "Both are import-time issues, but one concerns compiled Python files and the other noisy docstring lint output." }, { "left": "issue:36683", "right": "issue:43475", "accept": false, - "reason": "Missing vocab_size on Gemma3Config and missing fpn_position_embeddings on Sam3VisionEncoderOutput are different model attribute bugs." - }, - { - "left": "issue:43611", - "right": "issue:44655", - "accept": false, - "reason": "base_model_prefix loading regression and Pipeline save_pretrained failure do not share the same underlying path." + "reason": "Different model families and attributes: Gemma3 vocab_size vs SAM3 vision output fields." }, { - "left": "issue:43901", - "right": "issue:44655", + "left": "issue:44655", + "right": "issue:46050", "accept": false, - "reason": "A docs/behavior mismatch for return_all_scores is unrelated to pipeline serialization failure." + "reason": "Both are save/load related, but one is Pipeline.save_pretrained and the other is quantized sinq round-trip behavior." }, { - "left": "issue:43867", - "right": "issue:45216", + "left": "issue:36683", + "right": "issue:44077", "accept": false, - "reason": "State_dict sorting load error and Qwen3.5 checkpoint corruption after save_pretrained are not the same bug." + "reason": "No relation beyond being model-related issues; one is missing config attribute, the other is a post_init API constraint." }, { "left": "issue:43976", "right": "issue:44938", "accept": false, - "reason": "Different Python-version compatibility regressions with no shared concrete code path." + "reason": "Both are Python-version compatibility reports, but for different interpreter versions and likely unrelated causes." }, { - "left": "issue:36683", - "right": "issue:44077", + "left": "issue:44625", + "right": "issue:46077", "accept": false, - "reason": "Gemma3Config missing vocab_size and patchtsmixer post_init policy are unrelated." + "reason": "Different Gemma4/Qwen3.5 configuration bugs with no shared failing path." }, { - "left": "issue:44625", - "right": "issue:46077", + "left": "issue:43867", + "right": "issue:45216", "accept": false, - "reason": "Qwen3.5 num_labels propagation and Gemma4 attention mask behavior are unrelated." + "reason": "Both involve load/save behavior, but one is state_dict ordering during load and the other is Qwen3.5 checkpoint corruption on save." }, { "left": "issue:46077", "right": "issue:46082", "accept": false, - "reason": "Gemma4 bidirectional attention masking and LlamaConfig head_dim validation are different bugs." + "reason": "Different configuration/validation bugs in different models (Gemma4 attention mode vs Llama head_dim validation)." }, { "left": "issue:43618", "right": "issue:45216", "accept": false, - "reason": "CLIP output attentions regression and Qwen3.5 save_pretrained regression are unrelated." + "reason": "Different model families and different outputs; CLIP output fields vs Qwen3.5 checkpoint saving." }, { "left": "issue:43208", "right": "issue:46829", "accept": false, - "reason": "xLSTM training blockers and VideoPrism hidden_states misassignment are different models and code paths." + "reason": "Different models and domains: xLSTM training blockers vs VideoPrism output field mixup." }, { "left": "issue:36683", "right": "issue:43525", "accept": false, - "reason": "Gemma3Config vocab_size and Llama4Config pad_token_id are both missing-attribute reports, but on different configs and errors." + "reason": "Both are missing config attributes, but for unrelated models and different fields." }, { "left": "issue:36683", "right": "issue:44625", "accept": false, - "reason": "Different config propagation bugs for different models and fields." + "reason": "Different missing-attribute/config propagation bugs in different model families." }, { "left": "issue:29870", "right": "issue:36296", "accept": false, - "reason": "Bart generate decoder_attention_mask UnboundLocalError and tensor-parallel training bug are unrelated." + "reason": "Different training/runtime failures: BART generate decoder mask bug vs tensor-parallel training issue." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "Unrelated: cache update no_grad decoration vs broad import errors after version upgrade." }, { "left": "issue:43992", "right": "issue:45310", "accept": false, - "reason": "UMT5Encoder missing embed_tokens.weight and Qwen3.5 MoE from_pretrained failure are distinct loading issues." + "reason": "Different Qwen3.5 MoE issues: missing embed token weights vs from_pretrained error on newer transformers." }, { "left": "issue:43064", "right": "issue:44492", "accept": false, - "reason": "Distributed optimizer-state corruption and a cache-strategy typo are unrelated." + "reason": "Completely unrelated: optimizer state corruption under FSDP/PEFT vs a cache strategy typo." }, { - "left": "issue:43010", - "right": "issue:45278", + "left": "issue:44075", + "right": "issue:46365", "accept": false, - "reason": "torch.no_grad decoration for cache updates and broad import errors after upgrade are different problems." + "reason": "Different subsystems: SGD argument handling vs Trainer import failure due to datasets." }, { - "left": "issue:44075", - "right": "issue:46365", + "left": "issue:44908", + "right": "issue:46851", + "accept": false, + "reason": "Different features: scheduler kwargs ignored vs variable-length position_ids in Qwen3.5." + }, + { + "left": "issue:29870", + "right": "issue:33453", + "accept": false, + "reason": "Different issues: BART generation mask bug vs tokenizer loading regression." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Both are version/regression complaints, but one is Python 3.14 import failure and the other is PretrainedConfig type checking." + }, + { + "left": "issue:44355", + "right": "issue:44855", "accept": false, - "reason": "SGD argument handling and Trainer import resolution are unrelated." + "reason": "Different failure modes: compiled files import issues vs Python 3.13 IndentationError from TorchScript parsing." } ] }, @@ -17505,7 +17630,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 57, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -17514,7 +17639,7 @@ "cluster_id": "cluster-10000-502", "nodes": [ "issue:28282", - "issue:29870", + "issue:30333", "issue:33453", "issue:36296", "issue:36683", @@ -17523,8 +17648,10 @@ "issue:42898", "issue:43010", "issue:43064", + "issue:43065", "issue:43502", "issue:43575", + "issue:43576", "issue:43618", "issue:43723", "issue:43824", @@ -17543,11 +17670,8 @@ "issue:44485", "issue:44556", "issue:44734", - "issue:44855", - "issue:44908", - "issue:44938", + "issue:44749", "issue:44964", - "issue:45071", "issue:45125", "issue:45357", "issue:45375", @@ -17556,28 +17680,24 @@ "issue:45561", "issue:45600", "issue:45850", + "issue:46132", "issue:46202", "issue:46285", "issue:46617", - "issue:46664", - "issue:46851" + "issue:46664" ], "soft_pairs": [ - "issue:44908|issue:46851", - "issue:29870|issue:33453", - "issue:44938|issue:45071", - "issue:44355|issue:44855", "issue:43867|issue:46202", "issue:36683|issue:43064", - "issue:43867|issue:45375", "issue:42890|issue:45561", + "issue:43867|issue:45375", "issue:43010|issue:45850", "issue:41669|issue:44246", "issue:36296|issue:43575", "issue:43010|issue:43976", "issue:43618|issue:44077", - "issue:43992|issue:45478", "issue:43867|issue:45357", + "issue:43992|issue:45478", "issue:28282|issue:33453", "issue:43618|issue:46285", "issue:44360|issue:45440", @@ -17589,219 +17709,237 @@ "issue:44734|issue:46617", "issue:43723|issue:44556", "issue:44222|issue:45125", - "issue:44387|issue:44556", "issue:43824|issue:46617", + "issue:44387|issue:44556", "issue:43502|issue:44485", - "issue:43874|issue:46617" + "issue:43874|issue:46617", + "issue:43576|issue:46132", + "issue:43010|issue:43065", + "issue:45125|issue:45478", + "issue:30333|issue:33453", + "issue:44355|issue:44749", + "issue:45125|issue:45375" ], - "prepared_review_unit_hash": "eae5e0f7f755d184a0375f28ecfd0734f8145b1b7323c21242ba6cb243ff7e53", + "prepared_review_unit_hash": "3afccda443ab027e16ff2b2db3decdcc40519b8045d48346a3df977f928b53e3", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36467, - "estimated_input_tokens": 9117, - "estimated_eval_tokens": 18490 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36428, + "estimated_input_tokens": 9107, + "estimated_eval_tokens": 18470 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is heterogeneous: tokenizer loading, generation, tensor-parallel, multimodal config, CI, and performance issues are all mixed together. None of the soft pairs look like the same underlying bug or a mergeable PR-equivalent change.", - "confidence": 0.96, - "canonical_issue_reason": "No true canonical issue: the items span unrelated bug reports, and none of the soft pairs are duplicates of the same concrete failure mode.", + "summary": "The cluster is mostly a semantic neighborhood of Transformers regressions, but the soft pairs do not look like true duplicates of the same underlying bug/change. No PRs are present. The best representative issue is a generic loading-failure report, but it does not meaningfully subsume the rest.", + "confidence": 0.18, + "canonical_issue_reason": "Issue 43867 is the broadest and most generic loading-error report in the set, but the cluster is too heterogeneous for it to be a true canonical duplicate.", "canonical_pr_reason": null, - "best_issue_reason": "If one issue must represent the set, issue:33453 ('Regression in tokenizer loading') is the broadest and most generic loading-regression report, but it still does not cover the rest of the items.", + "best_issue_reason": "Issue 43867 is the most representative umbrella-style issue by title, but it should only be treated as a loose thematic anchor, not a duplicate target for the others.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44908", - "right": "issue:46851", - "accept": false, - "reason": "Both mention generation-related behavior, but one is a scheduler kwargs bug and the other is a variable-length position_ids issue; different code paths." - }, - { - "left": "issue:29870", - "right": "issue:33453", - "accept": false, - "reason": "Bart generate() decoder_attention_mask failure is unrelated to tokenizer-loading regression." - }, - { - "left": "issue:44938", - "right": "issue:45071", - "accept": false, - "reason": "Python 3.14 import failure and PretrainedConfig type-checking regression are separate compatibility issues." - }, - { - "left": "issue:44355", - "right": "issue:44855", - "accept": false, - "reason": "Compiled Python file errors and a TorchScript parsing IndentationError are different import-time failures." - }, { "left": "issue:43867", "right": "issue:46202", "accept": false, - "reason": "State_dict sorting load failure and Gemma pipeline roundtrip failure are distinct model-loading/serialization bugs." + "reason": "Both mention model/pipeline failure, but one is a sorted state_dict loading bug and the other is a Gemma roundtrip issue; different code paths." }, { "left": "issue:36683", "right": "issue:43064", "accept": false, - "reason": "Missing Gemma3Config vocab_size and FSDP2/PEFT optimizer-state corruption are unrelated training/config problems." + "reason": "Gemma3Config vocab_size missing and FSDP2/PEFT optimizer-state corruption are unrelated bugs." }, { - "left": "issue:43867", - "right": "issue:45375", + "left": "issue:42890", + "right": "issue:45561", "accept": false, - "reason": "Loading a model with sorted state_dict and missing deepstack_visual_indexes are not the same defect." + "reason": "One is flaky SamHQ integration tests from missing set_seed; the other is an xdist file race in testing utils." }, { - "left": "issue:42890", - "right": "issue:45561", + "left": "issue:43867", + "right": "issue:45375", "accept": false, - "reason": "Missing set_seed in SAM HQ tests is a flaky test issue; xdist race on captured_info.txt is a different test harness race." + "reason": "Sorted state_dict loading error and missing Qwen3.5 vision config field are distinct failures." }, { "left": "issue:43010", "right": "issue:45850", "accept": false, - "reason": "Decorating cache update with no_grad and a broad 'v5 breaks too many models' complaint are not the same concrete bug." + "reason": "torch.no_grad on cache updates and a broad 'many models break' complaint are not the same bug." }, { "left": "issue:41669", "right": "issue:44246", "accept": false, - "reason": "Removing import * for import-time cost and intermittent import slowness are related only at a high level, not the same bug." + "reason": "Both concern import time, but one attributes a specific import-star slowdown while the other is a general intermittent slow import report." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Tensor parallel training bug and Qwen2-57B-A14B-Instruct OOM on tp are different failure modes." + "reason": "Tensor-parallel training bug and tp-related OOM while loading Qwen2-57B are different symptoms and likely different paths." }, { "left": "issue:43010", "right": "issue:43976", "accept": false, - "reason": "Torch no_grad decoration for caches is unrelated to Python-version support for Transformers 5.1.0." + "reason": "Cache update no_grad and Python version compatibility are unrelated." }, { "left": "issue:43618", "right": "issue:44077", "accept": false, - "reason": "CLIPOutput attentions regression and patchtsmixer post_init allowance are separate model API issues." + "reason": "CLIPOutput attentions regression and patchtsmixer post_init validation are unrelated model issues." }, { - "left": "issue:43992", - "right": "issue:45478", + "left": "issue:43867", + "right": "issue:45357", "accept": false, - "reason": "UMT5Encoder missing embed_tokens.weight and Qwen3.5 Moe from_pretrained errors affect different model families and code paths." + "reason": "Loading sorted state_dict and incorrect visual encoder keys in save_pretrained are different bugs." }, { - "left": "issue:43867", - "right": "issue:45357", + "left": "issue:43992", + "right": "issue:45478", "accept": false, - "reason": "Sorted state_dict loading failure is not the same as incorrect visual encoder keys being saved." + "reason": "Missing embed_tokens.weight in UMT5Encoder and Qwen3.5 Moe from_pretrained failure are different model-loading issues." }, { "left": "issue:28282", "right": "issue:33453", "accept": false, - "reason": "Missing PyTorch dependency ImportError and tokenizer-loading regression are unrelated startup failures." + "reason": "PyTorch-missing import error and tokenizer-loading regression are unrelated." }, { "left": "issue:43618", "right": "issue:46285", "accept": false, - "reason": "Missing CLIPOutput attentions and CLIPTextModel structural inconsistency are distinct CLIP API regressions." + "reason": "These are both CLIP-related, but one is attentions assignment and the other is class-structure inconsistency." }, { "left": "issue:44360", "right": "issue:45440", "accept": false, - "reason": "A DSA indexer activation bug and DeepseekV3MoE divergence are different implementation concerns." + "reason": "DSA indexer ReLU discussion and DeepSeekV3MoE implementation divergence are unrelated." }, { "left": "issue:42890", "right": "issue:44964", "accept": false, - "reason": "Flaky test seeding and Phi-4-multimodal loading failure are unrelated." + "reason": "Test seeding flake and Phi-4 multimodal loading failure are unrelated." }, { "left": "issue:42890", "right": "issue:44112", "accept": false, - "reason": "Seed-related test flakiness and GraniteSpeech stale device override failure are separate CI issues." + "reason": "Different CI/test failures: missing seed versus stale device override." }, { "left": "issue:42890", "right": "issue:45600", "accept": false, - "reason": "Missing set_seed in a test and a broken auto_mappings reference are not the same bug." + "reason": "Test flakiness and auto_mappings references to removed configs are not the same issue." }, { "left": "issue:42898", "right": "issue:46664", "accept": false, - "reason": "Tokenization-space cleanup behavior changes and SmolVLM num_frames semantics are unrelated model/tokenizer issues." + "reason": "Tokenizer clean_up_tokenization_spaces behavior and SmolVLM num_frames semantics are unrelated." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "Unexpected API requests despite local_files_only and AddedToken special-arg duplication are different bug classes." + "reason": "local_files_only still triggering API calls and AddedToken keyword duplication are different subsystems." }, { "left": "issue:44734", "right": "issue:46617", "accept": false, - "reason": "KV-cache continuation tensor indexing and cache FileNotFoundError are unrelated serve/load problems." + "reason": "Serve KV-cache continuation indexing bug and trust_remote_code cache FileNotFoundError are different failure modes." }, { "left": "issue:43723", "right": "issue:44556", "accept": false, - "reason": "Tokenizer loading regression and checkpoint reload failure after version upgrade are not the same root cause." + "reason": "AutoTokenizer.from_pretrained regression and checkpoint reload incompatibility after upgrade are distinct load-path problems." }, { "left": "issue:44222", "right": "issue:45125", "accept": false, - "reason": "FP8 save_pretrained for moe and missing _tp_plan for tensor parallelism are separate model-export and parallelism issues." + "reason": "FP8 save_pretrained on MoE and missing _tp_plan for tensor parallelism are different issues." }, { - "left": "issue:44387", - "right": "issue:44556", + "left": "issue:43824", + "right": "issue:46617", "accept": false, - "reason": "Int4 quantization reserved-memory growth and checkpoint reload incompatibility are different regressions." + "reason": "ImportError for Qwen2_5_VLForConditionalGeneration and cache FileNotFoundError are unrelated." }, { - "left": "issue:43824", - "right": "issue:46617", + "left": "issue:44387", + "right": "issue:44556", "accept": false, - "reason": "Missing Qwen2_5_VLForConditionalGeneration import and trust_remote_code cache FileNotFoundError are distinct loading failures." + "reason": "Int4 quantization memory regression and model reload failure after version upgrade are not the same bug." }, { "left": "issue:43502", "right": "issue:44485", "accept": false, - "reason": "Local-files-only network leakage and GLM-5 RoPE implementation discussion do not describe the same bug." + "reason": "Remote-access leakage with local_files_only and GLM-5 RoPE implementation are unrelated." }, { "left": "issue:43874", "right": "issue:46617", "accept": false, - "reason": "Missing GLM46V image-processor method and trust_remote_code cache FileNotFoundError are unrelated errors." + "reason": "Missing image-patch helper in GLM46V and trust_remote_code cache loading failure are different code paths." + }, + { + "left": "issue:43576", + "right": "issue:46132", + "accept": false, + "reason": "Broken transformers env command and AttentionInterface.register behavior change are unrelated." + }, + { + "left": "issue:43010", + "right": "issue:43065", + "accept": false, + "reason": "Cache update no_grad and a dummy Conv2d in Sam3PixelDecoder are unrelated." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5 Moe, but one is missing _tp_plan and the other is a broader from_pretrained error; not enough to treat as the same bug." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLFlow reporting on failed training jobs and tokenizer loading regression are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Compiled Python file errors and post-upgrade filtering slowdown are different issues." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Missing _tp_plan for tensor parallelism and missing deepstack_visual_indexes in vision config are different config-field bugs." } ] }, @@ -17817,7 +17955,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 58, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -17837,7 +17975,6 @@ "issue:42548", "issue:42757", "issue:43010", - "issue:43065", "issue:43232", "issue:43299", "issue:43475", @@ -17849,6 +17986,7 @@ "issue:43901", "issue:44222", "issue:44230", + "issue:44273", "issue:44315", "issue:44351", "issue:44355", @@ -17856,16 +17994,17 @@ "issue:44485", "issue:44492", "issue:44556", - "issue:44749", + "issue:44623", "issue:44908", "issue:44938", "issue:44961", + "issue:45003", "issue:45071", - "issue:45125", - "issue:45375", + "issue:45230", + "issue:45362", "issue:45468", - "issue:45478", - "issue:46132", + "issue:45542", + "issue:45850", "issue:46143", "issue:46144", "issue:46209", @@ -17874,287 +18013,262 @@ "issue:46726" ], "soft_pairs": [ - "issue:43576|issue:46132", - "issue:43010|issue:43065", - "issue:45125|issue:45478", - "issue:30333|issue:33453", - "issue:42548|issue:44351", - "issue:44355|issue:44749", - "issue:45125|issue:45375", + "issue:30990|issue:31795", + "issue:43010|issue:43232", "issue:46664|issue:46726", - "issue:43502|issue:44222", + "issue:42548|issue:44351", "issue:43502|issue:44961", - "issue:43010|issue:43232", - "issue:30990|issue:31795", + "issue:43502|issue:44222", "issue:43827|issue:44961", "issue:30333|issue:36296", "issue:43299|issue:44556", "issue:29870|issue:31795", "issue:44355|issue:45071", "issue:33453|issue:34689", - "issue:42757|issue:43576", "issue:43576|issue:45071", + "issue:42757|issue:43576", "issue:44315|issue:44360", "issue:31795|issue:38617", - "issue:44961|issue:45468", "issue:43576|issue:43824", + "issue:44961|issue:45468", "issue:44485|issue:44961", "issue:44351|issue:46617", + "issue:36296|issue:43901", "issue:44908|issue:44961", "issue:43576|issue:44938", - "issue:36296|issue:43901", - "issue:46143|issue:46144", "issue:43475|issue:46617", + "issue:46143|issue:46144", "issue:44222|issue:44230", - "issue:41628|issue:42548", "issue:44315|issue:46209", + "issue:41628|issue:42548", "issue:44485|issue:46664", - "issue:43673|issue:44492" + "issue:43673|issue:44492", + "issue:45362|issue:45850", + "issue:44273|issue:45542", + "issue:44623|issue:45230", + "issue:44360|issue:46664", + "issue:44273|issue:45003" ], - "prepared_review_unit_hash": "04f26c5e8625ed334e560f67f49beef12674f5b59269f77fed1d4442c52a8507", + "prepared_review_unit_hash": "a52de9ea654dac04b21cd4e8ae507eec19b5be6c0cec18c61b9471051f141186", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 36, - "serialized_chars": 37270, - "estimated_input_tokens": 9318, - "estimated_eval_tokens": 18892 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 37295, + "estimated_input_tokens": 9324, + "estimated_eval_tokens": 18904 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of unrelated Transformers issues with superficial keyword overlap around loading, imports, caching, generation, and model families. I do not see any true duplicate-equivalent pair to merge.", - "confidence": 0.96, - "canonical_issue_reason": null, + "summary": "The set is very noisy and mostly contains distinct bugs rather than true duplicates; the closest shared theme is v5-era loading/import regressions, but there is no single strong duplicate cluster.", + "confidence": 0.74, + "canonical_issue_reason": "issue:45850 is the broadest umbrella report in the set and best matches the common v5 breakage/loading-regression theme, though it is still only a loose representative because the cluster is heterogeneous.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:45850 is the most representative issue for the overall cluster because it is the most general user-facing report of widespread v5 regressions.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43576", - "right": "issue:46132", + "left": "issue:30990", + "right": "issue:31795", "accept": false, - "reason": "Both mention API/behavior changes, but one is the v5 env CLI and the other is AttentionInterface.register; different code paths and symptoms." + "reason": "SentenceTransformers loading gets stuck is a runtime hang; the other is a documentation clarity issue about forward arguments." }, { "left": "issue:43010", - "right": "issue:43065", - "accept": false, - "reason": "Cache update decoration in generation is unrelated to the SAM3PixelDecoder dummy Conv2d issue." - }, - { - "left": "issue:45125", - "right": "issue:45478", - "accept": false, - "reason": "Both involve Qwen3.5 MoE loading, but one is missing tensor-parallel plan metadata while the other is a from_pretrained failure on transformers>=5.4.0; not the same bug." - }, - { - "left": "issue:30333", - "right": "issue:33453", - "accept": false, - "reason": "MLFlow reporting on failed training is unrelated to tokenizer loading regressions." - }, - { - "left": "issue:42548", - "right": "issue:44351", - "accept": false, - "reason": "These are different import failures for different symbols; same broad symptom, not the same underlying breakage." - }, - { - "left": "issue:44355", - "right": "issue:44749", - "accept": false, - "reason": "Compiled Python file errors and post-upgrade data-filtering slowdown are unrelated." - }, - { - "left": "issue:45125", - "right": "issue:45375", + "right": "issue:43232", "accept": false, - "reason": "Both are Qwen3.5 MoE-related, but one is tensor-parallel planning and the other is a missing config field; different fixes." + "reason": "Both mention generation/cache internals, but one is about no_grad decoration and the other about sync_gpus/model kwargs handling; different bugs." }, { "left": "issue:46664", "right": "issue:46726", "accept": false, - "reason": "Video num_frames semantics and missing-vision-backend processor errors are different problems." + "reason": "Different failures in different processor paths: video num_frames semantics vs missing vision-backend handling in AutoProcessor." }, { - "left": "issue:43502", - "right": "issue:44222", + "left": "issue:42548", + "right": "issue:44351", "accept": false, - "reason": "Local-files-only network leakage is unrelated to FP8 save_pretrained behavior." + "reason": "Both are import errors, but they concern different missing exports and likely different breakages." }, { "left": "issue:43502", "right": "issue:44961", "accept": false, - "reason": "The local_files_only bug is unrelated to the 'racoon' issue." - }, - { - "left": "issue:43010", - "right": "issue:43232", - "accept": false, - "reason": "Both touch generation/cache handling, but one is no_grad decoration for cache updates and the other is sync_gpus/_update_model_kwargs ordering; not the same concrete fix." + "reason": "API requests despite local_files_only is unrelated to the unrelated one-word issue 'racoon'." }, { - "left": "issue:30990", - "right": "issue:31795", + "left": "issue:43502", + "right": "issue:44222", "accept": false, - "reason": "Stuck loading Sentence Transformers is unrelated to confusing forward-method documentation." + "reason": "One is about unwanted network access during loading; the other is an FP8 save_pretrained issue for moe models." }, { "left": "issue:43827", "right": "issue:44961", "accept": false, - "reason": "Docs still referencing pipeline() has nothing to do with 'racoon'." + "reason": "Docs referencing pipeline() after removals is unrelated to the unrelated 'racoon' issue." }, { "left": "issue:30333", "right": "issue:36296", "accept": false, - "reason": "MLFlow failure reporting and tensor-parallel training bugs are unrelated." + "reason": "MLflow reporting on failed jobs is unrelated to a tensor-parallel training bug." }, { "left": "issue:43299", "right": "issue:44556", "accept": false, - "reason": "Both are version-upgrade loading regressions, but they affect different models and failure modes; too broad to be the same bug." + "reason": "Both are version-related loading regressions, but they affect different model families and code paths; not the same concrete bug." }, { "left": "issue:29870", "right": "issue:31795", "accept": false, - "reason": "Bart generate UnboundLocalError is unrelated to documentation confusion about input_ids/past_key_values." + "reason": "Generation UnboundLocalError is a runtime bug; the other is documentation confusion." }, { "left": "issue:44355", "right": "issue:45071", "accept": false, - "reason": "Compiled Python-file errors are unrelated to PretrainedConfig type-checking breakage." + "reason": "Compiled Python file loading errors are unrelated to PreTrainedModel type-checking regressions." }, { "left": "issue:33453", "right": "issue:34689", "accept": false, - "reason": "Tokenizer loading regression and Llama 3.2 vision model loading are both load failures, but not the same code path or fix." + "reason": "Both are loading regressions, but one is tokenizer loading and the other is a specific Llama 3.2 vision model-loading failure." }, { - "left": "issue:42757", - "right": "issue:43576", + "left": "issue:43576", + "right": "issue:45071", "accept": false, - "reason": "An import from huggingface_hub and the transformers env CLI are separate issues." + "reason": "Broken env command in v5 is unrelated to PreTrainedModel type checking." }, { - "left": "issue:43576", - "right": "issue:45071", + "left": "issue:42757", + "right": "issue:43576", "accept": false, - "reason": "The env CLI breakage is unrelated to PretrainedConfig typing." + "reason": "Missing is_offline_mode from huggingface_hub is a dependency/export issue, not the env command bug." }, { "left": "issue:44315", "right": "issue:44360", "accept": false, - "reason": "Liger Kernel/model_init behavior has nothing to do with the DSA indexer ReLU discussion." + "reason": "Liger kernel not applied in model_init and the DSA indexer ReLU discussion are unrelated code paths." }, { "left": "issue:31795", "right": "issue:38617", "accept": false, - "reason": "Model.forward docs confusion is unrelated to the layer_type_validation import error." + "reason": "Documentation clarification is unrelated to the layer_type_validation import error." }, { - "left": "issue:44961", - "right": "issue:45468", + "left": "issue:43576", + "right": "issue:43824", "accept": false, - "reason": "These are unrelated model-specific issues with no shared bug path." + "reason": "Broken env command and missing Qwen2_5_VLForConditionalGeneration import are different failures." }, { - "left": "issue:43576", - "right": "issue:43824", + "left": "issue:44961", + "right": "issue:45468", "accept": false, - "reason": "The env command regression is unrelated to missing Qwen2.5-VL exports." + "reason": "Unrelated issues with no shared code path or symptom." }, { "left": "issue:44485", "right": "issue:44961", "accept": false, - "reason": "GLM-5 RoPE implementation and 'racoon' are unrelated." + "reason": "GLM-5 RoPE implementation is unrelated to the unrelated 'racoon' report." }, { "left": "issue:44351", "right": "issue:46617", "accept": false, - "reason": "HybridCache import failure and trust_remote_code cache FileNotFoundError are both cache-related only at a very high level; different failures and fixes." + "reason": "Importing HybridCache and trust_remote_code cache FileNotFoundError are different loading failures with different causes." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor-parallel training and pipeline doc behavior are unrelated." }, { "left": "issue:44908", "right": "issue:44961", "accept": false, - "reason": "Scheduler kwargs handling is unrelated to 'racoon'." + "reason": "Scheduler kwargs handling is unrelated to the unrelated 'racoon' issue." }, { "left": "issue:43576", "right": "issue:44938", "accept": false, - "reason": "The env CLI bug is different from Python 3.14 compatibility/loading failure." + "reason": "Broken env command and Python 3.14 load failure are different environment-related problems." }, { - "left": "issue:36296", - "right": "issue:43901", + "left": "issue:43475", + "right": "issue:46617", "accept": false, - "reason": "Tensor parallel training and TextClassificationPipeline docs are unrelated." + "reason": "SAM3 video encoder attribute error is unrelated to trust_remote_code cache loading." }, { "left": "issue:46143", "right": "issue:46144", "accept": false, - "reason": "Both are RoFormer-related, but one is kwargs propagation and the other is attention-interface wiring; not the same concrete bug." - }, - { - "left": "issue:43475", - "right": "issue:46617", - "accept": false, - "reason": "SAM3 video output attributes and trust_remote_code cache loading are unrelated." + "reason": "Both concern RoFormer, but one is kwargs forwarding and the other is attention-interface usage; not the same concrete bug." }, { "left": "issue:44222", "right": "issue:44230", "accept": false, - "reason": "Both are FP8/MoE-related, but one is save_pretrained behavior and the other is inference support on specific Qwen3-VL/Qwen3.5 models." + "reason": "FP8 save_pretrained for moe and FP8 inference support are related but distinct save vs inference issues." }, { - "left": "issue:41628", - "right": "issue:42548", + "left": "issue:44315", + "right": "issue:46209", "accept": false, - "reason": "Different missing imports for different symbols; same broad packaging symptom, not the same underlying change." + "reason": "Liger kernel application and FP8 inference on B200 are unrelated." }, { - "left": "issue:44315", - "right": "issue:46209", + "left": "issue:41628", + "right": "issue:42548", "accept": false, - "reason": "Model_init/Liger Kernel behavior is unrelated to FP8 inference on B200." + "reason": "Different missing imports from transformers; no evidence they stem from one exact export regression." }, { "left": "issue:44485", "right": "issue:46664", "accept": false, - "reason": "GLM-5 RoPE implementation and SmolVLM num_frames semantics are unrelated." + "reason": "GLM-5 RoPE implementation and SmolVLM video frame semantics are unrelated." }, { "left": "issue:43673", "right": "issue:44492", "accept": false, - "reason": "Chunked_prefill cache regression and a typo in Cache strategies are not the same bug." + "reason": "Generation cache missing in chunked_prefill is unrelated to a cache-strategy typo." + }, + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "Qwen3.5 chat crash is a specific model/chat-path bug; 45850 is a broad meta complaint about many v5 regressions." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading behavior and missing TensorFlow backend handling are different problems." } ] }, @@ -18170,7 +18284,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 59, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -18185,6 +18299,7 @@ "issue:42890", "issue:43352", "issue:43408", + "issue:43519", "issue:43525", "issue:43572", "issue:43576", @@ -18201,334 +18316,331 @@ "issue:44351", "issue:44360", "issue:44393", + "issue:44485", "issue:44492", "issue:44556", - "issue:44623", + "issue:44655", "issue:44704", "issue:44829", "issue:44861", "issue:44908", - "issue:45003", "issue:45071", "issue:45092", "issue:45125", "issue:45230", "issue:45278", + "issue:45310", "issue:45341", - "issue:45362", "issue:45468", "issue:45478", - "issue:45542", "issue:45753", "issue:45850", "issue:46095", "issue:46218", "issue:46328", - "issue:46664" + "issue:46655", + "issue:46894" ], "soft_pairs": [ - "issue:45362|issue:45850", - "issue:44273|issue:45542", - "issue:44623|issue:45230", - "issue:44360|issue:46664", "issue:44038|issue:45125", - "issue:44273|issue:45003", "issue:44393|issue:45071", "issue:44393|issue:45753", - "issue:43576|issue:44556", "issue:41669|issue:43673", + "issue:43576|issue:44556", "issue:43525|issue:45850", - "issue:44556|issue:45092", "issue:43572|issue:45850", + "issue:44556|issue:45092", "issue:44492|issue:45850", "issue:45278|issue:45850", - "issue:44230|issue:46095", "issue:42371|issue:44492", - "issue:41669|issue:46218", + "issue:44230|issue:46095", "issue:43408|issue:44315", - "issue:42890|issue:44908", + "issue:41669|issue:46218", "issue:44246|issue:44273", + "issue:42890|issue:44908", "issue:43576|issue:44908", "issue:43352|issue:44393", "issue:43976|issue:44393", "issue:44908|issue:45071", "issue:44295|issue:45230", - "issue:44297|issue:45230", "issue:44704|issue:44908", + "issue:44297|issue:45230", "issue:44492|issue:46328", "issue:41669|issue:44351", "issue:41669|issue:45341", "issue:43976|issue:45850", - "issue:44393|issue:45478", "issue:43976|issue:46218", + "issue:44393|issue:45478", "issue:44077|issue:44861", "issue:31795|issue:33453", "issue:43576|issue:44273", "issue:44315|issue:44829", - "issue:44360|issue:45468" + "issue:43519|issue:46655", + "issue:44360|issue:45468", + "issue:44273|issue:44908", + "issue:45468|issue:46894", + "issue:45230|issue:45310", + "issue:43576|issue:46218", + "issue:44655|issue:44861", + "issue:43519|issue:44485", + "issue:43519|issue:44908" ], - "prepared_review_unit_hash": "13522b7cc5d95e87be18367472c393c795da0f735775c970a1db2d6e54506e05", + "prepared_review_unit_hash": "a3ff16ac0ae5b78f6a0dbfb83c29a086855dc70504176ef83f31885a78d02404", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 39, - "serialized_chars": 38226, - "estimated_input_tokens": 9557, - "estimated_eval_tokens": 19370 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 42, + "serialized_chars": 39666, + "estimated_input_tokens": 9917, + "estimated_eval_tokens": 20090 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of distinct Transformers v5 regressions and model-specific bugs, not a single duplicate thread. The soft pairs share only broad topical overlap (imports, configs, cache, model families) and should not be merged as duplicates.", + "summary": "This cluster is mostly heterogeneous: docs, tokenizer/loading, config-attribute, Qwen/vision, caching, and training issues mixed together. The only loose common thread is broad v5 regression complaints, so there is no strong duplicate set; issue:45850 is just the closest umbrella among them.", "confidence": 0.92, - "canonical_issue_reason": "No safe canonical issue: the items are largely unrelated, with only superficial overlap around v5 regressions and import/config issues.", + "canonical_issue_reason": "issue:45850 is the closest umbrella for the v5-breakage complaints, but the cluster is too heterogeneous for a true duplicate canonical.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45850 is the broadest umbrella here ('Since 5.0 version it breaks too many models'), so it is the closest representative of the cluster, but it is still too generic to absorb the other issues as duplicates.", + "best_issue_reason": "issue:45850 is the broadest and most representative issue here, but it is still too vague to act as a precise duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45362", - "right": "issue:45850", - "accept": false, - "reason": "Specific chat crash vs a vague general complaint about v5 breakage; not the same bug." - }, - { - "left": "issue:44273", - "right": "issue:45542", - "accept": false, - "reason": "Lazy loading regression and TensorBoard/TF backend error are unrelated failures." - }, - { - "left": "issue:44623", - "right": "issue:45230", - "accept": false, - "reason": "Processor save/export missing files vs a generic bug report; no shared code-path." - }, - { - "left": "issue:44360", - "right": "issue:46664", - "accept": false, - "reason": "DSA indexer ReLU discussion and video processor num_frames semantics are unrelated." - }, { "left": "issue:44038", "right": "issue:45125", "accept": false, - "reason": "Same model family, but Qwen3-VL-Moe bug and missing tensor-parallel plan are different concrete issues." - }, - { - "left": "issue:44273", - "right": "issue:45003", - "accept": false, - "reason": "Both touch loading/import behavior, but lazy loading failure and unsafe sys.modules access are different bugs." + "reason": "Different bugs: Qwen3-VL FP8/inference support vs missing `_tp_plan` for Qwen3.5 MoE tensor parallelism." }, { "left": "issue:44393", "right": "issue:45071", "accept": false, - "reason": "Model output hallucination and PretrainedConfig type-checking regression are unrelated." + "reason": "Qwen3-VL 2D bbox hallucination is unrelated to `PretrainedConfig` type-checking regressions." }, { "left": "issue:44393", "right": "issue:45753", "accept": false, - "reason": "Both are Qwen3-VL/Qwen3.5 issues, but one is bbox output hallucination and the other is an infinite loop on a specific image." + "reason": "Same family (Qwen3-VL/Qwen3.5) but different failures: bbox output issue vs infinite-loop on one image." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import-star cleanup/performance issue vs missing GenerationMixin cache in chunked prefill; different code paths." }, { "left": "issue:43576", "right": "issue:44556", "accept": false, - "reason": "v5 compatibility complaints, but 'transformers env' broken and checkpoint reload failure are different paths." + "reason": "Both are v5 complaints, but one is the `transformers env` command and the other is checkpoint reload compatibility." }, { - "left": "issue:41669", - "right": "issue:43673", + "left": "issue:43525", + "right": "issue:45850", "accept": false, - "reason": "Import-star slowdown and missing GenerationMixin cache during chunked prefill are unrelated." + "reason": "Specific `pad_token_id` attribute bug vs a vague broad v5-breakage report; not the same defect." }, { - "left": "issue:43525", + "left": "issue:43572", "right": "issue:45850", "accept": false, - "reason": "Llama4Config missing pad_token_id is a specific config bug, not the same as the broad v5-breakage complaint." + "reason": "Concrete StableLm `pad_token_idx` regression vs generic v5 breakage; too broad to be the same issue." }, { "left": "issue:44556", "right": "issue:45092", "accept": false, - "reason": "Both concern upgrade compatibility, but checkpoint reload and remote-code meta-init failures are distinct problems." + "reason": "Both involve v5 migration, but one is checkpoint reload failure and the other is old remote-code/meta-init incompatibility." }, { - "left": "issue:43572", + "left": "issue:44492", "right": "issue:45850", "accept": false, - "reason": "StableLmConfig pad_token_idx bug is specific and separate from the generic v5 regression report." + "reason": "Cache-strategy typo is unrelated to broad model breakage reports." }, { - "left": "issue:44492", + "left": "issue:45278", "right": "issue:45850", "accept": false, - "reason": "Cache-strategy typo is a narrow issue, not the same as broad model breakage." + "reason": "General import errors after upgrade is not the same as the broad 'too many models break' complaint." }, { - "left": "issue:45278", - "right": "issue:45850", + "left": "issue:42371", + "right": "issue:44492", "accept": false, - "reason": "Many import errors after upgrade is broader, but still not the same concrete issue as the generic v5 complaint." + "reason": "TF32 API warning is unrelated to a cache-strategy typo." }, { "left": "issue:44230", "right": "issue:46095", "accept": false, - "reason": "FP8 inference support request and weight conversion script question are different topics." + "reason": "FP8 Qwen3-VL/Qwen3.5 inference support vs DeepSeek weight-conversion tooling; different change requests." }, { - "left": "issue:42371", - "right": "issue:44492", + "left": "issue:43408", + "right": "issue:44315", "accept": false, - "reason": "TF32 settings guidance and cache-strategy typo are unrelated." + "reason": "Model-type warning for sam3_tracker vs Liger Kernel application with `model_init`; unrelated." }, { "left": "issue:41669", "right": "issue:46218", "accept": false, - "reason": "Both mention imports, but import-star cleanup and a missing Qwen3_5 symbol are different failure modes." + "reason": "Import-star performance cleanup vs missing `Qwen3_5ForCausalLM` import/export; different problems." }, { - "left": "issue:43408", - "right": "issue:44315", + "left": "issue:44246", + "right": "issue:44273", "accept": false, - "reason": "Wrong model type warning and Liger Kernel not applied with model_init are unrelated." + "reason": "Both mention import/lazy loading, but one is slow import sometimes and the other is a specific lazy-loading bug; not enough to unify." }, { "left": "issue:42890", "right": "issue:44908", "accept": false, - "reason": "Missing set_seed in tests and inverse_sqrt scheduler kwargs are unrelated." - }, - { - "left": "issue:44246", - "right": "issue:44273", - "accept": false, - "reason": "Import slowdown and lazy loading malfunction are different bugs." + "reason": "Integration-test seeding failures are unrelated to inverse_sqrt scheduler kwargs handling." }, { "left": "issue:43576", "right": "issue:44908", "accept": false, - "reason": "Broken env command and scheduler kwargs handling do not share a code path." + "reason": "Broken env CLI and scheduler kwargs are separate features with no shared code-path bug." }, { "left": "issue:43352", "right": "issue:44393", "accept": false, - "reason": "Flash Attention support error and Qwen3-VL bbox hallucination are unrelated." + "reason": "Flash Attention 2 support for Nemotron is unrelated to Qwen3-VL bounding-box output errors." }, { "left": "issue:43976", "right": "issue:44393", "accept": false, - "reason": "Python version compatibility and Qwen3-VL output errors are unrelated." + "reason": "Python-version compatibility regression vs Qwen3-VL output bug; different scope and cause." }, { "left": "issue:44908", "right": "issue:45071", "accept": false, - "reason": "Scheduler kwargs ignored and PretrainedConfig type checking are separate problems." + "reason": "Scheduler kwargs ignored is unrelated to `PretrainedConfig` type-checking breaks." }, { "left": "issue:44295", "right": "issue:45230", "accept": false, - "reason": "position_ids buffer access bug and generic bug report are not duplicates." + "reason": "Position_ids buffer access bug is too specific to match a generic 'Bug report' entry." }, { - "left": "issue:44297", - "right": "issue:45230", + "left": "issue:44704", + "right": "issue:44908", "accept": false, - "reason": "tokenizer_class mismatch on save and generic bug report are not the same issue." + "reason": "AutoProcessor kwargs forwarding issue vs scheduler kwargs bug; unrelated components." }, { - "left": "issue:44704", - "right": "issue:44908", + "left": "issue:44297", + "right": "issue:45230", "accept": false, - "reason": "AutoProcessor cached_file kwargs bug and scheduler kwargs bug affect different components." + "reason": "Tokenizer save metadata mismatch is not the same as an unspecified generic bug report." }, { "left": "issue:44492", "right": "issue:46328", "accept": false, - "reason": "Cache-strategy typo and feature-extractor padding inefficiency are unrelated." + "reason": "Cache-strategy typo and feature-extractor padding performance are different areas." }, { "left": "issue:41669", "right": "issue:44351", "accept": false, - "reason": "Import-star cleanup and cannot-import-HybridCache are both import-related but not the same bug." + "reason": "Import-star removal/performance vs missing `HybridCache` symbol export; different root causes." }, { "left": "issue:41669", "right": "issue:45341", "accept": false, - "reason": "Import performance issue and a testing_utils bug are unrelated." + "reason": "Import hygiene/performance issue vs a testing utility bug; no shared defect." }, { "left": "issue:43976", "right": "issue:45850", "accept": false, - "reason": "Python 3.9+/3.10+ incompatibility is a distinct release-compatibility bug, not the generic v5 breakage report." + "reason": "Python 3.9/3.10 compatibility is a platform issue, not the same as generic v5 model breakage." }, { - "left": "issue:44393", - "right": "issue:45478", + "left": "issue:43976", + "right": "issue:46218", "accept": false, - "reason": "Qwen3-VL bbox error and Qwen3.5-Moe from_pretrained error are different model-specific issues." + "reason": "Platform compatibility regression and missing model import are unrelated." }, { - "left": "issue:43976", - "right": "issue:46218", + "left": "issue:44393", + "right": "issue:45478", "accept": false, - "reason": "Python version issue and missing Qwen3_5 import are unrelated." + "reason": "Different Qwen issues: bbox hallucination in Qwen3-VL vs Qwen3.5 MoE `from_pretrained` error." }, { "left": "issue:44077", "right": "issue:44861", "accept": false, - "reason": "Optional post_init policy for patchtsmixer and tied-weights key crash are unrelated." + "reason": "Model `post_init` policy validation vs tied-weights helper crashing on list input; different code paths." }, { "left": "issue:31795", "right": "issue:33453", "accept": false, - "reason": "Documentation confusion and tokenizer loading regression are not duplicates." + "reason": "Documentation confusion about forward args is unrelated to tokenizer loading regression." }, { "left": "issue:43576", "right": "issue:44273", "accept": false, - "reason": "Broken env command and lazy loading regression are different failures, even if both are v5-related." + "reason": "Broken CLI command and lazy-loading failure are not the same bug." }, { "left": "issue:44315", "right": "issue:44829", "accept": false, - "reason": "Liger Kernel application issue and flash_attention_3 training degeneration are different code paths." + "reason": "Liger Kernel not applied in `model_init` is unrelated to flash_attention_3 causing degenerate training." + }, + { + "left": "issue:43519", + "right": "issue:46655", + "accept": false, + "reason": "Both are processor math bugs, but timestamps in Qwen3VL and padding in SmolVLM are different underlying defects." }, { "left": "issue:44360", "right": "issue:45468", "accept": false, - "reason": "DSA indexer discussion and Gemma-4 audio positional encoding bug are unrelated." + "reason": "DSA indexer ReLU discussion is unrelated to Gemma-4 audio positional encoding." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading/import slowness is unrelated to scheduler keyword forwarding." + }, + { + "left": "issue:45468", + "right": "issue:46894", + "accept": false, + "reason": "Different model families and failures: Gemma-4 audio positional encoding vs CohereASR training-loss shift bug." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "A generic bug placeholder is not enough to match the specific Qwen3.5 MoE `from_pretrained` failure." } ] }, @@ -18544,7 +18656,7 @@ "component_index": 1, "component_count": 24, "review_unit_index": 60, - "review_unit_count": 61, + "review_unit_count": 60, "status": "reviewed", "reason": null, "source": "llm", @@ -18559,35 +18671,37 @@ "issue:31795", "issue:34689", "issue:36296", + "issue:37934", "issue:38617", "issue:41669", "issue:42371", "issue:42913", + "issue:42936", "issue:43066", + "issue:43240", "issue:43298", "issue:43352", "issue:43502", "issue:43519", "issue:43525", "issue:43572", - "issue:43576", + "issue:43606", "issue:43618", + "issue:43668", "issue:43704", "issue:43717", + "issue:43813", + "issue:44031", "issue:44230", - "issue:44261", "issue:44263", - "issue:44273", "issue:44351", "issue:44485", "issue:44492", "issue:44623", - "issue:44655", "issue:44861", "issue:44908", "issue:44938", "issue:44998", - "issue:45230", "issue:45250", "issue:45290", "issue:45310", @@ -18597,33 +18711,24 @@ "issue:46095", "issue:46144", "issue:46172", - "issue:46218", - "issue:46655", - "issue:46894" + "issue:46218" ], "soft_pairs": [ - "issue:43519|issue:46655", - "issue:44273|issue:44908", - "issue:45230|issue:45310", - "issue:44655|issue:44861", - "issue:43576|issue:46218", - "issue:43519|issue:44485", - "issue:45468|issue:46894", - "issue:43519|issue:44908", "issue:41669|issue:43704", "issue:31795|issue:34689", - "issue:46018|issue:46172", "issue:43704|issue:44485", + "issue:46018|issue:46172", "issue:44485|issue:45468", - "issue:43502|issue:43519", "issue:42371|issue:43704", + "issue:43502|issue:43519", "issue:45850|issue:46218", - "issue:43519|issue:43704", + "issue:43519|issue:44623", "issue:44263|issue:45468", - "issue:44261|issue:45468", + "issue:43606|issue:45468", "issue:43298|issue:46095", - "issue:44351|issue:46218", + "issue:43519|issue:43704", "issue:43352|issue:46218", + "issue:44351|issue:46218", "issue:42913|issue:43066", "issue:28348|issue:45250", "issue:44908|issue:44998", @@ -18643,427 +18748,311 @@ "issue:31515|issue:31795", "issue:44230|issue:45310", "issue:43298|issue:44230", - "issue:43717|issue:46144" + "issue:43717|issue:46144", + "issue:42936|issue:46018", + "issue:43668|issue:44031", + "issue:42936|issue:43240", + "issue:43240|issue:46144", + "issue:42936|issue:44031", + "issue:42936|issue:46144", + "issue:42936|issue:43813", + "issue:43240|issue:43668", + "issue:43668|issue:43813", + "issue:37934|issue:43668", + "issue:42936|issue:43668" ], - "prepared_review_unit_hash": "56f919903fd23f535503059c5962627afe60a9e99a94128b1d42ceb830f89e5b", + "prepared_review_unit_hash": "b78e3a01bc88e9332a776708d7c312924f9bb366e58d74d968646dc734717f3c", "component_budget": { "node_count": 502, "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 42, - "serialized_chars": 39368, - "estimated_input_tokens": 9842, - "estimated_eval_tokens": 19940 + "soft_pair_count": 46, + "serialized_chars": 40600, + "estimated_input_tokens": 10150, + "estimated_eval_tokens": 20556 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a loose grab bag of mostly unrelated Transformer issues, with only broad thematic overlap around v5 regressions, model-loading problems, and a few model-specific bugs. None of the soft pairs are strong enough to merge as true duplicates under a strict triage standard.", - "confidence": 0.41, - "canonical_issue_reason": "Issue 45850 is the broadest cluster-wide regression complaint and best matches the recurring v5 breakage theme across several items.", + "summary": "The cluster is mostly heterogeneous regression/feature issues, not true duplicates. None of the soft edges look like the same underlying bug or change, so I reject all pairwise merges. If an umbrella issue is needed, the broad Transformers 5.0 breakage report is the best anchor.", + "confidence": 0.26, + "canonical_issue_reason": "issue:45850 is the broadest umbrella-style report in the set (\u201cSince 5.0 version it breaks too many models\u201d), so it is the most suitable canonical anchor among these otherwise unrelated issues.", "canonical_pr_reason": null, - "best_issue_reason": "45850 is the most representative umbrella issue in an otherwise heterogeneous set, though it is still too vague to be a true duplicate target for most others.", + "best_issue_reason": "issue:45850 best fits as the global representative because it is the only item that broadly describes the overall regression theme rather than a narrow model-specific or unrelated bug.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43519", - "right": "issue:46655", - "accept": false, - "reason": "Different model processors and different bugs: timestamp calculation vs video padding calculation." - }, - { - "left": "issue:44273", - "right": "issue:44908", - "accept": false, - "reason": "Lazy loading behavior and scheduler kwargs are unrelated code paths." - }, - { - "left": "issue:45230", - "right": "issue:45310", - "accept": false, - "reason": "Generic bug report versus a specific Qwen3.5 MoE from_pretrained failure; not enough evidence of the same bug." - }, - { - "left": "issue:44655", - "right": "issue:44861", - "accept": false, - "reason": "Saving Pipeline objects and tied-weight key handling are separate serialization issues." - }, - { - "left": "issue:43576", - "right": "issue:46218", - "accept": false, - "reason": "Broken env command and Qwen3_5 import failure are unrelated." - }, - { - "left": "issue:43519", - "right": "issue:44485", - "accept": false, - "reason": "Qwen3VL timestamp logic and GLM-5 RoPE implementation are different model-specific bugs." - }, - { - "left": "issue:45468", - "right": "issue:46894", - "accept": false, - "reason": "Audio positional encoding and training-loss double-shift are different problems, even if both are model-training related." - }, - { - "left": "issue:43519", - "right": "issue:44908", - "accept": false, - "reason": "Processor timestamp math and scheduler timescale handling do not match." - }, { "left": "issue:41669", "right": "issue:43704", "accept": false, - "reason": "Import-time performance regression and VRAM leak are unrelated." + "reason": "Import-star/model import performance is unrelated to a VRAM leak in Qwen3ForCausalLM." }, { "left": "issue:31795", "right": "issue:34689", "accept": false, - "reason": "Documentation confusion and a model loading regression are not the same underlying issue." + "reason": "Documentation confusion about input_ids/past_key_values is not the same as a Llama 3.2 Vision model loading regression." }, { - "left": "issue:46018", - "right": "issue:46172", + "left": "issue:43704", + "right": "issue:44485", "accept": false, - "reason": "Both mention DeepSeek-V4, but one is gating logic and the other is a tensor transpose/cast bug." + "reason": "VRAM leak in Qwen3ForCausalLM and GLM-5 RoPE behavior are different code paths and symptoms." }, { - "left": "issue:43704", - "right": "issue:44485", + "left": "issue:46018", + "right": "issue:46172", "accept": false, - "reason": "VRAM leak in Qwen3 and GLM-5 RoPE implementation are unrelated." + "reason": "DeepSeek-V4 shared expert gating and a DeepSeek-V4 tensor transpose issue are distinct bugs." }, { "left": "issue:44485", "right": "issue:45468", "accept": false, - "reason": "Different models and different implementation areas." + "reason": "GLM-5 RoPE implementation and Gemma-4 audio positional encoding are unrelated model-specific issues." }, { - "left": "issue:43502", - "right": "issue:43519", + "left": "issue:42371", + "right": "issue:43704", "accept": false, - "reason": "Local-files-only network requests and timestamp calculation are unrelated." + "reason": "TF32 settings behavior is unrelated to the VRAM leak report." }, { - "left": "issue:42371", - "right": "issue:43704", + "left": "issue:43502", + "right": "issue:43519", "accept": false, - "reason": "TF32 configuration guidance and VRAM leakage are unrelated." + "reason": "Local-files-only network access and Qwen3VL timestamp calculation are different problems." }, { "left": "issue:45850", "right": "issue:46218", "accept": false, - "reason": "A broad v5 complaint is not the same as a specific import failure for one module." + "reason": "A broad \u201cmany models break in 5.0\u201d complaint is not the same as a specific Qwen3_5 import failure." }, { "left": "issue:43519", - "right": "issue:43704", + "right": "issue:44623", "accept": false, - "reason": "Processor timestamp bug and dataloader-thread VRAM leak are different failures." + "reason": "Timestamp calculation in a processor and missing save_pretrained files are different processor bugs." }, { "left": "issue:44263", "right": "issue:45468", "accept": false, - "reason": "torch.split indexing issue and audio positional encoding are unrelated." + "reason": "A torch.split return-value issue in GlmMoeDsaIndexer is unrelated to Gemma4AudioRelPositionalEncoding." }, { - "left": "issue:44261", + "left": "issue:43606", "right": "issue:45468", "accept": false, - "reason": "Different model families and different numerical issues; not the same bug." + "reason": "CPU offload device mismatch in bark-small is not the same as a Gemma-4 audio encoding bug." }, { "left": "issue:43298", "right": "issue:46095", "accept": false, - "reason": "Continuous batching support and weight-conversion documentation are not the same change." + "reason": "Audio-model support for continuous batching and a DeepSeek-V4 weight-conversion question are unrelated." }, { - "left": "issue:44351", - "right": "issue:46218", + "left": "issue:43519", + "right": "issue:43704", "accept": false, - "reason": "Both are import errors, but for different symbols/modules and different root causes." + "reason": "Timestamp calculation in Qwen3VL and a Qwen3ForCausalLM VRAM leak are different failures." }, { "left": "issue:43352", "right": "issue:46218", "accept": false, - "reason": "Flash Attention support missing for one model is unrelated to an import failure for another." + "reason": "Flash Attention 2.0 support for NemotronH and a Qwen3_5 import error are unrelated." + }, + { + "left": "issue:44351", + "right": "issue:46218", + "accept": false, + "reason": "HybridCache import failure and Qwen3_5 module import failure are different import regressions." }, { "left": "issue:42913", "right": "issue:43066", "accept": false, - "reason": "Both concern tokenizer v5 behavior, but one is a broad regression report and the other is a specific decoder-type bug; not clearly the same underlying issue." + "reason": "Tokenizer behavior differences across versions and a wrong tokenizer decoder type are related in theme but not the same concrete bug." }, { "left": "issue:28348", "right": "issue:45250", "accept": false, - "reason": "Specific GPT2LMHeadModel support request versus a generic Flash Attention 2.0 topic." + "reason": "GPT2-specific Flash Attention 2.0 support and a generic Flash Attention 2.0 issue are not the same concrete change." }, { "left": "issue:44908", "right": "issue:44998", "accept": false, - "reason": "Scheduler kwargs bug and unrelated spam issue." + "reason": "Scheduler kwargs handling and the unrelated spam issue are not duplicates." }, { "left": "issue:44492", "right": "issue:44998", "accept": false, - "reason": "Cache-strategy typo and unrelated spam issue." + "reason": "Cache-strategy typo and the unrelated spam issue are not duplicates." }, { "left": "issue:43618", "right": "issue:44998", "accept": false, - "reason": "CLIPOutput attentions regression and unrelated spam issue." + "reason": "CLIPOutput attentions regression and the unrelated spam issue are not duplicates." }, { "left": "issue:44998", "right": "issue:45850", "accept": false, - "reason": "Unrelated spam-like issue versus a broad v5 regression complaint." + "reason": "An unrelated spam issue is not a duplicate of a broad Transformers 5.0 regression report." }, { "left": "issue:44938", "right": "issue:44998", "accept": false, - "reason": "Python 3.14 load failure and unrelated spam issue." + "reason": "Python 3.14 import failure and the unrelated spam issue are not duplicates." }, { "left": "issue:44861", "right": "issue:45290", "accept": false, - "reason": "Tied-weight key crash and chat-template tool-call crash are different code paths." + "reason": "Tied-weight-key handling and apply_chat_template tool-call crashes affect different code paths." }, { "left": "issue:31515", "right": "issue:38617", "accept": false, - "reason": "from_pretrained slowness and an import error are unrelated." + "reason": "from_pretrained slowness and a missing layer_type_validation import are unrelated." }, { "left": "issue:43572", "right": "issue:44861", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig and tied-weight key handling are different bugs." + "reason": "Missing pad_token_idx in StableLmConfig is unrelated to tied-weight-key attribute handling." }, { "left": "issue:43525", "right": "issue:44861", "accept": false, - "reason": "Llama4Config pad_token_id missing and tied-weight key handling are unrelated." + "reason": "Missing pad_token_id in Llama4Config is unrelated to tied-weight-key attribute handling." }, { "left": "issue:44230", "right": "issue:44623", "accept": false, - "reason": "FP8 inference support request and processor save_pretrained file omission are not the same issue." + "reason": "FP8 support for qwen3-vl/qwen3.5 and processor.save_pretrained missing files are different issues." }, { "left": "issue:28348", "right": "issue:43352", "accept": false, - "reason": "Both ask for Flash Attention 2.0 support, but for different models and not one concrete fix." + "reason": "Both mention Flash Attention 2.0, but they concern different models and separate support requests." }, { "left": "issue:29870", "right": "issue:31515", "accept": false, - "reason": "Generate-time decoder_attention_mask bug and checkpoint loading slowness are unrelated." + "reason": "BART generate() decoder_attention_mask bug and checkpoint loading slowness are unrelated." }, { "left": "issue:30333", "right": "issue:31515", "accept": false, - "reason": "MLflow reporting on failure and checkpoint loading performance are unrelated." + "reason": "MLflow job-reporting behavior and checkpoint loading slowness are unrelated." }, { "left": "issue:31515", "right": "issue:36296", "accept": false, - "reason": "from_pretrained slowness and tensor-parallel training bug are different areas." + "reason": "from_pretrained slowness and tensor-parallel training bugs are different subsystems." }, { "left": "issue:31515", "right": "issue:31795", "accept": false, - "reason": "Loading performance and forward-signature documentation are not the same bug." + "reason": "Loading slowness and documentation confusion are not the same bug." }, { "left": "issue:44230", "right": "issue:45310", "accept": false, - "reason": "FP8 support discussion and Qwen3.5 MoE from_pretrained error are related to the same model family but not the same concrete bug." + "reason": "FP8 support request for Qwen3-VL/Qwen3.5 and a Qwen3.5 from_pretrained error are not the same concrete issue." }, { - "left": "issue:43298", - "right": "issue:44230", + "left": "issue:43717", + "right": "issue:46144", "accept": false, - "reason": "Continuous batching for audio models and FP8 inference support are separate feature gaps." + "reason": "Mamba init_weights behavior and RoFormer attention-interface usage are unrelated model implementations." }, { - "left": "issue:43717", + "left": "issue:42936", + "right": "issue:46018", + "accept": false, + "reason": "Mask2former ignore_value handling and DeepSeek-V4 shared expert gating are unrelated." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "ModernBERTConfig type-hint correctness and tokenizer regex warnings are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Mask2former ignore_value handling and loss kwargs forwarding are different bugs." + }, + { + "left": "issue:43240", "right": "issue:46144", "accept": false, - "reason": "Mamba init_weights usage and RoFormer attention interface usage are unrelated implementation 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": 24, - "review_unit_index": 61, - "review_unit_count": 61, - "status": "reviewed", - "reason": null, - "source": "llm", - "cache_hit": false, - "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-10000-502", - "nodes": [ - "issue:37934", - "issue:42936", - "issue:43240", - "issue:43668", - "issue:43813", - "issue:44031", - "issue:46018", - "issue:46144" - ], - "soft_pairs": [ - "issue:42936|issue:46018", - "issue:43668|issue:44031", - "issue:42936|issue:43240", - "issue:43240|issue:46144", - "issue:42936|issue:44031", - "issue:42936|issue:46144", - "issue:42936|issue:43813", - "issue:43240|issue:43668", - "issue:43668|issue:43813", - "issue:37934|issue:43668", - "issue:42936|issue:43668" - ], - "prepared_review_unit_hash": "21c2d8f22638725eb79cf912e8fe0c7780a31e8e13ec218a45310ceb46d70675", - "component_budget": { - "node_count": 502, - "item_count": 502, - "soft_pair_count": 1843, - "serialized_chars": 956562, - "estimated_input_tokens": 239141, - "estimated_eval_tokens": 478538 - }, - "budget": { - "node_count": 8, - "item_count": 8, - "soft_pair_count": 11, - "serialized_chars": 7775, - "estimated_input_tokens": 1944, - "estimated_eval_tokens": 4144 - }, - "overflow_policy": "truncate_then_skip", - "trimmed": false, - "aggressively_trimmed": false, - "split": true, - "analyst_result": { - "summary": "These issues are not duplicates of one another; they span unrelated bugs in norms, tokenizers, segmentation loss handling, configuration typing, a typo, DeepSeek gating, and RoFormer attention plumbing. No soft pair looks like the same underlying problem.", - "confidence": 0.96, - "canonical_issue_reason": "No single canonical issue emerges because the set is heterogeneous and the titles point to different code paths and model families.", - "canonical_pr_reason": null, - "best_issue_reason": "issue:43240 is the strongest standalone representative: it describes a concrete, actionable bug, has the most discussion/reference activity, and is broader than the mostly narrow typo/type-hint reports.", - "best_pr_reason": null, - "soft_edge_verdicts": [ - { - "left": "issue:42936", - "right": "issue:46018", - "accept": false, - "reason": "Different model areas and different bugs: Mask2Former ignore_value handling vs DeepSeek-V4 shared expert gating." - }, - { - "left": "issue:43668", - "right": "issue:44031", - "accept": false, - "reason": "One is a config type-hint error; the other is a tokenizer regex warning regression. Unrelated code paths." - }, - { - "left": "issue:42936", - "right": "issue:43240", - "accept": false, - "reason": "Both involve training/loss behavior, but one is a missing ignore_value usage and the other is kwargs propagation to a loss function; not the same defect." - }, - { - "left": "issue:43240", - "right": "issue:46144", - "accept": false, - "reason": "Kwargs forwarding to loss calculation is unrelated to RoFormer attention interface usage." + "reason": "Loss kwargs forwarding and RoFormer attention-interface usage are unrelated." }, { "left": "issue:42936", "right": "issue:44031", "accept": false, - "reason": "Mask2Former ignore_value handling and tokenizer regex warnings are unrelated." + "reason": "Mask2former ignore_value handling and tokenizer regex warnings are unrelated." }, { "left": "issue:42936", "right": "issue:46144", "accept": false, - "reason": "Different models and unrelated behaviors: loss ignore handling vs attention interface plumbing." + "reason": "Mask2former ignore_value handling and RoFormer attention-interface usage are unrelated." }, { "left": "issue:42936", "right": "issue:43813", "accept": false, - "reason": "Mask2Former ignore_value bug is unrelated to a typo in a quantization_operation attribute name." - }, - { - "left": "issue:43240", - "right": "issue:43668", - "accept": false, - "reason": "A loss-function kwargs bug and an incorrect type hint are not the same underlying issue." + "reason": "Mask2former ignore_value handling and a typo in quantization_operation are unrelated." }, { "left": "issue:43668", "right": "issue:43813", "accept": false, - "reason": "Both are small code-quality issues, but one is a type hint and the other is a typo in a symbol name." + "reason": "ModernBERTConfig type-hint correctness and a quantization typo are unrelated." }, { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "RMS norm semantics for Llama4TextL2Norm are unrelated to a ModernBERTConfig type-hint mistake." + "reason": "Llama4TextL2Norm semantics and a ModernBERTConfig type hint are unrelated." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "These concern different components and failure modes; no shared bug." + "reason": "Mask2former ignore_value handling and ModernBERTConfig type-hint correctness are unrelated." } ] }, @@ -19096,52 +19085,52 @@ "issue:10519|issue:11005", "issue:11005|issue:11307" ], - "prepared_review_unit_hash": "2d5c21dd759fdf7d97d3d1c3824be60b29c60b8b6e6dc075bfc4e642a5831619", + "prepared_review_unit_hash": "1e109591f1c0208abf877dc1e810dbcaf685bc8d8f9ddcfe33c2b8e6f3c42958", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2877, - "estimated_input_tokens": 720, - "estimated_eval_tokens": 1696 + "serialized_chars": 2875, + "estimated_input_tokens": 719, + "estimated_eval_tokens": 1694 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2877, - "estimated_input_tokens": 720, - "estimated_eval_tokens": 1696 + "serialized_chars": 2875, + "estimated_input_tokens": 719, + "estimated_eval_tokens": 1694 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is heterogeneous: one issue asks for reverse truncation behavior, one requests ReduceLROnPlateau-like scheduling, and one asks for Wav2Vec2 word time offsets. They do not describe the same underlying bug or feature, so they should not be deduplicated.", + "summary": "These three issues are unrelated: token truncation direction, ReduceLROnPlateau-like scheduler behavior, and Wav2Vec2 word timestamp offsets. None should be clustered as duplicates.", "confidence": 0.99, - "canonical_issue_reason": null, + "canonical_issue_reason": "No single canonical issue fits the cluster because the topics are different and do not describe the same underlying bug or feature request.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 10519 is the clearest standalone request, but it is not a duplicate of the others; the cluster does not have a valid single canonical issue.", + "best_issue_reason": "Issue 10519 is the most self-contained and specific, but it is not a true representative for this cluster since the other issues are on unrelated topics.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:10519", "right": "issue:11307", "accept": false, - "reason": "Unrelated feature requests: truncation direction vs word-level time offsets in Wav2Vec2." + "reason": "Truncation-direction behavior and Wav2Vec2 word offset timing are unrelated features." }, { "left": "issue:10519", "right": "issue:11005", "accept": false, - "reason": "Different domains and code paths: token truncation behavior vs learning-rate scheduler functionality." + "reason": "Token truncation options and ReduceLROnPlateau-like scheduling address different code paths and problems." }, { "left": "issue:11005", "right": "issue:11307", "accept": false, - "reason": "No shared underlying bug/change; scheduler feature request is unrelated to speech offset extraction." + "reason": "A learning-rate scheduler feature request and word timestamp offsets are not the same underlying issue." } ] }, @@ -19173,12 +19162,12 @@ "issue:13244|issue:16618", "issue:15980|issue:16618" ], - "prepared_review_unit_hash": "6382919fb19dbced478a520b117bf562ebd56e96e4260f2b81acc21d0fe9d56b", + "prepared_review_unit_hash": "cc06f0792774d6a7e57bd662b0a3ba556a92acd5496225a2677d46bdab9c9ef7", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2478, + "serialized_chars": 2477, "estimated_input_tokens": 620, "estimated_eval_tokens": 1496 }, @@ -19186,7 +19175,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2478, + "serialized_chars": 2477, "estimated_input_tokens": 620, "estimated_eval_tokens": 1496 }, @@ -19195,24 +19184,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The three issues are not true duplicates. Two are about model-loading failures, but one is specifically an auth/error-message problem for private models and the other is a generic failure loading `bert-base-uncased`; the Tapas tokenization report is unrelated.", - "confidence": 0.91, - "canonical_issue_reason": "#16618 is the broadest and most discussed issue, and it best represents the generic model-loading failure theme in this small cluster.", + "summary": "These three issues are not duplicates: one is about Tapas tokenization differing from TensorFlow code, one is about an incorrect error message when downloading a private model while logged out, and one is a generic inability to load bert-base-uncased. The cluster is heterogeneous, with no safe canonical duplicate.", + "confidence": 0.88, + "canonical_issue_reason": "No single canonical issue: the titles describe different bug classes and subsystems, so there is no clear duplicate target.", "canonical_pr_reason": null, - "best_issue_reason": "#16618 is the most suitable representative issue because it is the most general user-facing problem and has the highest discussion activity.", + "best_issue_reason": "If a representative must be chosen, issue 16618 is the broadest model-loading report, but it still does not match the other issues closely enough to serve as a true duplicate canonical.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:13244", "right": "issue:16618", "accept": false, - "reason": "Different underlying bugs: Tapas tokenization mismatch vs a model-loading failure for `bert-base-uncased`. They do not describe the same code path or fix." + "reason": "Different problems: Tapas tokenization mismatch vs a model loading failure for bert-base-uncased; not the same underlying bug." }, { "left": "issue:15980", "right": "issue:16618", "accept": false, - "reason": "Both mention model download/load errors, but #15980 is explicitly about private-model auth and misleading login behavior, while #16618 is a generic load failure for a public model. Not the same bug." + "reason": "Related to model access, but one is a bad private-model download/login error and the other is a generic load failure; too different to merge as duplicates." } ] }, @@ -19242,29 +19231,29 @@ "soft_pairs": [ "issue:26350|issue:33260" ], - "prepared_review_unit_hash": "9c3d3796730a8d90e91ca15093bf4705d01e477a610635125b17efe12c235eaf", + "prepared_review_unit_hash": "f4528b4bb127de021bed169922a2fed1586c7b537aa1ef22530e6da4a14531e2", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1588, - "estimated_input_tokens": 397, - "estimated_eval_tokens": 1050 + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1588, - "estimated_input_tokens": 397, - "estimated_eval_tokens": 1050 + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate feature requests in the same \u201ccommunity contribution\u201d style, but they target different capabilities: Flash Attention 2 support vs GGUF support. They are not the same underlying change.", + "summary": "These are both architecture-support community-contribution issues, but they target different features: Flash Attention 2 vs GGUF. They do not describe the same underlying bug or change, so they should not be clustered as duplicates.", "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, @@ -19275,7 +19264,7 @@ "left": "issue:26350", "right": "issue:33260", "accept": false, - "reason": "Different features and code paths: one asks for Flash Attention 2 support across more architectures, the other asks for GGUF support across more architectures. Similar wording is not enough to treat them as duplicates." + "reason": "Same broad theme of expanding architecture support, but the concrete changes are unrelated (Flash Attention 2 support vs GGUF support) and would not be merged as one issue/change." } ] }, @@ -19340,215 +19329,218 @@ "issue:46222|issue:46931", "issue:46226|issue:46931" ], - "prepared_review_unit_hash": "99afcf069f18cf5d3bd2b292edb9b76bf292ee8265bd2aaee88da7a54dbc747f", + "prepared_review_unit_hash": "55aedb2f7895a058ebef72d32a7afbdd53f862064b554f6fc15c2094ebb988d0", "component_budget": { "node_count": 9, "item_count": 9, "soft_pair_count": 29, - "serialized_chars": 15981, - "estimated_input_tokens": 3996, - "estimated_eval_tokens": 8248 + "serialized_chars": 15977, + "estimated_input_tokens": 3995, + "estimated_eval_tokens": 8246 }, "budget": { "node_count": 9, "item_count": 9, "soft_pair_count": 29, - "serialized_chars": 15981, - "estimated_input_tokens": 3996, - "estimated_eval_tokens": 8248 + "serialized_chars": 15977, + "estimated_input_tokens": 3995, + "estimated_eval_tokens": 8246 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly a set of i18n translation-request issues, and the near-identical text appears to come from a shared template rather than the same underlying request. The language-specific tickets are distinct, so I would not merge them as duplicates.", - "confidence": 0.86, - "canonical_issue_reason": "Issue 43834 is the clearest generic/template-shaped representative, but the cluster is heterogeneous and the language-specific issues are not true duplicates of one another.", + "summary": "This cluster is mostly a mix of separate i18n translation requests by language, so most cross-language pairs are not duplicates. The only convincing duplicates are the identical placeholder-title issues among the English-template entries, plus the exact same Thai-title pair.", + "confidence": 0.81, + "canonical_issue_reason": "issue:43834 is the oldest clean representative of the exact-title placeholder duplicate subgroup and is the best single issue to anchor that duplicate set.", "canonical_pr_reason": null, - "best_issue_reason": "43834 is the best single representative if one must be chosen because it most clearly reflects the reusable template form shared by the others; still, it should not be treated as the same request as the language-specific tickets.", + "best_issue_reason": "issue:43834 is the strongest representative issue in the cluster because it sits in the clearest duplicate subgroup and predates the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27088", "right": "issue:44370", "accept": false, - "reason": "Different translation request context; the overlap is only the shared i18n template wording." + "reason": "Different language-specific translation request vs. generic placeholder entry; not the same issue." }, { "left": "issue:27088", "right": "issue:44365", "accept": false, - "reason": "Different translation request context; similar bodies do not make these the same underlying issue." + "reason": "Turkish docs translation request is distinct from the placeholder entry." }, { "left": "issue:27088", "right": "issue:43834", "accept": false, - "reason": "One is a specific Turkish docs translation request, the other is a generic template-shaped issue." + "reason": "Different language request; no evidence of the same translation task." }, { "left": "issue:27088", "right": "issue:38435", "accept": false, - "reason": "Different language translation requests; they are not the same task." + "reason": "Turkish and Romanian translation requests are separate issues." }, { "left": "issue:27088", "right": "issue:28936", "accept": false, - "reason": "Turkish and Spanish docs translation requests are distinct." + "reason": "Turkish and Spanish translation requests are different tasks." }, { "left": "issue:28936", "right": "issue:43834", "accept": false, - "reason": "Spanish request vs generic template issue; not the same underlying request." + "reason": "Spanish docs request does not match the generic placeholder issue." }, { "left": "issue:28936", "right": "issue:44365", "accept": false, - "reason": "Similar template language, but likely separate translation requests." + "reason": "Different language request; not a duplicate." }, { "left": "issue:28936", "right": "issue:38435", "accept": false, - "reason": "Spanish and Romanian docs translation requests are distinct." + "reason": "Spanish and Romanian translation requests are separate." }, { "left": "issue:28936", "right": "issue:44370", "accept": false, - "reason": "Same subsystem, but not the same translation request." + "reason": "Spanish docs request is unrelated to the placeholder entry." }, { "left": "issue:38435", "right": "issue:44370", "accept": false, - "reason": "Romanian request and template-shaped issue are not the same underlying change." + "reason": "Romanian request vs. generic placeholder; different underlying issue." }, { "left": "issue:38435", "right": "issue:44365", "accept": false, - "reason": "Similar wording only; different translation requests." + "reason": "Different translation request; no concrete duplicate evidence." }, { "left": "issue:38435", "right": "issue:43834", "accept": false, - "reason": "Generic template issue vs a concrete Romanian translation request." + "reason": "Romanian docs issue is distinct from the placeholder entry." }, { "left": "issue:38435", "right": "issue:46222", "accept": false, - "reason": "These look like separate language-request tickets, not one bug/change." + "reason": "Romanian request and Thai-titled request are different issues." }, { "left": "issue:38435", "right": "issue:46226", "accept": false, - "reason": "These look like separate language-request tickets, not one bug/change." + "reason": "Different language/title; not the same change." }, { "left": "issue:43834", "right": "issue:44370", - "accept": false, - "reason": "Near-identical template text, but likely distinct language requests." + "accept": true, + "reason": "Same placeholder title and nearly identical body; looks like a duplicate of the same translation request." }, { "left": "issue:43834", "right": "issue:44365", - "accept": false, - "reason": "Template similarity alone is not enough to merge them." + "accept": true, + "reason": "Same placeholder title and near-identical issue text; likely duplicate." }, { "left": "issue:43834", "right": "issue:46222", "accept": false, - "reason": "They share the same template, but appear to be separate issue instances." + "reason": "Different language/title content; no evidence these are the same issue." }, { "left": "issue:43834", "right": "issue:46226", "accept": false, - "reason": "They share the same template, but appear to be separate issue instances." + "reason": "Placeholder entry and Thai-titled entry are not the same request." }, { "left": "issue:43834", "right": "issue:46931", "accept": false, - "reason": "Generic translation template vs a different concrete issue title." + "reason": "Different titles and different underlying translation task." }, { "left": "issue:44365", "right": "issue:44370", - "accept": false, - "reason": "Highly similar template text, but not enough evidence they are the same request." + "accept": true, + "reason": "Same placeholder title with very similar body text; likely duplicate." }, { "left": "issue:44365", "right": "issue:46222", "accept": false, - "reason": "Likely separate translation-request issues sharing boilerplate." + "reason": "Different language/title; not a duplicate." }, { "left": "issue:44365", "right": "issue:46226", "accept": false, - "reason": "Likely separate translation-request issues sharing boilerplate." + "reason": "Different request; no same-issue evidence." }, { "left": "issue:44365", "right": "issue:46931", "accept": false, - "reason": "Shared template similarity is not sufficient to conclude duplication." + "reason": "Different translation issue; not the same underlying task." }, { "left": "issue:44370", "right": "issue:46222", "accept": false, - "reason": "Same template pattern, but no evidence of the same underlying request." + "reason": "Different language/title and no concrete overlap beyond broad i18n similarity." }, { "left": "issue:44370", "right": "issue:46226", "accept": false, - "reason": "Same template pattern, but no evidence of the same underlying request." + "reason": "Different request; not the same change." }, { "left": "issue:44370", "right": "issue:46931", "accept": false, - "reason": "These appear to be different issue instances with different titles." + "reason": "Different underlying translation task." }, { "left": "issue:46222", "right": "issue:46226", - "accept": false, - "reason": "Nearly identical text, but no proof they refer to one shared request." + "accept": true, + "reason": "Exact same Thai title and identical body size; these look like duplicates." }, { "left": "issue:46222", "right": "issue:46931", "accept": false, - "reason": "Different titles and no concrete evidence of a duplicate bug/change." + "reason": "Similar language family is not enough; the Thai titles describe different requests." }, { "left": "issue:46226", "right": "issue:46931", "accept": false, - "reason": "Different titles and no concrete evidence of a duplicate bug/change." + "reason": "Different Thai-titled request; no evidence it is the same issue." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: it distinguishes the generic English-template placeholder cluster from the separate language-specific translation requests, and it only elevates the exact-title pairs (43834/44365/44370 and 46222/46226) as duplicates. The canonical issue choice is also conservative and matches the oldest representative in the placeholder subgroup." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -19627,12 +19619,12 @@ "issue:42503|issue:46653", "issue:45027|issue:46653" ], - "prepared_review_unit_hash": "7ae2569fa8fee67638b2273633dcd37fea792789aac673f5c84a18fdb4c97e9a", + "prepared_review_unit_hash": "2770cd8f81b19707a4a3b8701b8cce596b92ca718dddfa807b892ecea7567981", "component_budget": { "node_count": 19, "item_count": 19, "soft_pair_count": 40, - "serialized_chars": 24673, + "serialized_chars": 24676, "estimated_input_tokens": 6169, "estimated_eval_tokens": 12594 }, @@ -19640,7 +19632,7 @@ "node_count": 19, "item_count": 19, "soft_pair_count": 40, - "serialized_chars": 24673, + "serialized_chars": 24676, "estimated_input_tokens": 6169, "estimated_eval_tokens": 12594 }, @@ -19649,120 +19641,120 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This is a loose thematic cluster of model-request issues (add/support requests across BERT-like, audio/TTS/ASR, vision-language, and detection models), not a set of true duplicates. The similarity links appear driven by the shared 'add model/support' wording rather than the same underlying change.", - "confidence": 0.63, - "canonical_issue_reason": "issue:36879 is the cleanest representative of the cluster\u2019s core theme: a straightforward 'add model' request with multiple similarity links to other model-addition tickets.", + "summary": "These are mostly unrelated model-request issues connected only by generic \"add model/support\" wording; I see no true duplicate groups or mergeable PRs in the provided set.", + "confidence": 0.98, + "canonical_issue_reason": "No true duplicate anchor exists; the cluster is overbroad. If one issue must stand in, 36879 is the most generic and central model-add request.", "canonical_pr_reason": null, - "best_issue_reason": "issue:36879 is the best single representative because its title is generic enough to stand in for the cluster\u2019s overall model-request pattern, while still being concrete and central among the linked issues.", + "best_issue_reason": "Issue 36879 is the best overall representative only as a broad model-addition request, but it should not absorb the other issues as duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36718", "right": "issue:41115", "accept": false, - "reason": "Different requested artifacts: RoMa keypoint matcher vs MiniCPM3 model architecture. Same broad 'add support' pattern only." + "reason": "Different model domains: keypoint matching vs MiniCPM3 model support; shared phrasing only." }, { "left": "issue:36817", "right": "issue:41211", "accept": false, - "reason": "EuroBert config addition and DEIMv2 model addition are different model requests, not the same change." + "reason": "Both are model-add requests, but EuroBert config and DEIMv2 are unrelated changes." }, { "left": "issue:36817", "right": "issue:43671", "accept": false, - "reason": "EuroBert config vs Qwen3-TTS support are unrelated model families and code paths." + "reason": "Different models and modalities; BERT config vs Qwen3-TTS support." }, { "left": "issue:36817", "right": "issue:36879", "accept": false, - "reason": "Different models entirely: EuroBert config vs RF-DETR model." + "reason": "Both are generic model additions, but EuroBert and RF-DETR are distinct models with separate integration work." }, { "left": "issue:36879", "right": "issue:41211", "accept": false, - "reason": "RF-DETR and DEIMv2 are distinct model requests with no shared underlying implementation." + "reason": "RF-DETR and DEIMv2 are different vision models; not the same underlying request." }, { "left": "issue:36879", "right": "issue:43909", "accept": false, - "reason": "RF-DETR detector support and LFM2.5 Audio support are different domains and changes." + "reason": "RF-DETR model addition is unrelated to adding LFM2.5 Audio 1.5B support." }, { "left": "issue:36879", "right": "issue:44121", "accept": false, - "reason": "RF-DETR model addition and OpenAI weight-sparse transformer request are unrelated." + "reason": "Different architectures and codepaths; RF-DETR vs weight-sparse transformer request." }, { "left": "issue:36879", "right": "issue:43837", "accept": false, - "reason": "RF-DETR and Qwen3-ASR are different models and different support paths." + "reason": "Different models and tasks; RF-DETR is not the same as Qwen3-ASR support." }, { "left": "issue:36879", "right": "issue:43671", "accept": false, - "reason": "RF-DETR model vs Qwen3-TTS support are not the same underlying change." + "reason": "Different model families and modalities; RF-DETR vs Qwen3-TTS." }, { "left": "issue:37015", "right": "issue:43837", "accept": false, - "reason": "NeoBERT and Qwen3-ASR are different models; similarity is only at the request level." + "reason": "NeoBERT and Qwen3-ASR are unrelated model requests." }, { "left": "issue:37015", "right": "issue:44121", "accept": false, - "reason": "NeoBERT vs circuit-sparsity/circuitgpt request are different architectures and code paths." + "reason": "NeoBERT vs weight-sparse transformer are distinct model additions, not one bug/change." }, { "left": "issue:37015", "right": "issue:43909", "accept": false, - "reason": "NeoBERT and LFM2.5 Audio 1.5B are unrelated model requests." + "reason": "Different requests: NeoBERT and LFM2.5 Audio 1.5B." }, { "left": "issue:36879", "right": "issue:37015", "accept": false, - "reason": "RF-DETR model and NeoBERT are distinct additions, not duplicate reports." + "reason": "RF-DETR and NeoBERT are separate model integrations with no shared concrete change." }, { "left": "issue:37015", "right": "issue:42503", "accept": false, - "reason": "NeoBERT and ModernVBERT are both BERT-ish names but still different model requests." + "reason": "NeoBERT and ModernVBERT are different models; only broad similarity in request type." }, { "left": "issue:41115", "right": "issue:43837", "accept": false, - "reason": "MiniCPM3 architecture request and Qwen3-ASR support target different models and capabilities." + "reason": "MiniCPM3 model architecture request and Qwen3-ASR support are distinct." }, { "left": "issue:41115", "right": "issue:44121", "accept": false, - "reason": "MiniCPM3 and circuit-sparsity transformer are different requests with no shared concrete change." + "reason": "MiniCPM3 support is unrelated to adding a weight-sparse transformer implementation." }, { "left": "issue:41115", "right": "issue:42503", "accept": false, - "reason": "MiniCPM3 and ModernVBERT are separate model additions." + "reason": "MiniCPM3 and ModernVBERT are separate model requests." }, { "left": "issue:41115", "right": "issue:45027", "accept": false, - "reason": "MiniCPM3 architecture request and Voxtral TTS support are different model domains." + "reason": "Different audio/text model support requests; no shared code-path problem." }, { "left": "issue:41211", @@ -19774,121 +19766,127 @@ "left": "issue:42503", "right": "issue:43837", "accept": false, - "reason": "ModernVBERT models and Qwen3-ASR support are different model families and changes." + "reason": "ModernVBERT models and Qwen3-ASR support are different integration tasks." }, { "left": "issue:42503", "right": "issue:43671", "accept": false, - "reason": "ModernVBERT and Qwen3-TTS are distinct model requests." + "reason": "ModernVBERT and Qwen3-TTS are unrelated model requests." }, { "left": "issue:42503", "right": "issue:42738", "accept": false, - "reason": "ModernVBERT models and generic BERT-like RoPE support are related only by broad naming, not the same change." + "reason": "ModernVBERT models and generic BERT-like RoPE support are not the same concrete change." }, { "left": "issue:42738", "right": "issue:45027", "accept": false, - "reason": "BERT-like RoPE support and Voxtral TTS support are different model concerns." + "reason": "BERT-like models with RoPE and Voxtral TTS support target different model families." }, { "left": "issue:42738", "right": "issue:46653", "accept": false, - "reason": "BERT-like models with RoPE and the unrelated 'Mirko Privitera' issue are clearly not duplicates." + "reason": "Generic model-architecture issue vs an unrelated personal-name placeholder issue." }, { "left": "issue:43671", "right": "issue:43837", "accept": false, - "reason": "Qwen3-TTS and Qwen3-ASR are closely related product line items, but they are different capabilities and would not be one PR." + "reason": "Same Qwen3 family, but TTS vs ASR are different tasks and likely separate changes." }, { "left": "issue:43671", "right": "issue:45027", "accept": false, - "reason": "Qwen3-TTS and Voxtral TTS support are different model-specific requests." + "reason": "Qwen3-TTS and Voxtral-4B-TTS are separate model-support requests." }, { "left": "issue:43671", "right": "issue:45177", "accept": false, - "reason": "Qwen3-TTS support and DeepSeek-OCR2 are unrelated model requests." + "reason": "Qwen3-TTS support is unrelated to DeepSeek-OCR2." }, { "left": "issue:43671", "right": "issue:44121", "accept": false, - "reason": "Qwen3-TTS and the weight-sparse transformer request are different underlying changes." + "reason": "Audio model support and a sparse-transformer proposal are different codepaths." }, { "left": "issue:43837", "right": "issue:43909", "accept": false, - "reason": "Qwen3-ASR and LFM2.5 Audio 1.5B are different audio models and code paths." + "reason": "Qwen3-ASR support and LFM2.5 Audio 1.5B are different model requests." }, { "left": "issue:43837", "right": "issue:45322", "accept": false, - "reason": "Qwen3-ASR and EUPE are different model requests in different domains." + "reason": "Qwen3-ASR and EUPE are unrelated model integrations." }, { "left": "issue:43837", "right": "issue:44121", "accept": false, - "reason": "Qwen3-ASR support and circuit-sparsity transformer support are unrelated changes." - }, - { - "left": "issue:43837", - "right": "issue:45027", - "accept": false, - "reason": "Qwen3-ASR and Voxtral TTS support are different speech-model requests." + "reason": "ASR model support is not the same change as adding a weight-sparse transformer." }, { "left": "issue:43909", "right": "issue:44205", "accept": false, - "reason": "LFM2.5 Audio 1.5B and SAM3-LiteText are different models and modalities." + "reason": "LFM2.5 Audio 1.5B and SAM3-LiteText are different model families." }, { "left": "issue:43909", "right": "issue:46645", "accept": false, - "reason": "LFM2.5 Audio 1.5B and SWA+sink Granite option are unrelated model changes." + "reason": "Audio model support and SWA+sink for Granite are unrelated changes." }, { "left": "issue:44121", "right": "issue:45322", "accept": false, - "reason": "Weight-sparse transformer request and EUPE are separate model requests with no shared concrete bug/change." + "reason": "Weight-sparse transformer proposal and EUPE model request are different." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral-4B-TTS are distinct audio-model support requests." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT and Voxtral-4B-TTS are unrelated model additions." }, { "left": "issue:43837", "right": "issue:45177", "accept": false, - "reason": "Qwen3-ASR support and DeepSeek-OCR2 are unrelated model additions." + "reason": "Qwen3-ASR support and DeepSeek-OCR2 are different model requests." }, { "left": "issue:45322", "right": "issue:46133", "accept": false, - "reason": "EUPE and TIPSv2 are different vision-language model requests, not the same underlying change." + "reason": "EUPE and TIPSv2 are different vision-language encoder requests from different projects." }, { "left": "issue:42503", "right": "issue:46653", "accept": false, - "reason": "ModernVBERT models and the unrelated author-name issue are clearly not duplicates." + "reason": "ModernVBERT models and the placeholder-name issue have no substantive overlap." }, { "left": "issue:45027", "right": "issue:46653", "accept": false, - "reason": "Voxtral TTS support and the unrelated author-name issue are clearly not the same change." + "reason": "Voxtral TTS support is unrelated to the placeholder-name issue." } ] }, @@ -19918,7 +19916,7 @@ "soft_pairs": [ "issue:38740|issue:46443" ], - "prepared_review_unit_hash": "2f5f5c6e5c3e7e7d1148275b297a86b41b082e7af66485ed6c36af865c049285", + "prepared_review_unit_hash": "9d3a1598745feb6ac90598e1963f56971048b875f94ed89685f01ec63cf6ca54", "component_budget": { "node_count": 2, "item_count": 2, @@ -19940,18 +19938,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are unrelated: one is a docs request about adding `pruna` as an optimization framework, and the other is a technical refactor for Vision RoPE. This is not a duplicate cluster.", + "summary": "The two issues are unrelated: one is a docs request to add Pruna as an optimization framework, the other is a technical Vision RoPE refactor. The soft similarity signal is not enough to treat them as duplicates.", "confidence": 0.98, - "canonical_issue_reason": "No canonical issue: the items do not describe the same problem or change.", + "canonical_issue_reason": "Prefer issue 46443 as the cluster representative: it is the substantive technical issue, while 38740 is a separate docs-oriented request with no overlap in problem statement.", "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a good cluster representative because there is no underlying duplicate relationship; if forced, the Vision RoPE refactor is the more concrete code task.", + "best_issue_reason": "Issue 46443 is the best issue in this set to surface as the representative artifact because it is a concrete engineering task and the only one plausibly tied to a code change.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:38740", "right": "issue:46443", "accept": false, - "reason": "Different domains and intents: documentation/framework addition versus a Vision RoPE refactor. They do not appear to be the same underlying bug or change." + "reason": "Different topics and artifact types: a documentation/addition request for Pruna vs. a Vision RoPE refactor. No shared targets, filenames, or concrete code-path overlap." } ] }, @@ -19981,7 +19979,7 @@ "soft_pairs": [ "issue:41084|issue:43037" ], - "prepared_review_unit_hash": "d8bdb1657942a7c7004dd9e282cfb45bb7bc0e5a3968e3d36ea7bc87b858a5c6", + "prepared_review_unit_hash": "c071042d5dc526504e268c89c3cf1b22f83f9191ff5b4dbc1af5fa1d00a8ab64", "component_budget": { "node_count": 2, "item_count": 2, @@ -20003,18 +20001,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: one is about block decoding behavior, the other about adding DeepSeek v3.2 model support. The soft similarity is too weak to justify merging the issues.", - "confidence": 0.97, - "canonical_issue_reason": "No true canonical issue: the two issues concern different problem spaces (decoding implementation vs model support).", + "summary": "The two issues are unrelated: one is about block decoding, the other requests DeepSeek v3.2 support. They do not describe the same bug or change, so the soft edge should not be merged.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a good duplicate target for the other. If a single issue must be kept as the cluster representative, 41084 is slightly more generic, but this cluster looks spurious.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41084", "right": "issue:43037", "accept": false, - "reason": "Different underlying requests: block decoding changes vs DeepSeek v3.2 support. They do not describe the same bug or change." + "reason": "Different topics and scopes: block decoding versus model/version support. No shared code path or concrete duplicate behavior is evident." } ] }, @@ -20044,7 +20042,7 @@ "soft_pairs": [ "issue:41306|issue:41826" ], - "prepared_review_unit_hash": "a2902481131662f12e355fed8c606c8cd9646a0b9d497e34f16bb31a5c955de9", + "prepared_review_unit_hash": "6de6e7015ee51e418111acc4051ac3f475e2f7a2fb1855af5e49174be5ba4b6d", "component_budget": { "node_count": 2, "item_count": 2, @@ -20066,18 +20064,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are both feature requests, but they target different model components and different changes: one asks for a `from_pretrained()` flag for TiledFusedLogitsLoss / LigerFusedLinearCrossEntropy, while the other asks for integrating TiledMLP for lower memory use. They are not the same underlying bug or change.", - "confidence": 0.96, - "canonical_issue_reason": null, + "summary": "The two issues are related only by a broad theme of tiled/memory-efficient layers, but they describe different components and different asks: one is about a `from_pretrained()` flag for TiledFusedLogitsLoss / LigerFusedLinearCrossEntropy, the other is about integrating TiledMLP for lower memory usage.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 41306 is the more concrete, self-contained request: it names the exact feature (`from_pretrained()` flag) and the specific fused loss path affected.", "canonical_pr_reason": null, - "best_issue_reason": "No single issue is a clear canonical duplicate of the other; both are distinct feature requests in different areas, so neither should be treated as the representative issue for a duplicate cluster.", + "best_issue_reason": "41306 is the best representative because it is narrowly scoped and likely resolvable without needing broader architectural changes.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41306", "right": "issue:41826", "accept": false, - "reason": "Different requested functionality and code paths: FLCE/from_pretrained configuration versus TiledMLP memory-footprint integration. They are not plausibly merged into one fix." + "reason": "Different underlying changes: one concerns pretrained loading behavior for a fused logits/loss implementation, the other concerns adding TiledMLP for memory reduction. Shared tiling vocabulary is too broad to treat as the same issue." } ] }, @@ -20107,7 +20105,7 @@ "soft_pairs": [ "issue:43598|issue:43600" ], - "prepared_review_unit_hash": "89d06a9a65b26ca3198c8b8e6393196ae3f64f5aa89b150a909fbc76389da94a", + "prepared_review_unit_hash": "75cfdbcf387dd81bea5a2a892eb599947140ea57e95fee6eb318487f106a1dee", "component_budget": { "node_count": 2, "item_count": 2, @@ -20129,18 +20127,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two closed issues are related only loosely by Trainer customization; they describe different problems and should not be deduplicated.", - "confidence": 0.96, - "canonical_issue_reason": "No canonical duplicate here: #43598 is about `remove_unused_column` behavior and customizability in `Trainer`, while #43600 is about FLOPS estimation for custom trainers.", + "summary": "These two issues are only loosely related through custom trainers, but they describe different problems: one is about `remove_unused_column` behavior, the other about FLOPS estimation. They should not be treated as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": "No canonical issue here: 43598 concerns Trainer column pruning/customizability, while 43600 concerns FLOPS counting for custom trainers.", "canonical_pr_reason": null, - "best_issue_reason": "#43598 is the better representative issue because it describes a concrete Trainer behavior change request; #43600 is a separate instrumentation/accounting concern.", + "best_issue_reason": "If forced to pick a representative issue, 43598 is the more concrete Trainer behavior request, but it still does not subsume 43600.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43598", "right": "issue:43600", "accept": false, - "reason": "Different underlying problems: one targets unused-column removal/custom training behavior, the other targets FLOPS estimation for custom trainers. Shared Trainer context is too broad to treat as a duplicate." + "reason": "They share the broad custom-trainer context, but the underlying changes are different: one targets unused-column removal, the other FLOPS estimation. Not the same bug or change." } ] }, @@ -20170,7 +20168,7 @@ "soft_pairs": [ "issue:43684|issue:44405" ], - "prepared_review_unit_hash": "f7e5b7d9ee2887da4002edcffbcd24fd3fa4b2a3861dfe63e1f4f2908d3a75bc", + "prepared_review_unit_hash": "65c53f4a0574159181a949dd2afbc64771282b466de2c2db42b687f7a575c077", "component_budget": { "node_count": 2, "item_count": 2, @@ -20192,18 +20190,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate feature requests: one for Qwen3-Omni registration in AutoModel/AutoModelForConditionalGeneration, and one for AutoModelForSequenceClassification support for Qwen3.5. They share the broad theme of Qwen auto-class support, but they do not describe the same underlying change.", - "confidence": 0.92, - "canonical_issue_reason": "Issue 43684 is the broader and more central request, covering model registration across two auto classes; 44405 is a narrower request for a different config and task head.", + "summary": "These are related Qwen-model support requests, but they target different model families and different auto-class APIs: Qwen3-Omni registration vs Qwen3.5 sequence-classification support. They are not the same underlying change.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "43684 is the best representative issue because it is the more general auto-model registration request and is closer to a core framework change.", + "best_issue_reason": "Issue 43684 is the broader foundational registration request, but it does not cover the separate AutoModelForSequenceClassification support requested in 44405.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43684", "right": "issue:44405", "accept": false, - "reason": "Different model families and different auto-class targets: Qwen3-Omni registration vs Qwen3.5 sequence-classification support. Related theme, but not the same bug/change." + "reason": "Different model/config and different API surface: Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration versus Qwen3.5 AutoModelForSequenceClassification support. Not the same bug or change." } ] }, @@ -20240,76 +20238,76 @@ "issue:43845|issue:43859", "issue:43846|issue:43859" ], - "prepared_review_unit_hash": "a5c680299fef804f4ebf650c4be9afd8b0845adf73a2f616b8c6002b0552f4df", + "prepared_review_unit_hash": "e40aa6528e1a0ca4e4097ad9358492bf81352b0d6e165ac56aaead653abb29b7", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4267, - "estimated_input_tokens": 1067, - "estimated_eval_tokens": 2390 + "serialized_chars": 4264, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4267, - "estimated_input_tokens": 1067, - "estimated_eval_tokens": 2390 + "serialized_chars": 4264, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The three \"huggingface \" issues look related and may be duplicate reports, but the packet only provides soft-similarity evidence and no deterministic match. \"Fett\" appears unrelated. No PRs are present.", - "confidence": 0.72, - "canonical_issue_reason": "issue:43845 is still the best anchor for the likely huggingface-related cluster because it is the earliest of the similar issues, but the duplicate relationship is only tentative from the available evidence.", + "summary": "The three \"huggingface \" issues are very similar, but the packet only provides soft_similarity and no deterministic evidence, so none of the pairings should be accepted as duplicates. \"Fett\" still looks unrelated. No PRs are present.", + "confidence": 0.89, + "canonical_issue_reason": "issue:43845 is the earliest of the three repeated \"huggingface \" issues, so it is the best canonical issue candidate.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43845 is the best issue candidate because it is the earliest and most representative of the repeated \"huggingface \" reports, even though the grouping is not confirmed as duplicate.", + "best_issue_reason": "issue:43845 is the safest cluster representative because it is the earliest of the repeated-title issues, even though the duplicate evidence is limited.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43835", "right": "issue:43845", "accept": false, - "reason": "Only soft similarity is available, and the titles are different; this is not enough to confirm the same underlying issue." + "reason": "Only weak similarity is present, with different titles and no shared targets, filenames, or content evidence tying them to the same bug." }, { "left": "issue:43835", "right": "issue:43846", "accept": false, - "reason": "Only soft similarity is available, with no shared targets/files or other deterministic evidence of the same bug." + "reason": "\"Fett\" does not match the repeated \"huggingface \" reports, and there is no concrete evidence of the same underlying issue." }, { "left": "issue:43835", "right": "issue:43859", "accept": false, - "reason": "Only soft similarity is available, and there is no concrete evidence that these reports describe the same problem." + "reason": "The title is unrelated to the other reports and the packet contains no shared issue-specific evidence." }, { "left": "issue:43845", "right": "issue:43846", "accept": false, - "reason": "The titles match, but the packet still provides only soft similarity and deterministic_accept is false, so this should be treated as a possible duplicate, not a confirmed one." + "reason": "These are the strongest match because they share the same title, but the evidence is still only soft similarity with no deterministic support." }, { "left": "issue:43845", "right": "issue:43859", "accept": false, - "reason": "Same broad title pattern, but only soft similarity is present and there is no deterministic evidence to justify merging them as the same issue." + "reason": "Same title, but the packet provides no concrete details beyond sparse metadata, so this is not enough to merge them confidently." }, { "left": "issue:43846", "right": "issue:43859", "accept": false, - "reason": "These look like likely similar reports, but the evidence is still limited to soft similarity, so confirmation is not warranted." + "reason": "Same-title similarity exists, but there is no additional evidence showing they describe the same underlying report." } ] }, "evaluator_result": { "accept": false, - "feedback": "The anchor choice is plausible and the note that no PRs are present is grounded, but the soft-edge verdicts are not conservative enough. In the packet, every pair is supported only by soft_similarity, and all pairs have deterministic_accept=false. Treating the huggingface issue pairs as accepted duplicates overstates the evidence; they should be described as likely similar/possible duplicates, not confirmed matches." + "feedback": "The cluster summary is mostly grounded, but the soft-edge judgments are too strong for the evidence in the packet. All pairings are supported only by soft_similarity, and the packet\u2019s deterministic_accept is false for every pair. In particular, the acceptance of 43845\u219443859 and 43846\u219443859 overstates what can be inferred from shared title plus sparse metadata; 43845\u219443846 is the strongest match, but the evidence is still limited. The canonical issue choice (43845 as earliest of the three 'huggingface ' issues) is fine." }, "evaluator_used": true, "retried": true, @@ -20347,7 +20345,7 @@ "issue:44572|issue:44937", "issue:44573|issue:44937" ], - "prepared_review_unit_hash": "22eb13e5f267cf3138b3099bafc98d0734038e753cbf6c5a84db1281fdb30a0b", + "prepared_review_unit_hash": "fe46d237ac961ca84d9955a2ae3b7e6da0a7a6f87bedbc1f2078ea5d59274fb9", "component_budget": { "node_count": 5, "item_count": 5, @@ -20369,66 +20367,66 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster looks like unrelated low-information/spam issues plus one generic documentation suggestion; there is no evidence they describe the same underlying bug or change.", - "confidence": 0.93, - "canonical_issue_reason": "Issue #44418 is the only one with a somewhat descriptive, non-placeholder title, so it is the best representative if a canonical issue must be chosen.", + "summary": "These look like unrelated low-information spam/vague issues, not a real duplicate cluster. There isn\u2019t enough evidence to merge any pair as the same underlying report.", + "confidence": 0.69, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "#44418 is the most substantive issue in the set; the others are spam-like placeholders or equally generic, so it is the least bad anchor for triage.", + "best_issue_reason": "issue:44418 is the only item with a somewhat descriptive, non-spam title, but it still lacks body detail and does not clearly match the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44371", "right": "issue:44572", "accept": false, - "reason": "Both are just '' placeholders; identical titles alone are not enough to establish the same underlying issue." + "reason": "Both are titled \"\" with no body or targets; that is not enough to prove the same underlying issue." }, { "left": "issue:44371", "right": "issue:44573", "accept": false, - "reason": "Both are spam placeholders with no body detail, code path, or target to support duplication." + "reason": "Both are titled \"\" with no substantive details, so they could be unrelated spam posts." }, { "left": "issue:44371", "right": "issue:44418", "accept": false, - "reason": "One is spam and the other is a generic documentation suggestion; these do not look like the same issue." + "reason": "One is spam-like, the other is a generic documentation suggestion; no evidence they report the same problem." }, { "left": "issue:44371", "right": "issue:44937", "accept": false, - "reason": "'' and 'Check out \"Google LLC\"' are unrelated and lack any shared technical context." + "reason": "Spam placeholder vs. \"Check out 'Google LLC'\" are not the same concrete issue." }, { "left": "issue:44418", "right": "issue:44572", "accept": false, - "reason": "A documentation enhancement request is not the same underlying report as a spam placeholder." + "reason": "A documentation suggestion and a spam placeholder do not indicate the same underlying bug/change." }, { "left": "issue:44418", "right": "issue:44573", "accept": false, - "reason": "Generic documentation feedback and a spam placeholder do not indicate the same bug or change." + "reason": "A documentation suggestion and a spam placeholder are too different to treat as duplicates." }, { "left": "issue:44572", "right": "issue:44573", "accept": false, - "reason": "They share only the spam placeholder title; there is no concrete evidence they refer to the same report." + "reason": "Same placeholder title, but no body or contextual evidence tying them to the same report." }, { "left": "issue:44572", "right": "issue:44937", "accept": false, - "reason": "A spam placeholder and a named-topic issue are not sufficiently specific to merge as duplicates." + "reason": "Spam placeholder vs. unrelated \"Google LLC\" title; no shared concrete issue is visible." }, { "left": "issue:44573", "right": "issue:44937", "accept": false, - "reason": "No shared subject matter beyond being low-information issues, so they should not be merged." + "reason": "No sign these are the same underlying issue; the titles are unrelated and the bodies are empty." } ] }, @@ -20458,12 +20456,12 @@ "soft_pairs": [ "issue:45145|issue:45151" ], - "prepared_review_unit_hash": "256b96b106b2dd48ea064078d8cad1b21fed3f3beb81b0a71be63c9b42a85594", + "prepared_review_unit_hash": "1bb9be3a43e20797a8835bc8d3cff7c03959b505968200314f6a2ff02ca619bd", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1646, + "serialized_chars": 1645, "estimated_input_tokens": 412, "estimated_eval_tokens": 1080 }, @@ -20471,7 +20469,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1646, + "serialized_chars": 1645, "estimated_input_tokens": 412, "estimated_eval_tokens": 1080 }, @@ -20480,27 +20478,27 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "#45145 is the best canonical issue candidate because it was created first and already has an inbound reference. The pair looks like a probable duplicate based on the identical title and soft-similarity signal, but the packet does not conclusively prove the bodies are the same.", - "confidence": 0.82, - "canonical_issue_reason": "#45145 is the earlier report and has an inbound reference, so it is the best canonical choice if this is a duplicate pair.", + "summary": "The two closed issues have the same title, highly similar body size, and strong textual overlap, so they appear to be duplicate reports of the same underlying energy-efficiency proposal. Issue #45145 is the better canonical pick because it has an inbound reference; no PRs are present.", + "confidence": 0.98, + "canonical_issue_reason": "#45145 and #45151 share the exact title and appear to describe the same proposal. #45145 is preferable as canonical because it has an inbound reference, suggesting it is the more-referenced duplicate target.", "canonical_pr_reason": null, - "best_issue_reason": "#45145 is the strongest issue to keep as the representative because it was created first and already has an inbound reference.", + "best_issue_reason": "#45145 is the best issue candidate: same subject as #45151, but with at least one inbound reference and no evidence that #45151 is more central.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45145", "right": "issue:45151", "accept": true, - "reason": "Identical title and strong soft-similarity make this a likely duplicate/probable same report, but the packet does not provide enough detail to claim certainty about identical body content." + "reason": "Same title and near-identical topic/body footprint; these look like duplicate issues for the same underlying request." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The canonical choice is grounded: #45145 is earlier and has an inbound reference. However, the duplicate verdict overstates the evidence. The packet shows only identical titles, similar body length, and a soft-similarity signal; it does not establish that the bodies are the same or that this is a clear duplicate with high certainty. Please phrase the conclusion more conservatively (e.g., likely duplicate / probable same report) and lower the confidence." + "accept": true, + "feedback": "The analyst stays grounded in the packet: the two issues have the exact same title, similar body lengths, and a soft-similarity signal, so treating them as likely duplicates is conservative. Choosing #45145 as canonical because it has the only inbound reference is also supported. No PR claims were made." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -20578,30 +20576,30 @@ "soft_pairs": [ "issue:45762|issue:45763" ], - "prepared_review_unit_hash": "1a6537d7b02c9e4d8c0c9ef2573dcad2eee7edc60724f85fc9164568c3eae781", + "prepared_review_unit_hash": "1fd044e2d53b4cb29e95bdcd307c5aa7e64ed8653376779ecea2cad58a34a755", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1477, - "estimated_input_tokens": 370, - "estimated_eval_tokens": 996 + "serialized_chars": 1476, + "estimated_input_tokens": 369, + "estimated_eval_tokens": 994 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1477, - "estimated_input_tokens": 370, - "estimated_eval_tokens": 996 + "serialized_chars": 1476, + "estimated_input_tokens": 369, + "estimated_eval_tokens": 994 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two closed issues are not duplicates; they appear to be unrelated identity-like entries with different names and dates, with no shared code context or linked targets.", - "confidence": 0.98, + "summary": "The two closed issues are just person-name/date entries and do not indicate the same underlying bug or request; this looks like unrelated content, not a duplicate pair.", + "confidence": 0.97, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, @@ -20611,7 +20609,7 @@ "left": "issue:45762", "right": "issue:45763", "accept": false, - "reason": "The titles identify different people and different dates, and there is no body, code, or target overlap suggesting the same underlying issue." + "reason": "Different names and birthdates with no shared issue target, body, or code context; they do not appear to describe the same problem or change." } ] }, @@ -20641,40 +20639,40 @@ "soft_pairs": [ "issue:45764|issue:45765" ], - "prepared_review_unit_hash": "84b8c1e55d9574aebc226b74922176c408dfbdcc478a2ffb20badff504d7be43", + "prepared_review_unit_hash": "e412a072a723cc057635f78755c179f4d30711a8e73f5ac620baa0b3b85824b2", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1481, - "estimated_input_tokens": 371, - "estimated_eval_tokens": 998 + "serialized_chars": 1479, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1481, - "estimated_input_tokens": 371, - "estimated_eval_tokens": 998 + "serialized_chars": 1479, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are empty, unrelated, and look like personal-name placeholder reports rather than the same bug. No duplicate relationship is supported.", - "confidence": 0.98, - "canonical_issue_reason": "No canonical issue: both issues are empty, unrelated, and lack any technical description or shared problem statement.", + "summary": "These two closed issues do not appear to describe the same bug or request. They are just short name/date titles with no body or shared technical context, so the soft similarity signal is not enough to treat them as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": "No clear canonical issue: both issues lack descriptions and appear to be unrelated name/date entries rather than a shared problem report.", "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a good canonical target because neither contains actionable bug information or evidence of duplication.", + "best_issue_reason": "Neither issue is a good canonical choice; both are equally non-technical and provide no actionable context.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45764", "right": "issue:45765", "accept": false, - "reason": "Only superficial similarity in title structure; the names and dates differ and there is no shared technical content, so they are not the same underlying issue." + "reason": "The titles only resemble personal names with dates and provide no evidence of the same underlying bug, task, or code path." } ] }, @@ -20704,7 +20702,7 @@ "soft_pairs": [ "issue:45810|issue:46295" ], - "prepared_review_unit_hash": "210615ddc7a5a52b72ed7152ff4eda7a6f8f74acd9bed5fed510e19a5bd4f789", + "prepared_review_unit_hash": "eff1fda2d23b228a1b250da814dd1884374379aacdc19a65a952557bc302bf38", "component_budget": { "node_count": 2, "item_count": 2, @@ -20726,18 +20724,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are not duplicates: one is a model-class feature request, the other is a training FP8 support request. The soft similarity is too weak to cluster them as the same underlying problem.", - "confidence": 0.98, - "canonical_issue_reason": "Issue 46295 is the better single representative if one must be chosen, because it is the broader product-level request and has more inbound attention. However, there is no true canonical issue here because the two issues are unrelated.", + "summary": "The cluster does not look like a duplicate set: one issue asks for a new Qwen3.5 token-classification model head, while the other requests FP8 training support. They are different feature areas and code paths.", + "confidence": 0.95, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 46295 is the most generally useful issue in this set to keep as the cluster representative, but only as a standalone issue; it does not duplicate issue 45810.", + "best_issue_reason": "Issue 46295 is the better cluster representative only because it appears more broadly discussed and referenced, but it is not a true duplicate of 45810.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45810", "right": "issue:46295", "accept": false, - "reason": "Different requests in different areas: adding Qwen3_5ForTokenClassification vs. FP8 training support. No shared code-path problem or concrete overlap." + "reason": "Different requests: model-classification support vs FP8 training support. No shared target, filename, or underlying bug/change path." } ] }, @@ -20777,91 +20775,91 @@ "issue:45871|issue:45873", "issue:45872|issue:45873" ], - "prepared_review_unit_hash": "1867395015888c3156b6c03cd77b70a61c5dd0efbe0bdd2cfc1dddb37860d53c", + "prepared_review_unit_hash": "4af850637523eaf55e5d3dc0940b980d3753fcfbde552c8ab3d571f1aa4a8618", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5721, - "estimated_input_tokens": 1431, - "estimated_eval_tokens": 3118 + "serialized_chars": 5727, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5721, - "estimated_input_tokens": 1431, - "estimated_eval_tokens": 3118 + "serialized_chars": 5727, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster appears to contain two strong duplicate candidates\u201445870/45871 and 45872/45873\u2014based on identical titles and matching body lengths, plus one separate dataset request (45869). The evidence supports soft duplicate grouping, but not a claim of exact textual identity because the body text is not shown.", - "confidence": 0.76, - "canonical_issue_reason": "45872 is the strongest representative of one likely duplicate subgroup ('DataForge: Financial time series + scraping service'), but the cluster also contains a separate likely duplicate pair (45870/45871) and an unrelated dataset request, so there is no single issue that cleanly represents the whole cluster.", + "summary": "The set is not one coherent duplicate cluster: it contains two separate exact-duplicate issue pairs (45870/45871 and 45872/45873), while 45869 is only loosely related by topic and should not be merged with either pair.", + "confidence": 0.98, + "canonical_issue_reason": "45870 is the cleanest canonical candidate for its subcluster: 45871 is an exact later duplicate, with identical title/body shape, and 45870 is the earliest instance. The other exact-duplicate pair (45872/45873) is a separate issue theme, so there is no single canonical issue for all five items.", "canonical_pr_reason": null, - "best_issue_reason": "45872 is the best overall issue representative by similarity strength, but only as a representative of one soft-duplicate subgroup; the cluster is not a single unified issue.", + "best_issue_reason": "45870 is the best representative within the 'Example: Financial time series prediction with DataForge' duplicate pair because it is an exact duplicate target and the earliest submission. That said, the overall set splits into two unrelated duplicate pairs, so no one issue cleanly represents the entire cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45869", "right": "issue:45870", "accept": false, - "reason": "Dataset bundle request vs. example-prediction request; related financial-data context, but not the same underlying issue." + "reason": "Related financial-data topic, but one is a dataset promo and the other is an example issue; not the same underlying bug/change." }, { "left": "issue:45869", "right": "issue:45871", "accept": false, - "reason": "Dataset bundle request vs. example-prediction request; same broad domain, different request." + "reason": "Same as above: similar theme, but different artifact and intent." }, { "left": "issue:45870", "right": "issue:45871", "accept": true, - "reason": "Identical title and matching body length make this a strong likely duplicate pair, though the body text itself is not available." + "reason": "Exact duplicate pair: same title and matching body size, with no evidence of a different underlying issue." }, { "left": "issue:45870", "right": "issue:45872", "accept": false, - "reason": "Different request types: example-prediction example vs. scraping-service request; only broad topic overlap." + "reason": "Both mention financial time series/DataForge, but one is an example and the other adds a scraping service; these are different issue requests." }, { "left": "issue:45870", "right": "issue:45873", "accept": false, - "reason": "Different request types and distinct titles; overlap is only in the DataForge/financial theme." + "reason": "Different underlying content and scope; not the same issue." }, { "left": "issue:45871", "right": "issue:45872", "accept": false, - "reason": "Same broad financial/DataForge context, but they describe different concrete requests." + "reason": "Different issue themes despite shared finance/DataForge keywords." }, { "left": "issue:45871", "right": "issue:45873", "accept": false, - "reason": "Same broad financial/DataForge context, but not the same underlying issue." + "reason": "Different issue themes despite shared finance/DataForge keywords." }, { "left": "issue:45872", "right": "issue:45873", "accept": true, - "reason": "Identical title and matching body length make this a strong likely duplicate pair, but this is still soft evidence rather than proven exact identity." + "reason": "Exact duplicate pair: same title and matching body size, indicating the same underlying issue." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The analysis overstates the evidence. The packet only shows identical titles and matching body lengths for 45870/45871 and 45872/45873; it does not provide body text or any deterministic-accept signal, so calling them exact duplicates is not grounded. More conservatively, these are strong soft-similarity candidates, not proven exact matches. The summary and canonical reasons should be downgraded to reflect that uncertainty." + "accept": true, + "feedback": "The conclusions are grounded in the packet: 45870/45871 and 45872/45873 are the only clearly duplicate-looking pairs by title/body length and creation order, while 45869 is only loosely related. The soft-edge rejections are conservative, and the note that there is no single canonical issue for all five items is supported by the split into two separate pairs." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -20885,7 +20883,7 @@ "soft_pairs": [ "issue:46439|issue:46441" ], - "prepared_review_unit_hash": "10957c69e37807aa99566580c314e2a153c5654193f85495ccfbb1f3fbcdc801", + "prepared_review_unit_hash": "24a1ba1b9ca4b5c680683fcdad2897c2e26be82c44d689e5962a1cb7435368d3", "component_budget": { "node_count": 2, "item_count": 2, @@ -20907,18 +20905,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two related static-cache/linear-attention issues, but they appear to describe different failure modes: one shape-corruption RuntimeError during early initialization, the other a Qwen3.5 KeyError for `linear_attention`.", - "confidence": 0.81, - "canonical_issue_reason": "Issue 46439 is the better representative because it names the underlying code path more concretely (`early_initialization` corrupting linear-attention shapes) and sounds broader than the Qwen3.5-specific crash report.", + "summary": "Both issues point at static-cache problems around linear-attention models, but they describe different concrete failures: one is shape corruption leading to a RuntimeError, the other is a KeyError for a missing `linear_attention` cache entry in Qwen3.5. They are related in subsystem, but not clearly the same underlying bug.", + "confidence": 0.67, + "canonical_issue_reason": "Issue 46439 is the better canonical issue because it describes the broader, more actionable failure mode in `early_initialization` and names the affected component (`linear-attention` layers) rather than a single model symptom.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 46439 is the strongest standalone issue to keep as the canonical representative for this cluster because it is more specific about the bug mechanism and likely applicable beyond one model.", + "best_issue_reason": "46439 is the strongest representative of the cluster: it captures the likely root-cause area and the generic breakage pattern, while 46441 reads as a specific manifestation for Qwen3.5.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46439", "right": "issue:46441", "accept": false, - "reason": "Both mention static cache and linear attention, but one is a shape-corruption RuntimeError and the other is a KeyError for a missing `linear_attention` entry; that looks like related subsystem fallout, not the same concrete bug." + "reason": "Related static-cache/linear-attention symptoms, but one is a shape-corruption RuntimeError and the other is a KeyError for `linear_attention`; not enough evidence they are the same exact bug or would be fixed by one change." } ] }, @@ -20948,7 +20946,7 @@ "soft_pairs": [ "issue:46551|issue:46552" ], - "prepared_review_unit_hash": "99bd32c38f378bf79d011211e99d2ecd105c4d7fe559109018c5fd6fee8cdf24", + "prepared_review_unit_hash": "ea98469aa76957dc679bcc105e9236fe0bb37b9e7c3556252bac4a330f777477", "component_budget": { "node_count": 2, "item_count": 2, @@ -20970,27 +20968,27 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are exact duplicates of the same Wav2Vec2CTCTokenizer v5 special-token bug: identical titles/body length and perfect soft-similarity evidence. #46552 is the better anchor because it is open and has inbound references.", - "confidence": 0.98, - "canonical_issue_reason": "Issue #46552 is the better canonical issue because it is still open and has inbound references.", + "summary": "These two issues are duplicates of the same Wav2Vec2CTCTokenizer v5 bug report; they have identical titles and bodies, differing only in issue number and activity state.", + "confidence": 0.99, + "canonical_issue_reason": "issue:46552 is the better canonical issue because it is the open, more-referenced copy of the same report and is more likely to be the active tracking item.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #46552 is the best issue to anchor the cluster because it is the open report and already has inbound references.", + "best_issue_reason": "issue:46552 is the best representative of the cluster since it is the same problem statement and has inbound references, making it the likelier primary tracker.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46551", "right": "issue:46552", "accept": true, - "reason": "Same title and same described tokenizer bug, supported by perfect soft-similarity evidence; these are duplicates of the same underlying issue." + "reason": "Same title, same body, and same underlying tokenizer bug; these are duplicate reports of the exact same issue." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The core duplicate call is grounded: both issues have identical titles/body length and a soft-similarity score of 1.0, and 46552 is the open issue with inbound references. However, the reasoning overstates evidence by saying 46552 has \u201cmore discussion traction\u201d than 46551; the packet shows discussion_activity is 1 for both. Please keep the canonical/best-issue rationale limited to what is explicitly supported." + "accept": true, + "feedback": "The conclusion is well grounded: the two issues have identical titles and bodies, and the packet\u2019s soft-similarity signal is maximal (Jaccard 1.0). The canonical-issue preference for issue:46552 is also supported by its open state and higher inbound references. The soft-edge verdict is conservative enough." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -21014,7 +21012,7 @@ "soft_pairs": [ "issue:46657|issue:46692" ], - "prepared_review_unit_hash": "4fd86173af0c3fb40ae1cbd033698e8543591765cd02cad2c63afe9c9e1d3d56", + "prepared_review_unit_hash": "853dc8c1addcfacb8462cbcd006c8a286402c58ac097c0c3daeab5219dfee814", "component_budget": { "node_count": 2, "item_count": 2, @@ -21036,18 +21034,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues do not look like duplicates: one reports a tensor size mismatch during object detection training, while the other reports out-of-memory after several epochs in RT-DETR v2 training.", - "confidence": 0.98, - "canonical_issue_reason": "No canonical issue selected because the reports describe different failure modes and likely different root causes.", + "summary": "These two issues are both about training-time failures, but they describe different underlying problems: a tensor size mismatch vs. out-of-memory during RT-DETR v2 training. They should not be merged as duplicates.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 46657 is the cleaner representative bug report: it names a concrete tensor-shape mismatch during object detection training. Issue 46692 is a different failure mode entirely (OOM after several epochs).", "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a good cluster representative because they are unrelated symptom reports rather than the same bug.", + "best_issue_reason": "46657 is the better cluster representative because it is a specific, actionable defect report with a clearer error class. 46692 is separate and should remain distinct.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46657", "right": "issue:46692", "accept": false, - "reason": "Different problems: shape/tensor mismatch vs memory exhaustion after epochs; not the same underlying bug." + "reason": "Different symptoms and likely different root causes: tensor size mismatch vs. memory exhaustion during RT-DETR v2 training." } ] }, @@ -21077,7 +21075,7 @@ "soft_pairs": [ "issue:46822|issue:46823" ], - "prepared_review_unit_hash": "1690885dfb0b2970aa57346c221f2c21781501524beb41afc40a3e69be756e5c", + "prepared_review_unit_hash": "5055e8d779e9d8d8ef1390c548c74427c4a1548ce06d92b9a4fe6c7e766cffa8", "component_budget": { "node_count": 2, "item_count": 2, @@ -21099,18 +21097,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two distinct loading/memory-placement bugs: one in ZeRO-3 zero.Init partitioning for a composite submodule, the other in 4-bit device_map/max_memory dispatch behavior. They share a broad theme but not the same code path or fix.", - "confidence": 0.95, - "canonical_issue_reason": "Issue 46822 is the more concrete and narrowly scoped report, with a specific ZeRO-3 zero.Init failure mode and a clear offending submodule path.", + "summary": "The two issues both concern model-loading memory/device placement, but they describe different failure modes in different pathways: ZeRO-3 partitioning of a composite submodule vs 4-bit auto device mapping ignoring max_memory and pushing modules to CPU. They are not duplicates.", + "confidence": 0.96, + "canonical_issue_reason": "Issue #46822 is the more specific and actionable report: it names the exact failing path (ZeRO-3 zero.Init), the affected composite submodule, and the observed OOM during multi-GPU load.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 46822 is the best representative because it names the exact mechanism, the composite submodule involved, and the observed OOM symptom.", + "best_issue_reason": "#46822 is the stronger canonical issue candidate because it describes a concrete partitioning bug with a clear model/component target and likely narrower root cause than #46823.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46822", "right": "issue:46823", "accept": false, - "reason": "Both concern model loading memory handling, but one is a ZeRO-3 partitioning/OOM bug and the other is a 4-bit device_map dispatch error. They do not appear to be the same underlying bug or change." + "reason": "Both are about loading-time memory placement, but #46822 is a ZeRO-3 partitioning/OOM bug while #46823 is a 4-bit device_map/max_memory dispatch bug. Different code paths and failure symptoms, so they should not be merged." } ] }, @@ -21143,12 +21141,12 @@ "issue:46906|issue:46948", "issue:46940|issue:46948" ], - "prepared_review_unit_hash": "d0fb8b9f538d6b0d7b8b131a2376ec3101cb2fde58b0dac862961af69c6b1063", + "prepared_review_unit_hash": "1113d0b2d4a906c429c481a4e5c6990d1f9d7270b1849ff02504620fbd6ea0c9", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2806, + "serialized_chars": 2807, "estimated_input_tokens": 702, "estimated_eval_tokens": 1660 }, @@ -21156,7 +21154,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2806, + "serialized_chars": 2807, "estimated_input_tokens": 702, "estimated_eval_tokens": 1660 }, @@ -21165,30 +21163,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three date-stamped Serge integration-failure triage issues, not the same bug report. They appear to be separate daily tracking threads with similar templates and wording.", - "confidence": 0.93, - "canonical_issue_reason": "Issue #46906 is the earliest of the set and has the highest inbound references, making it the best representative if one is needed for the cluster.", + "summary": "All three items are nearly identical day-stamped integration triage issues for Serge, but the differing dates strongly suggest separate failure instances rather than duplicates of the same bug report.", + "confidence": 0.95, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "#46906 is the strongest issue candidate because it is first chronologically and appears most referenced; the others look like later follow-up triage entries rather than duplicates.", + "best_issue_reason": "issue 46906 is the best representative/anchor because it is the earliest of the three and has the most inbound references, but it should not be treated as a duplicate-canonical merge target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46906", "right": "issue:46940", "accept": false, - "reason": "Same triage-series wording, but the different date stamp suggests a separate daily issue rather than the same underlying bug/change." + "reason": "Same triage template and subsystem, but the different date in the title indicates a different daily failure instance, not the same underlying issue." }, { "left": "issue:46906", "right": "issue:46948", "accept": false, - "reason": "Looks like another distinct daily triage issue; similar template text is not enough to treat it as a duplicate." + "reason": "High textual similarity, but this is a distinct date-stamped triage issue and likely corresponds to a separate integration run/failure." }, { "left": "issue:46940", "right": "issue:46948", "accept": false, - "reason": "Both are date-specific integration triage threads, so they should remain separate unless they identify the exact same failure instance." + "reason": "These are recurring daily triage reports with matching wording, yet the date-specific titles imply separate incidents rather than one shared bug." } ] }, @@ -21218,12 +21216,12 @@ "soft_pairs": [ "pull_request:40546|pull_request:42781" ], - "prepared_review_unit_hash": "64753067f063f644626a4cd414dd25696a7c3c631de8c3cc351c4fc7c7a434f9", + "prepared_review_unit_hash": "a7430a30ce33a0a618c32addac3de23617980e9efc9392f33234f2cd68a18e76", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7224, + "serialized_chars": 7223, "estimated_input_tokens": 1806, "estimated_eval_tokens": 3868 }, @@ -21231,7 +21229,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7224, + "serialized_chars": 7223, "estimated_input_tokens": 1806, "estimated_eval_tokens": 3868 }, @@ -21240,18 +21238,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two VibeVoice PRs share substantial file overlap, but they target different scopes: #40546 is the initial VibeVoice implementation, while #42781 is a later Realtime-focused expansion that also adds the acoustic tokenizer and broader integration work. They are related, but not the same underlying change.", - "confidence": 0.89, + "summary": "Two VibeVoice PRs overlap heavily in base model files, but they target different scopes: #40546 adds the initial VibeVoice implementation, while #42781 extends it to a separate realtime/acoustic-tokenizer feature set. They are related, but not duplicates.", + "confidence": 0.82, "canonical_issue_reason": null, - "canonical_pr_reason": "#42781 is the better canonical PR for this cluster because it is the more comprehensive and newer VibeVoice workstream, covering the realtime/tokenizer additions and broader API integration.", + "canonical_pr_reason": "#40546 is the better canonical PR for the cluster because it represents the original VibeVoice integration and has substantially more review activity; #42781 is a follow-up feature expansion rather than the same change.", "best_issue_reason": null, - "best_pr_reason": "#42781 is the best representative of the cluster since it appears to supersede the simpler base implementation in #40546 with a larger, more complete feature set.", + "best_pr_reason": "#40546 is the best representative of the underlying VibeVoice addition. It is the broader base implementation, whereas #42781 introduces a distinct realtime variant with extra tokenizer and generation changes.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Reject: despite overlapping VibeVoice files, #40546 implements the base model while #42781 adds Realtime and acoustic-tokenizer work; these are related but not the same concrete change and would not naturally merge into one PR." + "reason": "Reject: despite many shared VibeVoice files, #42781 is not the same concrete change. It adds realtime support and an acoustic tokenizer, plus extra generation/auto wiring, while #40546 is the initial VibeVoice implementation. They are related follow-ups, not mergeable duplicates." } ] }, @@ -21281,12 +21279,12 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "ed8d9dac04c7d6fb182af4f16c5ba81252de2836dfd93e03b7fd8513df08ad8d", + "prepared_review_unit_hash": "d866530508a04d7047f6e760f1d80ebc0fb3e0439534a123648620f65a62f96c", "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 }, @@ -21294,7 +21292,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4674, + "serialized_chars": 4675, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -21303,8 +21301,8 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two pull requests are unrelated feature additions in different model families: TimesFM covariate forecasting vs. DINOv3 image classification support. They should not be clustered as duplicates or merged into one change.", - "confidence": 0.98, + "summary": "The two pull requests are unrelated: one adds covariate forecasting support to TimesFM, while the other adds DINOv3 image-classification support. They touch different models, different code paths, and different documentation/tests, so they should not be clustered as duplicates.", + "confidence": 0.99, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, @@ -21314,7 +21312,7 @@ "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different underlying code paths and user-facing features in separate models; no evidence they fix the same bug or implement the same change." + "reason": "Different underlying changes: TimesFM forecasting with covariates vs DINOv3 image classification support. No shared bug or mergeable code-path overlap." } ] }, @@ -21399,7 +21397,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-41855-14", + "cluster_id": "cluster-41855-13", "nodes": [ "pull_request:41855", "pull_request:42668", @@ -21413,8 +21411,7 @@ "pull_request:45213", "pull_request:45655", "pull_request:46048", - "pull_request:46267", - "pull_request:46911" + "pull_request:46267" ], "soft_pairs": [ "pull_request:41855|pull_request:43823", @@ -21422,183 +21419,168 @@ "pull_request:42668|pull_request:43823", "pull_request:43251|pull_request:43823", "pull_request:43251|pull_request:44827", - "pull_request:43424|pull_request:46911", "pull_request:43424|pull_request:44827", "pull_request:43424|pull_request:43823", "pull_request:43823|pull_request:45655", "pull_request:43823|pull_request:43995", "pull_request:43995|pull_request:45213", "pull_request:43995|pull_request:44827", - "pull_request:44827|pull_request:46911", "pull_request:44827|pull_request:45170", "pull_request:44827|pull_request:46267", "pull_request:44827|pull_request:46048", + "pull_request:44827|pull_request:45655", "pull_request:44827|pull_request:45213", "pull_request:43823|pull_request:45213", - "pull_request:44827|pull_request:45655", - "pull_request:43823|pull_request:46911", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "07ee2f03855be806fd7ddb253d18575358257a220d378abd9fd8f17a5fb8f0ab", + "prepared_review_unit_hash": "0a0daf5d8432d703936e3c945841004ae98f3b047662915bee6dbec1b38233f8", "component_budget": { - "node_count": 14, - "item_count": 14, - "soft_pair_count": 21, - "serialized_chars": 38153, - "estimated_input_tokens": 9539, - "estimated_eval_tokens": 19334 + "node_count": 13, + "item_count": 13, + "soft_pair_count": 18, + "serialized_chars": 34782, + "estimated_input_tokens": 8696, + "estimated_eval_tokens": 17648 }, "budget": { - "node_count": 14, - "item_count": 14, - "soft_pair_count": 21, - "serialized_chars": 38153, - "estimated_input_tokens": 9539, - "estimated_eval_tokens": 19334 + "node_count": 13, + "item_count": 13, + "soft_pair_count": 18, + "serialized_chars": 34782, + "estimated_input_tokens": 8696, + "estimated_eval_tokens": 17648 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Only one true duplicate pair stands out: the two fixed_cross_entropy PRs both address the same cross_entropy kwargs passthrough regression. The rest are unrelated changes that merely share broad themes or files.", - "confidence": 0.98, + "summary": "This cluster is mostly false-positive similarity across unrelated PRs. The only true duplicate pair is the two `fixed_cross_entropy` PRs (`43251` and `43254`), which implement the same fix in the same function with nearly identical changes. All other pairs touch different subsystems and should stay separate.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43251 is the best canonical representative: it directly targets issue 43240 and provides the fuller fix by explicitly forwarding supported cross_entropy kwargs.", + "canonical_pr_reason": "PR #43251 is the best canonical representative for the duplicate pair because it has the explicit issue link (#43240) and the more complete PR framing for the `fixed_cross_entropy` kwargs fix.", "best_issue_reason": null, - "best_pr_reason": "PR 43251 is the strongest overall PR candidate because it cleanly fixes the regression at the loss helper level and is the most complete version of the duplicate pair.", + "best_pr_reason": "PR #43251 is the strongest global representative among the items because it directly addresses a tracked bug with a clear code-path fix; however, the rest of the cluster is too heterogeneous to treat as one duplicate family.", "soft_edge_verdicts": [ { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Mistral tokenizer helper additions vs adding a new MobileLLM model; different code paths and goals." + "reason": "Different changes: Mistral tokenizer helper methods vs adding MobileLLM model files. No shared bug or mergeable code-path overlap." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Tokenizer API additions vs an executorch dynamic-shape export test; unrelated changes." + "reason": "Tokenizer API additions are unrelated to Executorch dynamic-shape export tests." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor-from-pretrained robustness changes vs a new model implementation; not the same bug or fix." - }, - { - "left": "pull_request:43251", - "right": "pull_request:43254", - "accept": true, - "reason": "Both PRs fix the same loss-utils regression by passing supported kwargs through to nn.functional.cross_entropy for the same target issue 43240." + "reason": "Processor auto-loading robustness vs introducing a new model implementation are unrelated." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Cross-entropy kwargs plumbing vs MobileLLM model creation; unrelated." + "reason": "`fixed_cross_entropy` kwargs support is unrelated to adding MobileLLM." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "Loss-helper kwargs fix vs Mistral4 test/model work; different subsystems and problems." - }, - { - "left": "pull_request:43424", - "right": "pull_request:46911", - "accept": false, - "reason": "Executorch dynamic-shape export test vs OLMo3 RoPE configuration change; not the same change." + "reason": "Loss utility kwargs forwarding and Mistral4 test fixes are different code paths and bugs." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "Export test for dynamic shapes vs Mistral4 model/test fixes; unrelated." + "reason": "Executorch dynamic-shape export tests do not match Mistral4 test fixes." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "Executorch export test vs adding MobileLLM; different underlying work." + "reason": "Different areas: export test coverage vs new model addition." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "New MobileLLM model code vs config-resolution order fix; no shared concrete bug." + "reason": "New model support and config-resolution order fix are unrelated changes." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "MobileLLM model addition vs Falcon output-capture refactor; different code paths." + "reason": "MobileLLM model addition and Falcon output-collection refactor do not fix the same concrete problem." }, { - "left": "pull_request:43823", + "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "A new model implementation vs a model-creation skill/docs PR; not the same artifact or fix." - }, - { - "left": "pull_request:43823", - "right": "pull_request:46911", - "accept": false, - "reason": "MobileLLM model addition vs OLMo3 per-layer RoPE behavior; unrelated changes." + "reason": "Falcon model refactor is unrelated to the new-model creation skill PR." }, { "left": "pull_request:43995", "right": "pull_request:44827", "accept": false, - "reason": "Falcon standardized output-interface refactor vs Mistral4 fixes; different models and issues." + "reason": "Falcon interface refactor and Mistral4 test fixes are not the same underlying bug." }, { - "left": "pull_request:43995", - "right": "pull_request:45213", + "left": "pull_request:44827", + "right": "pull_request:45170", "accept": false, - "reason": "Falcon model refactor vs a large model-creation skill/doc PR; unrelated." + "reason": "Mistral4 test fixes and a LayerNorm typo rename across CLIP-like models are unrelated." }, { "left": "pull_request:44827", - "right": "pull_request:45170", + "right": "pull_request:46267", "accept": false, - "reason": "Mistral4 test/model fixes vs CLIP-family weight-renaming cleanup; different targets." + "reason": "Mistral4 tests do not overlap with chat response parsing updates." }, { "left": "pull_request:44827", - "right": "pull_request:45213", + "right": "pull_request:46048", "accept": false, - "reason": "Mistral4 fixes vs model-creation skill/docs; even with shared auto/modeling files, the underlying changes are unrelated." + "reason": "Mistral4 tests and TDT loss kernel integration are unrelated." }, { "left": "pull_request:44827", "right": "pull_request:45655", "accept": false, - "reason": "Mistral4 tests/model fix vs config resolution-order fix; not the same code-path problem." + "reason": "Mistral4 test work is unrelated to config resolution ordering." }, { "left": "pull_request:44827", - "right": "pull_request:46048", + "right": "pull_request:45213", "accept": false, - "reason": "Mistral4 fixes vs adding a TDT loss kernel; unrelated functionality." + "reason": "Mistral4 tests and the model-creation skill are unrelated." }, { - "left": "pull_request:44827", - "right": "pull_request:46267", + "left": "pull_request:43823", + "right": "pull_request:45213", "accept": false, - "reason": "Mistral4 fixes vs chat parsing/serve updates; different features." + "reason": "A new model implementation and an AI skill for model creation are not the same change." }, { - "left": "pull_request:44827", - "right": "pull_request:46911", + "left": "pull_request:43995", + "right": "pull_request:44827", "accept": false, - "reason": "Mistral4 fixes vs OLMo3 RoPE-per-layer work; different models and bugs." + "reason": "Falcon refactor and Mistral4 tests are different model-specific changes." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Both PRs fix the same concrete `fixed_cross_entropy` code path by adding support for passing supported kwargs (`weight`, `label_smoothing`) to `nn.functional.cross_entropy`; they are mergeable into one PR." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: only PR 43251/43254 share a concrete issue target and the same loss-utils file, so treating them as the duplicate/mergeable pair is conservative. The other pairwise rejections stay on the safe side and do not overclaim shared bug identity. " + "feedback": "The summary is well grounded in the packet and stays conservative: the non-43251/43254 pairs are plausibly unrelated, and the 43251/43254 pair is a legitimate same-function, same-bugfix duplicate that looks mergeable into one PR." }, "evaluator_used": true, "retried": false, @@ -21625,7 +21607,7 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "44896aedd3066ac62715b9db7a29d712d17a70ab2ce1225767bed53a23a179bc", + "prepared_review_unit_hash": "676121bb0415cc6e29d2be0633dd53177569852d2e2dfb9976424e114a003b39", "component_budget": { "node_count": 2, "item_count": 2, @@ -21647,18 +21629,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs both bump GitHub Actions, but they target different action families and different problems: one is a Node 24 compatibility sweep, the other is a Docker build-push action upgrade.", + "summary": "Two open PRs both update workflow action versions, but they are different upgrade efforts: one is a broad GitHub Actions/Node 24 compatibility sweep, the other is a docker/build-push-action version bump. They should not be merged as duplicates.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #42975 is the more representative PR because it is the broader, multi-workflow upgrade focused on Node 24 compatibility across several core actions.", + "canonical_pr_reason": "No single canonical PR: 42975 and 42976 are separate dependency upgrades affecting different action ecosystems and different code paths, so they are not the same change.", "best_issue_reason": null, - "best_pr_reason": "PR #42975 is the best single representative of the cluster because it covers the wider cross-workflow Actions compatibility update; #42976 is a narrower Docker-action version bump.", + "best_pr_reason": "PR 42975 is the better representative of the cluster because it is the broader GitHub Actions compatibility sweep across many workflows; 42976 is a narrower, separate build-action version bump.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "Both are workflow maintenance PRs, but they upgrade different actions and fix different compatibility paths: checkout/cache/upload-artifact for Node 24 versus docker/build-push-action for image builds. They are not the same underlying change and would not plausibly be merged into one PR." + "reason": "Reject: 42975 upgrades multiple GitHub Actions to newer versions for Node 24 compatibility, while 42976 only bumps docker/build-push-action in build-related workflows. Same broad area, but not the same concrete fix and not mergeable as one duplicate PR." } ] }, @@ -21699,12 +21681,12 @@ "pull_request:45490|pull_request:46562", "pull_request:46562|pull_request:46658" ], - "prepared_review_unit_hash": "cc0aa17e285a819852dab033a25a820183815338f670a75c5e4cfb3920128c0e", + "prepared_review_unit_hash": "840417c95fe3547e38f8df92c5cc6a89b58fc13d9eea9557f00360f983571ba1", "component_budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 8, - "serialized_chars": 21199, + "serialized_chars": 21197, "estimated_input_tokens": 5300, "estimated_eval_tokens": 10856 }, @@ -21712,7 +21694,7 @@ "node_count": 6, "item_count": 6, "soft_pair_count": 8, - "serialized_chars": 21199, + "serialized_chars": 21197, "estimated_input_tokens": 5300, "estimated_eval_tokens": 10856 }, @@ -21721,60 +21703,60 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are mostly unrelated model-addition changes that only overlap on generic docs/auto-registry scaffolding. The only strong-looking pair, Step3p7 Flash vs the WIP Step3p7 Flash repo, still does not look like the same mergeable Transformers change.", - "confidence": 0.96, + "summary": "These PRs are mostly unrelated model-addition efforts. The only near-match is Step3p7 Flash vs its mini-model/debugging companion, but that is still a different deliverable from the core Transformers implementation, so no duplicates are accepted.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46562 is the closest cluster representative: it is a substantive Transformers model integration with the clearest model-specific overlap to another artifact, though the cluster is still not duplicate-heavy overall.", + "canonical_pr_reason": "PR 46562 is the main implementation for Step 3.7 Flash in Transformers, with the broadest code surface across model, processors, auto mappings, docs, and tests.", "best_issue_reason": null, - "best_pr_reason": "PR 46562 is the best global representative of the cluster because it is the main, self-contained model-integration PR and has the strongest model-specific connection among the candidates.", + "best_pr_reason": "PR 46562 is the strongest standalone PR in this cluster: it is the substantive model integration, not just docs or support scripts, and has the clearest scope and review activity.", "soft_edge_verdicts": [ { "left": "pull_request:42978", "right": "pull_request:43448", "accept": false, - "reason": "Different models (ViT NEPA vs Molmo). Shared docs/auto files are generic integration boilerplate, not the same bug or change." + "reason": "Different models entirely: ViT NEPA vs Molmo. Shared docs/auto-registry edits are incidental." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "Different model additions (ViT NEPA vs Molmo2) with only shared registry/docs scaffolding." + "reason": "Different models entirely: ViT NEPA vs Molmo2. Similar repository scaffolding does not indicate the same change." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "Different domains and code paths (vision model vs time-series model); overlap is only generic auto/docs plumbing." + "reason": "Different models entirely: ViT NEPA vs CTSM. Only generic docs/auto files overlap." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are distinct model integrations; shared files are broad framework registration edits, not one concrete change." + "reason": "Molmo and Molmo2 are distinct model integrations, not one underlying bug or code-path fix." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Unrelated model families; common touches are only boilerplate in docs and auto mappings." + "reason": "Molmo and CTSM are unrelated model additions; shared framework files are not enough to merge them." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 and CTSM are unrelated additions; shared auto/docs files do not indicate the same underlying fix or feature." + "reason": "Molmo2 and CTSM are separate model additions with different architectures and files." }, { "left": "pull_request:45490", "right": "pull_request:46562", "accept": false, - "reason": "CTSM vs Step3p7 Flash are different model families and different implementations; only generic auto-registry files overlap." + "reason": "CTSM and Step3p7 Flash are unrelated model integrations; overlap is limited to generic registry/doc updates." }, { "left": "pull_request:46562", "right": "pull_request:46658", "accept": false, - "reason": "Both mention Step3p7 Flash, but 46562 is a Transformers model integration while 46658 is a mini-model/conversion/debug script repo; they are not the same concrete code-path change and would not plausibly merge as one PR." + "reason": "Both are Step 3.7 Flash related, but 46562 adds the Transformers model while 46658 adds mini-model/conversion/debug scripts and artifacts; they are companion efforts, not the same concrete change." } ] }, @@ -21823,7 +21805,7 @@ "pull_request:46539|pull_request:46922", "pull_request:46886|pull_request:46922" ], - "prepared_review_unit_hash": "06a7b6672419d33988eb740b1de428f38bdb6b5bdb01d57643a010659dd11ac4", + "prepared_review_unit_hash": "d86633df9a370d81601f7405b4e87868c8372ac848a13fa1d5506167422e4a23", "component_budget": { "node_count": 7, "item_count": 7, @@ -21845,108 +21827,108 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly unrelated PRs. The only clear duplicate pair is the two YOLOS fixes for the same device_map=\"auto\" mismatch; everything else targets different models, test hacks, or CI/config behavior.", - "confidence": 0.97, + "summary": "This cluster is mostly unrelated PRs across different models and subsystems. The only clear duplicate pair is the two YOLOS fixes, which target the same device-mismatch bug in the same code path.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "pull_request:46886 is the best canonical PR: it has the specific YOLOS device_map=\"auto\" title, an explicit linked issue, and a direct fix for the underlying device-mismatch path.", + "canonical_pr_reason": "PR 46922 is the strongest representative because it includes the concrete YOLOS runtime fix (`.to(hidden_states.device)`) and the same device-map stabilization change, making it the more complete version of the shared bug fix.", "best_issue_reason": null, - "best_pr_reason": "pull_request:46886 is the strongest representative for the cluster because it is the most descriptive and issue-linked YOLOS fix, while the other YOLOS PR is a narrower/alternate patch to the same bug.", + "best_pr_reason": "PR 46922 is the best overall representative: it directly fixes the YOLOS device mismatch and also keeps the supporting `_no_split_modules` adjustment, so it subsumes the narrower 46886 change.", "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "Different model test files and different symptoms; both are test-only changes, not the same underlying bug or change." + "reason": "Both are superficial test-fix PRs, but they touch different models (`whisper` vs `clip`/`vit`) and do not address the same underlying bug." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper tokenizer test fix vs. FP8/dequant loading fix; unrelated code paths." + "reason": "Whisper tokenizer/test changes are unrelated to the FP8 dequantization/loading fix in 44815." }, { "left": "pull_request:43291", "right": "pull_request:46380", "accept": false, - "reason": "Whisper decoding tests vs. generation watermarking validation/CI behavior; not the same issue." + "reason": "Whisper test/code changes are unrelated to the generation-config watermarking fix in 46380." }, { "left": "pull_request:43291", "right": "pull_request:46922", "accept": false, - "reason": "Whisper tokenizer/tests and YOLOS device-mismatch fix are unrelated." + "reason": "Whisper test changes and YOLOS device-mismatch fixes are different models and different bugs." }, { "left": "pull_request:43665", "right": "pull_request:46539", "accept": false, - "reason": "Both touch tests/CI, but one is model-test assertions and the other is Phimoe integration CI memory handling; different underlying problems." + "reason": "These are unrelated test changes for different areas: CLIP/ViT vs Phimoe CI/integration." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "Unrelated test-only edits versus FP8/loading code changes." + "reason": "The CLIP/ViT test-only PR is unrelated to the FP8 loading/dequant fix." }, { "left": "pull_request:43665", "right": "pull_request:46380", "accept": false, - "reason": "Different models and different fix targets; no shared concrete bug." + "reason": "Test-only changes in CLIP/ViT do not match the generation watermarking bugfix." }, { "left": "pull_request:43665", "right": "pull_request:46922", "accept": false, - "reason": "CLIP/ViT test modifications are unrelated to the YOLOS device-mismatch fix." + "reason": "Different models and different failure modes; no shared code-path bug." }, { "left": "pull_request:44815", "right": "pull_request:46539", "accept": false, - "reason": "FP8/dequantization logic versus Phimoe CI integration setup; unrelated." + "reason": "FP8 dequantization/loading changes are unrelated to Phimoe integration-test CI setup." }, { "left": "pull_request:44815", "right": "pull_request:46380", "accept": false, - "reason": "Different subsystems and problems: dequantization/loading versus generation config validation." + "reason": "These fixes target different subsystems: quantization/loading vs generation config validation." }, { "left": "pull_request:44815", "right": "pull_request:46922", "accept": false, - "reason": "FP8/dequant fix is unrelated to YOLOS device placement/splitting." + "reason": "No common bug: FP8 kernel/padding work is unrelated to YOLOS device placement." }, { "left": "pull_request:46380", "right": "pull_request:46539", "accept": false, - "reason": "Generation config validation fix versus CI memory/workflow patch; not the same code-path problem." + "reason": "Generation watermarking validation and Phimoe CI fixes are different issues in different paths." }, { "left": "pull_request:46380", "right": "pull_request:46922", "accept": false, - "reason": "Watermarking-config validation and YOLOS device mismatch are unrelated." + "reason": "Generation config watermarking validation is unrelated to YOLOS device-mismatch behavior." }, { "left": "pull_request:46539", "right": "pull_request:46922", "accept": false, - "reason": "CI instance-sharing workaround versus YOLOS model runtime bug; different concerns." + "reason": "Phimoe integration-test/CI adjustments do not share the same underlying bug as YOLOS device placement." }, { "left": "pull_request:46886", "right": "pull_request:46922", "accept": true, - "reason": "Both address the same YOLOS device_map=\"auto\" device mismatch in the same file and could plausibly be merged into one PR; the second adds a direct tensor-to-device fix to the same bug." + "reason": "Both PRs fix the same YOLOS device-mismatch problem on the same file and same issue target; 46922 is the more complete variant of the same change." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The only merged-duplicate callout is the YOLOS pair, which does look like a single bugfix PR: both touch the same file, share the same issue target, and the later patch is a plausible superset of the earlier one. The other soft-edge rejections are consistent with the diff previews and do not overstate relatedness." + "feedback": "The summary is well grounded in the packet. The only clearly mergeable duplicate pair is 46886/46922: both target the same YOLOS file, same issue target, and 46922 contains the same `_no_split_modules` change plus the additional device placement fix. The other pairwise rejections are conservative and consistent with the diffs, since they concern different models/subsystems and do not look like a single bugfix PR." }, "evaluator_used": true, "retried": false, @@ -21976,7 +21958,7 @@ "pull_request:43488|pull_request:43532", "pull_request:43532|pull_request:46747" ], - "prepared_review_unit_hash": "88ca2ad68d4082ae1f8d50480cf2236ec61c3a05378c7572513d35efb387c823", + "prepared_review_unit_hash": "3b526447904312f0aa6220e20920bf20e4628b27909b56d66496db7115e2b288", "component_budget": { "node_count": 3, "item_count": 3, @@ -21998,30 +21980,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three PRs are separate synthetic \"don't merge\" changes in different code paths: setup/dependency wiring, CI/comment workflow behavior, and a PR security-gate workflow trigger change. They are superficially similar as repo-bot/CI edits, but not the same underlying fix.", - "confidence": 0.91, + "summary": "The three PRs are all unrelated test/dummy changes and do not look like duplicates of one underlying bug or fix, even though two touch workflow-related files and one shares a Vitest test filename.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43532 is the broadest and most representative of the cluster: it spans multiple workflow/testing files and is the clearest example of the repo-bot/CI behavior change theme.", + "canonical_pr_reason": "PR #43532 is the most representative and broadest of the three, but it is still a distinct test-only workflow/diff-display change rather than the same fix as the others.", "best_issue_reason": null, - "best_pr_reason": "PR 43532 is the best cluster representative because it has the widest scope and touches the most related automation files, making it the most central exemplar among the three.", + "best_pr_reason": "PR #43532 has the largest and broadest diff surface, so it is the best single representative of the cluster if one must be chosen.", "soft_edge_verdicts": [ { "left": "pull_request:43488", "right": "pull_request:46747", "accept": false, - "reason": "Different changes: 43488 alters setup/dependencies and a Vit test, while 46747 changes the PR security-gate workflow trigger. No shared concrete bug or mergeable code-path fix." + "reason": "Different concrete changes: setup/dependency/version tweaks versus a workflow security-gate change. They do not fix the same code-path problem." }, { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "43488 is about package/version/setup and a test tweak; 43532 is about GitHub Actions/self-comment CI and notification/testing utilities. The shared Vit file is incidental, not the same underlying change." + "reason": "Only weak overlap via a shared Vitest test file; the actual changes are unrelated (dependency/version setup vs workflow/testing-utils edits)." }, { "left": "pull_request:43532", "right": "pull_request:46747", "accept": false, - "reason": "Both touch workflows, but they fix different mechanisms: 43532 rewires self-comment/self-scheduled/reporting behavior, while 46747 switches the security-gate from pull_request_target to pull_request. Too different to merge as one PR." + "reason": "Both touch workflows, but they alter different CI behaviors and files. They are not the same underlying bug/change and should not be merged as duplicates." } ] }, @@ -22051,7 +22033,7 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "ee4cdc1fd66c21b78cf1ecef380446cd545bdf16fdcb8708458cf828b4c925f3", + "prepared_review_unit_hash": "5f3cc201525aedeaa173bc1d8df92d97b0fee60dfa03bedbbc191467a2969bf7", "component_budget": { "node_count": 2, "item_count": 2, @@ -22073,18 +22055,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs add new segmentation pipelines, but they target different tasks and code paths: promptable concept segmentation vs promptable visual segmentation. They are related in theme, not duplicates.", - "confidence": 0.95, + "summary": "These PRs are related pipeline additions, but they implement different features: promptable concept segmentation vs promptable visual segmentation. They share scaffolding and docs structure, yet not the same concrete code path or change.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: these are distinct feature additions with different task definitions, model integrations, and pipeline implementations.", + "canonical_pr_reason": "No canonical PR: the two pull requests add different pipelines and should not be treated as duplicates.", "best_issue_reason": null, - "best_pr_reason": "No single best PR for deduping; each PR is the right place for its own feature, since they are not the same change.", + "best_pr_reason": "Neither PR is a duplicate candidate; if a representative is needed, 43613 is broader, but it is not a canonical match for 43612.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Different underlying features and concrete implementations; shared doc/pipeline scaffolding alone is not enough to treat them as duplicates." + "reason": "Different underlying features: concept segmentation vs visual segmentation. Shared docs/pipeline scaffolding is not enough to treat them as the same change." } ] }, @@ -22114,40 +22096,40 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "96a519f025af976e5d41d8802a286cdf8473b973a37485a87f8fc7eecbf219f2", + "prepared_review_unit_hash": "d04bdee0700e80e1910b21d2616383b7bc682bb950d08d4278696793f10490fa", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4293, - "estimated_input_tokens": 1074, - "estimated_eval_tokens": 2404 + "serialized_chars": 4291, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4293, - "estimated_input_tokens": 1074, - "estimated_eval_tokens": 2404 + "serialized_chars": 4291, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs both touch `src/transformers/trainer.py` but they implement unrelated trainer extension points: one changes gradient-accumulation loss scaling, the other adds a hook for signature-column filtering. They are not the same underlying bug or change.", + "summary": "The two pull requests both edit `trainer.py`, but they address different trainer extension points: one changes gradient-accumulation loss scaling, the other factors out signature-column filtering for PEFT/custom trainers. They are not the same underlying bug or mergeable change.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No true canonical duplicate here: the PRs address different trainer behaviors and should not be merged as one change.", + "canonical_pr_reason": "PR #43651 is the more concrete behavior change: it fixes/extends training loss scaling logic in a single code path and has a clear user-facing effect. PR #43663 is also a refactor/extension point, but it targets a different mechanism.", "best_issue_reason": null, - "best_pr_reason": "43663 is the cleaner representative of a standalone trainer refactor because it introduces a reusable override hook for column filtering, but it is still unrelated to 43651.", + "best_pr_reason": "PR #43651 is the best representative because it changes a specific runtime behavior in `training_step` with a clear problem statement, whereas PR #43663 is about exposing a helper for signature-column filtering and is orthogonal.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Same file and broad subsystem, but different code paths and goals: gradient-accumulation loss scaling vs. model-signature column filtering." + "reason": "Shared file only; the diffs touch different trainer subsystems and solve unrelated problems. They would not plausibly be merged as one PR." } ] }, @@ -22177,12 +22159,12 @@ "soft_pairs": [ "pull_request:43656|pull_request:43842" ], - "prepared_review_unit_hash": "62cc6b21795fecab4e2d2c652b1ff062cab14dfda38344467db1514b683f1832", + "prepared_review_unit_hash": "e37f4f10366f93a04966c26ef0c6298f34331224b45cc6539e441e204010bf2e", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 3664, + "serialized_chars": 3662, "estimated_input_tokens": 916, "estimated_eval_tokens": 2088 }, @@ -22190,7 +22172,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 3664, + "serialized_chars": 3662, "estimated_input_tokens": 916, "estimated_eval_tokens": 2088 }, @@ -22199,28 +22181,25 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests address the same TypeAdapter NameError in `src/transformers/cli/serve.py`, using different but compatible typing fixes for the same runtime annotation problem.", - "confidence": 0.73, + "summary": "Both pull requests touch `src/transformers/cli/serve.py` and are about a `TypeAdapter`-related NameError, but they fix different runtime paths: one is a broader annotation/type-hint cleanup, the other is specifically about pydantic not being installed. They are related, but not the same underlying change, so they should not be deduplicated as one PR.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43656 is the broader, more representative fix: it addresses the annotation-evaluation problem at the module level with postponed annotations and broader typing cleanup, and it has more discussion/activity.", + "canonical_pr_reason": "PR #43656 is the better canonical candidate because it is the earlier, more general fix for the CLI `TypeAdapter` NameError and has the explicit issue linkage plus more discussion activity.", "best_issue_reason": null, - "best_pr_reason": "PR #43656 is the best cluster representative because it is the more complete fix for the CLI annotation NameError and is tied to a concrete issue with more surrounding activity.", + "best_pr_reason": "PR #43656 is the strongest standalone representative of the cluster: it targets the concrete CLI `TypeAdapter` NameError directly, has the explicit issue reference, and appears to address the broader code-path problem rather than only a narrower optional-dependency typing case.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", - "accept": true, - "reason": "Same underlying bug in the same file: `TypeAdapter`-related runtime annotation evaluation causing a NameError in the CLI. The fixes are different but compatible ways to eliminate that failure and could plausibly be combined into one PR." + "accept": false, + "reason": "Do not merge as duplicates: #43656 appears to fix a general annotation/runtime NameError in `serve.py`, while #43842 specifically guards the `TypeAdapter` type for environments without pydantic. Same file and related symptom, but different concrete bug/patch." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The analyst\u2019s pairwise judgment is grounded in the packet: both PRs target the same NameError in `src/transformers/cli/serve.py`, and the fixes are compatible typing/annotation changes that could be combined into one PR. The summary stays conservative enough, despite the separate issue links and different implementation details." - }, - "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 }, @@ -22351,8 +22330,8 @@ "pull_request:44018|pull_request:44019", "pull_request:44019|pull_request:44071", "pull_request:44068|pull_request:44722", - "pull_request:44068|pull_request:44116", "pull_request:44018|pull_request:44116", + "pull_request:44068|pull_request:44116", "pull_request:44071|pull_request:44722", "pull_request:44018|pull_request:44085", "pull_request:44018|pull_request:44086", @@ -22367,8 +22346,8 @@ "pull_request:44017|pull_request:44722", "pull_request:44017|pull_request:44073", "pull_request:44068|pull_request:44071", - "pull_request:43997|pull_request:44073", "pull_request:44071|pull_request:44086", + "pull_request:43997|pull_request:44073", "pull_request:44017|pull_request:44071", "pull_request:44073|pull_request:44101", "pull_request:44073|pull_request:44161", @@ -22406,394 +22385,364 @@ "pull_request:44044|pull_request:44154", "pull_request:44024|pull_request:44076" ], - "prepared_review_unit_hash": "27579073e31adc90b58fe26d8fface90979fdc1b434d7ade665086bae110316f", + "prepared_review_unit_hash": "779efc73c98441f309806e2125ff1409eb438e2d1c8aa81802eaa92bd9c8ba83", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84517, - "estimated_input_tokens": 21130, - "estimated_eval_tokens": 42516 + "serialized_chars": 84505, + "estimated_input_tokens": 21127, + "estimated_eval_tokens": 42510 }, "budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 72, - "serialized_chars": 76950, - "estimated_input_tokens": 19238, - "estimated_eval_tokens": 38732 + "serialized_chars": 76937, + "estimated_input_tokens": 19235, + "estimated_eval_tokens": 38726 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a batch of independent model-by-model refactors to standardized output tracing/capture_outputs, not one single duplicate change. The only clear duplicate subcluster is the GPT-J/CodeGen work, plus the GPT-Neo pair.", - "confidence": 0.84, + "summary": "This cluster is mostly a batch of per-model output-tracing refactors, not a single duplicate. The only clear duplicate groups are the GPT-Neo pair and the GPT-J variants; most other pairs touch different model code paths and should remain separate.", + "confidence": 0.81, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44722 is the best representative of the duplicate subcluster: it is the latest, most complete GPT-J/CodeGen output-tracing refactor and still targets the same standardized decorator-based code path.", + "canonical_pr_reason": "PR 44722 is the best representative for the cluster\u2019s actual duplicate subgroup: it is a concrete GPT-J/codegen output-tracing refactor, explicitly targets the tracking issue, and is the most complete version of that change path.", "best_issue_reason": null, - "best_pr_reason": "PR 44722 is the strongest global representative because it subsumes the GPT-J work and is the broadest concrete implementation among the overlapping PRs.", + "best_pr_reason": "PR 44722 is the strongest global representative because it cleanly captures the shared GPT-J output-tracing refactor and overlaps with the closest duplicates while being more complete than the narrower variants.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different models and code paths (CVT/FNet vs GPT-J); same umbrella theme is not enough to make them duplicates." + "reason": "Different model families and code paths (CVT/FNet vs GPT-J); only share the broad output-tracing theme." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "Different model families (CVT/FNet vs DeBERTa); these are separate refactors, not the same underlying fix." + "reason": "CVT/FNet refactor vs DeBERTa refactor; same umbrella theme, but not the same concrete change." }, { "left": "pull_request:44066", "right": "pull_request:44085", "accept": true, - "reason": "Both touch GPT-J output tracing in the same file and implement the same decorator-based refactor; one is broader, but they are the same underlying change." + "reason": "Both are GPT-J output-tracing refactors against the same file and appear to be overlapping versions of the same change; 44066 simply also bundles CodeGen work." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "ResNet/RT-DETR ResNet vs EfficientNet are distinct implementations; only the refactor style matches, not the concrete bug/change." + "reason": "ResNet vs EfficientNet are separate model implementations; not the same underlying bug/change." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "Both are standardized output-tracing refactors, but for different model families and code paths; no shared concrete change." + "reason": "EfficientNet output tracing and GPT-J/codegen output tracing are unrelated code paths." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen vs EfficientNet are unrelated model implementations, so these are not duplicates." + "reason": "GPT-J/CodeGen refactor vs EfficientNet refactor; too different to be duplicates." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 vs DeBERTa are unrelated changes beyond the shared output-tracing pattern." + "reason": "MobileNetV2 vs DeBERTa; same refactor pattern, but different implementations and fixes." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J/CodeGen and MGP-STR are different model code paths; same refactor theme does not imply duplicate." + "reason": "GPT-J/CodeGen refactor vs MGP-STR refactor; not the same concrete change." }, { "left": "pull_request:44018", "right": "pull_request:44068", "accept": true, - "reason": "Both are GPT-Neo output-tracing refactors in the same file/test area and appear to implement the same standardized decorator migration." + "reason": "Both are GPT-Neo output-tracing refactors on the same model file and appear to be the same change expressed in two PRs." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "GPT-J/CodeGen and MPT are separate model implementations with different forward paths." + "reason": "GPT-J/CodeGen vs MPT; different model code paths." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPT-J/CodeGen vs GPT-Neo are different model-specific refactors, not the same change." + "reason": "GPT-J/CodeGen refactor and GPT-Neo refactor are related only by the general decorator migration, not the same code path." }, { "left": "pull_request:44018", "right": "pull_request:44066", "accept": false, - "reason": "GPT-Neo and GPT-J/CodeGen are different model families; they do not fix the same concrete code-path problem." + "reason": "GPT-Neo and GPT-J are different model implementations; not the same underlying change." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet output tracing and GPT-J/CodeGen output tracing are unrelated model-specific changes." + "reason": "ResNet output tracing is unrelated to GPT-J/codegen output tracing." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "MGP-STR and GPT-J/CodeGen are unrelated code paths despite both touching output collection." + "reason": "MGP-STR refactor vs GPT-J/codegen refactor; same style, different change." }, { "left": "pull_request:44019", "right": "pull_request:44085", "accept": false, - "reason": "ResNet and GPT-J are unrelated models; no shared concrete fix." + "reason": "ResNet vs GPT-J; different models and code paths." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "GPT-Neo and ResNet are unrelated model implementations." + "reason": "GPT-Neo and ResNet are separate implementations; not duplicates." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet and MPT are distinct architectures with different output-tracing changes." + "reason": "ResNet vs MPT refactors are unrelated aside from the shared tracing theme." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo and GPT-J/CodeGen are different model code paths; not the same underlying change." - }, - { - "left": "pull_request:44068", - "right": "pull_request:44116", - "accept": false, - "reason": "GPT-Neo and Flaubert/XLM are separate model families." + "reason": "GPT-Neo refactor vs GPT-J/codegen refactor; different model families." }, { "left": "pull_request:44018", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo vs Flaubert/XLM are unrelated refactors." + "reason": "GPT-Neo vs Flaubert/XLM; different code paths." }, { "left": "pull_request:44071", "right": "pull_request:44722", "accept": false, - "reason": "MPT and GPT-J/CodeGen are different model implementations; same pattern, different changes." + "reason": "MPT output tracing and GPT-J/codegen output tracing are distinct model changes." }, { "left": "pull_request:44018", "right": "pull_request:44085", "accept": false, - "reason": "GPT-Neo and GPT-J are different model code paths." + "reason": "GPT-Neo vs GPT-J; same pattern, different implementation." }, { "left": "pull_request:44018", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are unrelated implementations." + "reason": "GPT-Neo vs MGP-STR; not the same underlying bug/change." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are unrelated model code paths." }, { "left": "pull_request:44059", "right": "pull_request:44086", "accept": false, - "reason": "GPT2 and MGP-STR are different model families; these are not duplicates." + "reason": "GPT2 vs MGP-STR; both touch output handling, but not the same change." }, { "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "ViLT and Flaubert/XLM are separate model-specific output handling refactors." + "reason": "ViLT vs Flaubert/XLM; different model-specific refactors." }, { "left": "pull_request:44059", - "right": "pull_request:44722", + "right": "pull_request:44071", "accept": false, - "reason": "GPT2 and GPT-J/CodeGen are different concrete code paths." + "reason": "GPT2 and MPT are separate implementations." }, { "left": "pull_request:44017", "right": "pull_request:44161", "accept": false, - "reason": "SegFormer and LongT5 are unrelated models and changes." + "reason": "Segformer vs LongT5; unrelated model code paths." }, { "left": "pull_request:44017", "right": "pull_request:44722", "accept": false, - "reason": "SegFormer output tracing is not the same change as GPT-J/CodeGen output tracing." + "reason": "Segformer refactor is not the same concrete change as GPT-J/codegen." }, { "left": "pull_request:44017", "right": "pull_request:44073", "accept": false, - "reason": "SegFormer and VisualBert are separate model-specific refactors." - }, - { - "left": "pull_request:44068", - "right": "pull_request:44086", - "accept": false, - "reason": "GPT-Neo and MGP-STR are unrelated model code paths." + "reason": "Segformer vs VisualBert; different model families." }, { "left": "pull_request:44068", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo and MPT are distinct models." + "reason": "GPT-Neo and MPT are different model implementations." }, { "left": "pull_request:43997", "right": "pull_request:44073", "accept": false, - "reason": "RegNet and VisualBert are different architectures; no duplicate fix." + "reason": "RegNet vs VisualBert; same output-tracing theme, but different code paths." }, { - "left": "pull_request:44071", - "right": "pull_request:44086", + "left": "pull_request:44059", + "right": "pull_request:44068", "accept": false, - "reason": "MPT and MGP-STR are unrelated implementations." + "reason": "GPT2 vs GPT-Neo are separate model code paths." }, { - "left": "pull_request:44017", - "right": "pull_request:44071", + "left": "pull_request:44101", + "right": "pull_request:44161", "accept": false, - "reason": "SegFormer and MPT are different models with different output-tracing code." + "reason": "Flaubert/XLM vs LongT5; different underlying changes." }, { - "left": "pull_request:44073", - "right": "pull_request:44101", + "left": "pull_request:43997", + "right": "pull_request:44017", "accept": false, - "reason": "VisualBert and Flaubert/XLM are separate model-specific refactors." + "reason": "RegNet vs Segformer are unrelated model-specific refactors." }, { - "left": "pull_request:44073", - "right": "pull_request:44161", + "left": "pull_request:43997", + "right": "pull_request:44056", "accept": false, - "reason": "VisualBert and LongT5 are unrelated changes." + "reason": "RegNet and MPNet are different model implementations." }, { - "left": "pull_request:44059", - "right": "pull_request:44071", + "left": "pull_request:44071", + "right": "pull_request:44076", "accept": false, - "reason": "GPT2 and MPT are different model code paths." + "reason": "MPT vs ImageGPT; not the same code-path problem." }, { - "left": "pull_request:44056", + "left": "pull_request:44074", "right": "pull_request:44101", "accept": false, - "reason": "MPNet and Flaubert/XLM are unrelated model implementations." - }, - { - "left": "pull_request:44018", - "right": "pull_request:44059", - "accept": false, - "reason": "GPT-Neo and GPT2 are separate model families." + "reason": "TextNet vs Flaubert/XLM; different model-specific output changes." }, { - "left": "pull_request:44059", - "right": "pull_request:44068", + "left": "pull_request:43997", + "right": "pull_request:44018", "accept": false, - "reason": "GPT2 and GPT-Neo are different models; same refactor theme does not make them duplicates." + "reason": "RegNet vs GPT-Neo; same broad migration, not the same duplicate." }, { - "left": "pull_request:44101", + "left": "pull_request:44056", "right": "pull_request:44161", "accept": false, - "reason": "Flaubert/XLM and LongT5 are unrelated model-specific changes." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44017", - "accept": false, - "reason": "RegNet and SegFormer are different vision architectures with different forward/output code." + "reason": "MPNet vs LongT5 are unrelated output-tracing refactors." }, { "left": "pull_request:43997", - "right": "pull_request:44056", - "accept": false, - "reason": "RegNet and MPNet are unrelated model implementations." - }, - { - "left": "pull_request:44071", - "right": "pull_request:44076", + "right": "pull_request:44161", "accept": false, - "reason": "MPT and ImageGPT are different model code paths." + "reason": "RegNet vs LongT5; different model code paths." }, { "left": "pull_request:44074", - "right": "pull_request:44101", - "accept": false, - "reason": "TextNet and Flaubert/XLM are unrelated model-specific refactors." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44018", + "right": "pull_request:44161", "accept": false, - "reason": "RegNet and GPT-Neo are different model families." + "reason": "TextNet vs LongT5; not the same underlying change." }, { - "left": "pull_request:43997", + "left": "pull_request:44018", "right": "pull_request:44161", "accept": false, - "reason": "RegNet and LongT5 are unrelated changes." + "reason": "GPT-Neo vs LongT5; different model implementations." }, { - "left": "pull_request:44074", - "right": "pull_request:44161", + "left": "pull_request:44010", + "right": "pull_request:44071", "accept": false, - "reason": "TextNet and LongT5 are distinct models and code paths." + "reason": "SqueezeBERT vs MPT; different code paths." }, { "left": "pull_request:44018", "right": "pull_request:44129", "accept": false, - "reason": "GPT-Neo and SpeechT5 are unrelated model-specific changes." + "reason": "GPT-Neo vs SpeechT5; unrelated model-specific changes." }, { "left": "pull_request:44010", - "right": "pull_request:44071", + "right": "pull_request:44073", "accept": false, - "reason": "SqueezeBert and MPT are different model code paths." + "reason": "SqueezeBERT vs VisualBert; not the same concrete change." }, { - "left": "pull_request:44018", - "right": "pull_request:44129", + "left": "pull_request:43997", + "right": "pull_request:44074", "accept": false, - "reason": "GPT-Neo and SpeechT5 are separate refactors with no shared concrete bug." + "reason": "RegNet vs TextNet; separate model refactors." }, { - "left": "pull_request:44066", - "right": "pull_request:44722", - "accept": true, - "reason": "These are the same GPT-J/CodeGen output-tracing refactor; 44722 is the broader follow-up/subsuming version." + "left": "pull_request:44010", + "right": "pull_request:44018", + "accept": false, + "reason": "SqueezeBERT vs GPT-Neo; same category, different model." }, { - "left": "pull_request:44085", - "right": "pull_request:44722", - "accept": true, - "reason": "Both alter GPT-J output tracing in the same file and implement the same decorator-based migration." + "left": "pull_request:43997", + "right": "pull_request:44129", + "accept": false, + "reason": "RegNet vs SpeechT5; different implementations." }, { - "left": "pull_request:44085", + "left": "pull_request:44071", "right": "pull_request:44129", "accept": false, - "reason": "GPT-J and SpeechT5 are unrelated models." + "reason": "MPT vs SpeechT5; unrelated output-tracing changes." }, { "left": "pull_request:43996", - "right": "pull_request:44007", + "right": "pull_request:44161", "accept": false, - "reason": "CVT/FNet and ResNet/RT-DETR ResNet are different model implementations." + "reason": "CVT/FNet vs LongT5; same pattern, different model code." }, { "left": "pull_request:43996", - "right": "pull_request:44129", + "right": "pull_request:44017", "accept": false, - "reason": "CVT/FNet and SpeechT5 are unrelated model-specific refactors." + "reason": "CVT/FNet vs Segformer; not the same change." }, { - "left": "pull_request:44044", - "right": "pull_request:44129", + "left": "pull_request:44010", + "right": "pull_request:44044", "accept": false, - "reason": "DeBERTa and SpeechT5 are different model code paths." + "reason": "SqueezeBERT vs DeBERTa; different model-specific refactors." }, { "left": "pull_request:44044", - "right": "pull_request:44154", + "right": "pull_request:44074", "accept": false, - "reason": "DeBERTa and Vits are unrelated model-specific output tracing changes." + "reason": "DeBERTa vs TextNet; same output-tracing style, but distinct code paths." }, { "left": "pull_request:44024", "right": "pull_request:44076", "accept": false, - "reason": "FocalNet and ImageGPT are different model families; same standardized-output theme is too broad." + "reason": "FocalNet and ImageGPT are different models; not a duplicate." } ] }, "evaluator_result": { "accept": true, - "feedback": "The clustering summary is grounded in the packet and stays appropriately conservative. The accepted duplicate pairs are well-supported by shared files/model families and the same output-tracing refactor, and the negative verdicts correctly avoid over-merging unrelated model-specific changes. The representative PR choice for the GPT-J/CodeGen subcluster is also defensible from the packet." + "feedback": "The summary is well grounded in the packet and stays conservative about the soft edges. The only duplicate groupings it elevates are the GPT-Neo pair and the GPT-J/CodeGen variants, which are the clearest mergeable overlaps in the evidence. The many cross-model pairs are correctly left as non-duplicates." }, "evaluator_used": true, "retried": false, @@ -22828,20 +22777,20 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "6fcd47d47e4cff6c1bbc6ef56befb33dd80b3c7b6bd5c62b714bab8c7aaee0e6", + "prepared_review_unit_hash": "82b80c6765e1a80e337691a8976f9bbd062d63f4bf9e6d67c91940e31f266f84", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84517, - "estimated_input_tokens": 21130, - "estimated_eval_tokens": 42516 + "serialized_chars": 84505, + "estimated_input_tokens": 21127, + "estimated_eval_tokens": 42510 }, "budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 5, - "serialized_chars": 13067, + "serialized_chars": 13068, "estimated_input_tokens": 3267, "estimated_eval_tokens": 6790 }, @@ -22850,42 +22799,42 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These PRs are all related to the same broad refactor theme\u2014migrating model code to standardized output capture/tracing\u2014but they are model-specific changes, not duplicates of one another.", - "confidence": 0.96, + "summary": "These are all model-specific refactors to standardized output tracing/capture, but they target different architectures and code paths. They read as a family of similar cleanup PRs rather than duplicate fixes or mergeable single changes.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: each change targets a different model implementation and file, so they are parallel refactors rather than the same PR duplicated.", + "canonical_pr_reason": "PR 44073 is a good representative of the cluster because it is a straightforward, model-specific application of the standardized output tracing refactor and clearly reflects the shared theme across the set.", "best_issue_reason": null, - "best_pr_reason": "44129 is the most representative single PR of the theme because it has the most explicit scope/issue linkage and a substantial output-tracing refactor, but it is still not a duplicate of the others.", + "best_pr_reason": "PR 44073 is the best global representative because it is cleanly scoped, uses the new capture_outputs/can_return_tuple pattern, and most clearly captures the common refactor theme without extra model-specific complexity.", "soft_edge_verdicts": [ { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "Same migration theme, but different models (VisualBert vs TextNet) and different file/test scope; not the same concrete code-path change." + "reason": "Same refactor theme, but VisualBert and TextNet are different model implementations and the diffs touch different files/tests; not the same underlying change and not something to merge into one PR." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "Both refactor output tracing, but MPNet and VisualBert are distinct implementations with different layers/hooks; not mergeable as one duplicate PR." + "reason": "Both are standardized output tracing refactors, but MPNet and VisualBert have distinct model-specific output plumbing; these are separate PRs for different code paths, not duplicates." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "Different models and different refactor shapes (MPNet output recording setup vs TextNet forward simplification); not the same underlying fix." + "reason": "Same broad migration pattern, but MPNet and TextNet are different architectures with different tracing hooks; the changes are not the same concrete fix." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are unrelated model code paths; similar standardized-output migration does not make them duplicates." + "reason": "RegNet and SqueezeBert are unrelated model code paths; despite similar decorator/import changes, they are separate refactors and not a single mergeable PR." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "Both touch output tracing, but SqueezeBert and SpeechT5 are separate model-specific refactors with different mechanics and no shared concrete bug." + "reason": "Both modernize output handling, but SqueezeBert vs SpeechT5 are different models with different forward signatures and explicit targets; not the same bug or change." } ] }, @@ -22965,274 +22914,274 @@ "pull_request:44028|pull_request:44030", "pull_request:44029|pull_request:44030" ], - "prepared_review_unit_hash": "7bb0ee1c866a5f565cf141a3a950e8c41055788da3278852206064ff223cf66f", + "prepared_review_unit_hash": "6e5113c188ad6fbaae378d319ad0badca36ee7a3c2f7bc9346e1ccf3190fe773", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40163, - "estimated_input_tokens": 10041, - "estimated_eval_tokens": 20338 + "serialized_chars": 40167, + "estimated_input_tokens": 10042, + "estimated_eval_tokens": 20340 }, "budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40163, - "estimated_input_tokens": 10041, - "estimated_eval_tokens": 20338 + "serialized_chars": 40167, + "estimated_input_tokens": 10042, + "estimated_eval_tokens": 20340 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs share a broad \"output tracing\" theme, but they are model-specific refactors across unrelated code paths. I do not see any true duplicate PR pairs here.", - "confidence": 0.92, + "summary": "These PRs all follow the same broad \"refactor output tracing\" theme, but they apply to different model families and different forward paths. They are not duplicates of one another, so none of the soft edges should be merged as the same change.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44003 is the strongest representative of the cluster because it makes the most concrete output-tracing plumbing change across two closely related model files, but it is still not a duplicate of the others.", + "canonical_pr_reason": "No single canonical PR: the items are separate model-specific refactors, not one shared bug fix or implementation change.", "best_issue_reason": null, - "best_pr_reason": "PR 44003 is the best overall representative of the theme: it has the clearest, most substantive refactor around output tracing and touches a related model pair rather than a single isolated file.", + "best_pr_reason": "PR 44025 is the broadest single representative in the set (it touches two related Depth Anything files), but it is still just one of many independent refactors.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Different models and code paths; same refactor theme only." + "reason": "Different model families and forward paths; same refactor pattern, not the same concrete change." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "timm_backbone and vision_text_dual_encoder are unrelated architectures." + "reason": "timm_backbone vs vision_text_dual_encoder are unrelated code paths despite similar tracing cleanup." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "Different model families with no shared implementation." + "reason": "timm_backbone and univnet are separate models with different output behavior." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "Different subsystems; not the same concrete bug or change." + "reason": "Different architectures (timm_backbone vs mamba/falcon_mamba); not the same bug or fix." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "Separate model code paths; only the tracing refactor naming matches." + "reason": "Unrelated models; shared tracing refactor theme is too broad to merge as duplicates." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "mobilenet_v1 and univnet are unrelated models." + "reason": "mobilenet_v1 and univnet are unrelated forward implementations." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "Different architectures and different forward/output plumbing." + "reason": "Different model families and distinct output-tracing code paths." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "Separate model families; not one mergeable change." + "reason": "MobileNetV1 vs Depth Anything are unrelated model implementations." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "No shared code path; both are just output-tracing refactors." + "reason": "Separate model wrappers; same refactor pattern, not one shared change." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "Different model stacks and different affected files." + "reason": "mobilenet_v1 and speech_encoder_decoder are different code paths and model types." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "vision_text_dual_encoder and univnet are unrelated." + "reason": "vision_text_dual_encoder and univnet are unrelated models." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Distinct model families; no same underlying fix." + "reason": "Different model families; no shared concrete bug or fix." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "Different code paths and model types." + "reason": "vision_text_dual_encoder vs mamba is only a shared refactor theme, not a duplicate change." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "Separate model-specific refactors, not one bug." + "reason": "Different forward paths in unrelated model families." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "UnivNet and UperNet are unrelated implementations." + "reason": "univnet and upernet are separate models with separate output semantics." }, { "left": "pull_request:44001", "right": "pull_request:44003", "accept": false, - "reason": "Different model code paths; no mergeable overlap." + "reason": "Different architectures and code paths; not mergeable as one PR." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "No shared underlying change beyond generic tracing cleanup." + "reason": "UnivNet and Depth Anything are unrelated model implementations." }, { "left": "pull_request:44002", "right": "pull_request:44029", "accept": false, - "reason": "UperNet and RWKV are unrelated model families." + "reason": "upernet and rwkv are unrelated; same refactor pattern is not enough." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "Different subsystems and refactor targets." + "reason": "Different model families and different forward signatures." }, { "left": "pull_request:44003", "right": "pull_request:44028", "accept": false, - "reason": "Mamba/FalconMamba vs SuperPoint are unrelated code paths." + "reason": "mamba/falcon_mamba and superpoint are unrelated code paths." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "Output-tracing cleanup only; different model implementations." + "reason": "Different subsystems entirely: mamba vs codegen." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "timm_backbone and CodeGen are unrelated models." + "reason": "timm_backbone and codegen are unrelated models." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "UnivNet and CodeGen do not share the same concrete fix." + "reason": "UnivNet and CodeGen fix different forward/output behaviors." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "Different model families and different forward logic." + "reason": "upernet and codegen are separate changes in unrelated models." }, { "left": "pull_request:44000", "right": "pull_request:44004", "accept": false, - "reason": "vision_text_dual_encoder vs CodeGen are unrelated changes." + "reason": "vision_text_dual_encoder vs codegen are unrelated code paths." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "DepthAnything/PromptDepthAnything and VisionEncoderDecoder are separate implementations." + "reason": "Both are output-tracing refactors, but they target different wrapper classes and code paths." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "Different model families; not one underlying bug." + "reason": "Depth Anything and Speech Encoder-Decoder are separate model families." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "DepthAnything refactor vs SuperPoint refactor are unrelated." + "reason": "Different tasks and forward APIs; not the same underlying change." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "Different model code paths; same theme only." + "reason": "Depth Anything and RWKV are unrelated implementations." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "Separate models and separate output-capture behavior." + "reason": "Different model families with separate output handling." }, { "left": "pull_request:44026", - "right": "pull_request:44027", + "right": "pull_request:44029", "accept": false, - "reason": "VisionEncoderDecoder and SpeechEncoderDecoder are different stacks." + "reason": "vision_encoder_decoder and rwkv are unrelated code paths." }, { "left": "pull_request:44026", - "right": "pull_request:44028", + "right": "pull_request:44027", "accept": false, - "reason": "Different models with no shared concrete bug." + "reason": "Both are encoder-decoder wrappers, but they are separate classes and not one duplicate fix." }, { "left": "pull_request:44026", - "right": "pull_request:44029", + "right": "pull_request:44028", "accept": false, - "reason": "Different model families and call paths." + "reason": "vision_encoder_decoder and superpoint are unrelated models." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "VisionEncoderDecoder and DPR are unrelated." + "reason": "Different model families and distinct output APIs." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "SpeechEncoderDecoder and SuperPoint are unrelated." + "reason": "speech_encoder_decoder and superpoint are unrelated." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "Different implementations and output signatures." + "reason": "Separate model families with no shared concrete bug." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "SpeechEncoderDecoder vs DPR is not the same change." + "reason": "speech_encoder_decoder and dpr are unrelated changes." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "SuperPoint and RWKV do not share a code path." + "reason": "superpoint and rwkv are unrelated implementations." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "Different model families; no duplicate fix." + "reason": "superpoint and dpr fix different output paths." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "RWKV and DPR are unrelated refactors." + "reason": "rwkv and dpr are different model families with different code paths." } ] }, @@ -23262,7 +23211,7 @@ "soft_pairs": [ "pull_request:44662|pull_request:45401" ], - "prepared_review_unit_hash": "9d12b2e8952ae904169a268fee20217caa26952133195064e74f21b5c8516897", + "prepared_review_unit_hash": "e284d220010712ec4e4fe4135198590af701109725b59185177d86fd36fe81c9", "component_budget": { "node_count": 2, "item_count": 2, @@ -23284,18 +23233,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated model-addition PRs that only overlap on shared Transformers boilerplate and auto-mapping files. PenguinVL and Voxtral TTS target different models, different code paths, and different documentation/tests, so they should not be clustered as duplicates.", - "confidence": 0.96, + "summary": "Two unrelated model-integration pull requests: one adds PenguinVL, the other adds Voxtral TTS. They overlap only in shared Transformers registration/docs plumbing, not the same underlying feature or bug.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45401 is the cleaner representative: it has an explicit issue target and a focused, model-specific implementation for Voxtral TTS.", + "canonical_pr_reason": "Pick PR #45401 as the canonical PR: it is tied to an explicit issue (#45027), has a narrower and clearly scoped feature (Voxtral TTS), and is the more targetable integration request.", "best_issue_reason": null, - "best_pr_reason": "PR 45401 is the best cluster representative because it is explicitly linked to an issue and appears to be a self-contained support addition for one model.", + "best_pr_reason": "PR #45401 is the strongest standalone representative because it has an explicit issue link and a focused model-specific implementation. PR #44662 is also a valid model add, but it is not the better canonical choice for the cluster.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different models and different features: PenguinVL implementation versus Voxtral TTS support. Shared auto/config/doc scaffolding is generic and not enough to indicate the same underlying change." + "reason": "Different model additions with different implementation stacks and targets; shared auto/docs registration is incidental and not enough to indicate the same change." } ] }, @@ -23325,12 +23274,12 @@ "soft_pairs": [ "pull_request:44965|pull_request:46845" ], - "prepared_review_unit_hash": "d8d04aa838aaefa08a8eeda97e04cffb1194e3b057d607de164dde0a22578028", + "prepared_review_unit_hash": "73a7a5661400c66a132ba01e9542d93cb1eaa5f24e41a32c771c0e89fdb6b09c", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4219, + "serialized_chars": 4218, "estimated_input_tokens": 1055, "estimated_eval_tokens": 2366 }, @@ -23338,7 +23287,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4219, + "serialized_chars": 4218, "estimated_input_tokens": 1055, "estimated_eval_tokens": 2366 }, @@ -23347,18 +23296,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two pull requests are not duplicates: one is a workflow experiment around converting PRs to draft, while the other is a CI-runner/branch-change PR with an unrelated code edit. They do not fix the same bug or implement the same change.", - "confidence": 0.98, + "summary": "The two PRs are not duplicates: one appears to rewrite a docs-related workflow into a draft-conversion action, while the other changes CI routing plus a model utility file in an unrelated smaller-runner experiment.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: the changes target different code paths and different behaviors, so there is no single underlying PR to collapse them into.", + "canonical_pr_reason": "PR 46845 is the better representative of this cluster because it has a clearer, coherent CI-oriented change set, whereas PR 44965 looks like an unrelated workflow rewrite with incomplete/garbled diff context.", "best_issue_reason": null, - "best_pr_reason": "pull_request:46845 is the more coherent standalone PR because it has a clearer CI-related scope and title; pull_request:44965 appears incomplete and its workflow edit is unrelated.", + "best_pr_reason": "PR 46845 is the most suitable global representative only in the sense that it is the more substantive and internally consistent PR; however, it should not be treated as a duplicate of PR 44965.", "soft_edge_verdicts": [ { - "left": "pull_request:44965 (try)", - "right": "pull_request:46845 (Try smaller group)", + "left": "pull_request:44965", + "right": "pull_request:46845", "accept": false, - "reason": "Different goals and files: 44965 rewrites a docs workflow to change PR handling, while 46845 modifies CI caller configuration and includes an unrelated code tweak. They do not appear to be the same underlying change and should not be merged as duplicates." + "reason": "Different underlying changes: 44965 is about converting PRs to draft via a documentation workflow, while 46845 changes the PR CI caller and touches model code. They do not fix the same code-path problem and would not plausibly merge into one PR." } ] }, @@ -23388,12 +23337,12 @@ "soft_pairs": [ "pull_request:44979|pull_request:45363" ], - "prepared_review_unit_hash": "a570969325c95b6a1713b004b5be4db42fb031a60dfb7185644421f44e8a7966", + "prepared_review_unit_hash": "406d73e7b80d504452faa6e73e26ec995bdc753b1aa70bf1ef2d76779d7ab8e7", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4408, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -23401,7 +23350,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4408, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -23410,18 +23359,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two related but distinct feature PRs around fusion: one adds a generic module fusion API, the other wires up n-to-1 kernel fusion through KernelConfig. They are not duplicates of the same concrete change.", - "confidence": 0.84, + "summary": "These are related fusion-feature PRs, but they address different layers: PR #44979 introduces a generic Module Fusion API, while PR #45363 implements n-to-1 kernel fusion through KernelConfig and related integration points. I would not treat them as duplicates.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #44979 is the cleaner canonical representative because it introduces the foundational Module Fusion API and tests, whereas #45363 is a separate, more specific kernel-fusion integration built on that general idea.", + "canonical_pr_reason": "No single canonical PR stands out as a duplicate target: #44979 is the foundational module-fusion API, whereas #45363 is a separate KernelConfig-based fusion implementation with different files and scope.", "best_issue_reason": null, - "best_pr_reason": "PR #44979 is the best single representative of the cluster since it defines the core fusion abstraction; #45363 is a downstream application/integration rather than the same mergeable change.", + "best_pr_reason": "#45363 is the better representative for the cluster\u2019s concrete user-facing change because it wires fusion into model loading/integration code and includes an example; #44979 is more of an enabling API layer than the end-user feature.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "Both concern fusion, but #44979 adds a new generic module-fusion API while #45363 implements n-to-1 kernel fusion via KernelConfig and integration changes. Different code paths, different scope, and not the same underlying bug or change." + "reason": "They share the broad theme of fusion, but the code paths and scope differ: #44979 adds a generic Module Fusion API, while #45363 implements n-to-1 kernel fusion via KernelConfig and integration/model-loading changes. They could plausibly coexist in one repo without being the same change." } ] }, @@ -23451,12 +23400,12 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "35e0a5d84995bbb57f4a8434323263a96f683d15462d478004b895b62e8f69ad", + "prepared_review_unit_hash": "d7ce7bf95280189d61471821482be11022cbbf4d5e58dc0cc71556bca089cbc4", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4414, + "serialized_chars": 4413, "estimated_input_tokens": 1104, "estimated_eval_tokens": 2464 }, @@ -23464,7 +23413,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4414, + "serialized_chars": 4413, "estimated_input_tokens": 1104, "estimated_eval_tokens": 2464 }, @@ -23473,18 +23422,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs address the same Mistral regex patch AttributeError in `tokenization_utils_tokenizers.py`; 45317 is the more complete candidate because it includes the code fix plus a test update.", - "confidence": 0.96, + "summary": "Both PRs fix the same Mistral pre-tokenizer AttributeError by switching from `backend_tokenizer.pre_tokenizer` to `tokenizer.pre_tokenizer`; 45317 appears to be the more complete version because it adds tests.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "45317 is the better canonical PR: it has the same core fix as 45086 and also adds test coverage, making it the more complete representation of the change.", + "canonical_pr_reason": "45317 is the stronger canonical PR because it contains the same code fix as 45086 plus test coverage, making it the more complete change set.", "best_issue_reason": null, - "best_pr_reason": "45317 is the best overall PR because it fixes the same concrete bug as 45086 while also adding tests, so it is the most merge-ready version.", + "best_pr_reason": "45317 is the best PR to keep: it addresses the same concrete bug and includes validation in tests, while 45086 is the narrower earlier patch.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Same underlying bug fix in the same function/file, same issue target, and the second PR appears to be an expanded version of the first with tests added." + "reason": "Same underlying bug and same fix in the same file; 45317 is effectively a superset with tests, so these are duplicate PRs." } ] }, @@ -23514,12 +23463,12 @@ "soft_pairs": [ "pull_request:45115|pull_request:45569" ], - "prepared_review_unit_hash": "a6dc14f2d63a77c1de448e243a52f4682f26c1cca16ec553e3e863e7650d7bcb", + "prepared_review_unit_hash": "4104d51c2210dbe6e1cfd2fafa46f7bb5cd1314f48d67f367a1399b4165fa165", "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 }, @@ -23527,7 +23476,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6066, + "serialized_chars": 6065, "estimated_input_tokens": 1517, "estimated_eval_tokens": 3290 }, @@ -23536,18 +23485,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs touch overlapping NemotronH code, but they address different changes: one is a refactor/conversion-mapping update for inheriting GraniteMoeHybrid, while the other splits NemotronH into dedicated dense/sparse model families and updates docs/tests. They are not the same underlying change and should not be merged as duplicates.", + "summary": "Both PRs touch NemotronH internals, but they address different changes: #45115 is a narrow checkpoint-conversion/mapping refactor, while #45569 is a broader model split and documentation/auto-registry update. They do not look like the same underlying bug or mergeable single fix.", "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45569 is the broader, more concrete implementation for the NemotronH dense/sparse split, with dedicated model files, auto-mapping changes, docs, and tests.", + "canonical_pr_reason": "PR #45569 is the better canonical PR for the cluster because it represents the more complete NemotronH architecture change, including new dense/sparse model packages, auto mappings, docs, and tests. PR #45115 is a narrower conversion-mapping follow-up and does not subsume the broader change.", "best_issue_reason": null, - "best_pr_reason": "PR 45569 is the strongest representative of the cluster because it covers the larger end-user-facing model restructuring and has the clearest product scope.", + "best_pr_reason": "#45569 is the strongest representative of the cluster since it is the more substantial end-to-end NemotronH update; #45115 is too specialized and only overlaps on a few shared NemotronH files.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "Shared NemotronH filenames alone are not enough: 45115 is a refactor/conversion-mapping inheritance change, while 45569 introduces separate NemotronH dense/sparse model variants and documentation. Different concrete code-path problems." + "reason": "Shared NemotronH files are incidental. #45115 only changes conversion mappings, while #45569 restructures the model into dense/sparse variants plus docs and auto-mapping updates. They fix different problems and would not plausibly be merged as one PR." } ] }, @@ -23577,7 +23526,7 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "b1f30e70a3257879c30be0ae6253b7699f3f74d0ec7e3bc00a62b0cb266f2e7e", + "prepared_review_unit_hash": "f9aedad3ef3bebe9d928406e27061da2fc1b7dd751e2de8273edf9af0f47c9d1", "component_budget": { "node_count": 2, "item_count": 2, @@ -23599,18 +23548,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are closely related but not true duplicates: PR #45333 is a narrower per-layer configuration piece, while PR #45332 is a broader umbrella adding heterogeneous model support including modeling logic and other integration changes.", - "confidence": 0.89, + "summary": "Both PRs are about the same heterogeneous-model feature family; #45333 is a narrower configuration-only subset of the broader end-to-end support in #45332.", + "confidence": 0.79, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical duplicate candidate among the two; #45332 is the broader umbrella PR, while #45333 is a config-only subset rather than the same change.", + "canonical_pr_reason": "#45332 is the broader, more complete PR: it includes the per-layer configuration work plus modeling, cache/masking, and integration changes needed for heterogeneous model support.", "best_issue_reason": null, - "best_pr_reason": "#45332 is the best single representative because it covers the full heterogeneous-model feature set and subsumes the config work in #45333.", + "best_pr_reason": "#45332 is the best representative because it subsumes the config-only work in #45333 and covers the full concrete feature path rather than just one piece of it.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", - "accept": false, - "reason": "Shared files and theme, but the scopes differ materially: #45333 focuses on per-layer configuration, while #45332 also changes modeling/cache/masking paths. They are related PRs, not the same concrete change." + "accept": true, + "reason": "Accept: #45333 looks like a narrower split-out of the same heterogeneous configuration feature covered by #45332; they share the core config files and the latter adds the modeling implementation on top." } ] }, @@ -23640,7 +23589,7 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "909c4df211808e330cd38cc25f9b95254f802b02f72d0e81cd8ba824aca7e8fa", + "prepared_review_unit_hash": "863f9a1c10407319a49d3f66da962fab2c570f37cb400b9d0473a3113576ad31", "component_budget": { "node_count": 2, "item_count": 2, @@ -23662,18 +23611,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related typing-focused PRs, but they are not the same change: one is a broad type-checking sweep across many Transformers files, while the other adds/adjusts specific typing helpers and protocols around modeling utilities. The soft duplicate edge should be rejected.", - "confidence": 0.93, + "summary": "Two separate typing-focused PRs with some overlapping files, but they address different changes: one is a broad repo-wide type-checking sweep, the other adds/adjusts shared typing protocols and modeling-utils-related annotations. Not a duplicate pair.", + "confidence": 0.87, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45415 is the broader umbrella change: it enables type checking across many `src/transformers/*` modules and adjusts type-ignore policy, making it the more representative PR for the cluster topic.", + "canonical_pr_reason": "PR 45425 is the better representative because it is more concrete and narrowly scoped around shared typing helpers and model-utils integration, rather than a broad type-checking sweep.", "best_issue_reason": null, - "best_pr_reason": "PR #45415 is the best representative PR because it covers the larger cross-cutting typing cleanup, whereas PR #45425 is a narrower helper/type-definition follow-up.", + "best_pr_reason": "PR 45425 is the strongest standalone candidate: it has a clearer, more cohesive technical change set and is less diffuse than the repo-wide typing PR.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Different underlying changes: #45415 is a broad repo-wide type-checking sweep, while #45425 adds specific typing protocols/aliases and modeling-utils-related adjustments. They overlap on typing files but are not the same concrete fix and would not plausibly merge as one PR." + "reason": "They overlap in typing-related files, but 45415 is a broad type-checking pass across many modules while 45425 adds specific typing protocols and targeted modeling_utils/peft annotations. Different concrete changes, not one mergeable duplicate PR." } ] }, @@ -23703,7 +23652,7 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "dc797ce74ef2c86b15547bb2c9b5b563b26a9ce59215c67f83a64c946aab5ef4", + "prepared_review_unit_hash": "e5b7571779251ae3718d61ef10054500f4913bb6b00d13c43b3f17da7f423dac", "component_budget": { "node_count": 2, "item_count": 2, @@ -23725,18 +23674,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related only at a broad expert-parallelism level, but they address different loading bugs in different distributed stacks: ZeRO-3/accelerate vs FSDP2/rank-0 broadcast. They should not be merged as duplicates.", - "confidence": 0.96, + "summary": "These PRs are related by expert-parallel loading, but they fix different bugs: #45548 is about preserving/loading EP config for DeepSpeed ZeRO-3 via accelerate, while #45662 fixes expert weights being overwritten during FSDP2/rank-0 broadcast.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45548 is the cleaner representative of the cluster: it targets a specific EP loading path (DeepSpeed ZeRO-3 via accelerate) and has a narrower, self-contained fix around config serialization and model state detection.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR 45548 is the better cluster anchor because it is the earlier, more focused fix; PR 45662 is a separate EP bug in a different distributed code path and is not a duplicate.", + "best_pr_reason": "pull_request:45662 is the stronger representative of the cluster\u2019s EP-loading theme because it targets a concrete distributed-loading corruption bug with a specific code-path fix, whereas pull_request:45548 is a narrower compatibility/serialization patch.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "Both mention expert parallelism, but the concrete failures differ: 45548 fixes EP + DeepSpeed ZeRO-3 loading/accelerate config handling, while 45662 fixes EP + FSDP2 expert parameters being overwritten by rank-0 broadcast. They touch different mechanisms and are not one mergeable change." + "reason": "Same broad EP loading area, but not the same underlying bug: #45548 handles config serialization/ZeRO-3 launch compatibility, while #45662 prevents rank-0 broadcast from overwriting experts under FSDP2. They would not plausibly merge into one single fix." } ] }, @@ -23766,7 +23715,7 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "dc986869a5a66ec426dbcf75b99aa8ef362feadcd77ee4ba95e8f2a7db4db054", + "prepared_review_unit_hash": "a2fbb038975fb9eaf1f9f8ad4098b3b5f184793e395ab7c3c4563fd38a3fff1c", "component_budget": { "node_count": 2, "item_count": 2, @@ -23788,18 +23737,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests address the same xdist/captured_info collision bug in testing debug output; PR 45645 is the broader, more complete fix, while PR 45639 is a narrower subset of the same change.", - "confidence": 0.92, + "summary": "Two open PRs target the same xdist/captured_info problem and share the same testing utility files; one is a broader superset that also updates CI log collection for multiple output files.", + "confidence": 0.84, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45645 is the better canonical PR because it includes the core xdist-safe output fix plus the CI workflow change needed to collect multiple captured_info files, making it the more complete implementation of the bugfix.", + "canonical_pr_reason": "PR 45645 is the better canonical representative because it covers the same captured_info xdist collision fix and also updates the CI workflow to collect all split log files, making it the more complete change.", "best_issue_reason": null, - "best_pr_reason": "PR 45645 is the strongest representative of the cluster since it addresses the same concrete xdist collision problem and also preserves CI debug log collection, so it best captures the full intended change.", + "best_pr_reason": "PR 45645 is the strongest cluster member: it appears to subsume the narrower testing_utils-only fix in 45639 while preserving the same underlying debug-log issue.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", "accept": true, - "reason": "Both PRs target the same underlying xdist collision in captured_info/debug logging; 45645 appears to be a superset of 45639 rather than a different change, so they can be treated as duplicates/same fix." + "reason": "Both PRs fix the same captured_info/xdist artifact collision in the same testing utility path; 45645 is a plausible superset that adds the CI-side handling needed for the same change." } ] }, @@ -23829,7 +23778,7 @@ "soft_pairs": [ "pull_request:45720|pull_request:46259" ], - "prepared_review_unit_hash": "b5a5f02bc9b662e292e586ee2efc593d0e326325394a692a612117b82783ff93", + "prepared_review_unit_hash": "9f04de5e9eb363615620820b00ea32db9c7a858431670981cd4bcd149dfcc3f0", "component_budget": { "node_count": 2, "item_count": 2, @@ -23851,18 +23800,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both items are unrelated lint-rule PRs in transformers models. They share some model files because each rule is applied broadly, but they address different code paths: TRF003 adds `can_return_tuple`/`return_dict` handling in forward signatures, while TRF018 changes model weight initialization logic.", - "confidence": 0.97, + "summary": "These PRs are both linter-rule implementations in transformers modeling code, but they address different problems: TRF003 return-tuple/return_dict handling vs TRF018 model weight initialization. The overlap in filenames is incidental, not evidence of the same fix.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR for duplication: these are distinct changes serving different lint rules and different fixes, despite overlapping model files.", + "canonical_pr_reason": "PR #46259 is the better representative only because it is the larger, more central linter-rule change with much more review activity; it still does not duplicate #45720.", "best_issue_reason": null, - "best_pr_reason": "PR 45720 is the better representative artifact because it is narrower and has a more concrete, self-contained code-path change across a set of forward methods.", + "best_pr_reason": "PR #46259 is the stronger cluster representative due to broader diff scope and higher review activity, but it is not a duplicate of #45720.", "soft_edge_verdicts": [ { "left": "pull_request:45720", "right": "pull_request:46259", "accept": false, - "reason": "Different underlying changes: TRF003 forward-signature/tuple-return handling vs TRF018 weight initialization behavior. Shared files reflect broad lint sweeps, not the same bug or mergeable fix." + "reason": "Different concrete changes: #45720 adds can_return_tuple/return_dict handling across model forwards, while #46259 adds TRF018 weight-init logic. Shared model files and linter context are too broad to indicate the same underlying bug or mergeable PR." } ] }, @@ -23950,63 +23899,48 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-45932-3", + "cluster_id": "cluster-45932-2", "nodes": [ "pull_request:45932", - "pull_request:46100", - "pull_request:46891" + "pull_request:46100" ], "soft_pairs": [ - "pull_request:45932|pull_request:46100", - "pull_request:45932|pull_request:46891", - "pull_request:46100|pull_request:46891" + "pull_request:45932|pull_request:46100" ], - "prepared_review_unit_hash": "1818fafaa06d487d327cb21ad854da45171e5ae00c60fc5422959696ccee59e0", + "prepared_review_unit_hash": "2e2cf026d6a51ca47251e160f40334a15501cb8acef36ee73c5d88446af5b8b1", "component_budget": { - "node_count": 3, - "item_count": 3, - "soft_pair_count": 3, - "serialized_chars": 7882, - "estimated_input_tokens": 1971, - "estimated_eval_tokens": 4198 + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5161, + "estimated_input_tokens": 1291, + "estimated_eval_tokens": 2838 }, "budget": { - "node_count": 3, - "item_count": 3, - "soft_pair_count": 3, - "serialized_chars": 7882, - "estimated_input_tokens": 1971, - "estimated_eval_tokens": 4198 + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5161, + "estimated_input_tokens": 1291, + "estimated_eval_tokens": 2838 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates of one another: each PR targets a different area (model code/workflow secrets, test migration files, and PEFT CI integration). The only overlap is broad CI/testing infrastructure, which is too weak to merge them as the same change.", - "confidence": 0.95, + "summary": "The two pull requests are not duplicates: one is a broad testing/telemetry and model-mapping change, while the other is a test-migration PR adding kernel and vision tests. They touch different code paths and would not plausibly be merged into one PR.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46891 is the best anchor only because it is the most self-contained CI/testing change; however, it is still not a duplicate of the others, which modify different models and test-migration files.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR #46891 is the strongest representative of the cluster\u2019s CI/test theme, but it should not be treated as the same underlying change as #45932 or #46100.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:45932", "right": "pull_request:46100", "accept": false, - "reason": "Different concrete changes: #45932 edits model implementations and workflow secrets, while #46100 adds kernel/test migration files. Same broad testing context, but not the same bug or change." - }, - { - "left": "pull_request:45932", - "right": "pull_request:46891", - "accept": false, - "reason": "Only shared file is the workflow caller, but the actual changes differ: #45932 is a broad model/telemetry-related PR, while #46891 focuses on PEFT integration CI and helpers." - }, - { - "left": "pull_request:46100", - "right": "pull_request:46891", - "accept": false, - "reason": "#46100 is a test migration PR for AltCLIP/ViT with new DDP config, whereas #46891 is a PEFT CI integration update. They are unrelated concrete changes." + "reason": "Not the same underlying change. PR 45932 updates CI secrets plus several model files/configs; PR 46100 adds ddp config and test files for altclip/vit migration. Different goals, files, and code paths." } ] }, @@ -24036,12 +23970,12 @@ "soft_pairs": [ "pull_request:46126|pull_request:46269" ], - "prepared_review_unit_hash": "2e3583210651b76dea787700ac058feffcf987605622913c12de3853a1e63a5b", + "prepared_review_unit_hash": "3fba17f3e6637d34618d2b59c81b932a8ca392ee386fb3bc4cd4e767747db795", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5525, + "serialized_chars": 5526, "estimated_input_tokens": 1382, "estimated_eval_tokens": 3020 }, @@ -24049,7 +23983,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5525, + "serialized_chars": 5526, "estimated_input_tokens": 1382, "estimated_eval_tokens": 3020 }, @@ -24058,18 +23992,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates. PR #46126 is a focused EP/Trainer integration change around distributed tensor-parallel/MoE handling, while PR #46269 is a broad distributed branch base touching many unrelated files and subsystems.", - "confidence": 0.91, + "summary": "These are related distributed-training changes, but they are not the same concrete fix/change: one is a focused EP + Trainer integration PR, while the other is a broad distributed branch base touching many unrelated areas. The overlap in filenames is too broad to merge as duplicates.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46126 is the better canonical PR because it targets one concrete feature/change set with a narrow code path and clear intent, whereas #46269 is a large umbrella/base branch with mixed distributed updates.", + "canonical_pr_reason": "PR #46126 is the more canonical, self-contained change: it targets a specific EP/Trainer integration path with a focused set of files and behavior. PR #46269 is a large base branch that spans many distributed subsystems and is not a clean representative of a single underlying change.", "best_issue_reason": null, - "best_pr_reason": "PR #46126 is the best representative of the underlying change because it is specific and self-contained; #46269 is too broad and mixed to serve as a canonical duplicate target.", + "best_pr_reason": "PR #46126 is the best representative PR for this cluster because it has a narrower, well-defined purpose and concrete code-path changes. PR #46269 is too broad and includes many unrelated distributed updates, making it a poor canonical duplicate target.", "soft_edge_verdicts": [ { "left": "pull_request:46126", "right": "pull_request:46269", "accept": false, - "reason": "Shared distributed filenames are not enough here: #46269 is a broad branch-base PR with many unrelated modifications, not the same concrete bug/change as the focused EP/Trainer integration in #46126. They would not plausibly collapse into one PR." + "reason": "They share distributed filenames, but the PRs are not the same underlying change. #46126 is a specific EP + Trainer integration, while #46269 is a broad distributed branch base with many additional areas (FSDP, integrations, generation, CI, etc.). They would not plausibly be merged as one PR." } ] }, @@ -24135,196 +24069,208 @@ "pull_request:46504|pull_request:46563", "pull_request:46504|pull_request:46716" ], - "prepared_review_unit_hash": "243de21b8619642eccabf8d38e707551e6256f725dcf900c2804a276c3c04683", + "prepared_review_unit_hash": "bf005833a957cda75d64e376ef483b4abb11f2c09e4a4a14ef1ad40eefe215e1", "component_budget": { "node_count": 10, "item_count": 10, "soft_pair_count": 29, - "serialized_chars": 25929, - "estimated_input_tokens": 6483, - "estimated_eval_tokens": 13222 + "serialized_chars": 25937, + "estimated_input_tokens": 6485, + "estimated_eval_tokens": 13226 }, "budget": { "node_count": 10, "item_count": 10, "soft_pair_count": 29, - "serialized_chars": 25929, - "estimated_input_tokens": 6483, - "estimated_eval_tokens": 13222 + "serialized_chars": 25937, + "estimated_input_tokens": 6485, + "estimated_eval_tokens": 13226 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are mostly unrelated workflow-chore PRs across different CI files. The only dense cluster is several independent hardening tweaks to pr-ci-security-gate.yml, but they address different concerns rather than the same concrete change.", - "confidence": 0.97, + "summary": "This cluster is mostly a set of unrelated workflow-only PRs. The only cohesive subcluster is several incremental updates to .github/workflows/pr-ci-security-gate.yml, but they address different hardening concerns rather than the same exact bug/change, so none of the soft edges should be merged.", + "confidence": 0.82, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46554 is the best representative anchor only because it is the most substantive change in the largest repeated file family (.github/workflows/pr-ci-security-gate.yml); the other nearby PRs are narrower, distinct follow-ups rather than the same fix.", + "canonical_pr_reason": "PR 46554 is the most substantive representative of the pr-ci-security-gate subcluster because it introduces the actual security-gate logic change (checkout plus changed-file checks), whereas the other PRs on that workflow are smaller permission/approval/comment tweaks.", "best_issue_reason": null, - "best_pr_reason": "If a single PR must represent the cluster, 46554 is the strongest anchor, but the cluster is not truly duplicative and should probably stay split by workflow and change type.", + "best_pr_reason": "46554 is the strongest overall PR candidate: it makes the concrete security-gate behavior change and is more representative than the later incremental edits to the same workflow.", "soft_edge_verdicts": [ { "left": "pull_request:46138", "right": "pull_request:46554", "accept": false, - "reason": "Both touch CI workflows, but one changes self-comment input piping and the other adds pr-ci-security-gate checks; different files and different bugs." + "reason": "Different workflow files and different behavior: stdin handling for self-comment-ci vs security-gate checks for fork PRs." }, { "left": "pull_request:46138", "right": "pull_request:46558", "accept": false, - "reason": "Unrelated workflow changes: stdin handling for a comment-driven script vs PR CI security-gate permissions/approval logic." + "reason": "Unrelated workflow changes; one updates self-comment-ci argument passing, the other hardens pr-ci-security-gate permissions/approval flow." }, { "left": "pull_request:46138", "right": "pull_request:46563", "accept": false, - "reason": "Different workflow concerns; one fixes comment handling, the other only adds documentation/comments in the security gate workflow." + "reason": "Different CI workflows and different concerns; no shared underlying bug or change." }, { "left": "pull_request:46138", "right": "pull_request:46716", "accept": false, - "reason": "Different workflow files and different fixes; no shared concrete code-path problem." + "reason": "Different workflow files and different fixes; not the same code-path problem." }, { "left": "pull_request:46138", "right": "pull_request:46504", "accept": false, - "reason": "Both use printf, but they fix different shell-input handling issues in different workflows and scripts." + "reason": "Different workflows and distinct shell-input handling changes in separate jobs." }, { "left": "pull_request:46432", "right": "pull_request:46503", "accept": false, - "reason": "SSH runner hardening is unrelated to benchmark_v2 trigger logic." + "reason": "SSH runner hardening vs benchmark workflow trigger change; unrelated changes." }, { "left": "pull_request:46432", "right": "pull_request:46501", "accept": false, - "reason": "Different workflows and different changes: SSH runner secret handling vs anti-slop permissions." + "reason": "Different workflows and different fixes: SSH runner secret handling vs anti-slop permissions structure." }, { "left": "pull_request:46432", "right": "pull_request:46502", "accept": false, - "reason": "Unrelated workflow updates; no common bug or shared code path." + "reason": "Different workflow logic; no shared bug or concrete change." }, { "left": "pull_request:46432", "right": "pull_request:46554", "accept": false, - "reason": "SSH runner safety changes are separate from pr-ci-security-gate workflow checks." + "reason": "SSH runner workflow hardening is unrelated to pr-ci-security-gate security checks." }, { "left": "pull_request:46432", "right": "pull_request:46558", "accept": false, - "reason": "Different workflows and distinct security issues; not the same fix." + "reason": "Different workflow files and separate security concerns." }, { "left": "pull_request:46501", "right": "pull_request:46503", "accept": false, - "reason": "Separate workflow triggers/permissions updates; no same underlying bug." + "reason": "Anti-slop permissions vs benchmark dispatch trigger; unrelated workflow edits." }, { "left": "pull_request:46501", "right": "pull_request:46502", "accept": false, - "reason": "Anti-slop permissions change and assign-reviewers checkout ref change are unrelated." + "reason": "Different workflow files and different logic changes." }, { "left": "pull_request:46501", "right": "pull_request:46504", "accept": false, - "reason": "Both are workflow tweaks, but one is anti-slop permissions and the other is shell-safe commit-message handling." + "reason": "Both are workflow tweaks, but they fix different issues in different jobs and are not one mergeable change." }, { "left": "pull_request:46501", "right": "pull_request:46554", "accept": false, - "reason": "Different workflow files and different hardening goals; not the same concrete change." + "reason": "Different workflows and different security/permission concerns." }, { "left": "pull_request:46502", "right": "pull_request:46503", "accept": false, - "reason": "Checkout-base-ref fix for assign-reviewers is unrelated to benchmark dispatch logic." + "reason": "Assign-reviewers checkout ref change vs benchmark trigger change are unrelated." }, { "left": "pull_request:46502", "right": "pull_request:46504", "accept": false, - "reason": "Different workflows and different failure modes; no duplicate bug." + "reason": "Different workflows with separate behavior changes; not the same underlying bug." }, { "left": "pull_request:46502", "right": "pull_request:46554", "accept": false, - "reason": "Assign-reviewers checkout behavior is unrelated to pr-ci-security-gate security checks." + "reason": "Assign-reviewers base-ref checkout fix is unrelated to pr-ci-security-gate changes." }, { "left": "pull_request:46503", "right": "pull_request:46504", "accept": false, - "reason": "Benchmark dispatch trigger change is not the same as build-ci-docker-images commit-message parsing fix." + "reason": "Benchmark dispatch support vs build-ci-docker-images commit-message handling; unrelated." }, { "left": "pull_request:46503", "right": "pull_request:46554", "accept": false, - "reason": "Different workflow files and unrelated changes." + "reason": "Different workflow functionality and different code paths." }, { "left": "pull_request:46504", "right": "pull_request:46554", "accept": false, - "reason": "Both are shell-safety style edits, but they fix different workflows and different inputs, so not the same change." + "reason": "Same broad CI area, but one changes commit-message parsing for image tagging while the other changes fork PR security gating." }, { "left": "pull_request:46504", "right": "pull_request:46558", "accept": false, - "reason": "Build-image commit-message handling is unrelated to PR CI security-gate permissions/approval logic." + "reason": "Different workflow logic and different concrete fixes." }, { "left": "pull_request:46554", "right": "pull_request:46558", "accept": false, - "reason": "Same workflow file, but different hardening changes: one adds file-change security checks, the other tightens permissions and approval gating." + "reason": "Same workflow file, but one adds file-change security checks while the other changes permissions and approval gating; these are separate fixes." }, { "left": "pull_request:46554", "right": "pull_request:46563", "accept": false, - "reason": "Same file, but one adds executable security checks while the other only documents safety assumptions; not the same bug/change." + "reason": "The comment-only documentation change does not fix the same concrete behavior as the security checks added in 46554." }, { "left": "pull_request:46554", "right": "pull_request:46716", "accept": false, - "reason": "Same workflow family, but different fixes: security checks vs removing inherited secrets." + "reason": "Both touch pr-ci-security-gate.yml, but 46554 adds detection logic while 46716 only changes secret inheritance; different fixes." }, { "left": "pull_request:46558", "right": "pull_request:46563", "accept": false, - "reason": "Both touch the same workflow, but they address different concerns and could be merged only as separate edits, not duplicates." + "reason": "Permissions/approval gating vs explanatory comments in the same workflow; not the same bug or change." }, { "left": "pull_request:46558", "right": "pull_request:46716", "accept": false, - "reason": "Permissions/approval hardening is distinct from changing secrets inheritance." + "reason": "Different hardening changes to the same workflow, but not one concrete fix." }, { "left": "pull_request:46563", "right": "pull_request:46716", "accept": false, - "reason": "A safety note/comment is not the same change as replacing secrets: inherit with secrets: {}." + "reason": "Commenting on safety and changing secrets inheritance are different changes with different effects." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46563", + "accept": false, + "reason": "Unrelated workflow updates in different files and code paths." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46716", + "accept": false, + "reason": "Build-image tagging logic vs pr-ci-security-gate secret handling; unrelated changes." } ] }, @@ -24343,51 +24289,51 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "llm", - "cache_hit": false, + "source": "cache", + "cache_hit": true, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-46152-2", + "cluster_id": "cluster-46174-2", "nodes": [ - "pull_request:46152", - "pull_request:46207" + "pull_request:46174", + "pull_request:46780" ], "soft_pairs": [ - "pull_request:46152|pull_request:46207" + "pull_request:46174|pull_request:46780" ], - "prepared_review_unit_hash": "d41a23c80bb8f83130b5fa6a9a4386bd4ed4f8765897db436d786ddc0db5b9fe", + "prepared_review_unit_hash": "a48ce4ca162f555682f79411d7229ac1ccc08d7822dfab654bb881639fccddbc", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4276, - "estimated_input_tokens": 1069, - "estimated_eval_tokens": 2394 + "serialized_chars": 6030, + "estimated_input_tokens": 1508, + "estimated_eval_tokens": 3272 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4276, - "estimated_input_tokens": 1069, - "estimated_eval_tokens": 2394 + "serialized_chars": 6030, + "estimated_input_tokens": 1508, + "estimated_eval_tokens": 3272 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are thematically similar (both replace direct file-open source inspection), but they fix different code paths in different modules and are not safe duplicates.", + "summary": "Two PRs target the same reported MPS crash in the sinusoidal position embedding helper, but they implement different fixes and are not safe to treat as duplicates.", "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46207 is the better anchor for this cluster because it targets a specific, test-backed source-reading bug in _can_set_attn_implementation; PR 46152 is a separate import-utils hardening change in a different path.", + "canonical_pr_reason": "PR 46174 is the broader, more complete fix for the float64-on-MPS crash: it updates the helper across all affected model files and keeps the internal math on CPU float64 before casting back.", "best_issue_reason": null, - "best_pr_reason": "PR 46207 is the most representative because it has a concrete behavior change plus tests, while 46152 is a distinct utility refactor focused on resolving package resources.", + "best_pr_reason": "PR 46174 is the best representative of the cluster because it is the original, more comprehensive fix and has more discussion/review context than the narrower follow-up.", "soft_edge_verdicts": [ { - "left": "pull_request:46152", - "right": "pull_request:46207", + "left": "pull_request:46174", + "right": "pull_request:46780", "accept": false, - "reason": "Both avoid open() for source access, but they touch different modules and different behaviors: import_utils resource resolution vs modeling_utils attention-implementation detection. They do not look like the same underlying bug or one mergeable fix." + "reason": "They share the same issue target and helper, but the patches are different solutions: 46174 avoids float64 on MPS by doing the arithmetic on CPU first, while 46780 switches the helper to use dtype for internal arithmetic. Those are alternative implementations, not one mergeable duplicate PR." } ] }, @@ -24406,51 +24352,427 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-46174-2", + "cluster_id": "cluster-46196-18", "nodes": [ - "pull_request:46174", - "pull_request:46780" + "pull_request:46196", + "pull_request:46707", + "pull_request:46836", + "pull_request:46891", + "pull_request:46907", + "pull_request:46908", + "pull_request:46912", + "pull_request:46926", + "pull_request:46927", + "pull_request:46928", + "pull_request:46941", + "pull_request:46949", + "pull_request:46950", + "pull_request:46951", + "pull_request:46953", + "pull_request:46954", + "pull_request:46956", + "pull_request:46957" ], "soft_pairs": [ - "pull_request:46174|pull_request:46780" + "pull_request:46196|pull_request:46957", + "pull_request:46196|pull_request:46836", + "pull_request:46196|pull_request:46956", + "pull_request:46196|pull_request:46891", + "pull_request:46196|pull_request:46953", + "pull_request:46707|pull_request:46953", + "pull_request:46707|pull_request:46891", + "pull_request:46707|pull_request:46836", + "pull_request:46707|pull_request:46957", + "pull_request:46707|pull_request:46912", + "pull_request:46836|pull_request:46957", + "pull_request:46836|pull_request:46891", + "pull_request:46836|pull_request:46956", + "pull_request:46836|pull_request:46953", + "pull_request:46891|pull_request:46957", + "pull_request:46891|pull_request:46956", + "pull_request:46891|pull_request:46912", + "pull_request:46907|pull_request:46926", + "pull_request:46907|pull_request:46941", + "pull_request:46907|pull_request:46908", + "pull_request:46907|pull_request:46927", + "pull_request:46907|pull_request:46928", + "pull_request:46908|pull_request:46926", + "pull_request:46908|pull_request:46928", + "pull_request:46908|pull_request:46951", + "pull_request:46908|pull_request:46941", + "pull_request:46912|pull_request:46957", + "pull_request:46836|pull_request:46912", + "pull_request:46912|pull_request:46956", + "pull_request:46926|pull_request:46941", + "pull_request:46926|pull_request:46928", + "pull_request:46926|pull_request:46927", + "pull_request:46927|pull_request:46949", + "pull_request:46927|pull_request:46951", + "pull_request:46927|pull_request:46950", + "pull_request:46927|pull_request:46941", + "pull_request:46928|pull_request:46941", + "pull_request:46928|pull_request:46951", + "pull_request:46949|pull_request:46951", + "pull_request:46949|pull_request:46950", + "pull_request:46891|pull_request:46949", + "pull_request:46836|pull_request:46949", + "pull_request:46950|pull_request:46951", + "pull_request:46908|pull_request:46950", + "pull_request:46926|pull_request:46950", + "pull_request:46891|pull_request:46951", + "pull_request:46953|pull_request:46956", + "pull_request:46953|pull_request:46957", + "pull_request:46891|pull_request:46953", + "pull_request:46891|pull_request:46954", + "pull_request:46954|pull_request:46957", + "pull_request:46954|pull_request:46956", + "pull_request:46953|pull_request:46954", + "pull_request:46836|pull_request:46954", + "pull_request:46956|pull_request:46957" ], - "prepared_review_unit_hash": "a48ce4ca162f555682f79411d7229ac1ccc08d7822dfab654bb881639fccddbc", + "prepared_review_unit_hash": "bc87f73801481b6b039710b02dd83d3eac2d4cb935da3ec05c72fbb9af31bc35", "component_budget": { - "node_count": 2, - "item_count": 2, - "soft_pair_count": 1, - "serialized_chars": 6030, - "estimated_input_tokens": 1508, - "estimated_eval_tokens": 3272 + "node_count": 18, + "item_count": 18, + "soft_pair_count": 55, + "serialized_chars": 59205, + "estimated_input_tokens": 14802, + "estimated_eval_tokens": 29860 }, "budget": { - "node_count": 2, - "item_count": 2, - "soft_pair_count": 1, - "serialized_chars": 6030, - "estimated_input_tokens": 1508, - "estimated_eval_tokens": 3272 + "node_count": 18, + "item_count": 18, + "soft_pair_count": 55, + "serialized_chars": 59205, + "estimated_input_tokens": 14802, + "estimated_eval_tokens": 29860 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two PRs target the same reported MPS crash in the sinusoidal position embedding helper, but they implement different fixes and are not safe to treat as duplicates.", - "confidence": 0.89, + "summary": "This is not a single duplicate cluster; it mixes unrelated PRs across model bring-up, CI/test maintenance, distributed/FSDP infrastructure, generation/FP8 work, and many model-specific bug fixes.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46174 is the broader, more complete fix for the float64-on-MPS crash: it updates the helper across all affected model files and keeps the internal math on CPU float64 before casting back.", + "canonical_pr_reason": "No single canonical PR: the items span unrelated changes (new models, CI, FSDP/FP8 infra, test refactors, and separate per-model fixes).", "best_issue_reason": null, - "best_pr_reason": "PR 46174 is the best representative of the cluster because it is the original, more comprehensive fix and has more discussion/review context than the narrower follow-up.", + "best_pr_reason": "No global best PR for duplicate collapse, because there is no shared underlying bug or change that would make one PR the canonical representative.", "soft_edge_verdicts": [ { - "left": "pull_request:46174", - "right": "pull_request:46780", + "left": "pull_request:46196", + "right": "pull_request:46957", "accept": false, - "reason": "They share the same issue target and helper, but the patches are different solutions: 46174 avoids float64 on MPS by doing the arithmetic on CPU first, while 46780 switches the helper to use dtype for internal arithmetic. Those are alternative implementations, not one mergeable duplicate PR." + "reason": "wav2vec2 SDPA dispatch work vs finegrained FP8 decode optimization; unrelated subsystems." + }, + { + "left": "pull_request:46196", + "right": "pull_request:46836", + "accept": false, + "reason": "audio model SDPA/docs changes vs adding the Unlimited OCR model." + }, + { + "left": "pull_request:46196", + "right": "pull_request:46956", + "accept": false, + "reason": "model dispatch changes vs shared test-helper refactor." + }, + { + "left": "pull_request:46196", + "right": "pull_request:46891", + "accept": false, + "reason": "wav2vec2/audio model work vs PEFT CI/workflow changes." + }, + { + "left": "pull_request:46196", + "right": "pull_request:46953", + "accept": false, + "reason": "SDPA dispatch work vs FusedRMSNormGated dtype cleanup." + }, + { + "left": "pull_request:46707", + "right": "pull_request:46953", + "accept": false, + "reason": "FSDP2 migration vs RMSNorm module dtype removal; different code paths." + }, + { + "left": "pull_request:46707", + "right": "pull_request:46891", + "accept": false, + "reason": "FSDP runtime/migration work vs PEFT CI/test changes." + }, + { + "left": "pull_request:46707", + "right": "pull_request:46836", + "accept": false, + "reason": "distributed FSDP utilities vs new Unlimited OCR model bring-up." + }, + { + "left": "pull_request:46707", + "right": "pull_request:46957", + "accept": false, + "reason": "FSDP migration vs generation/FP8 optimization; not the same bug." + }, + { + "left": "pull_request:46707", + "right": "pull_request:46912", + "accept": false, + "reason": "FSDP infrastructure vs RoPE test coverage changes." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46957", + "accept": false, + "reason": "new OCR model addition vs generation decode optimization." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46891", + "accept": false, + "reason": "model bring-up vs PEFT CI workflow/test changes." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46956", + "accept": false, + "reason": "new model code vs shared floats_list test helper refactor." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46953", + "accept": false, + "reason": "Unlimited OCR addition vs RMSNorm dtype cleanup." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46957", + "accept": false, + "reason": "PEFT CI/workflow changes vs generation FP8 runtime changes." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46956", + "accept": false, + "reason": "PEFT CI work vs feature-extractor helper refactor." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46912", + "accept": false, + "reason": "PEFT CI changes vs RoPE test updates." + }, + { + "left": "pull_request:46907", + "right": "pull_request:46926", + "accept": false, + "reason": "EdgeTam kwarg cleanup vs MiniMax-M3-VL offload-folder fix." + }, + { + "left": "pull_request:46907", + "right": "pull_request:46941", + "accept": false, + "reason": "EdgeTam processor kwarg handling vs DAC distance-formula fix." + }, + { + "left": "pull_request:46907", + "right": "pull_request:46908", + "accept": false, + "reason": "EdgeTam import_or_config cleanup vs DeepSeek-V3.2 load/offload workaround." + }, + { + "left": "pull_request:46907", + "right": "pull_request:46927", + "accept": false, + "reason": "EdgeTam kwarg cleanup vs GLM46V input casting fix." + }, + { + "left": "pull_request:46907", + "right": "pull_request:46928", + "accept": false, + "reason": "EdgeTam model bug vs BLIP-2 tokenizer allowlist update." + }, + { + "left": "pull_request:46908", + "right": "pull_request:46926", + "accept": false, + "reason": "different model-specific load_error fixes (DeepSeek vs MiniMax)." + }, + { + "left": "pull_request:46908", + "right": "pull_request:46928", + "accept": false, + "reason": "load/offload workaround vs tokenizer allowlist change." + }, + { + "left": "pull_request:46908", + "right": "pull_request:46951", + "accept": false, + "reason": "DeepSeek loading workaround vs PEFT hotswap mismatch handling." + }, + { + "left": "pull_request:46908", + "right": "pull_request:46941", + "accept": false, + "reason": "load/offload issue vs DAC numerical fix." + }, + { + "left": "pull_request:46912", + "right": "pull_request:46957", + "accept": false, + "reason": "RoPE tests vs generation/FP8 optimization." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46912", + "accept": false, + "reason": "new OCR model bring-up vs RoPE test suite changes." + }, + { + "left": "pull_request:46912", + "right": "pull_request:46956", + "accept": false, + "reason": "RoPE test coverage vs shared float-list helper refactor." + }, + { + "left": "pull_request:46926", + "right": "pull_request:46941", + "accept": false, + "reason": "MiniMax load workaround vs DAC codebook-distance correction." + }, + { + "left": "pull_request:46926", + "right": "pull_request:46928", + "accept": false, + "reason": "MiniMax load issue vs BLIP-2 tokenizer allowlist." + }, + { + "left": "pull_request:46926", + "right": "pull_request:46927", + "accept": false, + "reason": "different model-specific fixes for unrelated failures." + }, + { + "left": "pull_request:46927", + "right": "pull_request:46949", + "accept": false, + "reason": "GLM46V input_id casting vs Whisper/AudioFlamingo cache-state fix." + }, + { + "left": "pull_request:46927", + "right": "pull_request:46951", + "accept": false, + "reason": "GLM46V fix vs PEFT hotswap loading behavior." + }, + { + "left": "pull_request:46927", + "right": "pull_request:46950", + "accept": false, + "reason": "GLM46V input/type fix vs Mamba2 OOM fix." + }, + { + "left": "pull_request:46927", + "right": "pull_request:46941", + "accept": false, + "reason": "GLM46V casting bug vs DAC numeric bug." + }, + { + "left": "pull_request:46928", + "right": "pull_request:46941", + "accept": false, + "reason": "tokenizer allowlist update vs DAC distance calculation fix." + }, + { + "left": "pull_request:46928", + "right": "pull_request:46951", + "accept": false, + "reason": "tokenizer allowlist vs PEFT hotswap mismatch handling." + }, + { + "left": "pull_request:46949", + "right": "pull_request:46951", + "accept": false, + "reason": "Whisper/AudioFlamingo cache initialization bug vs PEFT loading/mismatch logic." + }, + { + "left": "pull_request:46949", + "right": "pull_request:46950", + "accept": false, + "reason": "different model fixes: cache-state init vs Mamba2 OOM." + }, + { + "left": "pull_request:46950", + "right": "pull_request:46951", + "accept": false, + "reason": "Mamba2 kernel/OOM fix vs PEFT hotswap load behavior." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46949", + "accept": false, + "reason": "PEFT CI/workflow work vs Whisper/AudioFlamingo runtime fix." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46949", + "accept": false, + "reason": "new OCR model addition vs Whisper/AudioFlamingo bug fix." + }, + { + "left": "pull_request:46953", + "right": "pull_request:46956", + "accept": false, + "reason": "RMSNorm module cleanup vs shared test-helper refactor." + }, + { + "left": "pull_request:46953", + "right": "pull_request:46957", + "accept": false, + "reason": "RMSNorm dtype cleanup vs FP8 generation optimization." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46953", + "accept": false, + "reason": "PEFT CI changes vs RMSNorm cleanup." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46954", + "accept": false, + "reason": "PEFT CI work vs GlmAsr Pytest v9 fix." + }, + { + "left": "pull_request:46954", + "right": "pull_request:46957", + "accept": false, + "reason": "GlmAsr pytest-related fix vs generation/FP8 optimization." + }, + { + "left": "pull_request:46954", + "right": "pull_request:46956", + "accept": false, + "reason": "GlmAsr model/test fix vs shared floats_list helper refactor." + }, + { + "left": "pull_request:46953", + "right": "pull_request:46954", + "accept": false, + "reason": "distinct fixes in different models; not the same underlying bug." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46954", + "accept": false, + "reason": "Unlimited OCR model addition vs GlmAsr pytest fix." + }, + { + "left": "pull_request:46956", + "right": "pull_request:46957", + "accept": false, + "reason": "test helper refactor vs generation/FP8 runtime optimization." } ] }, @@ -24480,12 +24802,12 @@ "soft_pairs": [ "pull_request:46281|pull_request:46890" ], - "prepared_review_unit_hash": "d94770203f8c763450984a356666d858c9d0e838fa160eb741c308e3a4af79b6", + "prepared_review_unit_hash": "c246c6a23343ef9d1083f5d6d16f578c85cc6d038d98859cb6b64fa0ff1c62a9", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4270, + "serialized_chars": 4271, "estimated_input_tokens": 1068, "estimated_eval_tokens": 2392 }, @@ -24493,7 +24815,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4270, + "serialized_chars": 4271, "estimated_input_tokens": 1068, "estimated_eval_tokens": 2392 }, @@ -24502,18 +24824,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs harden checkpoint loading against path traversal, but they target different untrusted inputs and different code paths, so they are related security fixes rather than duplicates.", - "confidence": 0.94, + "summary": "These PRs are related by theme (path traversal hardening) but they fix different untrusted inputs in different code paths, so they are not duplicates and should not be merged together.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "46281 is the cleaner representative of the cluster because it fixes a concrete traversal bug in sharded checkpoint loading with focused containment checks and tests; 46890 targets a separate `transformers_weights` config-field path.", "best_issue_reason": null, - "best_pr_reason": "If a single representative is needed, 46281 is the cleaner choice because it fixes the shard-file lookup path in `get_checkpoint_shard_files` with a concrete containment check and the vulnerability is narrowly defined and well-scoped.", + "best_pr_reason": "46281 is the strongest standalone PR to keep as the cluster representative, but only because it is a tighter, more explicit fix; it does not subsume the distinct bug fixed by 46890.", "soft_edge_verdicts": [ { "left": "pull_request:46281", "right": "pull_request:46890", "accept": false, - "reason": "Reject: they address different path-traversal sources in different loaders. 46281 validates shard filenames coming from the checkpoint index `weight_map`; 46890 validates the `transformers_weights` config field in `modeling_utils.py`. Same security theme, but not the same underlying bug and not something that would naturally merge into one PR." + "reason": "Different vulnerable inputs and loaders: 46281 sanitizes shard filenames from the checkpoint index in `hub.py`, while 46890 validates `transformers_explicit_filename` from model config in `modeling_utils.py`. Same security class, but not the same underlying bug or mergeable fix." } ] }, @@ -24543,12 +24865,12 @@ "soft_pairs": [ "pull_request:46310|pull_request:46546" ], - "prepared_review_unit_hash": "0eaadeaab09eb2eeeb248a264a08a7969b74b0f84a71d2f37c8e675b200f0666", + "prepared_review_unit_hash": "3458e8d1723e4efac7081d210033aee4c2d629590e5139b77c3dab65e52c792e", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4849, + "serialized_chars": 4850, "estimated_input_tokens": 1213, "estimated_eval_tokens": 2682 }, @@ -24556,7 +24878,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4849, + "serialized_chars": 4850, "estimated_input_tokens": 1213, "estimated_eval_tokens": 2682 }, @@ -24565,18 +24887,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs fix the same off-by-one bug in the streaming sliding-window mask: the right window boundary used `<` instead of `<=`. PR 46546 is the broader, more complete fix and also adds a test; PR 46310 is a narrower earlier variant of the same change.", - "confidence": 0.96, + "summary": "Both PRs fix the same sliding-window right-boundary off-by-one in Moonshine streaming mask generation; PR 46546 is the broader, test-backed version and also applies the same fix to Gemma4.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46546 is the best canonical PR because it fixes the same boundary bug in moonshine_streaming, also applies the same correction in gemma4, and includes a test, making it the more complete representative change.", + "canonical_pr_reason": "PR 46546 is the best canonical PR because it contains the Moonshine streaming boundary fix plus tests, and also applies the same correction to Gemma4; it is the more complete and representative change set.", "best_issue_reason": null, - "best_pr_reason": "PR 46546 is the strongest triage target: it addresses the same concrete code-path bug as PR 46310, but with wider coverage and test support, so it is the best representative of the fix.", + "best_pr_reason": "PR 46546 is the best global PR because it is the more complete fix for the same underlying boundary bug, with validation coverage and additional affected model handling.", "soft_edge_verdicts": [ { "left": "pull_request:46310", "right": "pull_request:46546", "accept": true, - "reason": "Same underlying off-by-one right-window boundary bug in the streaming mask; 46546 is essentially a superset of 46310's moonshine fix, so these are duplicate/mergeable variants of the same change." + "reason": "Both PRs change the same Moonshine streaming sliding-window mask from '< right_window_size' to '<= right_window_size', fixing the same off-by-one boundary bug; PR 46546 simply expands the same fix to another model and adds tests." } ] }, @@ -24615,70 +24937,70 @@ "pull_request:46425|pull_request:46850", "pull_request:46850|pull_request:46917" ], - "prepared_review_unit_hash": "1006591b44b99b1f6fe947512f747bb038b41d55f26d322cf102c8c19ecfb027", + "prepared_review_unit_hash": "e7ffef7f5fa87a835c6ed97c3137c6f28567290d02b7e7ae70b646098030c586", "component_budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 6, - "serialized_chars": 16832, - "estimated_input_tokens": 4208, - "estimated_eval_tokens": 8672 + "serialized_chars": 16834, + "estimated_input_tokens": 4209, + "estimated_eval_tokens": 8674 }, "budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 6, - "serialized_chars": 16832, - "estimated_input_tokens": 4208, - "estimated_eval_tokens": 8672 + "serialized_chars": 16834, + "estimated_input_tokens": 4209, + "estimated_eval_tokens": 8674 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are mostly unrelated: two new model integrations, one continuous-batching multimodal feature, one Gemma masking fix, one Nemotron Omni change, and one deprecation cleanup. They share broad Transformers/modeling vocabulary but not the same underlying bug or code path, so no soft pair should be merged.", + "summary": "This cluster is not a duplicate set: it mixes unrelated PRs for model additions, a continuous batching multimodal feature, a Gemma masking bugfix, and a deprecation cleanup. None of the soft-similarity pairs look like the same concrete change.", "confidence": 0.97, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR 46417 is the strongest standalone PR in the set: it is a complete model integration with config/model/processor/docs/tests, making it the clearest representative artifact even though it is not a duplicate of the others.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:46376", "right": "pull_request:46417", "accept": false, - "reason": "Different changes: continuous-batching multimodal support vs. a new HunYuan VL model integration. No shared bug or mergeable code path." + "reason": "Different features and codepaths: continuous batching multimodal support vs. adding HunYuan VL model files and docs." }, { "left": "pull_request:46376", "right": "pull_request:46425", "accept": false, - "reason": "Continuous batching infrastructure change and RADIO model support are unrelated features; they only overlap at the high-level Transformers codebase." + "reason": "Unrelated model areas: continuous batching internals vs. RADIO model implementation and registration." }, { "left": "pull_request:46425", "right": "pull_request:46509", "accept": false, - "reason": "RADIO model addition and Nemotron Omni support touch different model families and different implementation paths." + "reason": "Separate model additions for different architectures (RADIO vs. Nemotron Omni), not the same bug or change." }, { "left": "pull_request:46425", "right": "pull_request:46917", "accept": false, - "reason": "RADIO model support is unrelated to removing deprecated training-args/image backend code; no shared underlying fix." + "reason": "RADIO model support and deprecated training-args cleanup are unrelated; shared repo touchpoints are too broad to merge." }, { "left": "pull_request:46425", "right": "pull_request:46850", "accept": false, - "reason": "New RADIO model support and Gemma sliding-window masking fix address different subsystems and different concrete problems." + "reason": "RADIO model addition vs. Gemma attention-masking fix are different subsystems and issues." }, { "left": "pull_request:46850", "right": "pull_request:46917", "accept": false, - "reason": "Gemma attention-mask bug fix and deprecated-args cleanup are unrelated; they do not fix the same code path or issue." + "reason": "Gemma sliding-window masking bugfix does not match the deprecation-removal PR; no shared concrete code-path problem." } ] }, @@ -24708,7 +25030,7 @@ "soft_pairs": [ "pull_request:46574|pull_request:46580" ], - "prepared_review_unit_hash": "4f21080a3b19a6f1a6eb2da37a64bf8af0cdb4b314a45159d8dcf5bb47a8b0fc", + "prepared_review_unit_hash": "43570064594deb03d556be6e2877a1da0ed12cb333cf4336a6f25c93d4433ccb", "component_budget": { "node_count": 2, "item_count": 2, @@ -24730,18 +25052,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs address different tokenizer problems in different code paths: CodeLlama decode stripping vs. AutoTokenizer class selection for DeepSeek checkpoints. They are not duplicates.", + "summary": "The two PRs are unrelated: one fixes CodeLlama decode-time leading-space stripping, the other fixes DeepSeek-Coder/DeepSeek-LLM being routed to the wrong tokenizer class. They touch different code paths and should not be merged as duplicates.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46574 is the more self-contained representative fix: it addresses a concrete decode regression in CodeLlama tokenizer handling and includes a direct implementation change in the model tokenizer path.", + "canonical_pr_reason": "No single canonical PR: PR #46574 changes CodeLlama decoding behavior, while PR #46580 changes AutoTokenizer checkpoint-to-class routing for DeepSeek checkpoints.", "best_issue_reason": null, - "best_pr_reason": "PR #46574 is the better standalone canonical example because it fixes a specific user-visible bug with a narrowly scoped code change; PR #46580 is a separate AutoTokenizer registry workaround for misclassified DeepSeek checkpoints.", + "best_pr_reason": "Neither PR is a good global best representative for a duplicate cluster because they solve different bugs in different subsystems.", "soft_edge_verdicts": [ { "left": "pull_request:46574", "right": "pull_request:46580", "accept": false, - "reason": "Different underlying bugs and code paths: #46574 changes CodeLlama decode behavior, while #46580 changes AutoTokenizer checkpoint-to-class resolution for DeepSeek models. They would not plausibly be merged into one PR." + "reason": "Different underlying problems: one is a decoder configuration bug in CodeLlama, the other is incorrect tokenizer-class selection for specific DeepSeek checkpoints." } ] }, @@ -24771,7 +25093,7 @@ "soft_pairs": [ "pull_request:46603|pull_request:46604" ], - "prepared_review_unit_hash": "931e26484a226dfb237a346dc8f50d495180142d15ff970f62a30a43cc1db10c", + "prepared_review_unit_hash": "2fa60193f6c0aa6713574aca7daf9632d3517d8e90d051e4f511f93de8c0becb", "component_budget": { "node_count": 2, "item_count": 2, @@ -24793,18 +25115,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two Mistral-related PRs, but they address different changes: one refactors/moves the existing converter into the integrations package, while the other adds tekken tokenizer detection/conversion/save utilities. They overlap in files and subsystem, but not in the same underlying bug or change.", - "confidence": 0.92, + "summary": "These are related Mistral tokenizer PRs, but they target different changes: one refactors/moves `MistralConverter` into the Mistral integration package, while the other adds tekken tokenizer detection, conversion, and save utilities. They share files, but not the same underlying fix/change.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46604 is the broader functional change and most representative of the cluster\u2019s concrete tokenizer work; PR #46603 is a supporting refactor/move rather than the main feature addition.", + "canonical_pr_reason": "PR #46604 is the better representative because it introduces the broader functional change around tekken tokenizer support; PR #46603 is mainly a code organization/refactor move for an existing converter.", "best_issue_reason": null, - "best_pr_reason": "PR #46604 is the best standalone PR in this cluster because it introduces the new tokenizer behavior and utilities, making it the more substantial and self-contained change.", + "best_pr_reason": "PR #46604 is the strongest standalone PR in the cluster since it implements the main Mistral tokenizer feature work rather than just relocating code.", "soft_edge_verdicts": [ { "left": "pull_request:46603", "right": "pull_request:46604", "accept": false, - "reason": "Reject: #46603 is a package move/refactor for MistralConverter, while #46604 adds tekken tokenizer detection, conversion, and save utilities. They touch the same area but are not the same concrete code-path problem and should not be merged as duplicates." + "reason": "Different underlying changes: #46603 moves `MistralConverter` into `integrations/mistral/`, while #46604 adds tekken tokenizer detection/conversion/save utilities. Shared files alone are not enough to treat them as one PR." } ] }, @@ -24826,310 +25148,63 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-46707-15", + "cluster_id": "cluster-46741-3", "nodes": [ - "pull_request:46707", - "pull_request:46895", - "pull_request:46898", - "pull_request:46907", - "pull_request:46908", - "pull_request:46916", - "pull_request:46926", - "pull_request:46927", - "pull_request:46928", - "pull_request:46941", - "pull_request:46947", - "pull_request:46949", - "pull_request:46950", - "pull_request:46951", - "pull_request:46953" + "pull_request:46741", + "pull_request:46773", + "pull_request:46781" ], "soft_pairs": [ - "pull_request:46707|pull_request:46953", - "pull_request:46707|pull_request:46947", - "pull_request:46707|pull_request:46949", - "pull_request:46707|pull_request:46951", - "pull_request:46707|pull_request:46927", - "pull_request:46895|pull_request:46898", - "pull_request:46895|pull_request:46916", - "pull_request:46898|pull_request:46916", - "pull_request:46898|pull_request:46947", - "pull_request:46907|pull_request:46926", - "pull_request:46907|pull_request:46941", - "pull_request:46907|pull_request:46908", - "pull_request:46907|pull_request:46927", - "pull_request:46907|pull_request:46928", - "pull_request:46908|pull_request:46926", - "pull_request:46908|pull_request:46928", - "pull_request:46908|pull_request:46951", - "pull_request:46908|pull_request:46941", - "pull_request:46926|pull_request:46941", - "pull_request:46926|pull_request:46928", - "pull_request:46926|pull_request:46927", - "pull_request:46927|pull_request:46949", - "pull_request:46927|pull_request:46951", - "pull_request:46927|pull_request:46950", - "pull_request:46927|pull_request:46941", - "pull_request:46928|pull_request:46941", - "pull_request:46928|pull_request:46951", - "pull_request:46947|pull_request:46953", - "pull_request:46949|pull_request:46951", - "pull_request:46949|pull_request:46950", - "pull_request:46949|pull_request:46953", - "pull_request:46950|pull_request:46951", - "pull_request:46908|pull_request:46950", - "pull_request:46926|pull_request:46950", - "pull_request:46951|pull_request:46953", - "pull_request:46895|pull_request:46947" + "pull_request:46741|pull_request:46773", + "pull_request:46741|pull_request:46781" ], - "prepared_review_unit_hash": "e13bf0e63c6ff05896ee17a7a672a5fa2833dc405e8154da180186129ec11d19", + "prepared_review_unit_hash": "beb55dd5fe8528ae2d9d6e41bbaeea06f6da2b88ad0273528a6adb93dbaf625f", "component_budget": { - "node_count": 15, - "item_count": 15, - "soft_pair_count": 36, - "serialized_chars": 41745, - "estimated_input_tokens": 10437, - "estimated_eval_tokens": 21130 + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 8452, + "estimated_input_tokens": 2113, + "estimated_eval_tokens": 4482 }, "budget": { - "node_count": 15, - "item_count": 15, - "soft_pair_count": 36, - "serialized_chars": 41745, - "estimated_input_tokens": 10437, - "estimated_eval_tokens": 21130 + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 8452, + "estimated_input_tokens": 2113, + "estimated_eval_tokens": 4482 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Only one real duplicate cluster stands out: the CohereASR/Florence2 training-loss double-shift fixes. The rest are mostly unrelated model-specific test, load, or integration fixes despite similar wording.", - "confidence": 0.86, + "summary": "Three open PRs address different bugs in adjacent model/cache code; despite some shared filenames, they are not duplicates. The strongest representative of the cluster is the Mamba2 chunked-prefill/speculative-decoding fix.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "46916 is the broadest representative of the genuine duplicate cluster, since it applies the same loss-shift fix across CohereASR and Florence2 and directly ties to the Florence2 issue.", + "canonical_pr_reason": "PR 46741 is the best canonical PR because it fixes a concrete runtime bug in the Mamba2-style cache path, with targeted code changes and tests across the affected model families.", "best_issue_reason": null, - "best_pr_reason": "46916 is the strongest single PR representative because it covers the shared double-shift bug in the most complete form and bridges the overlapping CohereASR/Florence2 fixes.", + "best_pr_reason": "PR 46741 is the best overall PR to represent the cluster: it has the clearest single bug, the most directly related fix path, and the broadest but still coherent affected surface.", "soft_edge_verdicts": [ { - "left": "pull_request:46707", - "right": "pull_request:46953", - "accept": false, - "reason": "Unrelated changes: native FSDP export/migration vs removing default dtype handling in fused RMSNorm modules." - }, - { - "left": "pull_request:46707", - "right": "pull_request:46947", - "accept": false, - "reason": "Different code paths and bugs: FSDP migration vs CohereASR training-loss fix." - }, - { - "left": "pull_request:46707", - "right": "pull_request:46949", - "accept": false, - "reason": "Unrelated subsystems: FSDP utilities vs Whisper cache/test fix." - }, - { - "left": "pull_request:46707", - "right": "pull_request:46951", - "accept": false, - "reason": "Unrelated changes: FSDP module work vs PEFT hotswap loading behavior." - }, - { - "left": "pull_request:46707", - "right": "pull_request:46927", - "accept": false, - "reason": "Unrelated model-level fixes: FSDP migration vs GLM46V input-id dtype handling." - }, - { - "left": "pull_request:46895", - "right": "pull_request:46898", - "accept": false, - "reason": "Same bug pattern, but different concrete code paths/models (CohereASR vs Florence2), so not a safe duplicate." - }, - { - "left": "pull_request:46895", - "right": "pull_request:46916", - "accept": true, - "reason": "Same CohereASR double-shift training-loss fix; 46916 includes the same CohereASR change." - }, - { - "left": "pull_request:46898", - "right": "pull_request:46916", - "accept": true, - "reason": "Same Florence2 double-shift training-loss fix; 46916 includes the same Florence2 change." - }, - { - "left": "pull_request:46898", - "right": "pull_request:46947", - "accept": false, - "reason": "Different models and implementations; Florence2 fix is not the same concrete change as CohereASR." - }, - { - "left": "pull_request:46907", - "right": "pull_request:46908", - "accept": false, - "reason": "Both are integration-test fixes, but for unrelated models and failure modes." - }, - { - "left": "pull_request:46907", - "right": "pull_request:46926", - "accept": false, - "reason": "Different model failures: EdgeTam import_or_config vs MiniMax-M3-VL load_error." - }, - { - "left": "pull_request:46907", - "right": "pull_request:46927", - "accept": false, - "reason": "Different concrete issues: EdgeTam kwarg filtering vs GLM46V input dtype conversion." - }, - { - "left": "pull_request:46907", - "right": "pull_request:46928", - "accept": false, - "reason": "Unrelated fixes: EdgeTam model kwarg cleanup vs BLIP-2 tokenizer registration." - }, - { - "left": "pull_request:46907", - "right": "pull_request:46941", - "accept": false, - "reason": "Different bugs in different models: EdgeTam import_or_config vs DAC output mismatch." - }, - { - "left": "pull_request:46908", - "right": "pull_request:46926", - "accept": false, - "reason": "Both are load_error test fixes, but for different models and causes." - }, - { - "left": "pull_request:46908", - "right": "pull_request:46928", - "accept": false, - "reason": "Unrelated failures: DeepSeekV32 offload test setup vs BLIP-2 tokenizer mapping." - }, - { - "left": "pull_request:46908", - "right": "pull_request:46941", - "accept": false, - "reason": "Different model issues: DeepSeekV32 load_error vs DAC output mismatch." - }, - { - "left": "pull_request:46908", - "right": "pull_request:46950", - "accept": false, - "reason": "Different failure modes and code paths: DeepSeekV32 load setup vs Mamba2 OOM behavior." - }, - { - "left": "pull_request:46908", - "right": "pull_request:46951", - "accept": false, - "reason": "Different subsystems: DeepSeekV32 test offload setup vs PEFT integration loading behavior." - }, - { - "left": "pull_request:46926", - "right": "pull_request:46941", - "accept": false, - "reason": "Unrelated model fixes: MiniMax-M3-VL load_error vs DAC output mismatch." - }, - { - "left": "pull_request:46926", - "right": "pull_request:46928", - "accept": false, - "reason": "Different areas: MiniMax-M3-VL test loading vs BLIP-2 tokenizer registration." - }, - { - "left": "pull_request:46926", - "right": "pull_request:46927", - "accept": false, - "reason": "Different bugs in different models: load_error workaround vs GLM46V input handling." - }, - { - "left": "pull_request:46927", - "right": "pull_request:46949", - "accept": false, - "reason": "Unrelated model issues: GLM46V input_id casting vs Whisper cache update logic." - }, - { - "left": "pull_request:46927", - "right": "pull_request:46951", - "accept": false, - "reason": "Different subsystems: GLM46V model input handling vs PEFT loading behavior." - }, - { - "left": "pull_request:46927", - "right": "pull_request:46950", - "accept": false, - "reason": "Unrelated bugs: GLM46V input dtype fix vs Mamba2 OOM-related test fix." - }, - { - "left": "pull_request:46927", - "right": "pull_request:46941", - "accept": false, - "reason": "Different concrete code paths: GLM46V input casting vs DAC distance computation." - }, - { - "left": "pull_request:46928", - "right": "pull_request:46941", - "accept": false, - "reason": "Unrelated changes: BLIP-2 tokenizer registry vs DAC vector quantization math." - }, - { - "left": "pull_request:46928", - "right": "pull_request:46951", - "accept": false, - "reason": "Different subsystems: BLIP-2 tokenizer auto-mapping vs PEFT loading." - }, - { - "left": "pull_request:46947", - "right": "pull_request:46953", - "accept": false, - "reason": "CohereASR loss-shift fix vs fused RMSNorm default-dtype cleanup are unrelated." - }, - { - "left": "pull_request:46949", - "right": "pull_request:46951", - "accept": false, - "reason": "Different failures: Whisper cache handling vs PEFT hotswap loading." - }, - { - "left": "pull_request:46949", - "right": "pull_request:46950", - "accept": false, - "reason": "Whisper cache fix and Mamba2 OOM/test behavior are unrelated." - }, - { - "left": "pull_request:46949", - "right": "pull_request:46953", - "accept": false, - "reason": "Whisper attention/cache fix vs fused RMSNorm dtype cleanup are unrelated." - }, - { - "left": "pull_request:46950", - "right": "pull_request:46951", + "left": "pull_request:46741", + "right": "pull_request:46773", "accept": false, - "reason": "Mamba2 OOM/test changes vs PEFT loading changes are unrelated." + "reason": "Both touch linear-attention model files, but 46741 fixes chunked-prefill/speculative-decoding cache handling while 46773 fixes padding-mask application for batch-size-1; different bugs and different code paths." }, { - "left": "pull_request:46951", - "right": "pull_request:46953", + "left": "pull_request:46741", + "right": "pull_request:46781", "accept": false, - "reason": "PEFT loading fix vs FusedRMSNorm dtype cleanup are unrelated." - }, - { - "left": "pull_request:46895", - "right": "pull_request:46947", - "accept": true, - "reason": "Same CohereASR training-loss double-shift bug and same model code path." + "reason": "46741 is a model-specific cache/state bug fix, whereas 46781 changes QuantizedCache construction for hybrid SSM+attention models in cache_utils; these are unrelated failures." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative on the soft-edge calls. The CohereASR/Florence2 cluster is the only clearly supported duplicate group, and the other pairings are reasonably rejected as unrelated or model-specific fixes. No evidence of 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 }, @@ -25144,64 +25219,48 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-46741-4", + "cluster_id": "cluster-46895-2", "nodes": [ - "pull_request:46741", - "pull_request:46773", - "pull_request:46781", - "pull_request:46937" + "pull_request:46895", + "pull_request:46958" ], "soft_pairs": [ - "pull_request:46741|pull_request:46773", - "pull_request:46741|pull_request:46781", - "pull_request:46773|pull_request:46937" + "pull_request:46895|pull_request:46958" ], - "prepared_review_unit_hash": "4372d057729a1deed20833e96809ff1a74d1c39a9f6dbb7bdae0ca40fbbdcc19", + "prepared_review_unit_hash": "fdb7d9b457e40416e9df22f14c67b53b3b5bddf0d768aede8e15d3ca7ba84e59", "component_budget": { - "node_count": 4, - "item_count": 4, - "soft_pair_count": 3, - "serialized_chars": 11148, - "estimated_input_tokens": 2787, - "estimated_eval_tokens": 5830 + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4790, + "estimated_input_tokens": 1198, + "estimated_eval_tokens": 2652 }, "budget": { - "node_count": 4, - "item_count": 4, - "soft_pair_count": 3, - "serialized_chars": 11148, - "estimated_input_tokens": 2787, - "estimated_eval_tokens": 5830 + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4790, + "estimated_input_tokens": 1198, + "estimated_eval_tokens": 2652 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All four pull requests are related by broad stateful/linear-attention model handling, but they fix different concrete bugs in different code paths, so none should be merged as duplicates.", - "confidence": 0.96, + "summary": "Both pull requests target the same CohereASR training-loss double-shift bug under issue #46894, with differing implementation details. PR #46895 looks like the stronger canonical fix; #46958 is likely an alternate/less complete implementation of the same change.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46773 is the cleanest representative of the cluster: it fixes a concrete, narrowly-scoped padding-mask bug with a clear condition change and test coverage across several linear-attention models.", + "canonical_pr_reason": "PR #46895 has the clearer, more complete fix path for the double-shift bug and more review activity, while still targeting the same issue as #46958.", "best_issue_reason": null, - "best_pr_reason": "PR 46773 is the best single PR to represent this group because it has the most direct and generalizable fix pattern, while still targeting one specific defect and validating it with tests.", + "best_pr_reason": "PR #46895 is the best representative of the underlying bug fix because it directly handles when labels should be shifted and appears to preserve the intended loss behavior.", "soft_edge_verdicts": [ { - "left": "pull_request:46741", - "right": "pull_request:46773", - "accept": false, - "reason": "Different bugs: 46741 changes cache/precomputed-state handling for chunked-prefill/speculative decoding, while 46773 fixes padding masks being skipped for batch size 1. Shared model filenames do not make the underlying change the same." - }, - { - "left": "pull_request:46741", - "right": "pull_request:46781", - "accept": false, - "reason": "46741 is about model forward-path cache state reuse; 46781 is about QuantizedCache construction for hybrid models in cache_utils. They affect different mechanisms and are not the same fix." - }, - { - "left": "pull_request:46773", - "right": "pull_request:46937", - "accept": false, - "reason": "46773 fixes padding-mask application; 46937 marks LFM2/LFM2-MoE as stateful to reject assisted generation. These are distinct behaviors and code paths, not one underlying bug." + "left": "pull_request:46895", + "right": "pull_request:46958", + "accept": true, + "reason": "Same explicit issue target and same underlying CohereASR double-shift training-loss bug; the implementation differs, but both are fix attempts for the same code-path problem." } ] }, @@ -25231,7 +25290,7 @@ "soft_pairs": [ "pull_request:46933|pull_request:46942" ], - "prepared_review_unit_hash": "3685cd71678df2f5529cdc75f08e01ef090039b5695e4b167b2818c1150a4186", + "prepared_review_unit_hash": "15db6f7145d3893446235c4936314bff1b7d12c0ca0e670bb0fb95d6de5243cb", "component_budget": { "node_count": 2, "item_count": 2, @@ -25253,18 +25312,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two unrelated PRs address the same class of quantization dtype-casting bug in different model families, but they are not duplicates of each other.", - "confidence": 0.93, + "summary": "Both PRs address the same general anti-pattern\u2014casting activations to a quantized module\u2019s weight dtype\u2014but they fix different model-specific code paths in different architectures, so they are not duplicates.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46933 is the clearest representative: it has an explicit issue target and patches both Gemma4 embedders involved in the dtype-cast fix.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR 46933 is the strongest standalone representative of the cluster because it has the explicit tracking issue and a slightly broader Gemma4 fix surface than the TimesFM change.", + "best_pr_reason": "PR 46933 is the better representative of the cluster because it has an explicit issue link and fixes two Gemma4 embedder paths, but it is still not the same change as PR 46942.", "soft_edge_verdicts": [ { "left": "pull_request:46933", "right": "pull_request:46942", "accept": false, - "reason": "Both PRs fix quantized-weight dtype casting, but they target different models and separate code paths (Gemma4 vision/audio embedders vs TimesFM 2.5 ResidualBlock). They are conceptually similar, not the same underlying change, and should not be merged into one PR." + "reason": "Similar fix pattern, but different concrete bugs and code paths: Gemma4 vision/audio embedders vs TimesFM 2.5 ResidualBlock. They would not plausibly be merged into one PR." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index 2c35a3546e59b48f42fc60a94aff1e639cc8528a..9da67d4fcb5f092a798f7978a73bcecfb99d0f2b 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260629t060016z", + "analysis_id": "hybrid-model-20260629t120018z", "archived_artifacts": { - "hybrid": "snapshots/20260629T060016Z/analysis-runs/hybrid-model-20260629t060016z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260629T060016Z/analysis-runs/hybrid-model-20260629t060016z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-06-29T06:10:10Z", + "published_at": "2026-06-29T12:11:21Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260629T060016Z", + "snapshot_id": "20260629T120018Z", "variant": "hybrid" } diff --git a/snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.json b/snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..f423eb4fba86a2afa7d0f533b68750a0da036cb3 --- /dev/null +++ b/snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.json @@ -0,0 +1,2650 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260629T120018Z", + "generated_at": "2026-06-29T12:11:18Z", + "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-46894-4", + "summary": "Cluster of 2 issues and 2 PRs centered on issue #46894.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 46894, + "canonical_pr_number": 46895, + "issue_numbers": [ + 46894, + 46897 + ], + "pr_numbers": [ + 46895, + 46958 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46895, + "right_pr_number": 46958, + "code_similarity": 0.715, + "size_similarity": 0.833, + "file_overlap": 1.0, + "area_overlap": 0.366, + "patch_similarity": 0.802, + "shared_filenames": [ + "src/transformers/models/cohere_asr/modeling_cohere_asr.py", + "src/transformers/models/cohere_asr/modular_cohere_asr.py", + "tests/models/cohere_asr/test_modeling_cohere_asr.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/cohere_asr/modeling_cohere_asr.py", + "left_ranges": [ + [ + 615, + 627 + ], + [ + 639, + 654 + ] + ], + "right_ranges": [ + [ + 22, + 28 + ], + [ + 638, + 645 + ] + ] + }, + { + "filename": "src/transformers/models/cohere_asr/modular_cohere_asr.py", + "left_ranges": [ + [ + 478, + 490 + ], + [ + 502, + 517 + ] + ], + "right_ranges": [ + [ + 16, + 22 + ], + [ + 501, + 508 + ] + ] + }, + { + "filename": "tests/models/cohere_asr/test_modeling_cohere_asr.py", + "left_ranges": [ + [ + 35, + 41 + ], + [ + 176, + 214 + ] + ], + "right_ranges": [ + [ + 15, + 21 + ], + [ + 180, + 221 + ] + ] + } + ] + } + ] + }, + { + "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-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-46869-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46869.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 46869, + "canonical_pr_number": 46886, + "issue_numbers": [ + 46869 + ], + "pr_numbers": [ + 46886, + 46922 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 46886, + "right_pr_number": 46922, + "code_similarity": 0.711, + "size_similarity": 0.857, + "file_overlap": 1.0, + "area_overlap": 0.5, + "patch_similarity": 0.429, + "shared_filenames": [ + "src/transformers/models/yolos/modeling_yolos.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/yolos/modeling_yolos.py", + "left_ranges": [ + [ + 446, + 452 + ] + ], + "right_ranges": [ + [ + 434, + 440 + ], + [ + 446, + 452 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46159-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46159.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 46159, + "canonical_pr_number": 46174, + "issue_numbers": [ + 46159 + ], + "pr_numbers": [ + 46174, + 46780 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 46174, + "right_pr_number": 46780, + "code_similarity": 0.841, + "size_similarity": 0.87, + "file_overlap": 1.0, + "area_overlap": 0.767, + "patch_similarity": 0.657, + "shared_filenames": [ + "src/transformers/models/aimv2/modeling_aimv2.py", + "src/transformers/models/d_fine/modeling_d_fine.py", + "src/transformers/models/deimv2/modeling_deimv2.py", + "src/transformers/models/pp_doclayout_v2/modeling_pp_doclayout_v2.py", + "src/transformers/models/pp_doclayout_v3/modeling_pp_doclayout_v3.py", + "src/transformers/models/rt_detr/modeling_rt_detr.py", + "src/transformers/models/rt_detr_v2/modeling_rt_detr_v2.py", + "src/transformers/models/vit_mae/modeling_vit_mae.py", + "src/transformers/models/vit_mae/modular_vit_mae.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/aimv2/modeling_aimv2.py", + "left_ranges": [ + [ + 134, + 141 + ], + [ + 144, + 154 + ], + [ + 157, + 165 + ] + ], + "right_ranges": [ + [ + 134, + 140 + ], + [ + 143, + 153 + ], + [ + 156, + 162 + ] + ] + }, + { + "filename": "src/transformers/models/d_fine/modeling_d_fine.py", + "left_ranges": [ + [ + 631, + 638 + ], + [ + 641, + 651 + ], + [ + 654, + 662 + ] + ], + "right_ranges": [ + [ + 631, + 637 + ], + [ + 640, + 650 + ], + [ + 653, + 659 + ] + ] + }, + { + "filename": "src/transformers/models/deimv2/modeling_deimv2.py", + "left_ranges": [ + [ + 703, + 710 + ], + [ + 713, + 723 + ], + [ + 726, + 734 + ] + ], + "right_ranges": [ + [ + 703, + 709 + ], + [ + 712, + 722 + ], + [ + 725, + 731 + ] + ] + }, + { + "filename": "src/transformers/models/pp_doclayout_v2/modeling_pp_doclayout_v2.py", + "left_ranges": [ + [ + 1551, + 1558 + ], + [ + 1561, + 1571 + ], + [ + 1574, + 1582 + ] + ], + "right_ranges": [ + [ + 1551, + 1557 + ], + [ + 1560, + 1570 + ], + [ + 1573, + 1579 + ] + ] + }, + { + "filename": "src/transformers/models/pp_doclayout_v3/modeling_pp_doclayout_v3.py", + "left_ranges": [ + [ + 804, + 811 + ], + [ + 814, + 824 + ], + [ + 827, + 835 + ] + ], + "right_ranges": [ + [ + 804, + 810 + ], + [ + 813, + 823 + ], + [ + 826, + 832 + ] + ] + }, + { + "filename": "src/transformers/models/rt_detr/modeling_rt_detr.py", + "left_ranges": [ + [ + 847, + 854 + ], + [ + 857, + 867 + ], + [ + 870, + 878 + ] + ], + "right_ranges": [ + [ + 847, + 853 + ], + [ + 856, + 866 + ], + [ + 869, + 875 + ] + ] + }, + { + "filename": "src/transformers/models/rt_detr_v2/modeling_rt_detr_v2.py", + "left_ranges": [ + [ + 976, + 983 + ], + [ + 986, + 996 + ], + [ + 999, + 1007 + ] + ], + "right_ranges": [ + [ + 976, + 982 + ], + [ + 985, + 995 + ], + [ + 998, + 1004 + ] + ] + }, + { + "filename": "src/transformers/models/vit_mae/modeling_vit_mae.py", + "left_ranges": [ + [ + 152, + 159 + ], + [ + 162, + 172 + ], + [ + 175, + 183 + ] + ], + "right_ranges": [ + [ + 152, + 158 + ], + [ + 161, + 171 + ], + [ + 174, + 180 + ] + ] + }, + { + "filename": "src/transformers/models/vit_mae/modular_vit_mae.py", + "left_ranges": [ + [ + 62, + 69 + ], + [ + 72, + 82 + ], + [ + 85, + 93 + ] + ], + "right_ranges": [ + [ + 62, + 68 + ], + [ + 71, + 81 + ], + [ + 84, + 90 + ] + ] + } + ] + } + ] + }, + { + "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": 0.721, + "size_similarity": 0.973, + "file_overlap": 0.75, + "area_overlap": 0.563, + "patch_similarity": 0.696, + "shared_filenames": [ + "src/transformers/testing_utils.py", + "tests/utils/test_testing_utils.py", + "utils/notification_service.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/testing_utils.py", + "left_ranges": [ + [ + 3611, + 3618 + ], + [ + 3835, + 3869 + ] + ], + "right_ranges": [ + [ + 3525, + 3558 + ], + [ + 3782, + 3788 + ] + ] + }, + { + "filename": "tests/utils/test_testing_utils.py", + "left_ranges": [ + [ + 1, + 144 + ] + ], + "right_ranges": [ + [ + 1, + 114 + ] + ] + }, + { + "filename": "utils/notification_service.py", + "left_ranges": [ + [ + 935, + 967 + ] + ], + "right_ranges": [ + [ + 935, + 967 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46305-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46305.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 46305, + "canonical_pr_number": 46546, + "issue_numbers": [ + 46305 + ], + "pr_numbers": [ + 46310, + 46546 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46310, + "right_pr_number": 46546, + "code_similarity": 0.566, + "size_similarity": 0.444, + "file_overlap": 0.5, + "area_overlap": 0.867, + "patch_similarity": 0.16, + "shared_filenames": [ + "src/transformers/models/moonshine_streaming/modeling_moonshine_streaming.py", + "src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/moonshine_streaming/modeling_moonshine_streaming.py", + "left_ranges": [ + [ + 356, + 362 + ] + ], + "right_ranges": [ + [ + 355, + 361 + ] + ] + }, + { + "filename": "src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py", + "left_ranges": [ + [ + 269, + 275 + ] + ], + "right_ranges": [ + [ + 269, + 275 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45081-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45081.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45081, + "canonical_pr_number": 45317, + "issue_numbers": [ + 45081 + ], + "pr_numbers": [ + 45086, + 45317 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45086, + "right_pr_number": 45317, + "code_similarity": 0.589, + "size_similarity": 0.34, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.136, + "shared_filenames": [ + "src/transformers/tokenization_utils_tokenizers.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/tokenization_utils_tokenizers.py", + "left_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ], + "right_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43698-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43698.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43698, + "canonical_pr_number": 43779, + "issue_numbers": [ + 43698 + ], + "pr_numbers": [ + 43779, + 43816 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43779, + "right_pr_number": 43816, + "code_similarity": 0.418, + "size_similarity": 0.538, + "file_overlap": 1.0, + "area_overlap": 0.02, + "patch_similarity": 0.02, + "shared_filenames": [ + "src/transformers/integrations/integration_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/integration_utils.py", + "left_ranges": [ + [ + 2227, + 2241 + ], + [ + 2303, + 2309 + ] + ], + "right_ranges": [ + [ + 2278, + 2291 + ], + [ + 2309, + 2322 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45332-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 45333, + "issue_numbers": [], + "pr_numbers": [ + 45332, + 45333 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45332, + "right_pr_number": 45333, + "code_similarity": 0.612, + "size_similarity": 0.35, + "file_overlap": 0.389, + "area_overlap": 0.988, + "patch_similarity": 0.531, + "shared_filenames": [ + "docs/source/en/_toctree.yml", + "docs/source/en/heterogeneous_configurations.md", + "src/transformers/configuration_utils.py", + "src/transformers/integrations/heterogeneity/__init__.py", + "src/transformers/integrations/heterogeneity/configuration_utils.py", + "tests/heterogeneity/__init__.py", + "tests/heterogeneity/test_configuration_utils.py" + ], + "shared_file_areas": [ + { + "filename": "docs/source/en/_toctree.yml", + "left_ranges": [ + [ + 16, + 23 + ] + ], + "right_ranges": [ + [ + 16, + 23 + ] + ] + }, + { + "filename": "docs/source/en/heterogeneous_configurations.md", + "left_ranges": [ + [ + 1, + 206 + ] + ], + "right_ranges": [ + [ + 1, + 206 + ] + ] + }, + { + "filename": "src/transformers/configuration_utils.py", + "left_ranges": [ + [ + 30, + 36 + ], + [ + 123, + 129 + ], + [ + 191, + 198 + ], + [ + 306, + 315 + ], + [ + 320, + 328 + ], + [ + 447, + 454 + ], + [ + 857, + 865 + ], + [ + 963, + 971 + ], + [ + 989, + 1004 + ], + [ + 1027, + 1035 + ], + [ + 1080, + 1088 + ], + [ + 1367, + 1376 + ] + ], + "right_ranges": [ + [ + 30, + 36 + ], + [ + 123, + 129 + ], + [ + 191, + 198 + ], + [ + 306, + 315 + ], + [ + 320, + 328 + ], + [ + 447, + 454 + ], + [ + 857, + 865 + ], + [ + 963, + 971 + ], + [ + 989, + 1004 + ], + [ + 1027, + 1035 + ], + [ + 1080, + 1088 + ], + [ + 1367, + 1376 + ] + ] + }, + { + "filename": "src/transformers/integrations/heterogeneity/__init__.py", + "left_ranges": [ + [ + 1, + 35 + ] + ], + "right_ranges": [ + [ + 1, + 22 + ] + ] + }, + { + "filename": "src/transformers/integrations/heterogeneity/configuration_utils.py", + "left_ranges": [ + [ + 1, + 392 + ] + ], + "right_ranges": [ + [ + 1, + 392 + ] + ] + }, + { + "filename": "tests/heterogeneity/test_configuration_utils.py", + "left_ranges": [ + [ + 1, + 343 + ] + ], + "right_ranges": [ + [ + 1, + 343 + ] + ] + } + ] + } + ] + }, + { + "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-46894-4", + "canonical_issue_number": 46894, + "duplicate_issue_numbers": [ + 46897 + ], + "reason": "Issues in cluster-46894-4 are treated as duplicates because they share closing_reference, shared_issue_target, soft_similarity evidence." + } + ], + "duplicate_prs": [ + { + "cluster_id": "cluster-41211-3", + "canonical_pr_number": 44339, + "duplicate_pr_numbers": [ + 41356 + ], + "target_issue_number": 41211, + "reason": "PRs in cluster-41211-3 are treated as duplicates because they converge on issue #41211 with closing_reference evidence." + }, + { + "cluster_id": "cluster-43240-3", + "canonical_pr_number": 43251, + "duplicate_pr_numbers": [ + 43254 + ], + "target_issue_number": 43240, + "reason": "PRs in cluster-43240-3 are treated as duplicates because they converge on issue #43240 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43698-3", + "canonical_pr_number": 43779, + "duplicate_pr_numbers": [ + 43816 + ], + "target_issue_number": 43698, + "reason": "PRs in cluster-43698-3 are treated as duplicates because they converge on issue #43698 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43979-11", + "canonical_pr_number": 44007, + "duplicate_pr_numbers": [ + 43996, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "target_issue_number": 43979, + "reason": "PRs in cluster-43979-11 are treated as duplicates because they converge on issue #43979 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-44018-2", + "canonical_pr_number": 44068, + "duplicate_pr_numbers": [ + 44018 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-45081-3", + "canonical_pr_number": 45317, + "duplicate_pr_numbers": [ + 45086 + ], + "target_issue_number": 45081, + "reason": "PRs in cluster-45081-3 are treated as duplicates because they converge on issue #45081 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45332-2", + "canonical_pr_number": 45333, + "duplicate_pr_numbers": [ + 45332 + ], + "target_issue_number": null, + "reason": "PRs in cluster-45332-2 are treated as duplicates because they share 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-46159-3", + "canonical_pr_number": 46174, + "duplicate_pr_numbers": [ + 46780 + ], + "target_issue_number": 46159, + "reason": "PRs in cluster-46159-3 are treated as duplicates because they converge on issue #46159 with closing_reference evidence." + }, + { + "cluster_id": "cluster-46305-3", + "canonical_pr_number": 46546, + "duplicate_pr_numbers": [ + 46310 + ], + "target_issue_number": 46305, + "reason": "PRs in cluster-46305-3 are treated as duplicates because they converge on issue #46305 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-46869-3", + "canonical_pr_number": 46886, + "duplicate_pr_numbers": [ + 46922 + ], + "target_issue_number": 46869, + "reason": "PRs in cluster-46869-3 are treated as duplicates because they converge on issue #46869 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-46894-4", + "canonical_pr_number": 46895, + "duplicate_pr_numbers": [ + 46958 + ], + "target_issue_number": 46894, + "reason": "PRs in cluster-46894-4 are treated as duplicates because they converge on issue #46894 with closing_reference, shared_issue_target, soft_similarity evidence." + } + ], + "best_issue": { + "cluster_id": "cluster-46894-4", + "issue_number": 46894, + "reason": "Issue #46894 is the strongest global issue candidate because it is open, belongs to a cluster with 4 artifacts, and carries 0 discussion comments plus 6 inbound references.", + "score": 146.0 + }, + "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/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..ad0ec39cd94baaca5c0772d3e9f8d7f6a21f99bb --- /dev/null +++ b/snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,25338 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260629T120018Z", + "generated_at": "2026-06-29T12:11:18Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 1, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:24643", + "issue:30277", + "issue:34634", + "issue:34928", + "issue:35545", + "issue:36010", + "issue:36331", + "issue:39290", + "issue:41720", + "issue:41762", + "issue:42915", + "issue:43039", + "issue:43425", + "issue:43541", + "issue:43596", + "issue:43638", + "issue:43716", + "issue:43827", + "issue:43828", + "issue:43975", + "issue:44292", + "issue:44322", + "issue:44457", + "issue:44509", + "issue:44512", + "issue:44560", + "issue:44805", + "issue:44841", + "issue:44918", + "issue:45081", + "issue:45161", + "issue:45237", + "issue:45464", + "issue:45507", + "issue:45685", + "issue:45800", + "issue:45823", + "issue:45859", + "issue:45874", + "issue:45901", + "issue:45902", + "issue:46015", + "issue:46165", + "issue:46294", + "issue:46468", + "issue:46515", + "issue:46661", + "issue:46869" + ], + "soft_pairs": [ + "issue:44322|issue:45464", + "issue:45237|issue:45685", + "issue:43716|issue:45237", + "issue:43828|issue:45237", + "issue:45237|issue:46468", + "issue:45237|issue:45507", + "issue:30277|issue:35545", + "issue:35545|issue:42915", + "issue:35545|issue:41720", + "issue:45901|issue:46015", + "issue:44509|issue:44512", + "issue:35545|issue:36010", + "issue:44918|issue:46165", + "issue:46165|issue:46869", + "issue:44805|issue:45823", + "issue:43638|issue:44805", + "issue:43638|issue:46165", + "issue:43596|issue:44805", + "issue:44805|issue:46294", + "issue:44805|issue:45161", + "issue:46165|issue:46468", + "issue:44560|issue:46165", + "issue:34634|issue:35545", + "issue:30277|issue:43638", + "issue:43596|issue:43638", + "issue:43039|issue:45161", + "issue:43975|issue:44457", + "issue:44292|issue:44457", + "issue:44841|issue:46661", + "issue:43716|issue:46869", + "issue:34928|issue:44805", + "issue:24643|issue:30277", + "issue:45685|issue:46869", + "issue:45859|issue:45902", + "issue:43638|issue:45161", + "issue:34928|issue:36331", + "issue:43596|issue:45161", + "issue:43716|issue:45507", + "issue:45874|issue:46294", + "issue:44292|issue:45237", + "issue:39290|issue:45081", + "issue:45823|issue:45874", + "issue:45800|issue:46515", + "issue:43425|issue:44292", + "issue:34928|issue:41720", + "issue:43828|issue:45161", + "issue:43828|issue:44292", + "issue:43541|issue:43716", + "issue:43716|issue:46468", + "issue:44292|issue:44918", + "issue:45507|issue:45685", + "issue:43716|issue:44292", + "issue:44918|issue:45161", + "issue:43716|issue:45685", + "issue:41762|issue:43638", + "issue:43039|issue:44918", + "issue:43827|issue:44512", + "issue:43828|issue:45685", + "issue:43039|issue:43828", + "issue:30277|issue:36331", + "issue:34928|issue:36010", + "issue:45237|issue:46015", + "issue:43039|issue:44805" + ], + "prepared_review_unit_hash": "0d38d2f8a2898a5e56f0042f457ce204b7262d5b8593c484ed6265c9a6f26983", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 63, + "serialized_chars": 46569, + "estimated_input_tokens": 11643, + "estimated_eval_tokens": 23542 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous, with a few true duplicate subclusters: DeepSpeed ZeRO-3 index-0-out-of-bounds reports around BERT/Gemma3, a Qwen3.5 MoE rotary-embedding issue, and a torch 2.4.1 compatibility pair. Most other soft edges are only loosely theme-similar and should stay separate.", + "confidence": 0.72, + "canonical_issue_reason": "Best anchor is issue 43638: it sits at the center of the clearest duplicate subcluster (43596 and 41762) and captures the shared ZeRO-3 index-0-out-of-bounds failure more generically than the model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43638 is the strongest representative of the concrete ZeRO-3/index-error bug family in this set, with clean duplicate evidence from both 43596 and 41762.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Different failures: Qwen3.5 config attribute error vs streaming chat/completions API failure." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Unrelated hardware-specific bugs; AMD GPU failure is not the same as an MPS histogram-op limitation." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Both mention runtime problems, but the models, platforms, and concrete failures differ." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Only share a broad dtype/device theme; the actual model and error paths are different." + }, + { + "left": "issue:45237", + "right": "issue:46468", + "accept": false, + "reason": "GPT-OSS on AMD GPUs vs a broken Nemotron example are different issues." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Different model families and symptoms; not the same underlying bug." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "DeepSpeed collective mismatch and ONNX export failure are unrelated." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "ModernBERT ONNX export and Qwen3Moe FP8 config failure are different bugs." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "ONNX export error vs Qwen3 auto-device-map CUDA assert; same general area, different bug." + }, + { + "left": "issue:45901", + "right": "issue:46015", + "accept": false, + "reason": "Table QA crash vs outdated RAG examples are not the same issue." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are v5 docs cleanup, but they target different removed content and would not be one duplicate bug." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "ModernBERT export failure and GenerationMixin import error are unrelated." + }, + { + "left": "issue:44918", + "right": "issue:46165", + "accept": false, + "reason": "TRL SFT embedding unpacking and resume/trust_remote_code are different training issues." + }, + { + "left": "issue:46165", + "right": "issue:46869", + "accept": false, + "reason": "Remote-code resume failure and YOLO multi-GPU device-map failure are unrelated." + }, + { + "left": "issue:44805", + "right": "issue:45823", + "accept": false, + "reason": "Mask-shape mismatch and Gemma4 PLE device-mapping failure are different forward-path bugs." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Both are runtime errors, but one is a ZeRO-3 init index error and the other is a mask/tensor shape mismatch." + }, + { + "left": "issue:43638", + "right": "issue:46165", + "accept": false, + "reason": "ZeRO-3 index error vs resume/trust_remote_code problem; no shared code-path." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "Different failure modes: ZeRO-3 index error versus mask/tensor shape mismatch." + }, + { + "left": "issue:44805", + "right": "issue:46294", + "accept": false, + "reason": "These are different model-input bugs; one is mask indexing, the other is video feature/token mismatch." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Mask-shape mismatch is not the same as GPT-OSS TP failure." + }, + { + "left": "issue:46165", + "right": "issue:46468", + "accept": false, + "reason": "Remote-code resume failure and broken Nemotron example are unrelated." + }, + { + "left": "issue:44560", + "right": "issue:46165", + "accept": false, + "reason": "Qwen3-vl video StopIteration and training-resume remote-code issue are different." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "BarkProcessor voice_preset bug and ModernBERT ONNX export are unrelated." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Both involve DeepSpeed, but one is collective mismatch and the other is ZeRO-3 index error; different bugs." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "Same ZeRO-3 index-0-out-of-bounds bug during Bert model loading/init; only the model wording differs." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger Kernel cross_entropy being called is not the same as GPT-OSS only-TP failure." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "Incorrect detokenization and LoRA merge/save inconsistency are different behaviors." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Qwen-3-8B NVFP4 runtime failure vs LoRA merge/save mismatch are unrelated." + }, + { + "left": "issue:44841", + "right": "issue:46661", + "accept": false, + "reason": "Processor failure for Voxtral and remote-code config_class error are different problems." + }, + { + "left": "issue:43716", + "right": "issue:46869", + "accept": false, + "reason": "Mistral dtype mismatch and YOLO multi-GPU device-map failure are not the same bug." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "Activation-checkpointing/FSDP recomputation mismatch is not the same as a mask indexing error." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "DeepSpeed training error about weight shape and DeepSpeed collective mismatch are different issues." + }, + { + "left": "issue:45685", + "right": "issue:46869", + "accept": false, + "reason": "MPS histogram-op limitation is unrelated to YOLO multi-GPU auto device mapping." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": true, + "reason": "Same Qwen3.5 MoE rotary-embedding bug family; both point to bad `inv_freq` handling across device/materialization/offload paths." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "ZeRO-3 index error and GPT-OSS TP issue do not share the same underlying failure." + }, + { + "left": "issue:43828", + "right": "issue:44292", + "accept": false, + "reason": "Both are model runtime failures, but the actual mechanisms and models differ." + }, + { + "left": "issue:43541", + "right": "issue:43716", + "accept": false, + "reason": "Mixtral grouped_mm tracing failure and Mistral-3 dtype mismatch are different bugs." + }, + { + "left": "issue:43716", + "right": "issue:46468", + "accept": false, + "reason": "Different model/example and different failure modes." + }, + { + "left": "issue:44292", + "right": "issue:44918", + "accept": false, + "reason": "Qwen NVFP4 runtime failure and TRL SFT embedding unpacking are unrelated." + }, + { + "left": "issue:45507", + "right": "issue:45685", + "accept": false, + "reason": "GraniteMoEHybrid invalid method and MPS histogram-op limitation are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:44292", + "accept": false, + "reason": "Mistral dtype mismatch vs Qwen NVFP4 failure; same broad area, different bugs." + }, + { + "left": "issue:44918", + "right": "issue:45161", + "accept": false, + "reason": "TRL SFT embedding issue and GPT-OSS tensor-parallel issue are not the same underlying bug." + }, + { + "left": "issue:43716", + "right": "issue:45685", + "accept": false, + "reason": "Mistral dtype mismatch and MPS histogram-op limitation are different failures." + }, + { + "left": "issue:41762", + "right": "issue:43638", + "accept": true, + "reason": "Same ZeRO-3 index-0-out-of-bounds loading/init failure; model names differ but the underlying bug is the same." + }, + { + "left": "issue:43039", + "right": "issue:44918", + "accept": false, + "reason": "Liger Kernel cross_entropy call and TRL SFT embedding unpacking are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs regressions after v5, but they refer to different removed docs content and should not be merged as duplicates." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Autocast dtype mismatch and MPS int histogram limitation are different bugs." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Liger Kernel cross_entropy being called and an autocast dtype mismatch are not the same code-path problem." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch and custom trainer compute_loss signature break are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP/checkpointing tensor-size mismatch and GenerationMixin import error are unrelated." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "AMD GPU failure and outdated RAG examples are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Liger Kernel cross_entropy call and mask-shape mismatch are different bugs." + }, + { + "left": "issue:45800", + "right": "issue:46515", + "accept": true, + "reason": "Same torch 2.4.1 incompatibility report; 45800 is the more specific versioned instance of the same issue." + }, + { + "left": "issue:43425", + "right": "issue:44292", + "accept": false, + "reason": "Torch 2.10 incompatibility and Qwen-3-8B NVFP4 failure are different compatibility/runtime issues." + }, + { + "left": "issue:34928", + "right": "issue:41720", + "accept": false, + "reason": "FSDP activation-checkpointing mismatch and Qwen3 auto-device-map CUDA assert are unrelated." + }, + { + "left": "issue:43828", + "right": "issue:45161", + "accept": false, + "reason": "Phi autocast dtype mismatch and GPT-OSS TP failure do not share the same bug." + }, + { + "left": "issue:43828", + "right": "issue:44292", + "accept": false, + "reason": "Different models and failure mechanisms; both are runtime errors only at a very broad level." + }, + { + "left": "issue:43541", + "right": "issue:43716", + "accept": false, + "reason": "Mixtral grouped_mm tracing and Mistral-3 dtype mismatch are distinct issues." + }, + { + "left": "issue:43716", + "right": "issue:46468", + "accept": false, + "reason": "Mistral-3 dtype mismatch and broken Nemotron example are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:44918", + "accept": false, + "reason": "Qwen NVFP4 runtime failure vs TRL SFT input-embedding unpacking are different bugs." + }, + { + "left": "issue:45507", + "right": "issue:45685", + "accept": false, + "reason": "GraniteMoEHybrid invalid method and MPS histogram-op error are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:44292", + "accept": false, + "reason": "Different model-specific failures; not the same underlying bug." + }, + { + "left": "issue:44918", + "right": "issue:45161", + "accept": false, + "reason": "TRl SFT trainer bug and GPT-OSS tensor-parallel bug are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:45685", + "accept": false, + "reason": "Mistral dtype mismatch and MPS histogram-op limitation are unrelated." + }, + { + "left": "issue:41762", + "right": "issue:43638", + "accept": true, + "reason": "Same ZeRO-3 index-0-out-of-bounds loading/init failure; model names differ but the underlying bug is the same." + }, + { + "left": "issue:43039", + "right": "issue:44918", + "accept": false, + "reason": "Liger Kernel cross_entropy call and TRL SFT embedding issue are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Related v5 docs cleanup, but different removed references; not the same duplicate change." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Autocast dtype mismatch and MPS int histogram limitation are different bugs." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Liger Kernel cross_entropy call and autocast dtype mismatch are different failure paths." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch and custom trainer signature error are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP/checkpointing mismatch and GenerationMixin import error are unrelated." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "AMD GPU runtime failure and outdated RAG docs are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Different bugs: Liger Kernel cross_entropy call versus mask indexing mismatch." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet and stays conservative overall. The three accepted subclusters are supported by the titles/body patterns: ZeRO-3 index-0-out-of-bounds loading errors, the Qwen3.5 MoE rotary/inv_freq bug family, and the torch 2.4.1 incompatibility pair. The rejected soft edges are mostly well-justified as different models, platforms, or failure modes, without overclaiming duplication." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 2, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:36010", + "issue:39290", + "issue:41093", + "issue:41720", + "issue:41762", + "issue:43039", + "issue:43296", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43638", + "issue:43792", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43872", + "issue:43901", + "issue:43975", + "issue:44292", + "issue:44512", + "issue:44560", + "issue:44589", + "issue:44841", + "issue:44863", + "issue:45020", + "issue:45070", + "issue:45084", + "issue:45237", + "issue:45362", + "issue:45399", + "issue:45464", + "issue:45507", + "issue:45685", + "issue:45800", + "issue:45812", + "issue:45823", + "issue:45901", + "issue:45920", + "issue:46015", + "issue:46294", + "issue:46304", + "issue:46402", + "issue:46661", + "issue:46772", + "issue:46869" + ], + "soft_pairs": [ + "issue:43866|issue:46402", + "issue:45800|issue:46304", + "issue:43296|issue:46304", + "issue:41762|issue:43596", + "issue:39290|issue:41720", + "issue:43901|issue:44512", + "issue:44589|issue:46304", + "issue:45237|issue:45901", + "issue:43975|issue:46015", + "issue:43039|issue:44560", + "issue:45823|issue:46294", + "issue:44841|issue:45823", + "issue:43866|issue:46772", + "issue:44863|issue:46015", + "issue:45020|issue:46304", + "issue:45812|issue:45920", + "issue:43866|issue:44863", + "issue:44292|issue:46015", + "issue:43872|issue:46304", + "issue:39290|issue:43531", + "issue:41093|issue:43638", + "issue:43792|issue:45901", + "issue:43854|issue:43866", + "issue:43425|issue:45070", + "issue:43975|issue:45901", + "issue:45362|issue:45464", + "issue:43792|issue:45070", + "issue:43596|issue:45800", + "issue:43541|issue:46869", + "issue:44292|issue:45901", + "issue:39290|issue:45823", + "issue:36010|issue:39290", + "issue:45464|issue:45685", + "issue:43541|issue:45685", + "issue:44841|issue:45084", + "issue:43541|issue:43828", + "issue:43404|issue:43866", + "issue:45507|issue:46869", + "issue:41093|issue:41762", + "issue:45399|issue:46661", + "issue:43572|issue:45070", + "issue:45800|issue:45823" + ], + "prepared_review_unit_hash": "ee4d1220f04e282440873d1fbdc953a3cfbf9d35f3d304cea6a909a19a2694dd", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 42, + "serialized_chars": 40040, + "estimated_input_tokens": 10010, + "estimated_eval_tokens": 20276 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly noisy and heterogeneous: it spans unrelated model-loading, device-mapping, docs, and runtime bugs. The only soft edge that looks like a likely true duplicate is the Granite/AutoTokenizer regression pair (45812 \u2194 45920).", + "confidence": 0.33, + "canonical_issue_reason": "Issue 45920 is the broadest representative of the tokenizer-ID regression subcluster, covering multiple affected models rather than just Granite.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45920 is the best single anchor for the cluster\u2019s only plausible duplicate theme because it describes the wider v5 tokenizer regression across several models.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43866", + "right": "issue:46402", + "accept": false, + "reason": "Different bugs: corrupted checkpoint vs. text-encoder weights not loading." + }, + { + "left": "issue:45800", + "right": "issue:46304", + "accept": false, + "reason": "Both are version-compatibility reports, but for different torch/transformers breakages." + }, + { + "left": "issue:43296", + "right": "issue:46304", + "accept": false, + "reason": "PaddleOCR-VL load failure is unrelated to USE_DISTRIBUTED=0 support." + }, + { + "left": "issue:41762", + "right": "issue:43596", + "accept": false, + "reason": "Same symptom shape, but different models and likely different zero3 code paths." + }, + { + "left": "issue:39290", + "right": "issue:41720", + "accept": false, + "reason": "Gemma3 config attribute error and Qwen3 device-map CUDA assert are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs issues, but they document different removed/deprecated behaviors." + }, + { + "left": "issue:44589", + "right": "issue:46304", + "accept": false, + "reason": "Float8 storage lookup failure is unrelated to distributed build support." + }, + { + "left": "issue:45237", + "right": "issue:45901", + "accept": false, + "reason": "AMD GPU model failure and table-question-answering crash are different problems." + }, + { + "left": "issue:43975", + "right": "issue:46015", + "accept": false, + "reason": "Detokenization regression is unrelated to outdated RAG examples." + }, + { + "left": "issue:43039", + "right": "issue:44560", + "accept": false, + "reason": "Liger Kernel cross-entropy fallback and Qwen3.5 video StopIteration are unrelated." + }, + { + "left": "issue:45823", + "right": "issue:46294", + "accept": false, + "reason": "Device-map PLE mismatch and chat-template video-token mismatch are different failures." + }, + { + "left": "issue:44841", + "right": "issue:45823", + "accept": false, + "reason": "Voxtral processor failure is unrelated to Gemma4 forward device mismatch." + }, + { + "left": "issue:43866", + "right": "issue:46772", + "accept": false, + "reason": "Corrupted checkpoint vs. ModelOpt/NVFP4 compatibility are different issues." + }, + { + "left": "issue:44863", + "right": "issue:46015", + "accept": false, + "reason": "NemotronH checkpoint loading is unrelated to RAG documentation examples." + }, + { + "left": "issue:45020", + "right": "issue:46304", + "accept": false, + "reason": "Remote-code model breakage and USE_DISTRIBUTED=0 support are separate regression classes." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same underlying tokenizer-ID regression; 45920 is the broader follow-up report across multiple models." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Corrupted checkpoint and NemotronH loading failure are not the same bug." + }, + { + "left": "issue:44292", + "right": "issue:46015", + "accept": false, + "reason": "Qwen-3 NVFP4 execution failure is unrelated to RAG docs." + }, + { + "left": "issue:43872", + "right": "issue:46304", + "accept": false, + "reason": "bitsandbytes Int8Params incompatibility is unrelated to distributed-build support." + }, + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Both mention sliding_window, but Gemma3 vLLM config and Qwen3-MoE sliding_window are distinct." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Both are zero3-related IndexErrors, but the error signatures and model contexts differ." + }, + { + "left": "issue:43792", + "right": "issue:45901", + "accept": false, + "reason": "Whisper run failure is unrelated to table-question-answering crashes." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "GLM-4.7 unit-test loading issue and corrupted Ovis2 checkpoint are unrelated." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Torch compatibility and PretrainedConfig pydantic regression are different breakages." + }, + { + "left": "issue:43975", + "right": "issue:45901", + "accept": false, + "reason": "Detokenization regression is unrelated to table-question-answering crashes." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Qwen3.5 chat crash and streaming inference API failure are different concrete issues." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "Whisper run failure and pydantic PretrainedConfig regression are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "Zero3 BertModel IndexError and torch/transformers version incompatibility are not the same bug." + }, + { + "left": "issue:43541", + "right": "issue:46869", + "accept": false, + "reason": "Mixtral grouped_mm tracing error and YOLO multi-GPU device-map failure are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:45901", + "accept": false, + "reason": "NVFP4 execution issue is unrelated to table QA crashes." + }, + { + "left": "issue:39290", + "right": "issue:45823", + "accept": false, + "reason": "Gemma3 sliding_window attribute error and Gemma4 device mismatch are different model bugs." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "GenerationMixin import error and Gemma3 config attribute error are unrelated." + }, + { + "left": "issue:45464", + "right": "issue:45685", + "accept": false, + "reason": "Streaming chat API failure and MPS histogram kernel error are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:45685", + "accept": false, + "reason": "Grouped_mm tracing error and MPS Int histogram error are different failures." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Processor failure and compile/non-template-node error are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Both involve MoE/runtime dtype problems, but they are different model paths and symptoms." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "lm_head weight tying bug and corrupted checkpoint are unrelated." + }, + { + "left": "issue:45507", + "right": "issue:46869", + "accept": false, + "reason": "GraniteMoEHybrid invalid method and YOLO device-map bug are different issues." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both are IndexErrors, but one is mask-shape mismatch and the other is zero-size tensor during loading." + }, + { + "left": "issue:45399", + "right": "issue:46661", + "accept": false, + "reason": "flash-attn fallback blocking and remote-code config_class errors are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "Missing pad_token_idx and PretrainedConfig pydantic regression are different breakages." + }, + { + "left": "issue:45800", + "right": "issue:45823", + "accept": false, + "reason": "Torch compatibility and Gemma4 device-map mismatch are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet and stays conservative. The only accepted soft edge, 45812 \u2194 45920, is well supported by the titles and discussion: both are AutoTokenizer wrong-token-ID regressions, with 45920 clearly the broader v5 regression report. The rest of the rejected pairs are appropriately treated as different bugs, often with different models or failure modes. No obvious overstatement stands out." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 3, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:39186", + "issue:40990", + "issue:41093", + "issue:42915", + "issue:43054", + "issue:43296", + "issue:43366", + "issue:43404", + "issue:43531", + "issue:43541", + "issue:43596", + "issue:43645", + "issue:43716", + "issue:43828", + "issue:43873", + "issue:43994", + "issue:44291", + "issue:44360", + "issue:44387", + "issue:44479", + "issue:44492", + "issue:44512", + "issue:44530", + "issue:44805", + "issue:44863", + "issue:44918", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45161", + "issue:45183", + "issue:45399", + "issue:45464", + "issue:45632", + "issue:45698", + "issue:45759", + "issue:45800", + "issue:45902", + "issue:46165", + "issue:46294", + "issue:46463", + "issue:46468", + "issue:46489", + "issue:46491", + "issue:46640", + "issue:46661", + "issue:46869" + ], + "soft_pairs": [ + "issue:43296|issue:43366", + "issue:44387|issue:45005", + "issue:44805|issue:44918", + "issue:43541|issue:45161", + "issue:43828|issue:45464", + "issue:43366|issue:45464", + "issue:44291|issue:45070", + "issue:43531|issue:45070", + "issue:41093|issue:43596", + "issue:43716|issue:46294", + "issue:45020|issue:45800", + "issue:43054|issue:43994", + "issue:44479|issue:46294", + "issue:46468|issue:46869", + "issue:45759|issue:45902", + "issue:43596|issue:46165", + "issue:46640|issue:46661", + "issue:43716|issue:44918", + "issue:39186|issue:44918", + "issue:40990|issue:44960", + "issue:43645|issue:45399", + "issue:45632|issue:45698", + "issue:43404|issue:45005", + "issue:42915|issue:45399", + "issue:43873|issue:45005", + "issue:44530|issue:45005", + "issue:44360|issue:44512", + "issue:44492|issue:44512", + "issue:44863|issue:45005", + "issue:46489|issue:46491", + "issue:45183|issue:46463" + ], + "prepared_review_unit_hash": "8902d863cadbb8d9fef1d7d81c2f341abfb4edefe256784d9a508cb40d75ccd1", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36890, + "estimated_input_tokens": 9223, + "estimated_eval_tokens": 18702 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are broad topical overlaps rather than true duplicates. The clearest duplicate candidates are the Qwen3.5 video regression pair and the input-validation error-message pair; the rest are different bugs in the same subsystem.", + "confidence": 0.61, + "canonical_issue_reason": "Issue 44479 is the broadest, most concrete regression report here: it names the affected model families and the specific video-input failure mode, making it the best anchor if these reports are grouped.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44479 is the strongest representative issue overall because it describes a specific, user-facing regression with multiple affected models and a clear failure path.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "Different problems: one is a PaddleOCR-VL load failure, the other is GGUF support for gpt-oss." + }, + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Unrelated bugs; int4 memory/OOM behavior is not the same as tied-weights handling." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Both are shape-related errors, but they occur in different code paths and model/training setups." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Different MoE/Tensor Parallel failures; no clear shared underlying bug." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "One is an autocast dtype mismatch, the other is a streaming chat/completions failure." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "GGUF gpt-oss support and Qwen3.5 streaming inference are unrelated." + }, + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "Both involve v5 compatibility, but they are distinct issues affecting different APIs." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior is not the same as the PretrainedConfig/pydantic breakage." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Both raise IndexError, but the tensor mismatch and zero-sized init failures are different bugs." + }, + { + "left": "issue:43716", + "right": "issue:46294", + "accept": false, + "reason": "Different multimodal failures: dtype mismatch in preprocessing versus video feature/token mismatch." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "Remote-code breakage and torch/transformers version incompatibility are separate problems." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but the reports describe different symptoms and likely different defects." + }, + { + "left": "issue:44479", + "right": "issue:46294", + "accept": true, + "reason": "Same video-input regression family; 46294 looks like a specific instance of the broader Qwen3.5-related bug in 44479." + }, + { + "left": "issue:46468", + "right": "issue:46869", + "accept": false, + "reason": "Official Nemotron example breakage and YOLO multi-GPU device-map failure are unrelated." + }, + { + "left": "issue:45759", + "right": "issue:45902", + "accept": false, + "reason": "Different Qwen 3.5/3.6 issues: composite config unwrapping versus uninitialized rotary embedding memory." + }, + { + "left": "issue:43596", + "right": "issue:46165", + "accept": false, + "reason": "Different loading/training failures with different causes; no common code-path evidence." + }, + { + "left": "issue:46640", + "right": "issue:46661", + "accept": false, + "reason": "Flash Attention fallback behavior and remote-code config_class errors are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Different multimodal/training bugs with no concrete shared mechanism." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "FSDP weight-shape error and TRL embedding unpacking failure are different issues." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "No evidence these perplexity/GLM5 reports share the same underlying defect." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Custom-model notebook initialization and Flash Attention fallback checks are unrelated." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": false, + "reason": "Both involve remote-code/custom module loading, but the reported failure modes are not concrete enough to treat as the same bug." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Tied lm_head weights in Mistral3 and translation-model tied-weight issues are different model families and fixes." + }, + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "FineGrainedFP8Config failure and Flash Attention fallback blocking are unrelated." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization offloading/OOM behavior is not the same as tied-weight handling." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash and tied-weight translation models are unrelated." + }, + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "A DSA indexer discussion and a docs typo are not duplicate bugs." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "Both are documentation/naming issues, but they are different changes." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint loading and tied-weights translation models are unrelated." + }, + { + "left": "issue:46489", + "right": "issue:46491", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and symptoms." + }, + { + "left": "issue:45183", + "right": "issue:46463", + "accept": true, + "reason": "Both describe the same validation/error-message bug class; 46463 appears to be a T5-specific manifestation of the broader XOR/input validation issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet and stays appropriately conservative. The two accepted soft pairs are well-supported as likely same-bug clusters, while the rejections avoid overclaiming despite topical similarity. The overall summary is careful and consistent with the issue titles/evidence provided." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 4, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:24643", + "issue:28530", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:41553", + "issue:41762", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43493", + "issue:43643", + "issue:43644", + "issue:43782", + "issue:43824", + "issue:43828", + "issue:43856", + "issue:43872", + "issue:43881", + "issue:43883", + "issue:43950", + "issue:43975", + "issue:44292", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44589", + "issue:44910", + "issue:44918", + "issue:44928", + "issue:44960", + "issue:44991", + "issue:45020", + "issue:45399", + "issue:45507", + "issue:45520", + "issue:45663", + "issue:45701", + "issue:45859", + "issue:45902", + "issue:45920", + "issue:45987", + "issue:46489", + "issue:46858", + "issue:46899", + "issue:46932" + ], + "soft_pairs": [ + "issue:43975|issue:44292", + "issue:44918|issue:45507", + "issue:43950|issue:44534", + "issue:39692|issue:43054", + "issue:43644|issue:43950", + "issue:36331|issue:41762", + "issue:43872|issue:44589", + "issue:46899|issue:46932", + "issue:43828|issue:44928", + "issue:45920|issue:46489", + "issue:45987|issue:46858", + "issue:43054|issue:45701", + "issue:38175|issue:43054", + "issue:43950|issue:45902", + "issue:41553|issue:43883", + "issue:43054|issue:43493", + "issue:44910|issue:45859", + "issue:42947|issue:44387", + "issue:45663|issue:45902", + "issue:44403|issue:44451", + "issue:43975|issue:45520", + "issue:41553|issue:43881", + "issue:42947|issue:43856", + "issue:24643|issue:28530", + "issue:43643|issue:45020", + "issue:44403|issue:44488", + "issue:41762|issue:42915", + "issue:44991|issue:45399", + "issue:43782|issue:44960", + "issue:43824|issue:43872" + ], + "prepared_review_unit_hash": "3bcee8f123458ecbff84421eba671b47d7c61d2bc9326f6f575e3d815458b721", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36794, + "estimated_input_tokens": 9199, + "estimated_eval_tokens": 18654 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated issue reports with a few small duplicate subgroups. The clearest cluster is the non-persistent-buffer regression anchored by #43950, with #43644 and #44534 as strong matches. #45902 is adjacent but not clearly the same bug, so it should not be treated as a confirmed duplicate of that subgroup. Smaller duplicate families appear around SigLIP2 discrepancies and static-cache seq-length behavior.", + "confidence": 0.66, + "canonical_issue_reason": "#43950 is still the best canonical issue for the non-persistent-buffer regression because it states the broad underlying bug clearly and matches the strongest confirmed duplicates (#43644, #44534). #45902 is related in theme but not safely the same failure mode, so it should not broaden the canonical cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "#43950 is the most representative issue overall because it captures the broadest confirmed regression in the packet and has the strongest supporting duplicate links.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "Different model-specific failures: tokenizer detokenization vs. Qwen3 NVFP4 loading error." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Unrelated code paths: TRL SFT embedding unpacking vs. GraniteMoEHybrid invalid method call." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "Same underlying regression: non-persistent buffers are being corrupted/junk-filled in Transformers v5." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "Docs example errors/quantization failure is not the same bug as degraded SigLIP2 text embeddings." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": true, + "reason": "Both report the same non-persistent buffer corruption regression, with #43644 phrased as the broader symptom." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "Different failures: custom loss signature mismatch vs. DeepSpeed ZeRO-3 loading index error." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "Both are low-level runtime compatibility issues, but they are different errors and different code paths." + }, + { + "left": "issue:46899", + "right": "issue:46932", + "accept": true, + "reason": "Same Gemma4 quantized vision/audio embedder dtype-casting bug; #46932 explicitly references the same fix." + }, + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Different symptoms and model paths: autocast dtype mismatch vs. Qwen3.5 position_ids/Flash Attention crash." + }, + { + "left": "issue:45920", + "right": "issue:46489", + "accept": true, + "reason": "Same tokenizer regression family: #46489 is a specific remaining DeepSeek-Coder v1 case of the broader v5 token-id bug." + }, + { + "left": "issue:45987", + "right": "issue:46858", + "accept": true, + "reason": "Same cache API contract bug: get_seq_length() returning a tensor instead of an int breaks downstream arithmetic." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "SigLIP2 embedding quality complaints are not the same as a generic tokenizer-version change report." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "Related model family, but the reported symptoms differ enough that this is not clearly the same underlying bug." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": false, + "reason": "#45902 is a Qwen3_5Moe rotary-embedding uninitialized-memory issue after meta -> to_empty(cuda) materialization; that is related in spirit but not clearly the same non-persistent-buffer corruption regression as #43950." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Bad AutoTokenizer error messaging for Voxtral is unrelated to Molmo missing tied-weights metadata." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": true, + "reason": "Both point to the same SigLIP2 implementation discrepancy relative to the reference/JAX behavior." + }, + { + "left": "issue:44910", + "right": "issue:45859", + "accept": false, + "reason": "One is a FlashAttention crash from 3D position_ids; the other is CPU-offload incompatibility in rotary embedding." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "Gradient checkpointing/LoRA behavior is unrelated to int4 quantization OOM from reserved CUDA memory growth." + }, + { + "left": "issue:45663", + "right": "issue:45902", + "accept": false, + "reason": "Different model subsystems: Gemma4 FSDP2 shared_kv_states rebuild vs. Qwen3_5Moe rotary embedding materialization." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic loading noise is not the same bug as a model actually failing to load." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "Tokenizer detokenization regression is unrelated to the Python 3.13 flash_attn import error." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Bad tokenizer error messaging for Voxtral is unrelated to glm-4v-9b loading failure." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "Different training inefficiency problems: PEFT gradient checkpointing vs. Qwen3 MoE memory usage." + }, + { + "left": "issue:24643", + "right": "issue:28530", + "accept": false, + "reason": "DeepSpeed weight-shape runtime error is unrelated to early-stopping metric selection." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "Missing fields from AutoConfig with trust_remote_code=True is a narrower config-loading bug, not the same as broad remote_code breakage." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "Generic loading noise is not the same underlying problem as failing to load a specific model." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Different model-loading failures: Gemma3 ZeRO-3 index error vs. Qwen3Moe FineGrainedFP8Config issue." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer load failure and flash-attn fallback logic are unrelated." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Qwen3VL weight_only=True load error is unrelated to GLM5." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "ImportError for Qwen2.5 VL is unrelated to the bitsandbytes Int8Params incompatibility." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Most of the pair judgments are conservative and grounded, but the cluster summary overreaches on the non-persistent-buffer subgroup. In particular, the acceptance of #43950 with #45902 is not well supported by the packet alone: #45902 is a Qwen3_5Moe rotary-embedding uninitialized-memory issue after `meta -> to_empty(cuda)` materialization, which is not clearly the same bug as the explicit non-persistent-buffer regression in #43950/#43644/#44534. Because that link is too strong, the summary\u2019s claim that #45902 is a narrower manifestation of the same regression is overstated. The rest of the summary is mostly fine, but this edge should be downgraded or marked uncertain." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 5, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:36331", + "issue:41093", + "issue:41553", + "issue:42491", + "issue:42947", + "issue:43257", + "issue:43404", + "issue:43792", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43872", + "issue:43881", + "issue:43883", + "issue:43940", + "issue:43950", + "issue:44220", + "issue:44291", + "issue:44368", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44560", + "issue:44661", + "issue:44805", + "issue:44863", + "issue:44928", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45305", + "issue:45313", + "issue:45356", + "issue:45399", + "issue:45488", + "issue:45588", + "issue:45685", + "issue:45715", + "issue:45759", + "issue:45823", + "issue:45923", + "issue:46315", + "issue:46489", + "issue:46661", + "issue:46869", + "issue:46894", + "issue:46897" + ], + "soft_pairs": [ + "issue:45685|issue:45823", + "issue:43881|issue:43883", + "issue:45081|issue:45399", + "issue:45020|issue:45823", + "issue:44661|issue:45020", + "issue:42491|issue:44368", + "issue:43828|issue:45823", + "issue:45020|issue:45356", + "issue:45715|issue:46315", + "issue:45488|issue:46489", + "issue:43827|issue:44509", + "issue:46894|issue:46897", + "issue:45759|issue:46661", + "issue:43872|issue:46869", + "issue:43404|issue:43950", + "issue:43872|issue:44291", + "issue:45020|issue:46661", + "issue:43792|issue:44220", + "issue:43257|issue:45305", + "issue:42947|issue:45923", + "issue:36331|issue:41093", + "issue:41553|issue:44991", + "issue:43854|issue:44863", + "issue:43940|issue:45313", + "issue:44928|issue:45588", + "issue:41093|issue:44805", + "issue:44479|issue:44560", + "issue:41553|issue:44488" + ], + "prepared_review_unit_hash": "bef3d89e437bd0880b77e6e30370eb8dd306e07fe3eb400677d778bf4839db8a", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 36189, + "estimated_input_tokens": 9048, + "estimated_eval_tokens": 18352 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab-bag of unrelated tokenizer/loading/regression issues. The clear duplicate groups are the _is_hf_initialized / init_empty_weights loading bug, the docs pipeline-removal cleanup, and the mask-shape IndexError; most other pairs are only loosely similar and should stay separate.", + "confidence": 0.76, + "canonical_issue_reason": "issue:44291 is the best issue anchor because it states the root cause and affected path for the _is_hf_initialized loading regression, while issue:43872 is the more symptom-focused variant.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44291 is the strongest representative issue in this cluster: concrete traceback, explicit version scope, and a clear fix target make it the most useful duplicate canonical.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45685", + "right": "issue:45823", + "accept": false, + "reason": "Different failures: MPS histogram/int handling vs device-map forward mismatch. Same broad area, not the same bug." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "Both are model-loading failures, but the models and failure modes differ; no shared concrete code-path is evident." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer regex crash vs flash-attn fallback gating. Unrelated fixes." + }, + { + "left": "issue:45020", + "right": "issue:45823", + "accept": false, + "reason": "Broad remote_code regression vs Gemma4 device mismatch. Too different to merge." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "One is an add-new-model-like tokenizer-mapping bug; the other is a broader remote_code regression. Related ecosystem, not same bug." + }, + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "Both mention tied embeddings, but one is a LoRA usability regression and the other is a warning/config issue. Not the same underlying defect." + }, + { + "left": "issue:43828", + "right": "issue:45823", + "accept": false, + "reason": "Autocast dtype mismatch and device-map mismatch are different code paths." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "Remote_code breakage vs Kimi tokenizer codec regression. Different tokenizer/model subsystems." + }, + { + "left": "issue:45715", + "right": "issue:46315", + "accept": false, + "reason": "Separate performance regressions in different conversion directions; not a single bug." + }, + { + "left": "issue:45488", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer output regressions, but they affect different models and likely different tokenizer overrides/fixes." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": true, + "reason": "Same documentation cleanup: both complain that removed v5 pipeline task references still remain in docs." + }, + { + "left": "issue:46894", + "right": "issue:46897", + "accept": true, + "reason": "Same training-loss double-shift bug pattern, reported for two model families and clearly pointing at the same underlying issue." + }, + { + "left": "issue:45759", + "right": "issue:46661", + "accept": false, + "reason": "Composite multimodal config handling vs remote-code config_class AttributeError. Similar surface area only." + }, + { + "left": "issue:43872", + "right": "issue:46869", + "accept": false, + "reason": "bitsandbytes init_empty_weights incompatibility vs YOLO multi-GPU device-map failure. Different problems." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "lm_head tying bug vs non-persistent buffer corruption. No shared fix path." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "These describe the same loading incompatibility around unexpected _is_hf_initialized during init_empty_weights/bitsandbytes-style model loading." + }, + { + "left": "issue:45020", + "right": "issue:46661", + "accept": false, + "reason": "The latter is a specific remote-code AttributeError, but the former is a broader regression report; not enough evidence they are the same bug." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": false, + "reason": "Whisper model failure vs a helper-function issue; could be related, but the underlying defect is not clearly the same." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Both involve DeepSpeed, but one is missing weight conversion and the other is gradient averaging. Different defects." + }, + { + "left": "issue:42947", + "right": "issue:45923", + "accept": false, + "reason": "Gradient checkpointing ineffective with PEFT LoRA is not the same as a missing supports_gradient_checkpointing flag on a remote-code variant." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "Unexpected compute_loss kwarg vs mask/tensor shape mismatch. Distinct failures." + }, + { + "left": "issue:41553", + "right": "issue:44991", + "accept": false, + "reason": "Different tokenizer-loading regressions for different models; only broad symptom overlap." + }, + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "Two separate model-loading/checkpoint problems with different model families and symptoms." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Both are ZeRO-3 loading issues, but the models and failure details differ; not enough to call them duplicates." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "Gradient explosion from Qwen3.5 SDPA/BF16 behavior vs flash-attention sinkless-model crash. Different bugs." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": true, + "reason": "Same mask-shape IndexError, with one issue giving a fuller reproduction of the same underlying tensor indexing bug." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": false, + "reason": "Broad video-input regression across several Qwen models vs a specific StopIteration on Qwen3-vl-embedding. Likely related area, not the same bug." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Both are model-loading complaints, but the models and likely root causes differ." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet, and the accepted duplicate pairs are reasonably supported by the issue titles/descriptions. The canonical choice for issue:44291 is conservative and well-justified, and the rejected soft-edge pairs mostly avoid overclaiming. No PR-mergeability concern applies here since this is an issue cluster." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 6, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36010", + "issue:38175", + "issue:42491", + "issue:42617", + "issue:42915", + "issue:43366", + "issue:43531", + "issue:43550", + "issue:43596", + "issue:43632", + "issue:43638", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43720", + "issue:43824", + "issue:43828", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44155", + "issue:44530", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44683", + "issue:44805", + "issue:44863", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:45030", + "issue:45325", + "issue:45584", + "issue:45663", + "issue:45672", + "issue:45685", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:45902", + "issue:46402", + "issue:46489", + "issue:46515", + "issue:46858", + "issue:46899" + ], + "soft_pairs": [ + "issue:24643|issue:30064", + "issue:44530|issue:44863", + "issue:44863|issue:46402", + "issue:43366|issue:43828", + "issue:45325|issue:45902", + "issue:43596|issue:46515", + "issue:43366|issue:43531", + "issue:44155|issue:45584", + "issue:45663|issue:45823", + "issue:43638|issue:46515", + "issue:43366|issue:44589", + "issue:43950|issue:45834", + "issue:43824|issue:43957", + "issue:42915|issue:43957", + "issue:44910|issue:46858", + "issue:43716|issue:44560", + "issue:43645|issue:43646", + "issue:43550|issue:44910", + "issue:44560|issue:45685", + "issue:42617|issue:43366", + "issue:44683|issue:46515", + "issue:36010|issue:42915", + "issue:44560|issue:44918", + "issue:45812|issue:46489", + "issue:45823|issue:46515", + "issue:44661|issue:45823", + "issue:38175|issue:43994", + "issue:44560|issue:44805", + "issue:45672|issue:46899", + "issue:43632|issue:45834", + "issue:44877|issue:45030", + "issue:42491|issue:43720", + "issue:43638|issue:44661", + "issue:43596|issue:44661" + ], + "prepared_review_unit_hash": "7ac4fa701a040373b3ae09262b43861e9fd51a9a3ebd46abd884b10f6914e145", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 38070, + "estimated_input_tokens": 9518, + "estimated_eval_tokens": 19292 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous; most pairs are clearly unrelated. I only see three likely duplicate subgroups: custom model initialization (43645/43646), SigLIP2 bad outputs (38175/43994), and Gemma4 4-bit multimodal dtype-cast failures (45672/46899).", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue cleanly covers the whole cluster because it is largely mixed; the strongest duplicate subgroup is the SigLIP2 regression, where 43994 is the broadest representative.", + "canonical_pr_reason": null, + "best_issue_reason": "43994 is the best overall issue representative among the accepted duplicates because it states the SigLIP2 regression more generally than 38175 and is easier to map to the underlying bad-output bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "Different subsystems and failure modes: DeepSpeed training weight-shape error vs void segmentation map preprocessing." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "Both are model-specific, but they describe different bugs and code paths: Qwen3.5 PagedAttentionCache crash vs NemotronH checkpoint loading." + }, + { + "left": "issue:44863", + "right": "issue:46402", + "accept": false, + "reason": "Both are loading issues, but for different models and different failures; not the same underlying bug." + }, + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "Unrelated: GGUF gpt-oss support vs Phi-tiny-MoE autocast dtype mismatch." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "Both touch rotary/position logic, but the models and concrete bugs differ: Qwen2.5-VL temporal scaling vs Qwen3_5Moe uninitialized memory after meta materialization." + }, + { + "left": "issue:43596", + "right": "issue:46515", + "accept": false, + "reason": "One is a DeepSpeed ZeRO3 Bert index error; the other is a broad torch version incompatibility claim." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "Different topics: GGUF architecture support vs Qwen3-MoE sliding_window behavior." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "Different audio/text generation bugs with different symptoms and code paths." + }, + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Same model family, but different failures: FSDP2 shared_kv_states logic vs device_map='auto' device mismatch." + }, + { + "left": "issue:43638", + "right": "issue:46515", + "accept": false, + "reason": "Different issues: ZeRO3 Bert index error vs general torch compatibility." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "Different models and failures; no shared underlying bug." + }, + { + "left": "issue:43950", + "right": "issue:45834", + "accept": false, + "reason": "Non-persistent buffer loading regression vs long OCR input index error are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "Different problems: missing Qwen2_5_VL import vs meta-device loading breakage." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "Different code paths and symptoms; FP8 config failure is not the same as meta-loading breakage." + }, + { + "left": "issue:44910", + "right": "issue:46858", + "accept": false, + "reason": "Both involve cache/position arithmetic, but one is Qwen3.5 flash-attention illegal memory access and the other is a static-cache seq-length tensor issue on non-CUDA backends." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Different multimodal bugs in different models: Mistral-3 dtype mismatch vs Qwen3-vl video StopIteration." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "Same Transformers v5 custom-model initialization regression; the notebook mention is just a context-specific report of the same bug." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different attention-path failures in different models: SDPA/torch.compile on Bamba vs flash-attention position_id handling on Qwen3.5." + }, + { + "left": "issue:44560", + "right": "issue:45685", + "accept": false, + "reason": "Unrelated: Qwen3-vl video processing vs MPS histogram/int dtype limitation in moe code." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "Different topics entirely: 3d_parallel.py runtime failure vs GGUF gpt-oss support." + }, + { + "left": "issue:44683", + "right": "issue:46515", + "accept": false, + "reason": "Both mention torch versions, but the concrete problems are different and not mergeable as one fix." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "ImportError for GenerationMixin is unrelated to the Qwen3Moe FP8 failure." + }, + { + "left": "issue:44560", + "right": "issue:44918", + "accept": false, + "reason": "Both are Qwen3.x-related, but the failures differ: video StopIteration vs SFT trainer embedding unpacking." + }, + { + "left": "issue:45812", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different model families and appear to be separate bugs." + }, + { + "left": "issue:45823", + "right": "issue:46515", + "accept": false, + "reason": "Gemma4 device-mapping bug is unrelated to a generic torch 2.4.1 incompatibility report." + }, + { + "left": "issue:44661", + "right": "issue:45823", + "accept": false, + "reason": "Tokenizer-mapping validation failure is unrelated to Gemma4 device mismatch." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": true, + "reason": "Same SigLIP2 model regression causing bad outputs; 43994 is just the broader description of the same underlying bug." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Both are index-like failures, but they involve different models and tensor shapes; not the same bug." + }, + { + "left": "issue:45672", + "right": "issue:46899", + "accept": true, + "reason": "Same Gemma4 4-bit multimodal quantization problem: unsafe dtype casting in audio/vision embedders." + }, + { + "left": "issue:43632", + "right": "issue:45834", + "accept": false, + "reason": "Transformers v5 _is_hf_initialized regression is unrelated to Kosmos2.5 OCR indexing." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": false, + "reason": "Different config-validation issues on different models; not the same underlying change." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "Different loading/compatibility failures with different model types and code paths." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "DeepSpeed/Bert training index error is unrelated to tokenizer-mapping model registration failure." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Different failures: ZeRO3/Bert index out-of-bounds vs add-new-model-like tokenizer mapping logic." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet and stays conservative about the soft edges. The three accepted subgroups are plausible duplicates from the titles/body themes (43645/43646, 38175/43994, 45672/46899), and the rejected pairs are generally distinguished by model, subsystem, or failure mode without overclaiming. I don\u2019t see a mergeability or evidence-grounding problem here." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 7, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:38175", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42915", + "issue:43257", + "issue:43317", + "issue:43644", + "issue:43645", + "issue:43708", + "issue:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43994", + "issue:44075", + "issue:44164", + "issue:44291", + "issue:44387", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44661", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44910", + "issue:44960", + "issue:44977", + "issue:44991", + "issue:45276", + "issue:45292", + "issue:45313", + "issue:45335", + "issue:45357", + "issue:45362", + "issue:45406", + "issue:45488", + "issue:45507", + "issue:45588", + "issue:45812", + "issue:46050", + "issue:46459", + "issue:46468", + "issue:46512", + "issue:46846", + "issue:46897", + "issue:46901" + ], + "soft_pairs": [ + "issue:43317|issue:44756", + "issue:44661|issue:44805", + "issue:38175|issue:39692", + "issue:43644|issue:44534", + "issue:41720|issue:42915", + "issue:42491|issue:45276", + "issue:42915|issue:44756", + "issue:44960|issue:44977", + "issue:44451|issue:44488", + "issue:44863|issue:46846", + "issue:46459|issue:46512", + "issue:45313|issue:45357", + "issue:44488|issue:44991", + "issue:45507|issue:46468", + "issue:43644|issue:43645", + "issue:39692|issue:43994", + "issue:43257|issue:43854", + "issue:43856|issue:43873", + "issue:42491|issue:45292", + "issue:43856|issue:44387", + "issue:43854|issue:43931", + "issue:43708|issue:44075", + "issue:42491|issue:45335", + "issue:44960|issue:46050", + "issue:46897|issue:46901", + "issue:45488|issue:45812", + "issue:45362|issue:45406", + "issue:44164|issue:44291", + "issue:45276|issue:45335", + "issue:44910|issue:45588" + ], + "prepared_review_unit_hash": "827488692e064c688ee9df435d20fb530ccbcb5151ca293110e5792be83db9c5", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36658, + "estimated_input_tokens": 9165, + "estimated_eval_tokens": 18586 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose pile of Transformers regression reports that only superficially resemble one another. I found one clear duplicate pair around the v5 non-persistent-buffer corruption bug; the rest are different model- or feature-specific issues and should stay separate.", + "confidence": 0.87, + "canonical_issue_reason": "issue:43644 is the clearest, broadest report of the shared v5 non-persistent-buffer corruption regression, and it predates the near-identical follow-up report.", + "canonical_pr_reason": null, + "best_issue_reason": "Same reason as canonical: it is the most representative and broadly phrased issue in this set, with the strongest evidence of a real underlying regression rather than a model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "Both are offload/memory-related, but they describe different failures in different contexts (dequantized loading on GPU+CPU offload vs Strix Halo mmap OOM)." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Completely different bugs: tokenizer mapping failure vs tensor/mask shape mismatch." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "Both mention SigLIP2, but one is inference probability weirdness and the other is documentation/example errors with quantization; not the same bug." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Near-identical wording and same underlying v5 bug: non-persistent buffers being filled with junk." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Different Qwen3 failure modes: CUDA assert under auto device map vs FineGrainedFP8Config failure." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "Different models and different symptoms; no shared code-path evidence." + }, + { + "left": "issue:42915", + "right": "issue:44756", + "accept": false, + "reason": "Unrelated bugs: Qwen3Moe FP8 failure vs mmap/OOM workaround on Strix Halo." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "Different model families and different flash-attention failures; not the same change." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": false, + "reason": "Both are model-loading failures for different BERT repos, but there is not enough evidence they stem from the same root cause." + }, + { + "left": "issue:44863", + "right": "issue:46846", + "accept": false, + "reason": "NemotronH checkpoint loading and Qwen3.5 tensor-parallel planning are separate issues." + }, + { + "left": "issue:46459", + "right": "issue:46512", + "accept": false, + "reason": "Both involve dtype handling, but one is composite-checkpoint loading and the other is from_config initialization/config staleness." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "Different regressions: ZeRO-3 weight loading vs save_pretrained key handling for visual encoder weights." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-load failures, but for different models and likely different tokenizer configs; too little to merge." + }, + { + "left": "issue:45507", + "right": "issue:46468", + "accept": false, + "reason": "GraniteMoEHybrid invalid method call vs broken Nemotron example are unrelated." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "Both are v5 regressions, but one corrupts non-persistent buffers and the other breaks custom model initialization in notebooks." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but one is a doc/example mismatch and quantization failure while the other is bad inference output from AutoModel/pipeline." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Different problems: Qwen3 MoE weight conversion vs GLM-4.7-Flash loading in tests." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "Both mention memory/offload, but the causes and affected workflows are different." + }, + { + "left": "issue:42491", + "right": "issue:45292", + "accept": false, + "reason": "LoRA/Qwen3-MoE branch incompatibility is unrelated to resize_token_embeddings not updating output embeddings." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both are memory regressions, but one is Qwen3 MoE training memory usage and the other is int4 quantization reserved-memory growth." + }, + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Different model-loading failures with different symptoms and model families." + }, + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Trainer resume/max_steps bug and SGD args not used are unrelated." + }, + { + "left": "issue:42491", + "right": "issue:45335", + "accept": false, + "reason": "Different model families and different resize_token_embeddings behavior." + }, + { + "left": "issue:44960", + "right": "issue:46050", + "accept": false, + "reason": "GLM5 issue and sinq save/load corruption are unrelated." + }, + { + "left": "issue:46897", + "right": "issue:46901", + "accept": false, + "reason": "Both describe training-loss double-shift symptoms, but for different models; not enough evidence they share the same implementation bug." + }, + { + "left": "issue:45488", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different model families and different tokenizer behaviors." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Qwen3.5 chat crash and Gemma4Processor missing _tokenizer are different code paths." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "Different save/load bugs: extra_state handling vs unexpected _is_hf_initialized argument during init_empty_weights." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": false, + "reason": "Related resize_token_embeddings theme, but different models and different affected embeddings." + }, + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Flash-attention crashes in Qwen3.5 are not the same as the s_aux=None crash in sink-less models." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. It identifies the one clearly supported duplicate pair around the v5 non-persistent-buffer corruption bug, and the remaining pairwise rejections avoid over-claiming sameness across different model/feature contexts. No soft-edge verdict appears overly aggressive." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 8, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:36010", + "issue:39404", + "issue:39692", + "issue:41720", + "issue:42796", + "issue:42947", + "issue:43335", + "issue:43381", + "issue:43404", + "issue:43425", + "issue:43454", + "issue:43828", + "issue:43854", + "issue:43873", + "issue:43957", + "issue:44079", + "issue:44155", + "issue:44164", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44488", + "issue:44756", + "issue:44863", + "issue:44877", + "issue:44912", + "issue:44928", + "issue:44960", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45588", + "issue:45663", + "issue:45684", + "issue:45698", + "issue:45704", + "issue:45759", + "issue:45823", + "issue:45859", + "issue:46326", + "issue:46468", + "issue:46829", + "issue:46894", + "issue:46901" + ], + "soft_pairs": [ + "issue:44402|issue:44488", + "issue:39692|issue:44960", + "issue:41720|issue:44155", + "issue:45663|issue:45759", + "issue:42947|issue:45704", + "issue:43404|issue:43454", + "issue:43425|issue:43957", + "issue:44291|issue:44756", + "issue:43873|issue:44387", + "issue:44079|issue:46829", + "issue:36010|issue:39404", + "issue:44292|issue:44912", + "issue:45663|issue:45859", + "issue:44387|issue:44756", + "issue:44155|issue:45823", + "issue:42947|issue:43381", + "issue:43335|issue:45663", + "issue:43957|issue:44756", + "issue:44863|issue:46468", + "issue:45276|issue:45292", + "issue:44479|issue:44877", + "issue:43381|issue:44928", + "issue:44164|issue:44756", + "issue:45684|issue:45698", + "issue:45588|issue:45663", + "issue:42796|issue:46326", + "issue:43854|issue:45216", + "issue:45292|issue:45335", + "issue:46894|issue:46901", + "issue:44402|issue:44451", + "issue:43828|issue:43957" + ], + "prepared_review_unit_hash": "e5fb3c6b2ae5fc7c1fd987eb2c159aa3e3748d470beb68792dbb4067014dacca", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37011, + "estimated_input_tokens": 9253, + "estimated_eval_tokens": 18762 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are only thematically similar, not the same concrete bug. The only clear duplicate cluster is the `resize_token_embeddings` regression across model variants; everything else should stay separate.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 45292 is the cleanest general statement of the shared `resize_token_embeddings` propagation bug, and it sits between the more model-specific reports, making it the best representative issue.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45292 best captures the underlying API regression without overfitting to one model family, so it is the strongest global issue representative here.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "Both are model-loading failures for different BERT checkpoints, but the titles point to different concrete symptoms and no shared code path is evident." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "SigLIP2 doc example errors and a GLM5 loading issue are unrelated beyond both being model-specific bug reports." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "A Qwen3 device-mapping CUDA assert and an AudioFlamingo batched-inference token leak are different failures in different subsystems." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "One is an FSDP2 attention KeyError in Gemma4; the other is `from_config` not unwrapping `text_config` for Qwen multimodal configs." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing with LoRA and a T5 RMSNorm memory leak are unrelated bugs." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both mention `lm_head`, but they concern different model families and likely different weight-tying paths." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Torch version incompatibility and meta-device model loading failures are not the same bug." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "Unexpected `_is_hf_initialized` handling and Strix Halo mmap OOM are separate loading/runtime issues." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Quantization offloading behavior and int4 CUDA memory growth may overlap broadly, but they are not the same concrete defect." + }, + { + "left": "issue:44079", + "right": "issue:46829", + "accept": false, + "reason": "Both involve output objects, but one is a `ModelOutput` key assignment bug and the other is a wrong tensor stored in `hidden_states`." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "An import error for `GenerationMixin` is unrelated to Whisper `return_language` pipeline behavior." + }, + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Both are quantization-related, but one is a Qwen-3 NVFP4 runtime error and the other is MXFP4 fallback to bf16." + }, + { + "left": "issue:45663", + "right": "issue:45859", + "accept": false, + "reason": "Gemma4 FSDP2 attention state handling and Qwen3.5-MoE CPU-offload rotary embedding incompatibility are different code paths." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "An int4 quantization memory regression is not the same as disabling mmap to avoid OOM on Strix Halo." + }, + { + "left": "issue:44155", + "right": "issue:45823", + "accept": false, + "reason": "AudioFlamingo batching corruption and Gemma4 PLE device mismatch under `device_map=\"auto\"` are unrelated." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Both mention gradient checkpointing, but one is ineffective under PEFT LoRA while the other forbids eval-mode use." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "A SwitchTransformers sparse-layer config bug is unrelated to Gemma4 shared KV state rebuilding." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "Meta-device loading regressions and Strix Halo mmap OOM are different failure modes." + }, + { + "left": "issue:44863", + "right": "issue:46468", + "accept": false, + "reason": "A NemotronH checkpoint-loading bug and a broken official Nemotron example are related by model family only, not by the same concrete defect." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Both report the same `resize_token_embeddings` regression where updated embeddings are not propagated to output embeddings; 45276 is a model-specific manifestation of the same bug." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "A video-input regression for Qwen VL models is unrelated to a Granite speech config loading failure." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Eval-mode gradient checkpointing restrictions and RLHF NaNs from 3D position IDs are different problems." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "`extra_state` serialization and mmap/OOM behavior are separate save/load vs memory issues." + }, + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "One is a permissions issue during save, the other is loading the wrong custom module after save; different bugs, same feature area." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "A flash-attention crash on `s_aux=None` and a Gemma4 FSDP2 `shared_kv_states` KeyError are unrelated." + }, + { + "left": "issue:42796", + "right": "issue:46326", + "accept": false, + "reason": "Tokenizer BPE-marker handling and `continue_final_message` behavior are both tokenizer-related but not the same underlying bug." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "GLM-4.7-Flash test loading and Qwen3.5 checkpoint saving regression are different code paths." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Both describe the same `resize_token_embeddings` propagation bug, one for `output_embeddings` and the other for `decoder.embed_tokens`." + }, + { + "left": "issue:46894", + "right": "issue:46901", + "accept": false, + "reason": "They share a generic double-shift pattern, but these are different model trainers and not clearly the same concrete implementation bug." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Both are model-loading failures for BERT-like models, but the symptoms and likely root causes differ." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "A dtype mismatch under autocast and meta-device loading failures are unrelated runtime issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall. The summary stays conservative, and the only duplicated cluster called out is well supported by the packet: 45276/45292 and 45292/45335 have strong title-level overlap around `resize_token_embeddings` propagation. The other pair judgments avoid overclaiming shared root cause and are appropriately cautious." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 9, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:36032", + "issue:39404", + "issue:43299", + "issue:43425", + "issue:43452", + "issue:43596", + "issue:43611", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43716", + "issue:43828", + "issue:43856", + "issue:43901", + "issue:43906", + "issue:43931", + "issue:43994", + "issue:44079", + "issue:44164", + "issue:44265", + "issue:44479", + "issue:44509", + "issue:44530", + "issue:44589", + "issue:44671", + "issue:44683", + "issue:44756", + "issue:44805", + "issue:44898", + "issue:44912", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45479", + "issue:45701", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:46050", + "issue:46315", + "issue:46326", + "issue:46491", + "issue:46665", + "issue:46821", + "issue:46869", + "issue:46894", + "issue:46897", + "issue:46899" + ], + "soft_pairs": [ + "issue:43646|issue:43708", + "issue:43645|issue:44164", + "issue:44589|issue:44756", + "issue:46315|issue:46491", + "issue:45812|issue:46315", + "issue:43425|issue:43828", + "issue:39404|issue:45701", + "issue:44530|issue:45084", + "issue:39404|issue:46665", + "issue:43611|issue:43645", + "issue:43425|issue:46869", + "issue:44683|issue:45823", + "issue:43425|issue:45823", + "issue:43425|issue:43716", + "issue:44479|issue:44805", + "issue:43906|issue:44530", + "issue:44589|issue:45084", + "issue:43901|issue:44509", + "issue:44671|issue:44991", + "issue:43994|issue:44079", + "issue:36032|issue:43452", + "issue:44265|issue:44898", + "issue:39404|issue:44912", + "issue:46050|issue:46326", + "issue:44164|issue:45759", + "issue:43906|issue:46326", + "issue:44991|issue:45081", + "issue:45823|issue:46899", + "issue:43299|issue:43931", + "issue:45479|issue:46894", + "issue:43856|issue:44756", + "issue:45479|issue:46897", + "issue:44164|issue:46821", + "issue:43596|issue:44683" + ], + "prepared_review_unit_hash": "b411695f0961b250623691bb8a8d68c8392a8c1d1d5f99b4b7b6e3c9d5aecdae", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 38130, + "estimated_input_tokens": 9533, + "estimated_eval_tokens": 19322 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is not a tight duplicate cluster; it\u2019s a loose set of Transformers v5 regression reports spanning tokenizer behavior, model loading, multimodal forwarding, training-loss bugs, and quantization/device issues. I picked the broad tokenization-regression issue as the best representative, but most soft edges are not true duplicates.", + "confidence": 0.72, + "canonical_issue_reason": "issue:45701 is the broadest umbrella in the set: it describes version-dependent tokenization changes, which is the closest common theme across several neighboring reports. It is still only a rough representative, not an exact duplicate of the whole cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45701 is the best single anchor because it is broad, version-wide, and centered on a core regression class that recurs in this cluster. It is not a good substitute for the many unrelated loading, multimodal, and training bugs here, so this is a weak-best choice.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model init in notebooks and Trainer checkpoint-resume max_steps are different code paths and failure modes." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Notebook custom-model initialization is unrelated to save/from_pretrained handling of extra_state." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Different bugs: missing Float8 storage lookup vs mmap-related OOM on Strix Halo." + }, + { + "left": "issue:46315", + "right": "issue:46491", + "accept": false, + "reason": "One is tokenizer performance regression; the other is a CodeLlama round-trip spacing bug." + }, + { + "left": "issue:45812", + "right": "issue:46315", + "accept": false, + "reason": "Wrong token IDs for Granite models and slow convert_tokens_to_ids are distinct tokenizer bugs." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Torch 2.10 incompatibility is broader and different from an autocast dtype mismatch in Phi-tiny-MoE." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper pipeline return_language regression is unrelated to general tokenization changes." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache linear_attention failure and a template-node compile error are unrelated." + }, + { + "left": "issue:39404", + "right": "issue:46665", + "accept": false, + "reason": "Whisper pipeline regression and from_pretrained-inside-classes behavior are different issues." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "base_model_prefix loading regression is not the same as custom model initialization in notebooks." + }, + { + "left": "issue:43425", + "right": "issue:46869", + "accept": false, + "reason": "Torch compatibility problem and YOLO multi-GPU device-map failure are not the same code-path bug." + }, + { + "left": "issue:44683", + "right": "issue:45823", + "accept": false, + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 PLE device mismatch are unrelated." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "General torch incompatibility does not match the Mistral-3 image-preprocessor dtype mismatch." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "A video-input regression for Qwen models is not the same as a mask-shape indexing error." + }, + { + "left": "issue:43906", + "right": "issue:44530", + "accept": false, + "reason": "Reproduction of a prior issue and Qwen3.5 PagedAttentionCache crash are different problems." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage lookup failure and template compile error are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "Both are docs issues, but they cover different removed/deprecated pipeline APIs and should not be merged." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "CamemBERT masked-LM prediction regression is not the same as tokenizer loading failure for est-roberta." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "Wrong SigLIP2 outputs and ModelOutput key assignment semantics are different bugs." + }, + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "T5 tokenizer method conflict and gguf_file loading breakage are unrelated loading failures." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "torch.export + torch_compilable_check and Perceiver interpolation failure are different export/inference issues." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper return_language regression does not match MXFP4 quantization fallback behavior." + }, + { + "left": "issue:46050", + "right": "issue:46326", + "accept": false, + "reason": "Quantized model save/load breakage and mistral_common continue_final_message behavior are unrelated." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "extra_state save/load handling is different from unwrapping text_config for composite multimodal configs." + }, + { + "left": "issue:43906", + "right": "issue:46326", + "accept": false, + "reason": "A reproduction issue and a mistral_common tokenizer behavior bug are not the same underlying change." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer loading failure for est-roberta and Mistral regex patch crash are different tokenizer-path bugs." + }, + { + "left": "issue:45823", + "right": "issue:46899", + "accept": false, + "reason": "Device-map mismatch in Gemma4 PLE and 4-bit dtype-cast issues in Gemma4 multimodal embedders are different failures." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Both involve Qwen3-VL family loading, but one is a dev-version MOE breakage and the other is a weight-shape mismatch for a specific checkpoint." + }, + { + "left": "issue:45479", + "right": "issue:46894", + "accept": false, + "reason": "Single-label classification zero-loss and CohereASR double-shift training loss bugs affect different model families and code paths." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "Qwen3 MoE training memory inefficiency and Strix Halo mmap OOM are unrelated." + }, + { + "left": "issue:45479", + "right": "issue:46897", + "accept": false, + "reason": "Zero-loss sequence-classification bug and Florence2 double-shift training-loss bug are not the same issue." + }, + { + "left": "issue:44164", + "right": "issue:46821", + "accept": false, + "reason": "extra_state save/load handling and TimesFM window_size AttributeError are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "deepspeed zero3 BertModel indexing failure and compiled flex_attention breakage on torch>=2.9 are different bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 10, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:34567", + "issue:36032", + "issue:39404", + "issue:40444", + "issue:40990", + "issue:42175", + "issue:42796", + "issue:42915", + "issue:43262", + "issue:43317", + "issue:43377", + "issue:43404", + "issue:43421", + "issue:43452", + "issue:43482", + "issue:43541", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43643", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43883", + "issue:44206", + "issue:44220", + "issue:44451", + "issue:44462", + "issue:44488", + "issue:44530", + "issue:44661", + "issue:44671", + "issue:44683", + "issue:44737", + "issue:44811", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45507", + "issue:45584", + "issue:45859", + "issue:46178", + "issue:46285", + "issue:46326", + "issue:46402", + "issue:46489" + ], + "soft_pairs": [ + "issue:44488|issue:44987", + "issue:44488|issue:45081", + "issue:43611|issue:43644", + "issue:44462|issue:46489", + "issue:42796|issue:43404", + "issue:36032|issue:42915", + "issue:46285|issue:46402", + "issue:39404|issue:44671", + "issue:43716|issue:45084", + "issue:43643|issue:44530", + "issue:44737|issue:45859", + "issue:44661|issue:44683", + "issue:42915|issue:44530", + "issue:40990|issue:43421", + "issue:44451|issue:44987", + "issue:43644|issue:43646", + "issue:34567|issue:40444", + "issue:43541|issue:45084", + "issue:40444|issue:46326", + "issue:42175|issue:46178", + "issue:43582|issue:43632", + "issue:44206|issue:44220", + "issue:44811|issue:45584", + "issue:36032|issue:43645", + "issue:43452|issue:43482", + "issue:43883|issue:44488", + "issue:44987|issue:44991", + "issue:43317|issue:43482", + "issue:45084|issue:45507", + "issue:43262|issue:43377" + ], + "prepared_review_unit_hash": "13ba43acb0bfc7db64aec49cdace3097e58680a25347075dfa4c1e090c992770", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36732, + "estimated_input_tokens": 9183, + "estimated_eval_tokens": 18622 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is a loose mix of Transformers v5/v5.1 regressions and tokenizer/model-loading bugs, but the paired items describe different concrete failures and should not be merged as duplicates. No soft edge is strong enough to accept.", + "confidence": 0.81, + "canonical_issue_reason": "issue:44488 is the broadest representative of the tokenizer-loading regression reports and has the strongest discussion/inbound signal among the near-matches.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44488 is the best single issue to anchor this cluster because it is a clear user-facing loading regression with broad applicability, while the other items are more model-specific or unrelated.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "Both are tokenizer-loading failures, but they affect different models and describe different breakages; not the same underlying bug." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Both involve Mistral-related tokenizer loading, but one is a repository/model loading failure and the other is a specific regex-patching crash path." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "Both are Transformers v5 regressions, but one is about base_model_prefix loading and the other about non-persistent buffers; different code paths." + }, + { + "left": "issue:44462", + "right": "issue:46489", + "accept": false, + "reason": "Both touch tokenizer loading, but 44462 is about ignoring tokenizer.json while 46489 is an output regression for a specific tokenizer/model after a prior fix." + }, + { + "left": "issue:42796", + "right": "issue:43404", + "accept": false, + "reason": "A tokenizer BPE-marker bug and a tied-weights/model-head bug are unrelated failures." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "T5 tokenizer method-name conflict and Qwen3Moe FP8 config failure are unrelated." + }, + { + "left": "issue:46285", + "right": "issue:46402", + "accept": false, + "reason": "Both concern model structure/loading, but they affect different model families and different missing components." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline return_language and CamemBERT masked-LM predictions are unrelated model-specific bugs." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Image preprocessor dtype mismatch and template compilation failure are different subsystems and failure modes." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "Missing fields in AutoConfig with trust_remote_code is unrelated to a PagedAttentionCache linear_attention crash." + }, + { + "left": "issue:44737", + "right": "issue:45859", + "accept": false, + "reason": "Both are device/offload-related, but they concern different models and different tensor construction paths." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "Tokenizer-mapping name handling and compiled flex_attention on torch >= 2.9 are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and Qwen3.5 linear_attention cache crash are different bugs in different model paths." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "High perplexity on a model and runtime special-token post-processor updates are not the same bug." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Both are model-loading complaints, but they target different repositories and likely different tokenizer assets." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 custom-model regressions, but one is about junk buffers and the other about initialization itself; too distinct to merge." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "TrainerState token counter and multi-image IterableDataset finetuning failure are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Mixtral grouped_mm tracing and non-template-node compilation errors are different runtime/compiler issues." + }, + { + "left": "issue:40444", + "right": "issue:46326", + "accept": false, + "reason": "Multi-image finetuning and Whisper continue_final_message behavior are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:46178", + "accept": false, + "reason": "Both are backend-installation issues, but one is missing tensorflow in the torch extra and the other is an unsupported tensorflow_text backend entry." + }, + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Apple Silicon warmup TypeError and _is_hf_initialized flag breakage are unrelated." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": false, + "reason": "Both are audio-feature related, but one is about an unsupported center argument and the other about a different fbank extraction issue." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Both involve Whisper, but decode skip_special_tokens handling and empty-transcription generation after align_special_tokens are different bugs." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "Tokenizer method conflict and Jupyter custom-model initialization failure are not the same issue." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": false, + "reason": "Both are GGUF/loading regressions, but one is a generic from_pretrained gguf_file breakage and the other is a specific Qwen2.5-GGUF failure; not enough evidence of one underlying bug." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Molmo tied-weights attribute error and tokenizer loading failures are unrelated." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer load failures, but for different models and likely different tokenizer implementations." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "Dequantized model offload loading and Qwen2.5-GGUF loading are different loading paths." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "Template compilation failure and GraniteMoEHybrid invalid method calls are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Audio chat-template sample-rate defaults and MIMI batched-vs-single padding-mask behavior are different bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 11, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28530", + "issue:34567", + "issue:36032", + "issue:41720", + "issue:42915", + "issue:43089", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43531", + "issue:43540", + "issue:43632", + "issue:43645", + "issue:43646", + "issue:43742", + "issue:43746", + "issue:43792", + "issue:43819", + "issue:43883", + "issue:43994", + "issue:44008", + "issue:44062", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44464", + "issue:44466", + "issue:44479", + "issue:44534", + "issue:44589", + "issue:44610", + "issue:44849", + "issue:44877", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45987", + "issue:46315", + "issue:46326", + "issue:46661", + "issue:46821", + "issue:46869", + "issue:46899" + ], + "soft_pairs": [ + "issue:43746|issue:46326", + "issue:43646|issue:44534", + "issue:44849|issue:45325", + "issue:44403|issue:45081", + "issue:43645|issue:46661", + "issue:42915|issue:45084", + "issue:43377|issue:43819", + "issue:43883|issue:44991", + "issue:43792|issue:46869", + "issue:43404|issue:44466", + "issue:44464|issue:45987", + "issue:43994|issue:44610", + "issue:44987|issue:45081", + "issue:28530|issue:34567", + "issue:44008|issue:46899", + "issue:42915|issue:44291", + "issue:44451|issue:45081", + "issue:36032|issue:43742", + "issue:44062|issue:44987", + "issue:44265|issue:44610", + "issue:43089|issue:45987", + "issue:43322|issue:43540", + "issue:44479|issue:45325", + "issue:42915|issue:44589", + "issue:36032|issue:41720", + "issue:43632|issue:46821", + "issue:44991|issue:46315", + "issue:43531|issue:44877", + "issue:45216|issue:45357" + ], + "prepared_review_unit_hash": "45eff05b7d86399e88a8518a718273b14d5796df77ea02bfb0e64974bf6ec6f4", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36615, + "estimated_input_tokens": 9154, + "estimated_eval_tokens": 18564 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous; nearly all pairs are adjacent-sounding but clearly different bugs. Only the two Qwen3.5 `save_pretrained` regressions look like a plausible duplicate pair.", + "confidence": 0.77, + "canonical_issue_reason": "issue:45216 is the clearest representative of the only strong duplicate-like area here: Qwen3.5 `save_pretrained` checkpoint serialization regressions.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45216 is the best standalone anchor because it describes a concrete, user-visible regression and overlaps directly with the closest soft match in the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43746", + "right": "issue:46326", + "accept": false, + "reason": "Different features: GraniteSpeech local checkpoint loading vs mistral_common tokenizer message handling." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Custom model init breakage is unrelated to non-persistent buffer corruption." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Different models and failures: Qwen3.5 hidden states vs Qwen2.5-VL rope/position-id scaling." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Different code paths: generic loading noise vs Mistral tokenizer regex patch crash." + }, + { + "left": "issue:43645", + "right": "issue:46661", + "accept": false, + "reason": "Both touch model loading, but one is notebook/custom-model init and the other is remote-code `config_class` handling." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "FineGrainedFP8 Qwen3Moe failure is unrelated to template-node compile errors." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "MIMI padding-mask batching bug is unrelated to DAC latent reconstruction mismatch." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Missing `all_tied_weights_keys` is unrelated to tokenizer loading for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43792", + "right": "issue:46869", + "accept": false, + "reason": "Whisper model loading failure is unrelated to YOLO auto device-map multi-GPU failure." + }, + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Both involve lm_head/tied weights, but one is a missing tie and the other is device-dependent serialization inconsistency." + }, + { + "left": "issue:44464", + "right": "issue:45987", + "accept": false, + "reason": "Chunked generation/compiled forward mismatch is unrelated to `StaticCache.get_seq_length()` returning a tensor." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "SigLIP2 bad outputs and OmDet processor size mismatch are separate preprocessing/model-shape bugs." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer loading regression vs Mistral regex patch crash; different failure modes and targets." + }, + { + "left": "issue:28530", + "right": "issue:34567", + "accept": false, + "reason": "Early stopping metric selection is unrelated to `num_input_tokens_seen` not updating." + }, + { + "left": "issue:44008", + "right": "issue:46899", + "accept": false, + "reason": "Gemma3n name collision bug is unrelated to Gemma4 4-bit multimodal dtype casting." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "Both are init/quantization related at a high level, but the concrete failures are different and not mergeable." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "Different models and root causes: ScandiBERT loading vs Mistral tokenizer regex patching." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "Tokenizer special-token conflict is unrelated to MobileLLM key error on load." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "AddedToken argument duplication is not the same bug as the physical-intelligence loading failure." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "`torch.export`/`torch_compilable_check` failure is unrelated to processor image-size mismatch." + }, + { + "left": "issue:43089", + "right": "issue:45987", + "accept": false, + "reason": "Generation performance overhead is unrelated to cache length return type." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Llava Next loading crash and Qwen3OmniMoe video-input error are different model pipelines." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Both involve Qwen VL variants, but one is a broad video regression and the other is a still-image RoPE scaling bug." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "FP8 config failure is unrelated to missing storage-object resolution for float8 tensors." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Tokenizer method-name conflict is unrelated to Qwen3 device-map CUDA assert." + }, + { + "left": "issue:43632", + "right": "issue:46821", + "accept": false, + "reason": "`_is_hf_initialized` breakage is unrelated to TimesFM `window_size` handling." + }, + { + "left": "issue:44991", + "right": "issue:46315", + "accept": false, + "reason": "Tokenizer load failure and slow `convert_tokens_to_ids` are different tokenizer problems." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior is unrelated to strict GraniteSpeech config loading." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Both report the same Qwen3.5 `save_pretrained` regression: incorrect checkpoint serialization, with the visual-encoder-key problem being a narrower manifestation." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The evaluation is grounded in the packet and stays conservative. The only accepted soft pair, issue:45216 vs issue:45357, is plausibly the same Qwen3.5 `save_pretrained` checkpoint-serialization bug with a narrower visual-encoder-key manifestation. The rejected soft pairs are explained with model/feature-level differences rather than overstated evidence. No clear precision issue stands out." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 12, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:33666", + "issue:35707", + "issue:40444", + "issue:40990", + "issue:41553", + "issue:41720", + "issue:42175", + "issue:42796", + "issue:43296", + "issue:43317", + "issue:43381", + "issue:43421", + "issue:43425", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43716", + "issue:43746", + "issue:43792", + "issue:43856", + "issue:44079", + "issue:44220", + "issue:44291", + "issue:44403", + "issue:44462", + "issue:44509", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44683", + "issue:44756", + "issue:44811", + "issue:44991", + "issue:45081", + "issue:45325", + "issue:45538", + "issue:45542", + "issue:45663", + "issue:45812", + "issue:45923", + "issue:46491", + "issue:46710", + "issue:46899", + "issue:46909" + ], + "soft_pairs": [ + "issue:40444|issue:45325", + "issue:43611|issue:44534", + "issue:43425|issue:44403", + "issue:43296|issue:43531", + "issue:43716|issue:46899", + "issue:45812|issue:46491", + "issue:44220|issue:44683", + "issue:42796|issue:43746", + "issue:43381|issue:45923", + "issue:44403|issue:44991", + "issue:43632|issue:44756", + "issue:33666|issue:40444", + "issue:40990|issue:44811", + "issue:43792|issue:44589", + "issue:44291|issue:44589", + "issue:42796|issue:46710", + "issue:44079|issue:45663", + "issue:35707|issue:40444", + "issue:43482|issue:43525", + "issue:43421|issue:44568", + "issue:43482|issue:45538", + "issue:43632|issue:44534", + "issue:43531|issue:44291", + "issue:41553|issue:43582", + "issue:42175|issue:45542", + "issue:41720|issue:45081", + "issue:43317|issue:43856", + "issue:44462|issue:45812", + "issue:44079|issue:44534", + "issue:44509|issue:46909" + ], + "prepared_review_unit_hash": "9e56bb5744a78f3b817f16213efd36f1d43d274d8a7c59fe72fa9de33e7f06be", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36000, + "estimated_input_tokens": 9000, + "estimated_eval_tokens": 18256 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No true duplicates here: the soft matches are mostly superficial overlaps within Transformers/v5, tokenizer, and multimodal areas, but each item describes a different concrete bug or regression. No PRs are present.", + "confidence": 0.93, + "canonical_issue_reason": "issue:45812 is the closest cluster representative: it describes a broad, concrete AutoTokenizer regression affecting all Granite models and has follow-up references; however, the set is still heterogeneous, so it is only a representative, not a duplicate target.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45812 is the best single issue to anchor the cluster because it is specific, user-impacting, and broadly scoped compared with the others; the rest are separate bugs across different models/code paths.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Both are Qwen2.5-VL-related, but one is a finetuning failure with multiple images per prompt while the other is a rope-index scaling bug for still-image position_ids; different failure modes and fixes." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "Both mention Transformers v5, but one is about loading models with base_model_prefix and the other about non-persistent buffers being filled with junk; unrelated code paths." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "One is a Torch 2.10 compatibility break, the other is a logging/noise problem when loading a transformer; not the same bug." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "PaddleOCR-VL model loading in vLLM and Qwen3-MoE sliding_window behavior are separate model-specific issues." + }, + { + "left": "issue:43716", + "right": "issue:46899", + "accept": false, + "reason": "Both involve multimodal dtype handling, but one is a Mistral-3 preprocessor/model dtype mismatch and the other is unsafe 4-bit cast in Gemma4 embedders; distinct paths." + }, + { + "left": "issue:45812", + "right": "issue:46491", + "accept": false, + "reason": "Granite tokenizer wrong IDs and CodeLlama leading-space round-trip are both tokenizer regressions, but they affect different tokenizers and different semantics." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "_torch_extract_fbank_features() and compiled flex_attention on torch>=2.9 are different low-level failures with no shared code-path evidence." + }, + { + "left": "issue:42796", + "right": "issue:43746", + "accept": false, + "reason": "Ministral tokenizer BPE-marker handling and GraniteSpeech PEFT local checkpoint loading are unrelated tokenizer vs model-loading bugs." + }, + { + "left": "issue:43381", + "right": "issue:45923", + "accept": false, + "reason": "Eval-mode gradient checkpointing and a missing supports_gradient_checkpointing flag are related only at a high level; one is a runtime mode restriction, the other is a config flag omission." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Unnecessary loading noise and tokenizer loading failure for EMBEDDIA/est-roberta are different issues; one is cosmetic, the other blocks loading." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "_is_hf_initialized regression and mmap OOM on Strix Halo are unrelated v5/runtime environment problems." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Qwen2-VL multi-GPU training and Qwen2.5-VL multi-image finetuning are both training topics, but they concern different models and different bottlenecks." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "High perplexity on gpt-oss-20b and Whisper batch_decode skip_special_tokens behavior are unrelated symptoms and code paths." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Whisper-large-v2 loading failure and missing Float8 storage support are distinct backend/runtime errors." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "Unexpected _is_hf_initialized argument during init_empty_weights and missing Float8 storage are unrelated transformer-v5 breakages." + }, + { + "left": "issue:42796", + "right": "issue:46710", + "accept": false, + "reason": "Tokenizer BPE-marker handling for Ministral and accuracy regression for DeepSeek-R1-Distill-Llama are different model-specific regressions." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key assignment bug and Gemma4 FSDP2 shared_kv_states KeyError are unrelated internals." + }, + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Progressive generation with inputs_embeds/past_key_values and Qwen2.5-VL multi-image finetuning are different generation vs training bugs." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "Qwen2.5-GGUF loading in v5 and Llama4Config missing pad_token_id are separate config/loading issues." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Runtime post-processor updates in TokenizersBackend and mdeberta-v3 add_special_tokens behavior are different tokenizer-stack bugs." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "Qwen2.5-GGUF loading failure and CLIPTokenizer model_max_length are unrelated model-loading vs tokenizer-constant issues." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "_is_hf_initialized flag breakage and non-persistent buffers junk are both v5 regressions, but they affect different internals and fixes." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Qwen3-MoE sliding_window and unexpected _is_hf_initialized argument are not the same bug; one is model config behavior, the other is init-time argument compatibility." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Bad AutoTokenizer error for Voxtral and caching_allocator_warmup TypeError on Apple Silicon are completely different issues." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": false, + "reason": "TensorFlow missing from `.[torch]` install and tensorboard-only-without-TensorFlow backend error are related packaging issues but not the same underlying fix." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "Qwen3 auto device mapping cudaErrorAssert and Mistral regex patch crash are distinct model/device vs tokenizer-patch failures." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Dequantized model offload failure with device_map=auto and Qwen3 MoE training memory inefficiency are different load-time vs training-memory problems." + }, + { + "left": "issue:44462", + "right": "issue:45812", + "accept": false, + "reason": "Ignoring tokenizer.json from repo and wrong Granite token IDs both involve AutoTokenizer, but they are different repository-selection vs token-construction bugs." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "ModelOutput key handling and non-persistent buffer junk are unrelated internal state-management regressions." + }, + { + "left": "issue:44509", + "right": "issue:46909", + "accept": false, + "reason": "Removed pipeline-task docs and a broken DIA documentation example are both docs issues, but they concern different pages and content." + } + ] + }, + "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": 24, + "review_unit_index": 13, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42175", + "issue:42796", + "issue:42915", + "issue:43262", + "issue:43296", + "issue:43377", + "issue:43454", + "issue:43482", + "issue:43493", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43644", + "issue:43646", + "issue:43701", + "issue:43746", + "issue:43818", + "issue:43819", + "issue:43828", + "issue:43873", + "issue:43994", + "issue:44079", + "issue:44186", + "issue:44265", + "issue:44291", + "issue:44466", + "issue:44479", + "issue:44829", + "issue:44877", + "issue:45072", + "issue:45325", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45491", + "issue:45584", + "issue:45701", + "issue:45823", + "issue:45834", + "issue:46402", + "issue:46821" + ], + "soft_pairs": [ + "issue:43377|issue:45491", + "issue:43611|issue:43646", + "issue:43296|issue:46821", + "issue:43818|issue:46402", + "issue:44829|issue:45479", + "issue:43644|issue:44079", + "issue:43818|issue:45357", + "issue:42175|issue:43296", + "issue:43493|issue:43994", + "issue:44265|issue:45072", + "issue:40444|issue:45381", + "issue:45325|issue:45834", + "issue:45325|issue:45381", + "issue:43454|issue:44466", + "issue:42915|issue:43482", + "issue:43873|issue:45701", + "issue:44291|issue:44877", + "issue:43701|issue:45325", + "issue:42796|issue:45381", + "issue:43606|issue:44265", + "issue:40990|issue:45584", + "issue:43262|issue:43746", + "issue:44479|issue:45823", + "issue:35707|issue:41720", + "issue:34634|issue:35707", + "issue:43828|issue:46821", + "issue:39692|issue:40444", + "issue:44186|issue:44265", + "issue:43454|issue:43550", + "issue:43606|issue:43819" + ], + "prepared_review_unit_hash": "df6d9f9670d1d6349e0ddcb6ad9496ba555abcd5019504bcbff5617db1ec3a21", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36374, + "estimated_input_tokens": 9094, + "estimated_eval_tokens": 18444 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a loose collection of mostly unrelated model/runtime issues, not a tight duplicate cluster. All soft pairs look like superficial similarity only, so none should be merged as the same underlying bug.", + "confidence": 0.84, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "Different bugs: MIMI padding-mask batching mismatch vs Gemma3 sliding-window/all-padding NaNs." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is custom model initialization." + }, + { + "left": "issue:43296", + "right": "issue:46821", + "accept": false, + "reason": "Unrelated models and failure modes: PaddleOCR-VL load error vs TimesFM window_size AttributeError." + }, + { + "left": "issue:43818", + "right": "issue:46402", + "accept": false, + "reason": "Different code paths: Video-LLaVA architecture/weights issue vs Sam3 text encoder weights not loading." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Different training regressions: flash_attention_3 degenerate training vs num_labels=1 zero-loss classification behavior." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "Non-persistent buffer serialization bug is unrelated to ModelOutput key assignment." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "Video-LLaVA tower bug is unrelated to Qwen3.5 save_pretrained visual-key regression." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "Packaging/backend dependency issue is unrelated to a specific model load failure." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": false, + "reason": "Both touch SigLIP2, but one is HF-vs-JAX implementation mismatch and the other is bad AutoModel/pipeline outputs." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "torch.export export/check failure is a different bug from bfloat16 inference dtype mismatches." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Multiple-images finetuning failure is not the same as Qwen2.5-VL video position-id regression." + }, + { + "left": "issue:45325", + "right": "issue:45834", + "accept": false, + "reason": "Different models and defects: Qwen2.5-VL rope scaling vs Kosmos2.5 long-OCR index error." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": false, + "reason": "Same subsystem, but distinct bugs: still-image temporal scaling vs video vision_position_ids." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "Tied-weight serialization inconsistency is not the same as failing to tie lm_head weights." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "FP8 config failure and GGUF loading failure are unrelated issues." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Quantization/offloading behavior is unrelated to tokenization-version changes." + }, + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "Unexpected init argument during init_empty_weights is unrelated to strict config loading for granite_speech." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint key mismatch is unrelated to Qwen2.5-VL rope-index scaling." + }, + { + "left": "issue:42796", + "right": "issue:45381", + "accept": false, + "reason": "Tokenizer BPE-marker handling is unrelated to video vision_position_ids." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to torch_compilable_check export failures." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "High perplexity on WikiText-2 and Whisper empty-transcription generation failure are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio chat-template sample-rate defaulting is unrelated to GraniteSpeech PEFT checkpoint loading." + }, + { + "left": "issue:44479", + "right": "issue:45823", + "accept": false, + "reason": "Qwen video-input regression and Gemma4 device-mismatch are different concrete bugs." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "inputs_embeds/past_key_values generation bug is unrelated to Qwen3 auto device-map CUDA asserts." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "BarkProcessor voice_preset and progressive generation are unrelated features." + }, + { + "left": "issue:43828", + "right": "issue:46821", + "accept": false, + "reason": "Autocast dtype mismatch and TimesFM window_size AttributeError are different failures." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "SigLIP2 doc/example issues are unrelated to Qwen2.5-VL multi-image finetuning failures." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Tokenizer padding/truncation crash is unrelated to torch.export/export-check behavior." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "AyaVision weight-tying bug is unrelated to Bamba torch.compile SDPA failure." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to DAC.from_latents vs forward mismatch." + } + ] + }, + "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": 24, + "review_unit_index": 14, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:26413", + "issue:28530", + "issue:36331", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42796", + "issue:42915", + "issue:42947", + "issue:43065", + "issue:43257", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43450", + "issue:43454", + "issue:43526", + "issue:43572", + "issue:43720", + "issue:43749", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44186", + "issue:44387", + "issue:44466", + "issue:44488", + "issue:44534", + "issue:44568", + "issue:44610", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45357", + "issue:45701", + "issue:45704", + "issue:45759", + "issue:45823", + "issue:45920", + "issue:45923", + "issue:46050", + "issue:46209", + "issue:46305", + "issue:46315", + "issue:46402", + "issue:46821" + ], + "soft_pairs": [ + "issue:43881|issue:44488", + "issue:41720|issue:45823", + "issue:43377|issue:44610", + "issue:43257|issue:45923", + "issue:46305|issue:46315", + "issue:44568|issue:45920", + "issue:42796|issue:46209", + "issue:43454|issue:45072", + "issue:45823|issue:46821", + "issue:40990|issue:42947", + "issue:43377|issue:44898", + "issue:43454|issue:43720", + "issue:42617|issue:42915", + "issue:43749|issue:44534", + "issue:43317|issue:43873", + "issue:43572|issue:45759", + "issue:43322|issue:46050", + "issue:41720|issue:43322", + "issue:36331|issue:42617", + "issue:44387|issue:45704", + "issue:41720|issue:44977", + "issue:44466|issue:45357", + "issue:36331|issue:42915", + "issue:43526|issue:43883", + "issue:40990|issue:45701", + "issue:43450|issue:45701", + "issue:43881|issue:44991", + "issue:44877|issue:44991", + "issue:43065|issue:46402", + "issue:26413|issue:28530", + "issue:41720|issue:42617", + "issue:44186|issue:45072" + ], + "prepared_review_unit_hash": "2051762cde6eb64631f75b775900e1e6cc740f0786d2b76112847237f1540eea", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37300, + "estimated_input_tokens": 9325, + "estimated_eval_tokens": 18906 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is largely heterogeneous: it spans tokenizer regressions, model-loading failures, device_map/offload issues, quantization bugs, and a few trainer/model-configuration problems. There is no strong duplicate pair; the closest theme is broad loading/offload/quantization regressions.", + "confidence": 0.39, + "canonical_issue_reason": "Issue 43873 is the broadest and most central report in the cluster, covering the shared offloading/quantization theme better than the more model-specific incidents.", + "canonical_pr_reason": null, + "best_issue_reason": "43873 is the best anchor for the cluster because it is general enough to represent the loading/offload/quantization subgroup, even though the overall set is not a clean duplicate cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "Both are loading failures, but they involve different models and likely different failure modes; too little evidence they are the same bug." + }, + { + "left": "issue:41720", + "right": "issue:45823", + "accept": false, + "reason": "Both mention device_map/auto, but one is a cudaErrorAssert on Qwen3 and the other is a device mismatch in Gemma4; not the same concrete bug." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "Different subsystems entirely: MIMI encoder batching/padding vs OmDet-Turbo processor image sizing." + }, + { + "left": "issue:43257", + "right": "issue:45923", + "accept": false, + "reason": "Qwen3 MoE loading/accelerate-deepspeed conversion issue is unrelated to a missing gradient_checkpointing flag." + }, + { + "left": "issue:46305", + "right": "issue:46315", + "accept": false, + "reason": "One is a sliding-window mask off-by-one bug, the other is a tokenizer performance regression; unrelated." + }, + { + "left": "issue:44568", + "right": "issue:45920", + "accept": false, + "reason": "Both are tokenizer regressions, but one is specific BOS/EOS insertion for a single tokenizer while the other is a broader AutoTokenizer ID mapping problem across several models." + }, + { + "left": "issue:42796", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer BPE-marker handling and FP8 inference failure on B200 are unrelated code paths." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "Weight-tying/generation corruption and bfloat16 dtype mismatches are different bugs." + }, + { + "left": "issue:45823", + "right": "issue:46821", + "accept": false, + "reason": "Device-mapping forward mismatch and a TimesFM window_size AttributeError do not appear to share a code path." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "Perplexity regression on a model is unrelated to gradient checkpointing ineffective with PEFT LoRA." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Missing padding-mask support in MIMI encoder is not the same as Perceiver non-default-resolution failures." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "Tied-weight generation issues and BitNet packed-weight loading failures are distinct problems." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "Unable to run a parallel script and a FineGrainedFP8Config failure are unrelated." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "Both involve loading behavior, but one is FSDP CPU RAM efficient loading and the other is non-persistent buffer initialization; not the same bug." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "These are the closest pair, but one is a specific dequantized model offload failure and the other is a broader offloading/quantization report; insufficient evidence to merge as the same bug." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "Missing pad_token_idx after a v5 update and failing to unwrap text_config in composite Qwen configs are different configuration issues." + }, + { + "left": "issue:43322", + "right": "issue:46050", + "accept": false, + "reason": "Segmentation fault loading Llava Next and broken save_pretrained/from_pretrained after sinq quantization are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Different symptoms and models: cudaErrorAssert with auto device mapping vs a segmentation fault while loading Llava Next." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "CustomTrainer compute_loss signature change is unrelated to a 3d_parallel.py runtime failure." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "Int4 CUDA reserved memory/OOM and an apex.FusedRMSNorm memory leak are distinct quantization/memory issues." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Both involve Qwen-family models, but one is a device-map assert and the other is a flash-attention generation issue." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Both are serialization-related, but one is general tied-weight serialization and the other is incorrect visual encoder keys for Qwen3.5; not the same concrete bug." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "Unexpected compute_loss keyword argument and FineGrainedFP8Config failure are unrelated trainer/model-loading issues." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and Molmo tied-weights attribute errors do not share a code path." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "High perplexity on a benchmark and tokenizer behavior changing across versions are not the same bug." + }, + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Video processor batched-shape bug and tokenization changes are different subsystems." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "Both are loading/tokenizer failures, but they concern different models and different failure modes." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Strict config blocking granite_speech and tokenizer loading failure for est-roberta are unrelated." + }, + { + "left": "issue:43065", + "right": "issue:46402", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder and Sam3Model text encoder weights not loading are different model components and bugs." + }, + { + "left": "issue:26413", + "right": "issue:28530", + "accept": false, + "reason": "Resume-from-checkpoint epoch iterator failure and early-stopping metric handling are separate trainer issues." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Device-map assert on a model and failure to run 3d_parallel.py do not appear related." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "LayoutLMv2Tokenizer batching/padding crashes and bf16 dtype mismatches in inference 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": 24, + "review_unit_index": 15, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28530", + "issue:33666", + "issue:36010", + "issue:36331", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:43295", + "issue:43316", + "issue:43317", + "issue:43322", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43646", + "issue:43708", + "issue:43720", + "issue:43873", + "issue:44008", + "issue:44186", + "issue:44361", + "issue:44423", + "issue:44464", + "issue:44610", + "issue:44671", + "issue:44734", + "issue:44743", + "issue:44756", + "issue:44898", + "issue:44913", + "issue:45161", + "issue:45357", + "issue:45406", + "issue:45632", + "issue:45684", + "issue:45685", + "issue:45701", + "issue:45823", + "issue:46291", + "issue:46612", + "issue:46665", + "issue:46821", + "issue:46858", + "issue:46869" + ], + "soft_pairs": [ + "issue:43873|issue:44756", + "issue:46612|issue:46858", + "issue:43550|issue:43720", + "issue:43577|issue:43708", + "issue:43295|issue:45701", + "issue:43550|issue:43606", + "issue:44186|issue:44610", + "issue:45685|issue:46821", + "issue:43646|issue:46665", + "issue:43404|issue:45357", + "issue:44610|issue:44898", + "issue:36010|issue:42617", + "issue:44734|issue:46858", + "issue:42915|issue:43317", + "issue:44423|issue:45406", + "issue:44008|issue:45823", + "issue:46821|issue:46869", + "issue:45632|issue:45684", + "issue:41720|issue:45161", + "issue:44464|issue:46858", + "issue:28530|issue:33666", + "issue:44671|issue:44743", + "issue:43531|issue:43572", + "issue:36331|issue:41720", + "issue:42915|issue:46291", + "issue:44186|issue:44361", + "issue:43316|issue:44913", + "issue:42915|issue:43322", + "issue:43425|issue:46821" + ], + "prepared_review_unit_hash": "287755d9279118edf189fd7a9bda4fa03ef5db46b6a5eab67e84f6b27ac8fd27", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36448, + "estimated_input_tokens": 9112, + "estimated_eval_tokens": 18480 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These issues span many unrelated bugs across offloading, generation/cache handling, tokenization, config serialization, and multimodal/model-loading paths. None of the suggested soft pairs look like safe duplicates of the same underlying bug or change.", + "confidence": 0.86, + "canonical_issue_reason": "issue:43873 is the broadest open issue in the set and has the strongest cross-references, making it the best umbrella candidate for the offloading/quantization-related cluster even though the other items are not true duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best single issue to anchor this set because it is open, broadly framed, and most likely to attract related reports about offloading behavior.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Both mention memory/offloading, but one is about quantization offload behavior and the other is a Strix Halo mmap/OOM workaround; different root causes and fixes." + }, + { + "left": "issue:46612", + "right": "issue:46858", + "accept": false, + "reason": "Both involve cache logic, but one is a beam-search cache reorder bug across specific models while the other is a static-cache prefill arithmetic bug on non-CUDA backends under torch.compile." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "Different failures: one is torch.compile with SDPA on Bamba-9B-v2, the other is accelerate loading of packed BitNet weights." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "Unrelated problems: dtype propagation for Blip2 versus Trainer checkpoint-resume max_steps calculation." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "Both touch tokenization behavior, but one is a regression in processor/tokenizer access for custom multimodal code and the other is a broader version-dependent tokenization change; not the same concrete bug." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Both are model-specific runtime failures, but the affected models and failure mechanisms differ (torch.compile/SDPA vs CPU-offload device mismatch)." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Different components and bugs: LayoutLMv2 tokenizer crashes on NER/batched padding versus OmDet-Turbo processor output-size mismatch." + }, + { + "left": "issue:45685", + "right": "issue:46821", + "accept": false, + "reason": "No shared underlying bug: an MPS histogram kernel limitation for MoE versus a TimesFM window_size AttributeError." + }, + { + "left": "issue:43646", + "right": "issue:46665", + "accept": false, + "reason": "Both concern model loading, but one is Transformers 5.0 custom-model initialization breakage and the other is inconsistent from_pretrained behavior inside Python classes; not the same code path." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Different serialization/model-head issues: untied lm_head in Mistral3ForConditionalGeneration versus incorrect visual encoder keys saved by Qwen3.5." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Both are vision-related, but one is a processor resize/output mismatch and the other is a non-default-resolution interpolation failure in Perceiver classification." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "Completely different failures: an import regression for GenerationMixin versus inability to run a 3D parallel script." + }, + { + "left": "issue:44734", + "right": "issue:46858", + "accept": false, + "reason": "Both involve cache handling, but the first is an API serving KV-cache indexing bug while the second is a static-cache prefill tensor/device arithmetic bug under compilation." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "Both are model-loading/runtime failures around quantization/offload, but one is FineGrainedFP8 on Qwen3Moe and the other is dequantized model loading with device_map auto; different concrete problems." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Both are serve/multimodal processor crashes, but one is a string-to-device conversion failure in continuous batching and the other is a missing _tokenizer attribute on Gemma4Processor." + }, + { + "left": "issue:44008", + "right": "issue:45823", + "accept": false, + "reason": "Different Gemma-family bugs: a variable-name collision in forward versus a device mismatch during forward with device_map=auto." + }, + { + "left": "issue:46821", + "right": "issue:46869", + "accept": false, + "reason": "Different models and symptoms: a TimesFM window_size AttributeError versus YOLO multi-GPU device_map auto failure." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Unrelated filesystem/permissions issue versus save_pretrained read-only permission propagation." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Both mention distributed/device mapping style concerns, but one is a Qwen3 auto-device-map CUDA assert and the other is GPT-OSS MoE tensor-parallel support; not the same bug." + }, + { + "left": "issue:44464", + "right": "issue:46858", + "accept": false, + "reason": "Both mention compiled generation/cache behavior, but one reports inconsistent chunked generation outputs while the other is a specific cache length tensor arithmetic bug." + }, + { + "left": "issue:28530", + "right": "issue:33666", + "accept": false, + "reason": "No overlap in underlying bug: early stopping metric selection versus Qwen2-VL multi-GPU training." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "Both are model-specific correctness bugs, but one is masked-LM prediction regression in CamemBERT and the other is Qwen3.5 recurrent state reset with cache." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Different config/model issues: sliding_window behavior in Qwen3-MoE versus missing pad_token_idx in StableLmConfig." + }, + { + "left": "issue:36331", + "right": "issue:41720", + "accept": false, + "reason": "Trainer custom compute_loss signature change versus Qwen3 device_map CUDA assert are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:46291", + "accept": false, + "reason": "Different subsystems: Qwen3Moe FP8 loading failure versus import-time pipeline failure due to kernels package requirements." + }, + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Both tokenizer-related, but on different tokenizers and different failure modes; not a single duplicate bug." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Both are config consistency problems, but one is a Gemma3TextConfig API discrepancy and the other is GPTNeoX rotary_pct not persisting on reload." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Different model families and runtime errors: Qwen3Moe FP8 failure versus Llava Next segmentation fault on load." + }, + { + "left": "issue:43425", + "right": "issue:46821", + "accept": false, + "reason": "Torch version compatibility and a TimesFM attribute error are not the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 16, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28530", + "issue:29159", + "issue:30064", + "issue:33357", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:36032", + "issue:41720", + "issue:42490", + "issue:43329", + "issue:43388", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43606", + "issue:43644", + "issue:43653", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43873", + "issue:43881", + "issue:44062", + "issue:44077", + "issue:44387", + "issue:44451", + "issue:44492", + "issue:44509", + "issue:44534", + "issue:44589", + "issue:44625", + "issue:44898", + "issue:44991", + "issue:45059", + "issue:45072", + "issue:45357", + "issue:45406", + "issue:45464", + "issue:45701", + "issue:45704", + "issue:45759", + "issue:45859", + "issue:46402", + "issue:46665", + "issue:46821", + "issue:46858" + ], + "soft_pairs": [ + "issue:28530|issue:30064", + "issue:43526|issue:43881", + "issue:33357|issue:35707", + "issue:29159|issue:36032", + "issue:44387|issue:45701", + "issue:43873|issue:45704", + "issue:43540|issue:43782", + "issue:30064|issue:34634", + "issue:43550|issue:43653", + "issue:43606|issue:43720", + "issue:44534|issue:46665", + "issue:44492|issue:44509", + "issue:43531|issue:45759", + "issue:41720|issue:43746", + "issue:45406|issue:45464", + "issue:42490|issue:43329", + "issue:43388|issue:45059", + "issue:43606|issue:43653", + "issue:34567|issue:35707", + "issue:34634|issue:41720", + "issue:43531|issue:46821", + "issue:43540|issue:46858", + "issue:45759|issue:45859", + "issue:45357|issue:46402", + "issue:44451|issue:44991", + "issue:43531|issue:44589", + "issue:44077|issue:44625", + "issue:43644|issue:46665", + "issue:44062|issue:44451", + "issue:44898|issue:45072" + ], + "prepared_review_unit_hash": "2d9b333363498906fb14e087f90c844685041e5730a2b5aebc217e995d3bb48c", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36783, + "estimated_input_tokens": 9196, + "estimated_eval_tokens": 18648 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The proposed soft edges are all false positives. The items span unrelated bugs across tokenizers, multimodal processors, loading/offload, training state, docs, and model-specific regressions, with no clear duplicate cluster.", + "confidence": 0.95, + "canonical_issue_reason": "No canonical issue: the set is too heterogeneous and the soft pairs do not describe one shared underlying bug.", + "canonical_pr_reason": "No PRs are present in this cluster.", + "best_issue_reason": "No best issue stands out; none of the issues is representative of the whole set.", + "best_pr_reason": "No PRs are present in this cluster.", + "soft_edge_verdicts": [ + { + "left": "issue:28530", + "right": "issue:30064", + "accept": false, + "reason": "Early-stopping metric lookup and void segmentation-map processing are unrelated bugs." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and glm-4v loading failure affect different code paths." + }, + { + "left": "issue:33357", + "right": "issue:35707", + "accept": false, + "reason": "MacOS CLIP bus error and progressive generation with inputs_embeds/past_key_values are unrelated." + }, + { + "left": "issue:29159", + "right": "issue:36032", + "accept": false, + "reason": "Slow/fast tokenizer inconsistency and T5Tokenizer add_special_tokens conflict are different tokenizer issues." + }, + { + "left": "issue:44387", + "right": "issue:45701", + "accept": false, + "reason": "Int4 quantization OOM and tokenization changes are not the same problem." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization/offloading behavior and T5 apex RMSNorm memory leak are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Qwen3OmniMoe video processing error and Qwen3VL weight_only load error are separate failures." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Void segmentation maps and BarkProcessor voice_preset are different processor bugs." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "Bamba torch.compile SDPA failure and BigBirdTokenizer special-token registration are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "CPU offload device mismatch and BitNet packed-weight unpacking are different loading/offload issues." + }, + { + "left": "issue:44534", + "right": "issue:46665", + "accept": false, + "reason": "Non-persistent buffer junk and from_pretrained behavior inside Python classes are not the same bug." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache strategies and removed pipeline-task docs are unrelated documentation issues." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "Qwen3-MoE sliding_window and multimodal text_config unwrapping are distinct configuration bugs." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Qwen3 auto device mapping cudaErrorAssert and GraniteSpeech PEFT local checkpoint loading are unrelated." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer in transformers serve and Qwen3.5 streaming chat/completions failure are different code paths." + }, + { + "left": "issue:42490", + "right": "issue:43329", + "accept": false, + "reason": "Trailing-slash local model import warning and VLM video token counting are unrelated." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "gather_for_metrics label truncation and SAM3 text+box behavior are different bugs." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "CPU offload device mismatch and tokenizer special-token decoding are not the same issue." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState token counter not updating and progressive generation are unrelated." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping affect different models and code paths." + }, + { + "left": "issue:43531", + "right": "issue:46821", + "accept": false, + "reason": "Qwen3-MoE sliding_window and TimesFM window_size AttributeError are only superficially similar." + }, + { + "left": "issue:43540", + "right": "issue:46858", + "accept": false, + "reason": "Qwen3OmniMoe video input failure and cache seq-length device tensor arithmetic are unrelated." + }, + { + "left": "issue:45759", + "right": "issue:45859", + "accept": false, + "reason": "Composite Qwen config unwrapping and CPU-offload rotary embedding behavior are different bugs." + }, + { + "left": "issue:45357", + "right": "issue:46402", + "accept": false, + "reason": "Incorrect visual encoder keys on save and Sam3 text encoder weights not loaded are separate serialization/loading issues." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "ScandiBERT model loading and est-roberta tokenizer loading are not the same bug." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Qwen3-MoE sliding_window and missing Float8 storage are unrelated." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "patchtsmixer post_init policy and Qwen3.5 num_labels propagation are separate model/config issues." + }, + { + "left": "issue:43644", + "right": "issue:46665", + "accept": false, + "reason": "Non-persistent buffers filled with junk and from_pretrained inconsistency inside Python classes are not the same underlying bug." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "AddedToken special-argument conflict and ScandiBERT loading failure are unrelated tokenizer/model issues." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Perceiver interpolation failure and bfloat16 dtype mismatches are different model/runtime problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 17, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:39692", + "issue:41720", + "issue:42886", + "issue:43381", + "issue:43421", + "issue:43493", + "issue:43653", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43819", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44186", + "issue:44368", + "issue:44514", + "issue:44545", + "issue:44610", + "issue:44625", + "issue:44671", + "issue:44843", + "issue:44918", + "issue:44936", + "issue:44987", + "issue:44995", + "issue:45042", + "issue:45081", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45704", + "issue:45759", + "issue:45907", + "issue:45923", + "issue:45987", + "issue:46032", + "issue:46050", + "issue:46463", + "issue:46655", + "issue:46693", + "issue:46829", + "issue:46832", + "issue:46858" + ], + "soft_pairs": [ + "issue:44987|issue:45042", + "issue:44625|issue:45479", + "issue:43381|issue:43844", + "issue:44368|issue:45357", + "issue:39692|issue:44610", + "issue:41720|issue:45704", + "issue:43746|issue:46050", + "issue:43421|issue:43927", + "issue:46655|issue:46693", + "issue:43381|issue:45704", + "issue:45704|issue:45923", + "issue:42886|issue:44843", + "issue:44918|issue:44936", + "issue:45759|issue:45923", + "issue:43746|issue:43873", + "issue:44514|issue:44545", + "issue:43994|issue:44671", + "issue:43421|issue:43653", + "issue:39692|issue:43493", + "issue:46032|issue:46463", + "issue:43421|issue:44186", + "issue:43782|issue:45381", + "issue:45081|issue:45356", + "issue:30064|issue:36010", + "issue:44995|issue:46858", + "issue:45356|issue:45907", + "issue:39692|issue:43873", + "issue:43720|issue:43819", + "issue:43653|issue:43720", + "issue:45987|issue:46829", + "issue:46829|issue:46832" + ], + "prepared_review_unit_hash": "d5974906f2726f0612a983d5b847c69aabd01d6a4fbd5030e1061291ad5fc09f", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37548, + "estimated_input_tokens": 9387, + "estimated_eval_tokens": 19030 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly heterogeneous and does not form one duplicate cluster. The only clear duplicate pair is the Qwen2.5-VL batched `apply_chat_template` crash; the rest are separate issues that only share broad subsystem similarity.", + "confidence": 0.86, + "canonical_issue_reason": "Issue 44545 is the cleanest canonical for the only true duplicate pair (44514/44545) and has slightly more inbound/reference activity.", + "canonical_pr_reason": null, + "best_issue_reason": "44545 is the best representative issue in this set, but only for the one exact duplicate pair; it is not a good global representative of the full mixed cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "Different bugs: one is a model-loading failure for `physical-intelligence/fast`, the other is an image processor dependency regression requiring `torchvision`." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "Both are classification-related, but one is `num_labels` propagation in Qwen3.5 configs and the other is a generic single-label zero-loss bug across models." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Both involve training behavior, but eval-mode gradient checkpointing and ZeRO-3 gradient growth are different code paths and failure modes." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "Different Qwen issues: one is a LoRA warning about `tie_word_embeddings`, the other is incorrect visual encoder keys on `save_pretrained`." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "Different models and symptoms: SigLIP2 doc example mistakes vs OmDet-Turbo processor output size mismatch." + }, + { + "left": "issue:41720", + "right": "issue:45704", + "accept": false, + "reason": "Unrelated failures: Qwen3 auto device mapping CUDA assert vs a T5 Apex RMSNorm memory leak." + }, + { + "left": "issue:43746", + "right": "issue:46050", + "accept": false, + "reason": "Different loading/saving paths and different model families; PEFT local checkpoint loading is not the same as sinq quantized save/load breakage." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "Both are token-related, but one is runtime special-token post-processing and the other is config persistence of custom token IDs." + }, + { + "left": "issue:46655", + "right": "issue:46693", + "accept": false, + "reason": "Different performance/processing bugs: video processor padding calculation vs FA2 long-context generation regression." + }, + { + "left": "issue:43381", + "right": "issue:45704", + "accept": false, + "reason": "No shared bug: gradient checkpointing eval-mode failure is unrelated to the T5 fused RMSNorm leak." + }, + { + "left": "issue:45704", + "right": "issue:45923", + "accept": false, + "reason": "Different concerns entirely: Apex memory leak vs a missing `supports_gradient_checkpointing` flag on a remote-code model." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": false, + "reason": "Both mention offline tokenizer behavior, but one is generic cache loading and the other is a specific `_patch_mistral_regex` unconditional `model_info()` call; not enough to treat as the same bug." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "Distinct trainer issues: Qwen3.5 embedding unpacking with TRL SFT vs `trainer.evaluate()` failing after `trainer.train()`." + }, + { + "left": "issue:45759", + "right": "issue:45923", + "accept": false, + "reason": "Different fixes: unwrapping `text_config` for composite Qwen multimodal configs is unrelated to a checkpointing flag on Nemotron." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "Both touch loading/offload, but PEFT checkpoint loading and quantization offloading behavior are different bugs." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Same concrete bug: identical Qwen2.5-VL batched `apply_chat_template` crash when `padding=False`." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "Different model-specific correctness bugs: SigLIP2 bad outputs vs CamemBERT masked LM predictions in v5." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Both involve special tokens, but one is a tokenizer backend update bug and the other is BigBirdTokenizer mask-token registration/decoding behavior." + }, + { + "left": "issue:46032", + "right": "issue:46463", + "accept": false, + "reason": "Unrelated: Mamba2 cache/use_cache correctness is not the same as a T5 decoder-input error message issue." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Different tokenizer problems with different triggers; runtime special-token updates are not the same as LayoutLMv2 NER padding/truncation crashes." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Different Qwen/VL bugs: weight_only loading failure vs wrong vision_position_ids for video input." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Both mention `fix_mistral_regex`, but one is a crash in `_patch_mistral_regex` and the other is a Kimi-K2.5 tokenizer codec regression with a misleading warning; too different to merge." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Completely unrelated: image processor void segmentation maps vs GenerationMixin import failure." + }, + { + "left": "issue:44995", + "right": "issue:46858", + "accept": false, + "reason": "Both mention cache/indexing, but the stale indexer cache on second forward is not the same bug as `get_seq_length()` returning a device tensor in `_prefill`." + }, + { + "left": "issue:45356", + "right": "issue:45907", + "accept": false, + "reason": "Kimi tokenizer codec/fix_mistral_regex regression is unrelated to `list_repo_files` not forwarding `revision`." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "Different areas and symptoms: SigLIP2 documentation/model mismatch vs quantization offloading not working as expected." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "No shared underlying change: packed-weight loading failure in BitNet vs DAC latent reconstruction mismatch from missing STE." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "Different models and bugs: BigBird special-token decoding vs BitNet accelerate-loading unpacking." + }, + { + "left": "issue:45987", + "right": "issue:46829", + "accept": false, + "reason": "Different cache/output bugs: static cache sequence length type mismatch vs VideoPrism output field wiring." + }, + { + "left": "issue:46829", + "right": "issue:46832", + "accept": false, + "reason": "Different model heads and failure modes: VideoPrism hidden_states misassignment vs RT-DETR feature-level mismatch in forward." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The analysis is grounded in the packet and stays conservative on the soft edges. The only clearly duplicate pair in the provided titles is 44514/44545, and the choice of 44545 as canonical is supported by the slightly higher inbound/reference activity. The remaining pairwise rejections are appropriately cautious given the limited evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 18, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:30064", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:39692", + "issue:41720", + "issue:43295", + "issue:43335", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43596", + "issue:43611", + "issue:43618", + "issue:43643", + "issue:43653", + "issue:43697", + "issue:43756", + "issue:43819", + "issue:43927", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44220", + "issue:44361", + "issue:44442", + "issue:44448", + "issue:44488", + "issue:44661", + "issue:44671", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45183", + "issue:45381", + "issue:45479", + "issue:45834", + "issue:45987", + "issue:46294", + "issue:46665", + "issue:46693", + "issue:46858", + "issue:46932" + ], + "soft_pairs": [ + "issue:43295|issue:45834", + "issue:43550|issue:43927", + "issue:43550|issue:44361", + "issue:44062|issue:44488", + "issue:43540|issue:46294", + "issue:45381|issue:46693", + "issue:43531|issue:43643", + "issue:43596|issue:44220", + "issue:44008|issue:46858", + "issue:43611|issue:46665", + "issue:43653|issue:43819", + "issue:44186|issue:44898", + "issue:41720|issue:44991", + "issue:45987|issue:46932", + "issue:30064|issue:36331", + "issue:44220|issue:44661", + "issue:43643|issue:44877", + "issue:43618|issue:43697", + "issue:43335|issue:43756", + "issue:43550|issue:43819", + "issue:44843|issue:45081", + "issue:44448|issue:44671", + "issue:39692|issue:45834", + "issue:44186|issue:44442", + "issue:44821|issue:45084", + "issue:43540|issue:44008", + "issue:34634|issue:36010", + "issue:45183|issue:45479" + ], + "prepared_review_unit_hash": "5444e746e36ab4e8cd575036d9e382c930d584f3a43f77a3a996706a749ec46e", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35781, + "estimated_input_tokens": 8946, + "estimated_eval_tokens": 18148 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No duplicate cluster here: the items span unrelated regressions across tokenizers, configs, vision/audio models, generation caches, and docs. All soft edges are rejected.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue fits this heterogeneous set; the titles indicate many distinct code paths and failure modes with little or no overlap.", + "canonical_pr_reason": null, + "best_issue_reason": "If forced to choose one representative, issue 43611 is the broadest core regression around model loading, but it still does not meaningfully represent the rest of the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43295", + "right": "issue:45834", + "accept": false, + "reason": "Different subsystems and failures: processor/tokenizer regression vs static cache sequence-length shape bug." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "Unrelated models and code paths: torch.compile/SDPA issue vs config save/load token-ID regression." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "Different model-specific runtime errors with no shared bug evidence." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "Token construction error vs tokenizer-loading regression for a specific model; not the same bug." + }, + { + "left": "issue:43540", + "right": "issue:46294", + "accept": false, + "reason": "Both involve video inputs, but the model families and concrete failures differ too much to be one fix." + }, + { + "left": "issue:45381", + "right": "issue:46693", + "accept": false, + "reason": "Vision position IDs bug vs performance regression in FA2 long-context generation; unrelated." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "MoE sliding-window bug vs AutoConfig trust_remote_code field-loss issue; different paths." + }, + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "Deepspeed/BertModel init error vs audio fbank extraction issue; no overlap." + }, + { + "left": "issue:44008", + "right": "issue:46858", + "accept": false, + "reason": "Gemma3n variable collision vs cache.get_seq_length tensor-return bug; distinct code paths." + }, + { + "left": "issue:43611", + "right": "issue:46665", + "accept": false, + "reason": "Model-loading regression in v5 vs from_pretrained behavior inside Python classes; not the same defect." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "BigBird tokenizer special-token registration bug vs DAC latents/STE mismatch; unrelated." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "LayoutLMv2 tokenizer crashes vs Perceiver image classification failure; different models and causes." + }, + { + "left": "issue:41720", + "right": "issue:44991", + "accept": false, + "reason": "CUDA/device-mapping crash vs tokenizer loading failure for a different model; no shared bug." + }, + { + "left": "issue:45987", + "right": "issue:46932", + "accept": false, + "reason": "Static cache sequence-length contract bug vs Gemma4 quantized embedder dtype casting; unrelated." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Image processor segmentation-map bug vs CustomTrainer compute_loss signature mismatch; different areas." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Audio feature extraction bug vs tokenizer mapping-name model-addition bug; no common code path." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "Both are config-loading related, but one is missing fields with trust_remote_code and the other is strict config rejection; not the same issue." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "CLIPOutput attentions regression vs RTDetrV2 output mismatch; different models and symptoms." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug vs Smollm3 RoPE-layer count bug; unrelated." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "torch.compile/SDPA failure vs DAC forward/from_latents mismatch; no shared concrete defect." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Both touch _patch_mistral_regex, but one is offline model_info access and the other is a backend_tokenizer attribute crash; different bugs." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Different model regressions with separate outputs/predictions issues; not mergeable as one fix." + }, + { + "left": "issue:39692", + "right": "issue:45834", + "accept": false, + "reason": "Docs example errors vs OCR long-input index error; unrelated." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "Tokenizer crash vs AutoTokenizer loading failure for a different tokenizer family." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "ImageProcessor-from-URL loading issue vs compilation/template-node error; unrelated." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Two different multimodal model bugs with different failure modes and code paths." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "Voice preset bug in BarkProcessor vs GenerationMixin import error; unrelated." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "Input validation error messaging vs zero-loss classification bug; different model logic." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 19, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28218", + "issue:29159", + "issue:34634", + "issue:36331", + "issue:43322", + "issue:43388", + "issue:43421", + "issue:43540", + "issue:43550", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43644", + "issue:43645", + "issue:43716", + "issue:43720", + "issue:43749", + "issue:43782", + "issue:43818", + "issue:43825", + "issue:43827", + "issue:43931", + "issue:43994", + "issue:44008", + "issue:44291", + "issue:44479", + "issue:44561", + "issue:44610", + "issue:44671", + "issue:44805", + "issue:44811", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45020", + "issue:45072", + "issue:45381", + "issue:45584", + "issue:45685", + "issue:45701", + "issue:45834", + "issue:46032", + "issue:46291", + "issue:46294", + "issue:46463", + "issue:46512", + "issue:46665", + "issue:46846" + ], + "soft_pairs": [ + "issue:43644|issue:43749", + "issue:45584|issue:45834", + "issue:43322|issue:45685", + "issue:43606|issue:45072", + "issue:43388|issue:46032", + "issue:43825|issue:43994", + "issue:44479|issue:45834", + "issue:43550|issue:45072", + "issue:43782|issue:44977", + "issue:43421|issue:44811", + "issue:44561|issue:45020", + "issue:43577|issue:43818", + "issue:44991|issue:46291", + "issue:46294|issue:46463", + "issue:43582|issue:46512", + "issue:43716|issue:44008", + "issue:43931|issue:46846", + "issue:28218|issue:29159", + "issue:44610|issue:45072", + "issue:43540|issue:45381", + "issue:44805|issue:44936", + "issue:43749|issue:44291", + "issue:43720|issue:45072", + "issue:43825|issue:43827", + "issue:44671|issue:45701", + "issue:43645|issue:46665", + "issue:34634|issue:36331", + "issue:43577|issue:43582" + ], + "prepared_review_unit_hash": "78c65142b2062abfd77938b6cdabeca50f51400daa3bf74fed04fb4a98ba7654", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36324, + "estimated_input_tokens": 9081, + "estimated_eval_tokens": 18418 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: most items are unrelated regressions across tokenizers, video/image models, loading, training, and dtype/device issues. None of the soft pairs looks like the same concrete bug/change, so I would not merge any of them as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the broadest representative of the version/regression theme, but it is still too generic to serve as a true duplicate canonical for this cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Different loading/initialization failures: non-persistent buffer junk vs FSDP CPU RAM efficient loading." + }, + { + "left": "issue:45584", + "right": "issue:45834", + "accept": false, + "reason": "Different models and bugs: Whisper empty transcription after align_special_tokens vs Kosmos2.5 long OCR index error." + }, + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Unrelated paths: Llava Next load segfault vs MPS histogram Int error in moe." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "Both mention mismatches, but they are different concrete problems in different models/code paths." + }, + { + "left": "issue:43388", + "right": "issue:46032", + "accept": false, + "reason": "Different bugs: last-batch label dropping vs Mamba2Mixer use_cache correctness." + }, + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "One is an error-message/docs issue; the other is a model output quality regression." + }, + { + "left": "issue:44479", + "right": "issue:45834", + "accept": false, + "reason": "Different models and failure modes: Qwen video regression vs Kosmos2.5 OCR long-input error." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Different concrete issues: torch.compile/SDPA failure vs bfloat16 dtype mismatches." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Different failure surfaces: weight_only loading error vs flash-attention generation bug." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Both are tokenizer-related, but one is post-processor runtime updates and the other is batch_decode skip_special_tokens behavior." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "Related theme (remote_code breakage) but not the same concrete underlying bug or code path." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "Different model-specific issues: Blip2 dtype state vs Video-LLaVA temporal-attention/weight-sharing bug." + }, + { + "left": "issue:44991", + "right": "issue:46291", + "accept": false, + "reason": "Different breakages: tokenizer loading for a specific repo vs pipelines import failing due to kernels dependency requirements." + }, + { + "left": "issue:46294", + "right": "issue:46463", + "accept": false, + "reason": "Different domains: Qwen3.5 video/chat-template mismatch vs T5 decoder input validation message." + }, + { + "left": "issue:43582", + "right": "issue:46512", + "accept": false, + "reason": "Different dtype/config issues on unrelated code paths and platforms." + }, + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Different causes: image preprocessor/model dtype mismatch vs Gemma3n variable-name collision AttributeError." + }, + { + "left": "issue:43931", + "right": "issue:46846", + "accept": false, + "reason": "Both are Qwen/VL shape problems, but one is checkpoint weight-shape mismatch and the other is TP plan coverage/shape error." + }, + { + "left": "issue:28218", + "right": "issue:29159", + "accept": false, + "reason": "The second issue is much broader; the titles do not clearly identify the same tokenizer-space bug." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Different model/input-size issue vs dtype mismatch issue." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Both concern video inputs, but they affect different Qwen model variants and different failure modes." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "Unrelated: mask/index shape mismatch vs trainer lifecycle failure after train()." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "Both involve loading, but the error mechanics are different: FSDP RAM-efficient loading vs unexpected _is_hf_initialized argument." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Different failures: packed-weight unpacking during accelerate loading vs bfloat16 dtype mismatches." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "Same general pipeline-removal theme, but one is an error message and the other is stale docs." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "Different problems: CamemBERT masked LM regression vs tokenizer version-dependent tokenization changes." + }, + { + "left": "issue:43645", + "right": "issue:46665", + "accept": false, + "reason": "Both mention model initialization, but they are distinct context-specific behaviors with no shared concrete bug." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "Completely different subsystems: BarkProcessor voice_preset vs Trainer compute_loss signature change." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "Unrelated bugs: Blip2 dtype state vs Apple Silicon caching_allocator_warmup TypeError." + } + ] + }, + "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": 24, + "review_unit_index": 20, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:36010", + "issue:36331", + "issue:42175", + "issue:42490", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43452", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43582", + "issue:43638", + "issue:43645", + "issue:43653", + "issue:43697", + "issue:43742", + "issue:43782", + "issue:43819", + "issue:43906", + "issue:43927", + "issue:44060", + "issue:44368", + "issue:44554", + "issue:44589", + "issue:44857", + "issue:44912", + "issue:44936", + "issue:44991", + "issue:45003", + "issue:45020", + "issue:45042", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45372", + "issue:45406", + "issue:45538", + "issue:45600", + "issue:46175", + "issue:46209", + "issue:46285", + "issue:46291", + "issue:46326", + "issue:46697", + "issue:46728", + "issue:46932" + ], + "soft_pairs": [ + "issue:42175|issue:45081", + "issue:46175|issue:46932", + "issue:45003|issue:46285", + "issue:46697|issue:46728", + "issue:43653|issue:44912", + "issue:43782|issue:45325", + "issue:43262|issue:43819", + "issue:45372|issue:46326", + "issue:45081|issue:45406", + "issue:43582|issue:44857", + "issue:44589|issue:45042", + "issue:43742|issue:45020", + "issue:43550|issue:44912", + "issue:43452|issue:44991", + "issue:43329|issue:45600", + "issue:43257|issue:43329", + "issue:44060|issue:44368", + "issue:43526|issue:43697", + "issue:36010|issue:36331", + "issue:43906|issue:46209", + "issue:45084|issue:45538", + "issue:36010|issue:43645", + "issue:43526|issue:44554", + "issue:43819|issue:43927", + "issue:43531|issue:45042", + "issue:43638|issue:44936", + "issue:42175|issue:46291", + "issue:42490|issue:43257" + ], + "prepared_review_unit_hash": "6642d6a9f31d0e0e5e06642f456175eccadc0ca46e13571d7b4b474cd31458c8", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 36030, + "estimated_input_tokens": 9008, + "estimated_eval_tokens": 18272 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated closed issues from Transformers. None of the soft pairs looks like the same underlying bug or change; the closest thing to an umbrella topic is loading/regression behavior, but it is still too broad for duplicate merging.", + "confidence": 0.93, + "canonical_issue_reason": "issue:45020 is the broadest umbrella-style report in the set and the closest thing to a hub for loading/regression breakages, though it is not a true duplicate canonical.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the best overall representative of the cluster because it has the widest scope and strongest \u201crecent versions break models\u201d framing, but the cluster is too heterogeneous to treat it as a duplicate anchor.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Different bugs: backend dependency packaging vs Mistral regex/tokenizer attribute handling." + }, + { + "left": "issue:46175", + "right": "issue:46932", + "accept": false, + "reason": "Both are Gemma4-related, but one is a huge one-hot position embedding issue and the other is quantized embedder dtype casting; not the same fix." + }, + { + "left": "issue:45003", + "right": "issue:46285", + "accept": false, + "reason": "Unrelated: unsafe sys.modules access vs CLIP text model structural inconsistency." + }, + { + "left": "issue:46697", + "right": "issue:46728", + "accept": false, + "reason": "Both concern image sizing/padding, but they affect different models and different code paths." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "Tokenizer special-token registration vs quantization fallback behavior are unrelated." + }, + { + "left": "issue:43782", + "right": "issue:45325", + "accept": false, + "reason": "Both are VL model issues, but the failure modes are different and not mergeable as one concrete bug." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting is unrelated to DAC latent reconstruction/STE mismatch." + }, + { + "left": "issue:45372", + "right": "issue:46326", + "accept": false, + "reason": "Both mention mistral_common, but one is an import failure for Gemma4 processor loading and the other is continue_final_message behavior." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Different attribute errors in different components: tokenizer backend patching vs Gemma4Processor state." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "Apple Silicon warmup TypeError and CUDA AMP loss crash are unrelated runtime failures." + }, + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Storage-object lookup failure is unrelated to PIL backend torchvision dependency handling." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "A specific model load KeyError is not the same concrete bug as the broader remote_code regression report." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "torch.compile SDPA failure and MXFP4 quantization fallback are different code paths." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "Both are loading failures, but for different tokenizer/model formats and different root causes." + }, + { + "left": "issue:43329", + "right": "issue:45600", + "accept": false, + "reason": "Multimodal token counting bug is unrelated to auto_mappings referencing removed configs." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is unrelated to a multimodal token-counting undefined-name bug." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Both involve tied-weights warnings in Qwen variants, but they describe different warnings and different root causes." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "Beit label reduction bug and RTDetr output divergence are unrelated." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "ImportError for GenerationMixin and compute_loss signature mismatch are unrelated trainer/import issues." + }, + { + "left": "issue:43906", + "right": "issue:46209", + "accept": false, + "reason": "An isolated reproduction of an old issue is unrelated to FP8 inference on B200." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "Template compilation error and CLIPTokenizer max_length bug are unrelated." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "Import path breakage and Jupyter custom-model initialization are different bugs." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "Label reduction in BeitImageProcessorFast is unrelated to MPS attention correctness." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "DAC forward/latent mismatch and DiaConfig token ID loss are different model/config problems." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to PIL backend torchvision requirements." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Both involve training/eval, but the failures are different and not the same underlying bug." + }, + { + "left": "issue:42175", + "right": "issue:46291", + "accept": false, + "reason": "TensorFlow extra packaging and kernels version/revision validation are unrelated dependency issues." + }, + { + "left": "issue:42490", + "right": "issue:43257", + "accept": false, + "reason": "Trailing-slash local model path warning is unrelated to Qwen3 MoE weight conversion." + } + ] + }, + "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": 24, + "review_unit_index": 21, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28282", + "issue:29159", + "issue:29870", + "issue:34567", + "issue:34689", + "issue:38175", + "issue:39692", + "issue:40990", + "issue:42915", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43450", + "issue:43540", + "issue:43606", + "issue:43643", + "issue:43650", + "issue:43696", + "issue:43720", + "issue:43742", + "issue:43746", + "issue:43749", + "issue:43854", + "issue:43906", + "issue:43927", + "issue:43931", + "issue:44291", + "issue:44387", + "issue:44479", + "issue:44568", + "issue:44610", + "issue:44617", + "issue:44756", + "issue:44811", + "issue:44821", + "issue:44912", + "issue:44977", + "issue:44987", + "issue:45042", + "issue:45356", + "issue:45459", + "issue:45538", + "issue:45759", + "issue:45803", + "issue:45859", + "issue:46421", + "issue:46424", + "issue:46701" + ], + "soft_pairs": [ + "issue:43746|issue:44811", + "issue:43720|issue:44610", + "issue:43854|issue:44977", + "issue:39692|issue:44912", + "issue:43643|issue:45759", + "issue:38175|issue:44610", + "issue:43257|issue:43696", + "issue:43262|issue:44811", + "issue:43257|issue:43931", + "issue:44387|issue:44617", + "issue:43329|issue:43450", + "issue:43329|issue:44479", + "issue:46421|issue:46424", + "issue:45459|issue:45803", + "issue:42915|issue:43742", + "issue:34567|issue:40990", + "issue:43749|issue:44756", + "issue:44291|issue:45042", + "issue:29159|issue:34567", + "issue:43650|issue:43906", + "issue:43606|issue:45859", + "issue:42915|issue:44821", + "issue:43927|issue:44912", + "issue:43540|issue:46701", + "issue:28282|issue:29870", + "issue:44568|issue:45356", + "issue:44987|issue:45538", + "issue:28282|issue:34689" + ], + "prepared_review_unit_hash": "eb9a1b81900132798ba1ff1e882d18c8ab88c7e3eb0c8796558969108eb30183", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36222, + "estimated_input_tokens": 9056, + "estimated_eval_tokens": 18368 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: it spans unrelated issues in loading, tokenization, multimodal/video processing, cache behavior, and memory/offload problems. None of the soft-edge pairs look like the same underlying bug or change, so I would not consolidate them.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue is suitable: the items are from different subsystems and failure modes, with no clear shared underlying defect across the set.", + "canonical_pr_reason": null, + "best_issue_reason": "No global best issue: there isn\u2019t one representative bug that covers the rest of the cluster without losing important distinctions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "Different code paths: local PEFT adapter loading versus Whisper batch decoding behavior." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "Unrelated problems: BitNet packed-weight loading on accelerate versus OmDet-Turbo processor image-size mismatch." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "Both are model-specific, but one is a unit-test loading failure and the other is a flash-attention generation regression; not the same bug." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "Documentation/example errors are distinct from MXFP4 quantization fallback/loading behavior." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "Both concern config loading, but they affect different APIs and different missing-field/unwrap behaviors." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Different symptoms and subsystems: zero probabilities in SigLIP2 versus processor output size mismatch in OmDet-Turbo." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "Qwen3 MoE weight-conversion failure is unrelated to a generic CUDA OOM in GPT-oss." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting is unrelated to Whisper's skip_special_tokens handling." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Both are Qwen3-related, but one is MoE weight conversion and the other is a Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both mention OOM, but the causes are different: int4 quantization memory regression versus a model-specific video OOM." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Video-related, but one is an undefined helper in multimodal token counting and the other is a batched video processor shape bug." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "One is a broken internal video token-count path; the other is a broader v5.3.0 video-input regression across models." + }, + { + "left": "issue:46421", + "right": "issue:46424", + "accept": false, + "reason": "Both are cache-related, but one is lazy-init recompilation and the other is max_cache_length being honored only for quantized cache." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Both involve exception handling, but they are different swallows: protobuf decode errors versus a bare except in FuyuBatchFeature." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Different failures: FineGrainedFP8Config support in Qwen3Moe versus a key error while loading MobileLLM." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "Trainer token-count bookkeeping is unrelated to GPT-oss perplexity quality issues." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "Both touch loading/memory, but one is FSDP CPU RAM efficient loading and the other is mmap behavior on Strix Halo." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "Different APIs and failure modes: init_empty_weights argument handling versus PIL image processors requiring torchvision." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency is unrelated to TrainerState token counting." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "No substantive overlap; one is a nonsensical placeholder title and the other is a reproduction issue." + }, + { + "left": "issue:43606", + "right": "issue:45859", + "accept": false, + "reason": "Both involve CPU offload, but the failing code paths and models are different." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3Moe FP8 config failure is unrelated to AutoImageProcessor loading from URL." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "Config save/load token IDs bug is unrelated to MXFP4 quantization fallback." + }, + { + "left": "issue:43540", + "right": "issue:46701", + "accept": false, + "reason": "Both are multimodal, but they concern different models, modalities, and failure points." + }, + { + "left": "issue:28282", + "right": "issue:29870", + "accept": false, + "reason": "ImportError for missing PyTorch is unrelated to Bart generate's decoder_attention_mask assignment bug." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different tokenizers and different mechanisms." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Transformers loading failure for a model is unrelated to CLIPTokenizer's model_max_length setting." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "Missing PyTorch ImportError and Llama 3.2 vision model-loading regression are distinct problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 22, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:24643", + "issue:27957", + "issue:28282", + "issue:29127", + "issue:29159", + "issue:33357", + "issue:34567", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43404", + "issue:43425", + "issue:43452", + "issue:43454", + "issue:43525", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43824", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44560", + "issue:44568", + "issue:44849", + "issue:44857", + "issue:44936", + "issue:44987", + "issue:44991", + "issue:45127", + "issue:45137", + "issue:45357", + "issue:45381", + "issue:45405", + "issue:45488", + "issue:45538", + "issue:45701", + "issue:45907", + "issue:46291", + "issue:46512", + "issue:46693", + "issue:46821", + "issue:46832", + "issue:46869" + ], + "soft_pairs": [ + "issue:43454|issue:43577", + "issue:43525|issue:46291", + "issue:43697|issue:45357", + "issue:44857|issue:46832", + "issue:33357|issue:34567", + "issue:43404|issue:45127", + "issue:45538|issue:46869", + "issue:43425|issue:46291", + "issue:45137|issue:45405", + "issue:42915|issue:45538", + "issue:28282|issue:29127", + "issue:42175|issue:43824", + "issue:44479|issue:46693", + "issue:44062|issue:44991", + "issue:43742|issue:44291", + "issue:42915|issue:43749", + "issue:44936|issue:46821", + "issue:44849|issue:45701", + "issue:24643|issue:29159", + "issue:43452|issue:43582", + "issue:43577|issue:46512", + "issue:44987|issue:45907", + "issue:27957|issue:28282", + "issue:41720|issue:43540", + "issue:43540|issue:44560", + "issue:44479|issue:45381", + "issue:44568|issue:45488" + ], + "prepared_review_unit_hash": "6c25d5bac050358ee0772fb10288da196e0eae916f4d87e8fabf481ab80b085a", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35181, + "estimated_input_tokens": 8796, + "estimated_eval_tokens": 17848 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is not a duplicate set: the issues span unrelated subsystems (tokenizers, multimodal models, training, device mapping, packaging, and version compatibility). The soft pairs are only superficially similar and do not describe the same underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "Both mention weight tying, but one is about Mistral3 lm_head tying and the other about BLIP2 dtype propagation; different models and different failure modes." + }, + { + "left": "issue:43525", + "right": "issue:46291", + "accept": false, + "reason": "Unrelated errors: missing pad_token_id on Llama4Config vs kernels package requiring version/revision metadata during import." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "RTDetrV2 output regression is unrelated to Qwen3.5 save_pretrained writing wrong visual encoder keys." + }, + { + "left": "issue:44857", + "right": "issue:46832", + "accept": false, + "reason": "Different RT-DETR bugs: AMP/CUDA crash in LwDetrImageLoss vs forward TypeError when feature levels exceed backbone outputs." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "MacOS bus error in CLIP community model loading is unrelated to TrainerState.num_input_tokens_seen not updating." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both involve tied weights, but one is a Mistral3 generation bug and the other is LoRA merge with extended vocabulary causing model collapse; not the same code path." + }, + { + "left": "issue:45538", + "right": "issue:46869", + "accept": false, + "reason": "CLIPTokenizer model_max_length constant is unrelated to YOLO multi-GPU auto device mapping failures." + }, + { + "left": "issue:43425", + "right": "issue:46291", + "accept": false, + "reason": "Torch 2.10 compatibility and kernels import requirements are different dependency issues with no shared bug mechanism." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque underflow is unrelated to an unreleased PEFT version bump." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "Qwen3Moe FineGrainedFP8Config failure is unrelated to CLIPTokenizer's max length constant." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "AutoModel ImportError for missing PyTorch is unrelated to LayoutLMv3 box-validation error messaging." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "TensorFlow missing from a torch-only install is unrelated to the Qwen2_5_VL import/export issue." + }, + { + "left": "issue:44479", + "right": "issue:46693", + "accept": false, + "reason": "Both are video/FA2 related at a high level, but one is a v5 video input regression across Qwen variants and the other is a long-context performance regression with FA2; not the same defect." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "AddedToken special-argument duplication is a tokenizer construction bug, while EMBEDDIA/est-roberta loading failure is a separate tokenizer compatibility issue." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "MobileLLM key mismatch and init_empty_weights unexpected _is_hf_initialized argument are different loading-time failures." + }, + { + "left": "issue:42915", + "right": "issue:43749", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and FSDP_CPU_RAM_EFFICIENT_LOADING breakage are unrelated model/config vs distributed-loading issues." + }, + { + "left": "issue:44936", + "right": "issue:46821", + "accept": false, + "reason": "Trainer evaluate-after-train failure is unrelated to TimesFM 2.5 window_size AttributeError." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Qwen3.5 output_hidden_states bug is a model-specific issue, not the broad tokenization behavior change described in issue 45701." + }, + { + "left": "issue:24643", + "right": "issue:29159", + "accept": false, + "reason": "DeepSpeed training weight shape error is unrelated to slow vs fast tokenizer inconsistency." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "gguf_file loading failures and Apple Silicon caching_allocator_warmup TypeError are separate APIs and code paths." + }, + { + "left": "issue:43577", + "right": "issue:46512", + "accept": false, + "reason": "BLIP2 dtype not propagating on load is related only superficially to from_config(dtype=...) leaving config.dtype stale; one is load-time model object state, the other is constructor-time config synchronization." + }, + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "Loading physical-intelligence/fast and forwarding revision in list_repo_files are unrelated repository/tooling bugs." + }, + { + "left": "issue:27957", + "right": "issue:28282", + "accept": false, + "reason": "XLMRoberta Flash Attention 2 support is unrelated to AutoModel failing when PyTorch is absent." + }, + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "Qwen3 auto device mapping CUDA assert and Qwen3OmniMoe video ValueError are different multimodal/device issues." + }, + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "Both concern Qwen3 video handling, but one is a processing ValueError in Qwen3OmniMoe and the other is a StopIteration in Qwen3-vl-embedding; different model families and failure points." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": false, + "reason": "The v5 video input regression and wrong vision_position_ids are both Qwen VL-related, but the reported bugs are distinct and target different internal computations." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "mdeberta-v3-base add_special_tokens behavior and LlamaTokenizer ByteLevel/Metaspace override are different tokenizer regressions with different tokenization stacks." + } + ] + }, + "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": 24, + "review_unit_index": 23, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:24643", + "issue:29870", + "issue:34567", + "issue:36331", + "issue:38175", + "issue:42175", + "issue:43257", + "issue:43452", + "issue:43531", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43716", + "issue:43742", + "issue:43756", + "issue:43906", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44568", + "issue:44589", + "issue:44625", + "issue:44737", + "issue:44792", + "issue:44821", + "issue:44877", + "issue:44898", + "issue:44964", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45405", + "issue:45600", + "issue:45663", + "issue:45672", + "issue:45674", + "issue:45823", + "issue:46399", + "issue:46463", + "issue:46468", + "issue:46491", + "issue:46693", + "issue:46701", + "issue:46821", + "issue:46846" + ], + "soft_pairs": [ + "issue:44792|issue:44964", + "issue:34567|issue:36331", + "issue:42175|issue:44488", + "issue:38175|issue:44898", + "issue:43257|issue:43756", + "issue:43540|issue:44479", + "issue:44062|issue:45081", + "issue:44291|issue:44821", + "issue:45042|issue:45081", + "issue:43756|issue:46399", + "issue:43756|issue:45325", + "issue:43582|issue:46821", + "issue:43742|issue:44589", + "issue:43716|issue:46701", + "issue:44568|issue:46491", + "issue:45405|issue:45600", + "issue:45672|issue:45674", + "issue:43906|issue:46468", + "issue:43531|issue:44821", + "issue:45823|issue:46463", + "issue:43577|issue:45072", + "issue:43906|issue:46693", + "issue:44737|issue:46463", + "issue:24643|issue:29870", + "issue:43452|issue:44488", + "issue:44625|issue:44877", + "issue:45663|issue:46846" + ], + "prepared_review_unit_hash": "15937fd2e9b9b52376e6ac398a8249f56a1dac73c0e39845e9819035aa9c1194", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35534, + "estimated_input_tokens": 8884, + "estimated_eval_tokens": 18024 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose collection of unrelated transformer bug reports, not a duplicate set. No soft-edge pair is strong enough to merge under strict triage; the closest pair is the Qwen video-input regression reports, but they are still too model/regression-specific to confidently treat as one issue.", + "confidence": 0.89, + "canonical_issue_reason": "issue:44479 is the broadest, most central report in the set: it describes a concrete v5.3.0 regression affecting multiple Qwen multimodal video models, making it the best representative issue if one must be chosen.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44479 has the clearest cross-model scope and strongest duplicate-triage value among the issues here; it reads like the most actionable umbrella report for a real regression.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Different models and failures: janus test failure vs Phi-4 multimodal load error. No shared code-path bug is evident." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "TrainerState token counting bug vs custom compute_loss signature incompatibility. Related to training, but not the same defect." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "Packaging/backend extra issue vs tokenizer/model loading failure for a specific model. Different surfaces and causes." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Both are vision-model issues, but one is zero-probabilities on SigLIP2 and the other is interpolation failure on Perceiver. Not the same bug." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Different models and failure modes: DeepSpeed/accelerate MoE weight conversion vs SmolLM3 RoPE-layer mismatch." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Both involve video inputs in Qwen-family models, but one is Qwen3OmniMoe-specific and the other is a broader v5.3.0 regression across several Qwen VL variants. Similar area, not confidently the same underlying bug." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer initialization collision vs Mistral regex patch crash. Both tokenizer-adjacent, but distinct code paths and symptoms." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "init_empty_weights constructor incompatibility vs AutoImageProcessor URL loading. Unrelated failures." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "Image processor torchvision dependency regression vs Mistral tokenizer regex crash. Different subsystems." + }, + { + "left": "issue:43756", + "right": "issue:46399", + "accept": false, + "reason": "Different Qwen/SmolLM issues: RoPE layer omission vs missing decorator for kernel registration. Not mergeable as one bug." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Both mention rotary-position logic, but one is missing RoPE layers in SmolLM3 and the other is a Qwen2.5-VL temporal position scaling bug. Different fixes." + }, + { + "left": "issue:43582", + "right": "issue:46821", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError vs TimesFM window_size attribute error. No shared underlying defect." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "MobileLLM load-key issue vs float8 storage lookup failure. Different loading/runtime problems." + }, + { + "left": "issue:43716", + "right": "issue:46701", + "accept": false, + "reason": "Mistral-3 image preprocessor dtype mismatch vs DiffusionGemma pipeline breakage. Different models and code paths." + }, + { + "left": "issue:44568", + "right": "issue:46491", + "accept": false, + "reason": "Tokenizer regressions, but one is BOS/EOS insertion behavior and the other is encode-decode whitespace preservation. Not the same bug." + }, + { + "left": "issue:45405", + "right": "issue:45600", + "accept": false, + "reason": "Dependency version bump problem vs auto_mappings referencing removed configs. Different maintenance issues." + }, + { + "left": "issue:45672", + "right": "issue:45674", + "accept": false, + "reason": "Both concern 4-bit inference, but one is a Gemma4 audio-module torch.finfo uint8 issue and the other is BitsAndBytesConfig skip-module handling. Not the same concrete bug." + }, + { + "left": "issue:43906", + "right": "issue:46468", + "accept": false, + "reason": "An isolated reproduction of an old issue vs a broken Nemotron example. No duplicate signal." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior vs image processor URL loading. Unrelated." + }, + { + "left": "issue:45823", + "right": "issue:46463", + "accept": false, + "reason": "Gemma4 device-placement bug vs T5 decoder-input validation message. Different model families and defects." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Both are dtype-related, but one is Blip2 dtype propagation and the other is CI inference mismatches in other models. Not the same issue." + }, + { + "left": "issue:43906", + "right": "issue:46693", + "accept": false, + "reason": "Reproduction of a prior issue vs FA2 long-context performance regression. Different problems." + }, + { + "left": "issue:44737", + "right": "issue:46463", + "accept": false, + "reason": "XLNet CPU placement bug vs T5 error-message issue. Not related." + }, + { + "left": "issue:24643", + "right": "issue:29870", + "accept": false, + "reason": "DeepSpeed training runtime error vs Bart generate-time UnboundLocalError. Different failures and subsystems." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "gguf_file loading breakage for tokenizer/model APIs vs a model loading failure for sleng-bert. Distinct root causes." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Config propagation bug in Qwen3.5 vs strict config rejection for granite_speech. Different config paths." + }, + { + "left": "issue:45663", + "right": "issue:46846", + "accept": false, + "reason": "Gemma4 FSDP2 attention state bug vs Qwen3.5 TP-plan missing linear_attn entries. Both are parallelization-related, but not the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 24, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28282", + "issue:32090", + "issue:38175", + "issue:41762", + "issue:43295", + "issue:43452", + "issue:43525", + "issue:43526", + "issue:43540", + "issue:43643", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43749", + "issue:43761", + "issue:43819", + "issue:43994", + "issue:44361", + "issue:44442", + "issue:44466", + "issue:44493", + "issue:44737", + "issue:44811", + "issue:44843", + "issue:44898", + "issue:44913", + "issue:44936", + "issue:45005", + "issue:45200", + "issue:45216", + "issue:45356", + "issue:45357", + "issue:45372", + "issue:45375", + "issue:45479", + "issue:45588", + "issue:45920", + "issue:46050", + "issue:46121", + "issue:46170", + "issue:46463", + "issue:46693", + "issue:46701", + "issue:46829", + "issue:46832", + "issue:46899", + "issue:46932" + ], + "soft_pairs": [ + "issue:43697|issue:46829", + "issue:43540|issue:46693", + "issue:45357|issue:45920", + "issue:44913|issue:46121", + "issue:28282|issue:32090", + "issue:46463|issue:46899", + "issue:44737|issue:45588", + "issue:43697|issue:45216", + "issue:43526|issue:45479", + "issue:43720|issue:44898", + "issue:43643|issue:45375", + "issue:44466|issue:45005", + "issue:45200|issue:46463", + "issue:44493|issue:46050", + "issue:44361|issue:44442", + "issue:43761|issue:46829", + "issue:43653|issue:44361", + "issue:41762|issue:44936", + "issue:43452|issue:44843", + "issue:43688|issue:43749", + "issue:38175|issue:43295", + "issue:43819|issue:44811", + "issue:43994|issue:44898", + "issue:45372|issue:46170", + "issue:46832|issue:46932", + "issue:43525|issue:46701", + "issue:43295|issue:45356" + ], + "prepared_review_unit_hash": "d4e1bcf03dd5de3a5d72d1f77a3c0e32891a9fc4df732ef12319f6bdc1426886", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 36474, + "estimated_input_tokens": 9119, + "estimated_eval_tokens": 18494 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These issues are mostly unrelated regression reports across different models/features. I found no soft-edge pairs that are strong enough to merge as duplicates based on title-level evidence alone.", + "confidence": 0.88, + "canonical_issue_reason": "issue:46832 is the most concrete and actionable report here: it names a specific model, a specific failing path (`RTDetrModel.forward`), and a precise trigger (`num_feature_levels > len(backbone outputs)`).", + "canonical_pr_reason": null, + "best_issue_reason": "issue:46832 is the best representative issue because it is specific, reproducible, and currently open, making it a cleaner anchor than the broader regression reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43697", + "right": "issue:46829", + "accept": false, + "reason": "Different models and failures: RTDetrV2 output divergence vs VideoPrism hidden_states misassignment." + }, + { + "left": "issue:43540", + "right": "issue:46693", + "accept": false, + "reason": "Unrelated components: Qwen3OmniMoe video-input validation vs a FlashAttention long-context performance regression." + }, + { + "left": "issue:45356", + "right": "issue:45920", + "accept": false, + "reason": "Both are tokenizer regressions, but they describe different failure modes and likely different code paths." + }, + { + "left": "issue:44913", + "right": "issue:46121", + "accept": false, + "reason": "GPTNeoX config reload issue is unrelated to rope-validation type handling." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "PyTorch import missing vs Trainer GPU broadcast TypeError are different problems." + }, + { + "left": "issue:46463", + "right": "issue:46899", + "accept": false, + "reason": "T5 decoder-input validation message vs Gemma4 quantized multimodal embedder dtype casting are unrelated." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "XLNet CPU tensor allocation bug is unrelated to flash_attention s_aux handling." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "Different symptoms and subsystems: RTDetrV2 output mismatch vs Qwen3.5 save_pretrained regression." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug is unrelated to sequence-classification loss degeneracy." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "BitNet packed-weight loading bug vs Perceiver interpolation failure are different code paths." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "AutoConfig remote-code field loss and Qwen3_5MoeVisionConfig strict-field dropping are similar in theme but not the same concrete bug." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": false, + "reason": "Both mention tied weights, but one is device-dependent lm_head serialization and the other is translation-model behavior; not enough to merge as the same bug." + }, + { + "left": "issue:45200", + "right": "issue:46463", + "accept": false, + "reason": "Gemma 4 multimodal token-type IDs are unrelated to T5 decoder-input validation." + }, + { + "left": "issue:44493", + "right": "issue:46050", + "accept": false, + "reason": "Position-id key warnings and sinq save/load quantization breakage are different issues." + }, + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Different tokenizer failures affecting different tokenizers and APIs." + }, + { + "left": "issue:43761", + "right": "issue:46829", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression is not the same as VideoPrism output field mispopulation." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "BigBird special-token registration bug and MLukeTokenizer task AttributeError are unrelated." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "Gemma3 ZeRO-3 loading crash and trainer.evaluate-after-train failure are not the same code-path." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "gguf_file handling and offline model_info() lookup in Mistral regex patch are different tokenizer-loading regressions." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Aux-loss normalization bug in MoE models is unrelated to FSDP CPU RAM efficient loading." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Siglip2 zero-probability issue and processor.tokenizer regression are different multimodal bugs." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "DAC latent reconstruction mismatch is unrelated to Whisper batch_decode skip_special_tokens behavior." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Siglip2 nonsensical outputs vs Perceiver interpolation failure are different models and failure modes." + }, + { + "left": "issue:45372", + "right": "issue:46170", + "accept": false, + "reason": "Gemma 4 processor import breakage is unrelated to Windows ROCm torch.distributed dependency issues." + }, + { + "left": "issue:46832", + "right": "issue:46932", + "accept": false, + "reason": "RT-DETR forward TypeError and Gemma4 quantized embedders are completely different bugs." + }, + { + "left": "issue:43525", + "right": "issue:46701", + "accept": false, + "reason": "Llama4Config pad_token_id and DiffusionGemma generation pipeline failure are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Processor.tokenizer regression and Kimi tokenizer codec/regex regression are different problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 25, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:29159", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:36010", + "issue:41720", + "issue:42898", + "issue:43452", + "issue:43504", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43927", + "issue:44464", + "issue:44466", + "issue:44479", + "issue:44484", + "issue:44554", + "issue:44568", + "issue:44625", + "issue:44704", + "issue:44734", + "issue:44743", + "issue:44821", + "issue:44843", + "issue:44977", + "issue:44991", + "issue:45440", + "issue:45491", + "issue:45542", + "issue:45701", + "issue:45704", + "issue:46002", + "issue:46050", + "issue:46178", + "issue:46402", + "issue:46491", + "issue:46612", + "issue:46693", + "issue:46821" + ], + "soft_pairs": [ + "issue:29159|issue:33357", + "issue:43531|issue:43742", + "issue:43653|issue:44568", + "issue:45542|issue:46178", + "issue:43646|issue:43749", + "issue:42898|issue:46491", + "issue:43688|issue:45704", + "issue:43756|issue:46402", + "issue:44843|issue:44991", + "issue:44464|issue:44743", + "issue:43653|issue:43927", + "issue:43452|issue:43577", + "issue:43688|issue:45701", + "issue:44479|issue:44625", + "issue:45491|issue:46002", + "issue:44484|issue:44704", + "issue:44821|issue:46821", + "issue:43688|issue:45440", + "issue:33666|issue:36010", + "issue:32090|issue:33666", + "issue:44466|issue:46050", + "issue:44734|issue:46612", + "issue:44977|issue:46693", + "issue:43504|issue:43550", + "issue:44464|issue:44977", + "issue:43504|issue:43720", + "issue:41720|issue:44464", + "issue:43577|issue:44554" + ], + "prepared_review_unit_hash": "5097b97dc08548b296d88493c6cf95cccf7187d3d160360dad1601d84261f467", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35566, + "estimated_input_tokens": 8892, + "estimated_eval_tokens": 18040 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: it mixes unrelated tokenizer regressions, model-loading failures, generation bugs, training/dispatch issues, and backend/device-specific crashes. None of the proposed soft pairs look like the same underlying bug or change closely enough to merge.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45701 is the broadest tokenizer-regression report and the closest thing to a representative item, but it still only covers one slice of the cluster and does not subsume the many unrelated loading/training/generation bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:29159", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency vs a MacOS CLIP bus error are different failure modes and code paths." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and MobileLLM key error are unrelated model-specific loading/runtime issues." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and mdeberta add_special_tokens BOS/EOS behavior are different tokenizer bugs." + }, + { + "left": "issue:45542", + "right": "issue:46178", + "accept": false, + "reason": "TensorBoard/TensorFlow backend mismatch is unrelated to the tensorflow_text backend mapping error." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Custom model init breakage in v5 is not the same as FSDP CPU RAM efficient loading being broken." + }, + { + "left": "issue:42898", + "right": "issue:46491", + "accept": false, + "reason": "clean_up_tokenization_spaces behavior changes and CodeLlama leading-space round-trip regression are distinct tokenizer semantics bugs." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "Auxiliary-loss normalization in MoE models is unrelated to T5's apex FusedRMSNorm memory leak." + }, + { + "left": "issue:43756", + "right": "issue:46402", + "accept": false, + "reason": "Smollm3 RoPE-layer mismatch and Sam3Model text-encoder weight loading are different model implementation defects." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Offline model_info() call in AutoTokenizer is unrelated to est-roberta tokenizer loading failure." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Chunked generation with compiled forward and Qwen3.5 recurrent-state reset are different generation/cache bugs." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "BigBird mask-token registration and DiaConfig custom token ID loss are unrelated tokenizer/config serialization issues." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "gguf_file loading failure and Blip2 dtype propagation to qformer are different model-loading bugs." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "MoE auxiliary-loss normalization is not the same as a general tokenization-version regression." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Video input regression for Qwen VL models and Qwen3.5 num_labels propagation affect different subsystems." + }, + { + "left": "issue:45491", + "right": "issue:46002", + "accept": false, + "reason": "Gemma3 NaN embeddings from sliding-window attention and classification-head NaNs after checkout/init are distinct numerical issues." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "save_pretrained shard-size behavior and AutoProcessor kwargs forwarding are unrelated API concerns." + }, + { + "left": "issue:44821", + "right": "issue:46821", + "accept": false, + "reason": "AutoImageProcessor URL loading and TimesFM window_size AttributeError are unrelated loader/config bugs." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "MoE auxiliary-loss normalization does not match DeepseekV3MoE divergence from remote implementation." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Qwen2-VL multi-GPU training and GenerationMixin import failure are different issues with no shared code path." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "Trainer _gpu_broadcast_one NoneType error and Qwen2-VL multi-GPU training are separate distributed-training failures." + }, + { + "left": "issue:44466", + "right": "issue:46050", + "accept": false, + "reason": "v5 lm_head.weight serialization inconsistency and sinq quantized save/load breakage are different persistence bugs." + }, + { + "left": "issue:44734", + "right": "issue:46612", + "accept": false, + "reason": "KV-cache continuation crash in serve and skipped beam-search cache reorder for certain models are not the same generation-path problem." + }, + { + "left": "issue:44977", + "right": "issue:46693", + "accept": false, + "reason": "Qwen3.5 flash-attention output bug and FA2 long-context performance regression differ in symptom and root cause." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Beit legacy-field loading failure and Bamba torch.compile SDPA failure are unrelated model-specific bugs." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Compiled chunked generation inconsistency and Qwen3.5 flash-attention generation failure affect different generation mechanisms." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Legacy field load failure for Beit and BitNet packed-weight unpacking during accelerate loading are different load-time defects." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "Qwen3 auto device mapping cuda assert and compiled chunked generation inconsistency are unrelated runtime issues." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "Blip2 dtype propagation and MPS attention correctness with mismatched value/query heads are different bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 26, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28218", + "issue:33357", + "issue:33666", + "issue:36331", + "issue:39692", + "issue:42796", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43452", + "issue:43493", + "issue:43504", + "issue:43525", + "issue:43572", + "issue:43606", + "issue:43611", + "issue:43643", + "issue:43646", + "issue:43688", + "issue:43720", + "issue:43782", + "issue:43825", + "issue:43844", + "issue:43927", + "issue:43994", + "issue:44361", + "issue:44451", + "issue:44466", + "issue:44514", + "issue:44554", + "issue:44561", + "issue:44568", + "issue:44610", + "issue:44625", + "issue:44964", + "issue:44977", + "issue:45290", + "issue:45405", + "issue:45584", + "issue:45823", + "issue:46050", + "issue:46192", + "issue:46209", + "issue:46365", + "issue:46489", + "issue:46693", + "issue:46710", + "issue:46846" + ], + "soft_pairs": [ + "issue:43611|issue:46846", + "issue:43643|issue:44561", + "issue:43720|issue:43927", + "issue:43782|issue:46693", + "issue:33666|issue:36331", + "issue:43278|issue:43688", + "issue:43606|issue:43927", + "issue:43504|issue:44610", + "issue:43262|issue:43295", + "issue:43720|issue:44361", + "issue:43493|issue:43825", + "issue:44964|issue:45405", + "issue:44514|issue:45290", + "issue:44977|issue:46209", + "issue:44554|issue:46365", + "issue:44625|issue:45584", + "issue:43525|issue:43572", + "issue:43452|issue:44451", + "issue:28218|issue:33357", + "issue:45823|issue:46050", + "issue:43994|issue:46192", + "issue:42796|issue:44977", + "issue:43493|issue:43844", + "issue:39692|issue:43295", + "issue:43278|issue:43844", + "issue:44568|issue:46489", + "issue:43646|issue:44466", + "issue:46209|issue:46710" + ], + "prepared_review_unit_hash": "ef99d0dbf73af3ffb5d295d5069ebb5bb6137310788e27c63df82b6acc80d5f5", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36263, + "estimated_input_tokens": 9066, + "estimated_eval_tokens": 18388 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous: most soft pairs share only broad subsystem labels (tokenizer, config loading, multimodal processors, training) rather than the same concrete bug. I would not merge any of the proposed soft edges; they are too coarse to be true duplicates.", + "confidence": 0.22, + "canonical_issue_reason": "If forced to pick one issue, #28218 is the most generic and discussion-heavy tokenizer bug in the set, but it is only a weak representative because the cluster does not form a single duplicate family.", + "canonical_pr_reason": null, + "best_issue_reason": "#28218 is the best issue candidate by age and generality, but the overall cluster is too mixed for it to be a strong canonical duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43611", + "right": "issue:46846", + "accept": false, + "reason": "Different problems: base_model_prefix model loading vs Qwen3.5 tensor-parallel planning/OOM and shape errors." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Both involve trust_remote_code, but the concrete failures differ: missing returned fields vs removal of is_torch_fx_available." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "Unrelated bugs: BitNet packed-weight loading vs DiaConfig losing custom token IDs." + }, + { + "left": "issue:43782", + "right": "issue:46693", + "accept": false, + "reason": "Different code paths and symptoms: weight_only loading error vs FA2 long-context performance regression." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Multi-GPU training Qwen2-VL vs a Trainer.compute_loss signature regression; not the same underlying bug." + }, + { + "left": "issue:43278", + "right": "issue:43688", + "accept": false, + "reason": "Different numerical issues: dtype drift between train/eval vs auxiliary-loss normalization." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "Different failures: CPU offload device mismatch vs config token ID persistence bug." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "Different models and bugs: legacy field load failure vs processor image-size mismatch." + }, + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Both are processor/chat-template regressions, but one is audio sampling-rate defaulting and the other is tokenizer/processor API breakage." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "BitNet loading issue and MLukeTokenizer AttributeError are unrelated." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 implementation discrepancy vs pipeline error-message wording; not the same bug." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Model loading failure vs a dependency-version bump issue; unrelated." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Both are apply_chat_template crashes, but the triggering inputs and affected logic differ." + }, + { + "left": "issue:44977", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug vs FP8 inference on B200; different runtime issues." + }, + { + "left": "issue:44554", + "right": "issue:46365", + "accept": false, + "reason": "MPS attention correctness bug vs Trainer import conflict with datasets; unrelated." + }, + { + "left": "issue:44625", + "right": "issue:45584", + "accept": false, + "reason": "Different regressions: num_labels propagation vs Whisper empty-transcription generation failure." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both are missing token-id fields after v5, but they affect different configs and different missing attributes." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "Both concern model loading, but one is gguf_file behavior and the other is a specific model compatibility failure." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer spacing bug and a Mac bus error on a CLIP model are unrelated." + }, + { + "left": "issue:45823", + "right": "issue:46050", + "accept": false, + "reason": "Device-map forward mismatch vs quantized save/load breakage; not the same defect." + }, + { + "left": "issue:43994", + "right": "issue:46192", + "accept": false, + "reason": "SigLIP2 wrong outputs vs return_tensors being ignored in text kwargs; unrelated." + }, + { + "left": "issue:42796", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer BPE-marker handling vs Qwen3.5 flash-attention generation failure; different bugs." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and DeepSpeed gradient growth are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "Docs/example errors for SigLIP2 vs processor.tokenizer / image-passing regression; not the same issue." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Evaluation dtype change vs gradient explosion under ZeRO-3 are distinct numerical problems." + }, + { + "left": "issue:44568", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer special-token regressions, but they affect different tokenizers and different concrete behaviors." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Custom model initialization breakage vs inconsistent lm_head serialization; different code paths." + }, + { + "left": "issue:46209", + "right": "issue:46710", + "accept": false, + "reason": "FP8 inference failure on B200 and accuracy regression after upgrade are not the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 27, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:33357", + "issue:33666", + "issue:43257", + "issue:43278", + "issue:43299", + "issue:43334", + "issue:43504", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43653", + "issue:43701", + "issue:43824", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43937", + "issue:44112", + "issue:44291", + "issue:44361", + "issue:44464", + "issue:44488", + "issue:44493", + "issue:44568", + "issue:44589", + "issue:44792", + "issue:44843", + "issue:44849", + "issue:44964", + "issue:45072", + "issue:45200", + "issue:45229", + "issue:45245", + "issue:45357", + "issue:45405", + "issue:45520", + "issue:45759", + "issue:45812", + "issue:46050", + "issue:46209", + "issue:46285", + "issue:46399", + "issue:46424", + "issue:46701", + "issue:46710", + "issue:46821", + "issue:46832", + "issue:46846" + ], + "soft_pairs": [ + "issue:44849|issue:46209", + "issue:33357|issue:33666", + "issue:44568|issue:45812", + "issue:43278|issue:43701", + "issue:44488|issue:44843", + "issue:44792|issue:45405", + "issue:43927|issue:44361", + "issue:43575|issue:45405", + "issue:44849|issue:45357", + "issue:43504|issue:43606", + "issue:43824|issue:44589", + "issue:43257|issue:43299", + "issue:44493|issue:46285", + "issue:45245|issue:46710", + "issue:46399|issue:46846", + "issue:46821|issue:46832", + "issue:43334|issue:45759", + "issue:43606|issue:44112", + "issue:43334|issue:43572", + "issue:43504|issue:43653", + "issue:44589|issue:45520", + "issue:45072|issue:46832", + "issue:43572|issue:44291", + "issue:44464|issue:46424", + "issue:43937|issue:46701", + "issue:43844|issue:44964", + "issue:43873|issue:46050", + "issue:45200|issue:45229" + ], + "prepared_review_unit_hash": "bc315ebd26f3a895244a1bc95e748798be2819ffbe4f70b42a351fb7bbd27ec8", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36146, + "estimated_input_tokens": 9037, + "estimated_eval_tokens": 18330 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The soft pairs are mostly unrelated issues that only share a model family or a broad subsystem (tokenization, multimodal configs, quantization, offload, etc.). I found no pair that clearly describes the same underlying bug/change, so none should be merged as duplicates.", + "confidence": 0.11, + "canonical_issue_reason": "Issue 45812 is the strongest standalone candidate: it describes a broad, concrete regression with clear user impact (wrong token IDs across all Granite models) and a focused failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45812 is the most representative and actionable report in the set because it is broad, specific, and directly tied to a reproducible regression.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44849", + "right": "issue:46209", + "accept": false, + "reason": "Different failures: Qwen3.5 hidden-state behavior vs FP8 inference on B200. Same general area, not the same bug." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "Completely different models and problems: MacOS bus error on CLIP vs Qwen2-VL multi-GPU training." + }, + { + "left": "issue:44568", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different model families and symptoms; not one concrete code-path bug." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "BF16/FP32 dtype mismatch in evaluation is unrelated to resume_from_checkpoint key mismatch." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "Model loading failure for cjvt/sleng-bert is unrelated to an offline model_info() call in AutoTokenizer." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "Janus test failure and a PEFT version bump are unrelated." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenization/config bugs with different failure modes: custom token IDs loss vs MLukeTokenizer AttributeError." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "OOM during tensor-parallel loading is unrelated to the PEFT minimum-version change." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Both mention Qwen3.5, but one is output_hidden_states behavior and the other is incorrect save_pretrained visual keys." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Legacy field loading for BEiT is unrelated to Bark CPU-offload device mismatch." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "ImportError for a missing model class is unrelated to an FP8 storage TypeError." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Both concern Qwen MoE loading, but one is accelerate+deepspeed conversion and the other is a transformers dev compatibility break." + }, + { + "left": "issue:44493", + "right": "issue:46285", + "accept": false, + "reason": "Unexpected key warnings with position IDs are not the same as the CLIP text-model structure inconsistency." + }, + { + "left": "issue:45245", + "right": "issue:46710", + "accept": false, + "reason": "Category cardinality limit error is unrelated to an accuracy regression after upgrade." + }, + { + "left": "issue:46399", + "right": "issue:46846", + "accept": false, + "reason": "Different Qwen3.5 issues: kernel registration crash vs tensor-parallel plan omissions causing OOM/shape errors." + }, + { + "left": "issue:46821", + "right": "issue:46832", + "accept": false, + "reason": "TimesFM window_size AttributeError and RT-DETR forward shape mismatch are unrelated." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "Both involve Qwen multimodal configs, but the specific bug differs: missing pad_token_id vs not unwrapping text_config." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "CPU offload device mismatch in Bark is unrelated to a stale GraniteSpeech CI test." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Different config fields and models: pad_token_id in Qwen3-VL vs pad_token_idx in StableLmConfig." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "BEiT legacy-field loading and BigBirdTokenizer special-token registration are unrelated." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "FP8 storage lookup failure is unrelated to flash_attn import_utils behavior on Python 3.13." + }, + { + "left": "issue:45072", + "right": "issue:46832", + "accept": false, + "reason": "dtype mismatches in inference are unrelated to RT-DETR's num_feature_levels/backbone-output error." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "StableLm missing field after 5.0 update is unrelated to the _is_hf_initialized argument error with init_empty_weights." + }, + { + "left": "issue:44464", + "right": "issue:46424", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency is unrelated to max_cache_length behavior in quantized cache." + }, + { + "left": "issue:43937", + "right": "issue:46701", + "accept": false, + "reason": "GLM-5 generation config validation is unrelated to the DiffusionGemma image-text-to-text pipeline failure." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "Training instability with ZeRO-3 is unrelated to Phi-4 multimodal loading failure." + }, + { + "left": "issue:43873", + "right": "issue:46050", + "accept": false, + "reason": "Offloading/quantization mismatch is not the same as save/load breakage for sinq-quantized models." + }, + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Missing mm_token_type_ids defaulting is unrelated to Gemma4 multi-GPU inference OOM." + } + ] + }, + "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": 24, + "review_unit_index": 28, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:15354", + "issue:28218", + "issue:29159", + "issue:29942", + "issue:30333", + "issue:32090", + "issue:34567", + "issue:37428", + "issue:38175", + "issue:41720", + "issue:42490", + "issue:43065", + "issue:43295", + "issue:43404", + "issue:43450", + "issue:43493", + "issue:43540", + "issue:43696", + "issue:43701", + "issue:43873", + "issue:43957", + "issue:43994", + "issue:44186", + "issue:44423", + "issue:44493", + "issue:44554", + "issue:44610", + "issue:44734", + "issue:44869", + "issue:44977", + "issue:45072", + "issue:45084", + "issue:45198", + "issue:45200", + "issue:45229", + "issue:45406", + "issue:45446", + "issue:45520", + "issue:45588", + "issue:45759", + "issue:45823", + "issue:46050", + "issue:46121", + "issue:46192", + "issue:46369", + "issue:46701", + "issue:46710", + "issue:46752" + ], + "soft_pairs": [ + "issue:44869|issue:46752", + "issue:45229|issue:45823", + "issue:43994|issue:44493", + "issue:45588|issue:46050", + "issue:29942|issue:30333", + "issue:42490|issue:43696", + "issue:29159|issue:32090", + "issue:28218|issue:34567", + "issue:44977|issue:46701", + "issue:45084|issue:45520", + "issue:43065|issue:46369", + "issue:44554|issue:45446", + "issue:44423|issue:44734", + "issue:15354|issue:38175", + "issue:43493|issue:43701", + "issue:41720|issue:46710", + "issue:45406|issue:46121", + "issue:43404|issue:46710", + "issue:45072|issue:45198", + "issue:43295|issue:43450", + "issue:45588|issue:45759", + "issue:37428|issue:43957", + "issue:43540|issue:45200", + "issue:43295|issue:43873", + "issue:44186|issue:46192", + "issue:45823|issue:46192", + "issue:43295|issue:44610" + ], + "prepared_review_unit_hash": "4d264ab493e3a76e344d7e462ec2606e314320ba4699f92399d89c2b5500390c", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35866, + "estimated_input_tokens": 8967, + "estimated_eval_tokens": 18190 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is a loose collection of unrelated transformer issues rather than a duplicate set. None of the proposed soft pairs look like the same underlying bug/change. The most representative issue is the broad, well-referenced offloading/quantization report.", + "confidence": 0.94, + "canonical_issue_reason": "issue:43873 is the broadest and most centrally referenced report in the set (open, 6 inbound references), so it is the best representative issue for the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 has the strongest signal for being the cluster anchor: it is general, actively discussed, and more reusable as a canonical placeholder than the model-specific or narrow bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44869", + "right": "issue:46752", + "accept": false, + "reason": "Different code paths and failures: Whisper timestamp decoding vs chat template handling on empty input." + }, + { + "left": "issue:45229", + "right": "issue:45823", + "accept": false, + "reason": "Both mention Gemma4, but one is a CUDA OOM and the other is a forward device-mismatch bug; not the same change." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 bad outputs and unexpected position-id keys are unrelated symptoms with different likely causes." + }, + { + "left": "issue:45588", + "right": "issue:46050", + "accept": false, + "reason": "Flash-attention sink handling and sinq save/load breakage are distinct bugs in different subsystems." + }, + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention 2 test failures are unrelated to MLflow job-status reporting on training failure." + }, + { + "left": "issue:42490", + "right": "issue:43696", + "accept": false, + "reason": "Import warning for a trailing-slash path is unrelated to GPT-oss CUDA out-of-memory." + }, + { + "left": "issue:29159", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency does not match the Trainer _gpu_broadcast_one NoneType error." + }, + { + "left": "issue:28218", + "right": "issue:34567", + "accept": false, + "reason": "Extra space after added token is a tokenizer formatting issue, not a TrainerState token-count update bug." + }, + { + "left": "issue:44977", + "right": "issue:46701", + "accept": false, + "reason": "Qwen3.5 flash-attention generation failure and DiffusionGemma image-text pipeline breakage are different problems." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Template-node compilation error and flash_attn import KeyError are separate failures." + }, + { + "left": "issue:43065", + "right": "issue:46369", + "accept": false, + "reason": "A dummy Conv2d in Sam3PixelDecoder is not the same as a memory overflow in sam3-litetext." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "MPS attention correctness and a PyTorch version-check bug in flex_attention are unrelated." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Continuous-batching multimodal crash and KV-cache continuation indexing error are different serve-path bugs." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "TorchScript GeneratorExp export failure and SigLIP2 zero-probability outputs are unrelated." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 HF-vs-JAX discrepancy and resume_from_checkpoint key mismatch do not share a concrete failure mode." + }, + { + "left": "issue:41720", + "right": "issue:46710", + "accept": false, + "reason": "Qwen3 auto device-mapping CUDA assert and DeepSeek accuracy regression after upgrade are different issues." + }, + { + "left": "issue:45406", + "right": "issue:46121", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer in serve and rope-param conversion TypeError are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:46710", + "accept": false, + "reason": "Mistral3 weight tying and DeepSeek accuracy regression are separate model-behavior problems." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "BF16 dtype mismatches in CI and Wav2Vec2 save/tokenization failure are not the same bug." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Processor.tokenizer regression and batched video-shape handling are different multimodal processing bugs." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention sink-less crash and Qwen multimodal text_config unwrapping are unrelated code paths." + }, + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "ImportError for _flash_supports_window_size and meta-device model loading breakage are distinct issues." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Video-input processing failure in Qwen3OmniMoe is unrelated to Gemma4 text-only mm_token_type_ids defaults." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor.tokenizer regression and offloading/quantization behavior are different subsystems and failure modes." + }, + { + "left": "issue:44186", + "right": "issue:46192", + "accept": false, + "reason": "LayoutLMv2Tokenizer NER/padding crash and return_tensors being ignored are separate tokenizer bugs." + }, + { + "left": "issue:45823", + "right": "issue:46192", + "accept": false, + "reason": "Gemma4 device mismatch during forward is unrelated to return_tensors handling." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Processor API regression and OmDet-Turbo image-size mismatch are different model-processing 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": 24, + "review_unit_index": 29, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28218", + "issue:29870", + "issue:29942", + "issue:35141", + "issue:37428", + "issue:39692", + "issue:42490", + "issue:42915", + "issue:43232", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43450", + "issue:43452", + "issue:43493", + "issue:43550", + "issue:43572", + "issue:43575", + "issue:43653", + "issue:43824", + "issue:43844", + "issue:43856", + "issue:43950", + "issue:43975", + "issue:44112", + "issue:44442", + "issue:44466", + "issue:44521", + "issue:44554", + "issue:44737", + "issue:44849", + "issue:44987", + "issue:45127", + "issue:45198", + "issue:45200", + "issue:45290", + "issue:45356", + "issue:45381", + "issue:45520", + "issue:45823", + "issue:46178", + "issue:46192", + "issue:46209", + "issue:46285", + "issue:46512", + "issue:46821", + "issue:46894" + ], + "soft_pairs": [ + "issue:43572|issue:46821", + "issue:43550|issue:43844", + "issue:45200|issue:45823", + "issue:29870|issue:29942", + "issue:43450|issue:43575", + "issue:45520|issue:46178", + "issue:43572|issue:44987", + "issue:44554|issue:45356", + "issue:42490|issue:43493", + "issue:43232|issue:44849", + "issue:37428|issue:43824", + "issue:39692|issue:46192", + "issue:43295|issue:43493", + "issue:44466|issue:45127", + "issue:44521|issue:45290", + "issue:43653|issue:44112", + "issue:45381|issue:46209", + "issue:43278|issue:43856", + "issue:28218|issue:35141", + "issue:44737|issue:46894", + "issue:43950|issue:46285", + "issue:42915|issue:43824", + "issue:43824|issue:43975", + "issue:43452|issue:46512", + "issue:43262|issue:43450", + "issue:44442|issue:45198" + ], + "prepared_review_unit_hash": "0a39e980791d03adf183122d31531721f680f06f37012f2f0d203e1a2e95f8c1", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 35249, + "estimated_input_tokens": 8813, + "estimated_eval_tokens": 17882 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items do not form a coherent duplicate cluster: the soft pairs mostly share only broad themes like tokenizers, generation, or import errors, but point to different concrete bugs and code paths. I would not merge any of the proposed soft edges, and I would not choose a single canonical issue for this set.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43572", + "right": "issue:46821", + "accept": false, + "reason": "Unrelated config bugs: missing pad_token_idx vs window_size AttributeError." + }, + { + "left": "issue:43550", + "right": "issue:43844", + "accept": false, + "reason": "Different failures and subsystems: torch.compile+SDPA vs DeepSpeed ZeRO-3 gradient growth." + }, + { + "left": "issue:45200", + "right": "issue:45823", + "accept": false, + "reason": "Same model family, but different bugs: mm_token_type_ids defaulting vs device_map forward mismatch." + }, + { + "left": "issue:29870", + "right": "issue:29942", + "accept": false, + "reason": "Bart generate() decoder_attention_mask error is unrelated to Flash Attention 2 test failures." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Video batch shape bug vs tensor-parallel OOM are different problems." + }, + { + "left": "issue:45520", + "right": "issue:46178", + "accept": false, + "reason": "Different import/backend issues: flash_attn on Python 3.13 vs tensorflow_text backend mapping." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "Both are loading-related, but one is StableLm config and the other is a model load regression." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "MPS attention correctness issue vs Kimi tokenizer regression; no shared code path." + }, + { + "left": "issue:42490", + "right": "issue:43493", + "accept": false, + "reason": "Local-path import warning is unrelated to SigLIP2 HF/JAX discrepancy." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation kwargs sync bug vs Qwen3.5 hidden-states bug are distinct." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "Different missing symbols and different affected models; both are import errors only at a broad level." + }, + { + "left": "issue:39692", + "right": "issue:46192", + "accept": false, + "reason": "Docs/example failure vs silent return_tensors handling bug; not the same issue." + }, + { + "left": "issue:43295", + "right": "issue:43493", + "accept": false, + "reason": "Processor.tokenizer regression is unrelated to SigLIP2 implementation mismatch." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "lm_head serialization inconsistency and LoRA-collapse after vocab extension are different tied-weight bugs." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Multimodal assistant masks vs tool-call tokenization crash are different apply_chat_template paths." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "Tokenizer special-token bug is unrelated to a GraniteSpeech CI test failure." + }, + { + "left": "issue:45381", + "right": "issue:46209", + "accept": false, + "reason": "Qwen2.5-VL vision_position_ids bug vs FP8 inference failure on B200 are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Embedding dtype regression vs Qwen3 MoE memory inefficiency are different training issues." + }, + { + "left": "issue:28218", + "right": "issue:35141", + "accept": false, + "reason": "Tokenizer spacing after added token is unrelated to output embedding reinit on resize." + }, + { + "left": "issue:44737", + "right": "issue:46894", + "accept": false, + "reason": "XLNet CPU device bug is unrelated to CohereASR training-loss shift bug." + }, + { + "left": "issue:43950", + "right": "issue:46285", + "accept": false, + "reason": "Non-persistent buffer corruption in from_pretrained is unrelated to CLIP text-model structure mismatch." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "Qwen3Moe FP8Config failure is unrelated to Qwen2_5_VL import errors." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "ImportError for a VL class and DeepSeek detokenization are different bugs entirely." + }, + { + "left": "issue:43452", + "right": "issue:46512", + "accept": false, + "reason": "gguf_file loading breakage is unrelated to from_config dtype/config stale-state behavior." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sampling-rate bug and video batched-shape bug are separate processor issues." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Different speech model/tokenization failures; not the 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": 24, + "review_unit_index": 30, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28282", + "issue:29942", + "issue:32090", + "issue:33357", + "issue:37428", + "issue:38175", + "issue:41720", + "issue:43334", + "issue:43335", + "issue:43388", + "issue:43425", + "issue:43525", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43819", + "issue:43827", + "issue:43901", + "issue:44112", + "issue:44206", + "issue:44479", + "issue:44561", + "issue:44610", + "issue:44792", + "issue:44857", + "issue:44898", + "issue:45083", + "issue:45200", + "issue:45325", + "issue:45412", + "issue:45520", + "issue:45563", + "issue:45584", + "issue:45593", + "issue:45759", + "issue:46170", + "issue:46192", + "issue:46209", + "issue:46234", + "issue:46285", + "issue:46424", + "issue:46512", + "issue:46821", + "issue:46858", + "issue:46899", + "issue:46932" + ], + "soft_pairs": [ + "issue:43334|issue:43525", + "issue:45200|issue:46899", + "issue:45200|issue:46932", + "issue:44792|issue:45563", + "issue:43819|issue:44112", + "issue:44898|issue:45412", + "issue:44479|issue:46192", + "issue:45200|issue:45325", + "issue:41720|issue:46209", + "issue:43827|issue:43901", + "issue:43335|issue:43550", + "issue:44561|issue:46234", + "issue:43819|issue:45083", + "issue:43388|issue:43526", + "issue:44857|issue:45593", + "issue:44610|issue:44857", + "issue:45759|issue:46285", + "issue:46424|issue:46858", + "issue:43606|issue:45083", + "issue:46512|issue:46821", + "issue:45520|issue:46170", + "issue:44206|issue:45584", + "issue:38175|issue:45412", + "issue:43335|issue:43653", + "issue:29942|issue:32090", + "issue:43550|issue:44112", + "issue:28282|issue:33357", + "issue:37428|issue:43425" + ], + "prepared_review_unit_hash": "099fbe9a3a7f3880d5b1985795187dd3fbc38a6d690c020d4615f8c6c4aeb072", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35853, + "estimated_input_tokens": 8964, + "estimated_eval_tokens": 18184 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated bug reports; the only clear duplicate pair is the Gemma4 4-bit quantized vision/audio embedder dtype-cast regression (#46899/#46932). The rest share broad themes but not the same concrete bug path.", + "confidence": 0.89, + "canonical_issue_reason": "#46899 is the more complete symptom report for the Gemma4 quantization regression; #46932 is a narrower fix-oriented restatement of the same dtype-casting bug in vision/audio embedders.", + "canonical_pr_reason": null, + "best_issue_reason": "#46899 is the clearest representative of the only true duplicate theme in this set and has the most informative problem description.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43334", + "right": "issue:43525", + "accept": false, + "reason": "Same symptom family (missing pad_token_id) but different configs/models and likely separate model-specific regressions." + }, + { + "left": "issue:45200", + "right": "issue:46899", + "accept": false, + "reason": "Both involve Gemma4, but one is about mm_token_type_ids defaults and the other about 4-bit quantized embedder dtype casting; different bugs." + }, + { + "left": "issue:45200", + "right": "issue:46932", + "accept": false, + "reason": "Different failure modes: text-only mm_token_type_ids handling vs quantized vision/audio embedder casting." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "Different generation paths and symptoms; one is image generation test failure, the other is a stale warning about num_return_sequences." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "Unrelated subsystems: DAC latent reconstruction mismatch vs GraniteSpeech CI device override flake." + }, + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Perceiver resolution handling and RT-DETR memory release are unrelated issues." + }, + { + "left": "issue:44479", + "right": "issue:46192", + "accept": false, + "reason": "Video input regression vs return_tensors being ignored; both are input-processing issues but not the same bug." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Both touch multimodal models, but the bugs differ: missing default token ids vs temporal position scaling in rope index." + }, + { + "left": "issue:41720", + "right": "issue:46209", + "accept": false, + "reason": "Different hardware/model-specific inference failures with different root causes and environments." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs regressions, but they concern different pipeline API deprecations and are not one underlying defect." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "Different models and different code paths: sparse-layer config creation vs torch.compile/SDPA failure." + }, + { + "left": "issue:44561", + "right": "issue:46234", + "accept": false, + "reason": "Both mention import_utils, but they are missing different symbols with different breakage patterns and no shared concrete bug path." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "DAC latent conversion and qwen3_omni_moe feature-length helper behavior are unrelated." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Different label-processing bugs in different components; not the same code path." + }, + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "Different object-detection losses and different failure conditions." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Processor input-size mismatch vs AMP/CUDA crash; no shared underlying bug." + }, + { + "left": "issue:45759", + "right": "issue:46285", + "accept": false, + "reason": "Composite Qwen text_config unwrapping and CLIP text model structure inconsistency are different API breakages." + }, + { + "left": "issue:46424", + "right": "issue:46858", + "accept": false, + "reason": "Both are cache-related, but one is about cache length configuration and the other about device tensors breaking slice arithmetic." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "CPU offload device mismatch and qwen3_omni_moe feature-length helper behavior are unrelated." + }, + { + "left": "issue:46512", + "right": "issue:46821", + "accept": false, + "reason": "Stale model.config.dtype after from_config(dtype=...) is unrelated to TimesFM window_size AttributeError." + }, + { + "left": "issue:45520", + "right": "issue:46170", + "accept": false, + "reason": "Different platform/import failures with different missing dependencies and error modes." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Unsupported feature-arg crash in LasrFeatureExtractor vs Whisper empty-transcription generation failure." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "Siglip2 zero probabilities and RT-DETR memory retention are unrelated." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug and BigBirdTokenizer special-token registration are separate issues." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Failing Flash Attention tests and Trainer gpu_broadcast_one NoneType error are different bugs." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "Different code paths and symptoms: compile/SDPA model failure vs stale CI device-override test." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "PyTorch-not-found ImportError and MacOS bus error on a CLIP model are unrelated." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Missing flash attention helper import and torch 2.10 incompatibility are too broad to be the same concrete bug." + }, + { + "left": "issue:46899", + "right": "issue:46932", + "accept": true, + "reason": "Same Gemma4 quantized vision/audio embedder dtype-casting regression; #46932 is essentially the fix-oriented restatement of #46899." + } + ] + }, + "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": 24, + "review_unit_index": 31, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:26413", + "issue:28218", + "issue:29127", + "issue:31501", + "issue:32090", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:39692", + "issue:42490", + "issue:43278", + "issue:43335", + "issue:43504", + "issue:43526", + "issue:43606", + "issue:43611", + "issue:43701", + "issue:43720", + "issue:43819", + "issue:43825", + "issue:43854", + "issue:43931", + "issue:44112", + "issue:44351", + "issue:44448", + "issue:44493", + "issue:44514", + "issue:44521", + "issue:44617", + "issue:44743", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:45083", + "issue:45127", + "issue:45245", + "issue:45290", + "issue:45310", + "issue:45459", + "issue:45478", + "issue:45588", + "issue:45593", + "issue:45859", + "issue:46234", + "issue:46326", + "issue:46431", + "issue:46566", + "issue:46697" + ], + "soft_pairs": [ + "issue:43611|issue:43931", + "issue:44448|issue:44743", + "issue:43335|issue:43504", + "issue:44855|issue:45083", + "issue:43278|issue:43825", + "issue:43335|issue:45083", + "issue:43335|issue:45593", + "issue:29127|issue:32090", + "issue:43335|issue:44112", + "issue:43720|issue:44112", + "issue:44514|issue:44521", + "issue:45588|issue:45859", + "issue:45290|issue:46326", + "issue:42490|issue:43854", + "issue:37428|issue:44351", + "issue:34689|issue:35141", + "issue:45127|issue:45245", + "issue:46431|issue:46566", + "issue:43819|issue:44493", + "issue:39692|issue:43526", + "issue:44448|issue:44779", + "issue:45310|issue:45478", + "issue:26413|issue:31501", + "issue:43606|issue:44857", + "issue:28218|issue:32090", + "issue:45459|issue:46234", + "issue:43701|issue:45083", + "issue:39692|issue:44493", + "issue:44617|issue:46697" + ], + "prepared_review_unit_hash": "2e5bcc17c51512fd970d872e7bf1827040476f362e04568647b60efc04f1d74d", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36623, + "estimated_input_tokens": 9156, + "estimated_eval_tokens": 18568 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Only one soft-edge pair is a true duplicate: the two Qwen3.5 MoE from_pretrained reports (45310/45478). The rest are conservative false matches across different bugs or code paths.", + "confidence": 0.89, + "canonical_issue_reason": "issue:45310 is the better canonical issue because it is the earlier report of the same Qwen3.5 MoE from_pretrained failure and can serve as the shared reference for the later near-duplicate issue:45478.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is the best issue overall since it is the earliest and clearest representative of the only confirmed duplicate pair, so it is the most suitable consolidation target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43611", + "right": "issue:43931", + "accept": false, + "reason": "Both are model-loading failures, but they describe different failure modes and models; not the same bug." + }, + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Different problems: Pegasus output mismatch vs Qwen recurrent state reset in cache handling." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "Unrelated model/config bugs in different components; no shared concrete code path." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Python 3.13 import/torchscript parsing error vs qwen3_omni_moe helper length behavior." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Training/eval dtype drift is unrelated to a pipeline error message regression." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "Different models and different bugs; only broad subsystem overlap." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "Sparse-layer config bug and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "LayoutLMv3 error clarity and Trainer NoneType broadcast are different issues." + }, + { + "left": "issue:43335", + "right": "issue:44112", + "accept": false, + "reason": "Different model/config bug vs CI device-override test failure." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "Packed-weight loading failure is not the same as a stale CI test." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Both touch apply_chat_template, but one is a batch-padding crash and the other is incorrect assistant masks." + }, + { + "left": "issue:45588", + "right": "issue:45859", + "accept": false, + "reason": "Flash attention None aux handling and CPU-offload rotary embedding incompatibility are separate code paths." + }, + { + "left": "issue:45290", + "right": "issue:46326", + "accept": false, + "reason": "Tool-call/no-content crash is distinct from continue_final_message behavior with mistral_common tokenizers." + }, + { + "left": "issue:42490", + "right": "issue:43854", + "accept": false, + "reason": "Local-path trailing-slash warning and GLM-4.7-Flash model loading are unrelated." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Missing flash-support helper and missing HybridCache export are different import regressions." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Model-loading breakage for Llama 3.2 Vision is unrelated to embedding reinitialization after resize." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA merge collapse and the 2^24 category limit are unrelated bugs." + }, + { + "left": "issue:46431", + "right": "issue:46566", + "accept": false, + "reason": "Missing chat template on a processor and meta-device weights with device_map='auto' are different failures." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "DAC latent mismatch and unexpected position-id keys are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "SigLIP2 doc/example issues and BeitImageProcessorFast label reduction bug are different problems." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Different output on a summarization model vs tokenizer regression in Deepseek; not the same change." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Same title, same reported Qwen3.5 MoE from_pretrained error, and the later issue is a near-duplicate of the same bug." + }, + { + "left": "issue:26413", + "right": "issue:31501", + "accept": false, + "reason": "Checkpoint resume iterator failure and FineWeb training not starting do not match the same root cause." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU-offload device mismatch in bark-small is distinct from LwDetrImageLoss AMP/CUDA crashing." + }, + { + "left": "issue:28218", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer added-space behavior is unrelated to the Trainer NoneType broadcast error." + }, + { + "left": "issue:45459", + "right": "issue:46234", + "accept": false, + "reason": "Tokenizer protobuf error masking and a missing availability helper are unrelated." + }, + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Resume-from-checkpoint key mismatch is unrelated to qwen3_omni_moe feature-length behavior." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "Doc/example model-processor mismatch is unrelated to position-id key warnings." + }, + { + "left": "issue:44617", + "right": "issue:46697", + "accept": false, + "reason": "CUDA OOM in Sam3Video and over-padding in Swin2SR are not the same bug." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Reject: the duplicate call on 45310/45478 is plausible, but the canonical/best-issue rationale overstates the packet. Both issues are shown as closed, so calling 45310 the \"primary open report\" or \"clearest live tracker\" is not grounded. The rest of the soft-edge rejections are generally conservative, but this factual error is enough to fail precision." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 32, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:10000", + "issue:15354", + "issue:25251", + "issue:27957", + "issue:29870", + "issue:29942", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:40444", + "issue:42175", + "issue:42490", + "issue:43278", + "issue:43329", + "issue:43381", + "issue:43526", + "issue:43550", + "issue:43575", + "issue:43873", + "issue:43927", + "issue:44206", + "issue:44360", + "issue:44448", + "issue:44466", + "issue:44493", + "issue:44561", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45083", + "issue:45229", + "issue:45290", + "issue:45381", + "issue:45588", + "issue:45698", + "issue:45823", + "issue:45920", + "issue:46192", + "issue:46399", + "issue:46421", + "issue:46655", + "issue:46858" + ], + "soft_pairs": [ + "issue:44206|issue:45823", + "issue:32090|issue:33357", + "issue:42490|issue:43575", + "issue:44857|issue:45072", + "issue:43550|issue:45083", + "issue:44360|issue:44561", + "issue:29942|issue:35141", + "issue:27957|issue:34689", + "issue:38175|issue:44493", + "issue:43278|issue:43329", + "issue:29870|issue:32090", + "issue:45920|issue:46192", + "issue:44448|issue:44977", + "issue:45290|issue:45381", + "issue:10000|issue:15354", + "issue:44977|issue:45229", + "issue:43927|issue:45698", + "issue:44855|issue:46399", + "issue:43278|issue:43381", + "issue:29870|issue:33290", + "issue:40444|issue:43526", + "issue:45290|issue:46655", + "issue:44857|issue:45588", + "issue:32090|issue:33290", + "issue:37428|issue:42175", + "issue:46421|issue:46858", + "issue:44561|issue:44779", + "issue:43873|issue:45229", + "issue:42490|issue:43873", + "issue:44448|issue:44466", + "issue:25251|issue:32090" + ], + "prepared_review_unit_hash": "89ee10672877a69d9791d6206084616bef6f9f8ca8ea7790402e292882bf6146", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37237, + "estimated_input_tokens": 9310, + "estimated_eval_tokens": 18876 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the items cover unrelated bugs in generation, tokenization, loading, Flash Attention, dtype/device handling, and training. None of the soft pairs look like the same underlying issue, so no duplicate consolidation is justified.", + "confidence": 0.98, + "canonical_issue_reason": "No single canonical issue fits this cluster; the issues are thematically broad but describe different failures and code paths.", + "canonical_pr_reason": null, + "best_issue_reason": "No best issue candidate stands out because the cluster is not a duplicate set and lacks one representative underlying bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44206", + "right": "issue:45823", + "accept": false, + "reason": "Both mention model/input handling, but one is a LasrFeatureExtractor crash and the other is a Gemma4 device-mismatch during forward; different bugs and fixes." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "NoneType GPU broadcast in Trainer vs MacOS bus error on a community CLIP model; different subsystems and failure modes." + }, + { + "left": "issue:42490", + "right": "issue:43575", + "accept": false, + "reason": "Local-path import warning on trailing slash vs OOM when loading a huge Qwen model with tensor parallelism; not the same bug." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "AMP/CUDA loss crash in LwDetrImageLoss vs bfloat16 dtype mismatch in SwitchTransformers/TimmWrapperModel; distinct code paths." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "torch.compile+SDPA failure for Bamba-9B-v2 vs helper-length calculation behavior in qwen3_omni_moe; unrelated issues." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "DSA indexer activation-function complaint vs removal of is_torch_fx_available breaking trust_remote_code models; no shared underlying defect." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention 2 test failures vs token-embedding resize reinitializing output embeddings; different components and fixes." + }, + { + "left": "issue:27957", + "right": "issue:34689", + "accept": false, + "reason": "XLMRoberta Flash Attention 2 support vs Llama 3.2 90B Vision model-loading regression; not the same concrete problem." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 vs unexpected position-id keys in many models; distinct symptoms and likely causes." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "BF16-to-FP32 embedding dtype regression during eval vs undefined video-branch variables in multimodal token counting; unrelated." + }, + { + "left": "issue:29870", + "right": "issue:32090", + "accept": false, + "reason": "BART generate decoder_attention_mask bug vs Trainer gpu_broadcast NoneType error; different call sites and failure mechanisms." + }, + { + "left": "issue:45920", + "right": "issue:46192", + "accept": false, + "reason": "AutoTokenizer producing wrong token IDs vs return_tensors being ignored when text_kwargs is passed; both tokenizer-related but not the same bug." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Pegasus v4/v5 output difference vs Qwen3.5 flash-attention generation failure; different models and code paths." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "apply_chat_template crash with tool-call assistant messages vs Qwen2.5-VL video vision_position_ids bug; both multimodal but distinct failures." + }, + { + "left": "issue:10000", + "right": "issue:15354", + "accept": false, + "reason": "German DistilBertModel issue vs torch.jit.script GeneratorExp unsupported for ViT export; unrelated historical issues." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Flash-attention generation bug in Qwen3.5 vs multi-GPU CUDA OOM for Gemma4; not mergeable into one fix." + }, + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "DiaConfig losing custom token IDs on save/load vs wrong custom module loaded after save_pretrained; both serialization-related but different objects and bugs." + }, + { + "left": "issue:44855", + "right": "issue:46399", + "accept": false, + "reason": "Python 3.13 indentation/JIT parsing failure in DebertaV2Model vs missing decorator causing Qwen3.5 kernel registration error; separate root causes." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Embedding dtype regression during eval vs gradient checkpointing disallowed in eval mode; different behaviors." + }, + { + "left": "issue:29870", + "right": "issue:33290", + "accept": false, + "reason": "generate() decoder_attention_mask bug vs Adafactor+DeepSpeed OOM; one is correctness, the other memory usage." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "Qwen2.5-VL iterable dataset with multiple images failing vs BeitImageProcessorFast reduce_labels returning one label; different multimodal/preprocessing issues." + }, + { + "left": "issue:45290", + "right": "issue:46655", + "accept": false, + "reason": "Chat template crash on tool-call messages vs wrong padding calculation in video processor; different layers of the stack." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "LwDetrImageLoss fp16 crash vs flash_attention.py AttributeError on s_aux=None; separate bugs in different modules." + }, + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Trainer gpu_broadcast NoneType error vs deepspeed Adafactor OOM; not the same underlying problem." + }, + { + "left": "issue:37428", + "right": "issue:42175", + "accept": false, + "reason": "Missing _flash_supports_window_size import vs TensorFlow not included in torch-only install; unrelated packaging/API issues." + }, + { + "left": "issue:46421", + "right": "issue:46858", + "accept": false, + "reason": "Cache lazy initialization causing recompilation vs seq_length returning a device tensor and breaking arithmetic; both cache-related but different bugs." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "Removal of is_torch_fx_available breaking remote-code models vs DeepSeek tokenizer regression in v5; different functionality and fixes." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Quantization/offloading not working as expected vs Gemma4 multi-GPU CUDA OOM; shared memory theme only, not the same bug." + }, + { + "left": "issue:42490", + "right": "issue:43873", + "accept": false, + "reason": "Trailing-slash local-model import warning vs quantization offloading behavior; unrelated issues." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "Pegasus output divergence across versions vs inconsistent lm_head.weight serialization depending on device; different model behaviors and code paths." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "pipeline top_k changing output shape vs Trainer NoneType GPU broadcast error; no common 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": 24, + "review_unit_index": 33, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:29127", + "issue:29870", + "issue:29942", + "issue:30064", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:39692", + "issue:41720", + "issue:42915", + "issue:42994", + "issue:43065", + "issue:43278", + "issue:43381", + "issue:43412", + "issue:43575", + "issue:43611", + "issue:43653", + "issue:43701", + "issue:43756", + "issue:43976", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44361", + "issue:44442", + "issue:44479", + "issue:44568", + "issue:44743", + "issue:44871", + "issue:44977", + "issue:45198", + "issue:45229", + "issue:45245", + "issue:45372", + "issue:45406", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45563", + "issue:45588", + "issue:45712", + "issue:45803", + "issue:45823", + "issue:46121", + "issue:46693" + ], + "soft_pairs": [ + "issue:43976|issue:45440", + "issue:44568|issue:45491", + "issue:44871|issue:44977", + "issue:45588|issue:45823", + "issue:41720|issue:45229", + "issue:29127|issue:34567", + "issue:43381|issue:43701", + "issue:43278|issue:43575", + "issue:33357|issue:39692", + "issue:44977|issue:45563", + "issue:45198|issue:45803", + "issue:44186|issue:45198", + "issue:44361|issue:45198", + "issue:43611|issue:46121", + "issue:43653|issue:44743", + "issue:29870|issue:34689", + "issue:45372|issue:45588", + "issue:43756|issue:45440", + "issue:42915|issue:42994", + "issue:29942|issue:30064", + "issue:44206|issue:44479", + "issue:43653|issue:44442", + "issue:45372|issue:45406", + "issue:45459|issue:45712", + "issue:45491|issue:46693", + "issue:44265|issue:45198", + "issue:43412|issue:45588", + "issue:43065|issue:43994", + "issue:29127|issue:33357", + "issue:45245|issue:45563", + "issue:43065|issue:45412", + "issue:43575|issue:43701" + ], + "prepared_review_unit_hash": "b2c3b36edc6a4c879816c3cbf85dbad4dd6fc883b8e656fd0934e2ffbe9e126b", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37412, + "estimated_input_tokens": 9353, + "estimated_eval_tokens": 18962 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the items span unrelated bugs in tokenizers, generation, model loading, quantization, device mapping, and documentation. None of the soft pairs look like the same underlying issue, so all are rejected.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Different failure classes: a Python version compatibility break vs a DeepSeek implementation divergence." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Tokenizer special-token behavior is unrelated to Gemma3 NaN embeddings in mixed-length batching." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "EOS token config mismatch is distinct from a flash-attention generation bug." + }, + { + "left": "issue:45588", + "right": "issue:45823", + "accept": false, + "reason": "Different code paths: flash_attention integration crash vs Gemma4 PLE device placement mismatch." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Both are inference failures, but one is a CUDA assert during auto device mapping and the other is multi-GPU OOM." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error messaging is unrelated to TrainerState token-count tracking." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Eval-mode gradient checkpointing incompatibility is separate from checkpoint resume key mismatch." + }, + { + "left": "issue:43278", + "right": "issue:43575", + "accept": false, + "reason": "Embedding dtype drift in eval is unrelated to tensor-parallel OOM on a large Qwen model." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "MacOS bus error on CLIP loading is not the same bug as SigLIP2 documentation example errors." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Generation malfunction with flash-attention is distinct from a stale warning in paged generate()." + }, + { + "left": "issue:45198", + "right": "issue:45803", + "accept": false, + "reason": "Wav2Vec2 save/tokenization failures do not match a bare-except bug in Fuyu tensor conversion." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "Different models and symptoms: LayoutLMv2 tokenizer crashes vs Wav2Vec2 save/tokenization failures." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "MLuke task-time AttributeError is unrelated to Wav2Vec2 save/tokenization behavior." + }, + { + "left": "issue:43611", + "right": "issue:46121", + "accept": false, + "reason": "Base-model-prefix loading breakage is not the same as rope-parameter list handling." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and Qwen3.5 recurrent-state reset are unrelated." + }, + { + "left": "issue:29870", + "right": "issue:34689", + "accept": false, + "reason": "BART generate() UnboundLocalError differs from a Llama 3.2 Vision loading regression." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "Gemma4 processor import failure is a different bug than sink-less flash_attention None handling." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Smollm3 RoPE-layer truncation is a different model behavior than DeepSeekV3MoE divergence." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "FineGrainedFP8Config model failure and quantized model saving failure are distinct quantization issues." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "Flash Attention 2 test failures are unrelated to void segmentation-map processing." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Unsupported center arg in LasrFeatureExtractor is not the same as the Qwen video-input regression." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "BigBirdTokenizer empty decode issue is unrelated to AutoTokenizer loading FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Gemma4 processor import error and missing _tokenizer on Gemma4Processor are different failure points." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "Tokenizer error masking with missing protobuf is unrelated to leftover dummy class cleanup." + }, + { + "left": "issue:45491", + "right": "issue:46693", + "accept": false, + "reason": "NaN embeddings in Gemma3 are a correctness bug, not the same as a performance regression in FA2 long-context generation." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export failures from torch_compilable_check are unrelated to Wav2Vec2 save/tokenization bugs." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Executorch export guard/erfinv support problems do not match the flash_attention sink-less-model crash." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "A dummy Conv2d in Sam3PixelDecoder is unrelated to SigLIP2 AutoModel/pipeline nonsense outputs." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "LayoutLMv3 box-information errors are not the same as the CLIP MacOS bus error." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Category-count overflow is unrelated to the stale num_return_sequences warning." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "A dummy convolution placeholder is unrelated to RT-DETR memory not being released." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "Tensor-parallel OOM and resume_from_checkpoint key mismatch are different problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 34, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:27957", + "issue:29127", + "issue:33290", + "issue:34689", + "issue:35141", + "issue:42490", + "issue:42915", + "issue:43066", + "issue:43334", + "issue:43381", + "issue:43421", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43646", + "issue:43761", + "issue:43824", + "issue:43873", + "issue:43931", + "issue:43994", + "issue:44077", + "issue:44186", + "issue:44279", + "issue:44387", + "issue:44442", + "issue:44462", + "issue:44464", + "issue:44514", + "issue:45072", + "issue:45250", + "issue:45372", + "issue:45381", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45584", + "issue:45588", + "issue:45593", + "issue:45600", + "issue:45834", + "issue:45859", + "issue:46121", + "issue:46369", + "issue:46399", + "issue:46640", + "issue:46693" + ], + "soft_pairs": [ + "issue:45440|issue:45593", + "issue:42490|issue:43381", + "issue:29127|issue:35141", + "issue:43066|issue:44462", + "issue:45372|issue:45459", + "issue:43421|issue:45491", + "issue:43994|issue:45412", + "issue:43575|issue:44387", + "issue:45412|issue:46369", + "issue:44442|issue:45072", + "issue:44464|issue:45584", + "issue:42915|issue:43931", + "issue:43526|issue:45834", + "issue:43824|issue:44279", + "issue:43526|issue:46693", + "issue:27957|issue:29127", + "issue:43994|issue:46369", + "issue:27957|issue:33290", + "issue:43526|issue:44186", + "issue:33290|issue:34689", + "issue:33290|issue:35141", + "issue:46121|issue:46399", + "issue:43334|issue:43531", + "issue:45250|issue:46640", + "issue:43381|issue:43873", + "issue:45588|issue:45600", + "issue:43761|issue:45859", + "issue:43646|issue:44077", + "issue:44514|issue:45381" + ], + "prepared_review_unit_hash": "1262e30054e7228d92dd2a55bc391ea83a9ca6efb203fc57a08b34b4cbe01e49", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35674, + "estimated_input_tokens": 8919, + "estimated_eval_tokens": 18094 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The candidate pairs are largely unrelated and should all be rejected as duplicates. If a representative issue is needed, #43873 is the strongest anchor because it is open, broad, and has the most inbound references/discussion.", + "confidence": 0.42, + "canonical_issue_reason": "#43873 is the most broadly referenced and actively discussed open issue in the set, so it is the best representative anchor if one must be chosen.", + "canonical_pr_reason": null, + "best_issue_reason": "#43873 is the strongest issue candidate overall: open, high discussion/inbound references, and framed as a broader regression rather than a narrow model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "Different model families and different failure modes: DeepseekV3MoE divergence vs D-FINE auxiliary loss behavior." + }, + { + "left": "issue:42490", + "right": "issue:43381", + "accept": false, + "reason": "Local-path import warning and gradient-checkpointing eval-mode error are unrelated bugs." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "LayoutLMv3 error messaging and embedding reinitialization on resize are separate code paths." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Both touch tokenizer loading, but one is a decoder-type mismatch in v5 and the other is AutoTokenizer ignoring tokenizer.json; not the same bug." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "Gemma processor import failure due to mistral_common differs from hidden tokenizer errors when protobuf is missing." + }, + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Runtime special-token post-processor updates and Gemma3 sliding-window NaNs are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 bad outputs and RT-DETR memory release issues are different problems." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both involve OOMs, but one is tensor-parallel model loading and the other is int4 quantization memory regression." + }, + { + "left": "issue:45412", + "right": "issue:46369", + "accept": false, + "reason": "Memory leak on deletion and memory overflow during SAM3 LiteText call are not the same underlying issue." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer loading failure and bf16 dtype mismatch in inference are unrelated." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Compiled-forward generation inconsistency and Whisper empty-transcription failure after special-token alignment are different code paths." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and Qwen3-VL shape mismatch are separate model-loading bugs." + }, + { + "left": "issue:43526", + "right": "issue:45834", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug and Kosmos2.5 OCR long-input index error are not the same defect." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Missing Qwen2_5_VL import and a generic dependency issue are too vague/different to merge." + }, + { + "left": "issue:43526", + "right": "issue:46693", + "accept": false, + "reason": "Label reduction in an image processor and FA2 long-context performance regression are unrelated." + }, + { + "left": "issue:27957", + "right": "issue:29127", + "accept": false, + "reason": "Flash Attention 2 on XLMRoberta and LayoutLMv3 box-validation messaging are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:46369", + "accept": false, + "reason": "SigLIP2 inference outputs and SAM3 LiteText memory overflow are different issues." + }, + { + "left": "issue:27957", + "right": "issue:33290", + "accept": false, + "reason": "Flash Attention 2 support and Adafactor+DeepSpeed OOM are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and LayoutLMv2Tokenizer NER/padding crashes are different tokenizer/processor bugs." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "Adafactor OOM in DeepSpeed and Llama 3.2 Vision loading breakage are unrelated." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "OOM during optimizer use and embedding reinitialization after resize are not the same problem." + }, + { + "left": "issue:46121", + "right": "issue:46399", + "accept": false, + "reason": "Rope-parameter validation TypeError and missing decorator registration for apply_rotary_pos_emb are different rope-related failures." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Qwen3-VL pad_token_id import failure and Qwen3-MoE sliding_window behavior are distinct issues." + }, + { + "left": "issue:45250", + "right": "issue:46640", + "accept": false, + "reason": "Both mention Flash Attention 2, but the titles are too generic and provide no evidence of the same concrete bug." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Gradient checkpointing in eval mode and offloading with quantization are separate training/runtime behaviors." + }, + { + "left": "issue:45588", + "right": "issue:45600", + "accept": false, + "reason": "Flash-attention sinkless-model crash and auto_mappings CI breakage are unrelated." + }, + { + "left": "issue:43761", + "right": "issue:45859", + "accept": false, + "reason": "CLIP hidden_states regression and Qwen3_5Moe CPU-offload incompatibility are different model-specific bugs." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "Custom model initialization breakage in Transformers 5.0 and patchtsmixer's post_init policy are not the same underlying issue." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Qwen2.5-VL batched chat-template crash and video position-id bug are related only by model family, not the same code-path problem." + } + ] + }, + "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": 24, + "review_unit_index": 35, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:27957", + "issue:29870", + "issue:42175", + "issue:42490", + "issue:43232", + "issue:43262", + "issue:43316", + "issue:43388", + "issue:43493", + "issue:43525", + "issue:43643", + "issue:43720", + "issue:43749", + "issue:43761", + "issue:43827", + "issue:43873", + "issue:43931", + "issue:44062", + "issue:44077", + "issue:44079", + "issue:44188", + "issue:44190", + "issue:44206", + "issue:44442", + "issue:44479", + "issue:44496", + "issue:44743", + "issue:44811", + "issue:44855", + "issue:44871", + "issue:44898", + "issue:44913", + "issue:44977", + "issue:44987", + "issue:45003", + "issue:45137", + "issue:45216", + "issue:45245", + "issue:45325", + "issue:45440", + "issue:45759", + "issue:45803", + "issue:45907", + "issue:46178", + "issue:46234", + "issue:46697", + "issue:46829", + "issue:46932" + ], + "soft_pairs": [ + "issue:44977|issue:45245", + "issue:44743|issue:44811", + "issue:44442|issue:44898", + "issue:43262|issue:44206", + "issue:42490|issue:44743", + "issue:45003|issue:45440", + "issue:43720|issue:44442", + "issue:42175|issue:46234", + "issue:43493|issue:43827", + "issue:44077|issue:44479", + "issue:27957|issue:29870", + "issue:44062|issue:45907", + "issue:43316|issue:43525", + "issue:45325|issue:46697", + "issue:43873|issue:44871", + "issue:44855|issue:45137", + "issue:43749|issue:43931", + "issue:43643|issue:44913", + "issue:43316|issue:45759", + "issue:44190|issue:44496", + "issue:44442|issue:45803", + "issue:43761|issue:44188", + "issue:43761|issue:44079", + "issue:44987|issue:46178", + "issue:43931|issue:45216", + "issue:43232|issue:43388", + "issue:46829|issue:46932" + ], + "prepared_review_unit_hash": "8345dc218082e15e323ddae14fbb7b7bfbd615b30a2321055f4d0f5c675f2588", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 36183, + "estimated_input_tokens": 9046, + "estimated_eval_tokens": 18348 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is a collection of largely unrelated issue reports across many different models and subsystems; none of the soft pairs look like the same underlying bug or change.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Different failures in different areas: Qwen3.5 flash-attn generation vs a categories cardinality runtime error." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Unrelated subsystems: Qwen recurrent cache state reset vs Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer loading failure for FastSpeech2Conformer is unrelated to Perceiver image-classification interpolation/padding behavior." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting is not the same bug as LastFeatureExtractor passing an unsupported center arg." + }, + { + "left": "issue:42490", + "right": "issue:44743", + "accept": false, + "reason": "Local-path import warning and Qwen recurrent-state reset are different code paths and symptoms." + }, + { + "left": "issue:45003", + "right": "issue:45440", + "accept": false, + "reason": "Both mention modeling code, but one is unsafe sys.modules access and the other is DeepseekV3MoE behavior divergence." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "BitNet accelerate-loading packed-weight bug is unrelated to tokenizer loading for FastSpeech2Conformer." + }, + { + "left": "issue:42175", + "right": "issue:46234", + "accept": false, + "reason": "Package/backend availability issues, but for different backends and different missing functionality." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "SigLIP2 implementation mismatch is unrelated to docs still referencing removed pipeline() APIs." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "Different models and bugs: PatchTSMixer post_init policy vs Qwen video-input regression." + }, + { + "left": "issue:27957", + "right": "issue:29870", + "accept": false, + "reason": "XLMRoberta flash-attention support and BART decoder_attention_mask generation error are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:45907", + "accept": false, + "reason": "AddedToken special-argument conflict is unrelated to list_repo_files failing to forward revision." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Both are config-related, but they concern different models and different missing/incorrect fields." + }, + { + "left": "issue:45325", + "right": "issue:46697", + "accept": false, + "reason": "Qwen2.5-VL rope-index temporal scaling and Swin2SR over-padding are distinct image/video preprocessing bugs." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization/offloading behavior is unrelated to Gemma-3 eos_token_id configuration mismatch." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 decorator parsing issue is unrelated to DeepSpeed ZeRO3 deque underflow." + }, + { + "left": "issue:43749", + "right": "issue:43931", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and Qwen3-VL weight-shape mismatch are different loading problems." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "Remote-code config field loss and GPTNeoX rotary_pct reload reset are separate config serialization issues." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Different composite-config problems: Gemma3TextConfig API discrepancy vs AutoModelForCausalLM not unwrapping text_config for Qwen 3.5/3.6." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Dataset loading in example script is unrelated to AutoConfig failing on a model without model_type." + }, + { + "left": "issue:44442", + "right": "issue:45803", + "accept": false, + "reason": "Tokenizer loading failure is unrelated to FuyuBatchFeature swallowing exceptions in tensor conversion." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression is unrelated to attention-kernel divergence under torch.compile." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "Hidden_states being None and ModelOutput key assignment with prior None are different output-handling bugs." + }, + { + "left": "issue:44987", + "right": "issue:46178", + "accept": false, + "reason": "Loading a specific model with transformers>=5.1.0 is unrelated to a missing BACKENDS_MAPPING entry for tensorflow_text." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Both involve Qwen models, but one is a load-time shape mismatch and the other is a save_pretrained checkpoint regression." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation kwargs update after sync_gpus and gather_for_metrics label truncation are unrelated distributed-training bugs." + }, + { + "left": "issue:46829", + "right": "issue:46932", + "accept": false, + "reason": "VideoPrism output field mixup and Gemma4 quantized embedders dtype casting are unrelated model 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": 24, + "review_unit_index": 36, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:25251", + "issue:27957", + "issue:28282", + "issue:32090", + "issue:42617", + "issue:42907", + "issue:43295", + "issue:43316", + "issue:43404", + "issue:43441", + "issue:43452", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43650", + "issue:43701", + "issue:43716", + "issue:43742", + "issue:43825", + "issue:44077", + "issue:44162", + "issue:44242", + "issue:44464", + "issue:44496", + "issue:44534", + "issue:44610", + "issue:44792", + "issue:44871", + "issue:44913", + "issue:45584", + "issue:45632", + "issue:45698", + "issue:45803", + "issue:46050", + "issue:46170", + "issue:46202", + "issue:46234", + "issue:46291", + "issue:46365", + "issue:46369", + "issue:46421", + "issue:46424", + "issue:46506", + "issue:46821", + "issue:46894" + ], + "soft_pairs": [ + "issue:43701|issue:46506", + "issue:42907|issue:45698", + "issue:42617|issue:43575", + "issue:43550|issue:43701", + "issue:43575|issue:43742", + "issue:27957|issue:32090", + "issue:43606|issue:43701", + "issue:25251|issue:28282", + "issue:46234|issue:46291", + "issue:43295|issue:43825", + "issue:16998|issue:22355", + "issue:43295|issue:44871", + "issue:44871|issue:45584", + "issue:44464|issue:46421", + "issue:44077|issue:44534", + "issue:43316|issue:43716", + "issue:43316|issue:46821", + "issue:44871|issue:46050", + "issue:43646|issue:44913", + "issue:43452|issue:44496", + "issue:45803|issue:46365", + "issue:44162|issue:44792", + "issue:43404|issue:43441", + "issue:45632|issue:46424", + "issue:44242|issue:46894", + "issue:45803|issue:46170", + "issue:44610|issue:46369", + "issue:43650|issue:46202" + ], + "prepared_review_unit_hash": "3d771bf70d637dd17ff9b8f344b658d6b5db8cec16d037e7d140220d30600d5f", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 35721, + "estimated_input_tokens": 8931, + "estimated_eval_tokens": 18118 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated Transformers issues; the soft-similarity pairs share broad themes (loading, cache, generation, model configs) but not the same concrete bug or change. No soft pair is a safe duplicate merge candidate.", + "confidence": 0.71, + "canonical_issue_reason": "issue:45698 is the broadest and most active report here: it is open, has the most discussion/inbound references, and describes a concrete `from_pretrained`/`save_pretrained` custom-module loading failure that could serve as the cluster\u2019s representative loading bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45698 is the strongest standalone issue for follow-up because it is specific, reproducible, and has the clearest user impact among the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43701", + "right": "issue:46506", + "accept": false, + "reason": "Different failures: checkpoint key mismatch vs CPU/CUDA result divergence." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "Both involve saving/loading, but one is dequantized model save failure and the other is wrong custom module loaded after save/load." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "Parallel script startup failure and tensor-parallel OOM are distinct problems." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "torch.compile + SDPA bug is unrelated to resume-from-checkpoint key mismatch." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "OOM during tensor-parallel load is not the same as a key error while loading a model." + }, + { + "left": "issue:27957", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention support issue and Trainer broadcast TypeError are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU offload device mismatch is not the same as checkpoint key mismatch." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline output-shape regression and missing PyTorch import are unrelated." + }, + { + "left": "issue:46234", + "right": "issue:46291", + "accept": false, + "reason": "Both are import-time dependency issues, but one is a missing utility symbol and the other is a kernels version/revision requirement." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Processor/tokenizer regression and misleading pipeline error messaging are different bugs." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "model_max_length behavior and missing `transformers.onnx` module are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Processor API regression and eos_token_id config mismatch are not the same underlying issue." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Both involve special tokens, but one is a config inconsistency and the other is a Whisper empty-transcription generation failure." + }, + { + "left": "issue:44464", + "right": "issue:46421", + "accept": false, + "reason": "Chunked generation inconsistency and cache lazy-init recompilation are related to compilation, but not the same bug." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "These are separate v5 issues: post_init policy vs non-persistent buffer initialization." + }, + { + "left": "issue:43316", + "right": "issue:43716", + "accept": false, + "reason": "Config API discrepancy and image-preprocessor dtype mismatch are unrelated." + }, + { + "left": "issue:43316", + "right": "issue:46821", + "accept": false, + "reason": "Gemma3TextConfig API mismatch and TimesFM window_size AttributeError are different model-specific issues." + }, + { + "left": "issue:44871", + "right": "issue:46050", + "accept": false, + "reason": "eos_token_id mismatch is not the same as quantized model save/load breakage." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "Custom model initialization breakage and GPTNeoX rotary_pct reload regression are distinct configuration problems." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "From-pretrained loader issues, but one is gguf handling and the other is missing model_type metadata." + }, + { + "left": "issue:45803", + "right": "issue:46365", + "accept": false, + "reason": "Bare except swallowing errors in Fuyu preprocessing is unrelated to importing Trainer with datasets installed." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 workflow breakage and Janus image-generation test failure are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Weight tying bug and FlashAttention incompatibility in the Mistral/Ministral family are different code paths." + }, + { + "left": "issue:45632", + "right": "issue:46424", + "accept": false, + "reason": "Cache-path collision for trust_remote_code is unrelated to max_cache_length only applying to quantized cache." + }, + { + "left": "issue:44242", + "right": "issue:46894", + "accept": false, + "reason": "Load-balancing loss omission and training-loss double-shift are not the same bug." + }, + { + "left": "issue:45803", + "right": "issue:46170", + "accept": false, + "reason": "Tensor conversion exception handling and Windows/ROCm distributed import failure are unrelated." + }, + { + "left": "issue:44610", + "right": "issue:46369", + "accept": false, + "reason": "Processor image-size mismatch and SAM memory overflow are different failures." + }, + { + "left": "issue:43650", + "right": "issue:46202", + "accept": false, + "reason": "One issue contains no technical bug content; the other is a Gemma pipeline roundtrip failure." + } + ] + }, + "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": 24, + "review_unit_index": 37, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:35141", + "issue:41950", + "issue:42490", + "issue:43065", + "issue:43352", + "issue:43388", + "issue:43475", + "issue:43504", + "issue:43525", + "issue:43531", + "issue:43540", + "issue:43572", + "issue:43673", + "issue:43688", + "issue:43761", + "issue:43827", + "issue:43994", + "issue:44190", + "issue:44479", + "issue:44514", + "issue:44534", + "issue:44617", + "issue:44734", + "issue:44743", + "issue:44792", + "issue:44855", + "issue:44869", + "issue:44913", + "issue:44964", + "issue:44977", + "issue:45127", + "issue:45200", + "issue:45245", + "issue:45406", + "issue:45542", + "issue:45588", + "issue:45600", + "issue:45632", + "issue:45698", + "issue:46082", + "issue:46178", + "issue:46234", + "issue:46617", + "issue:46640", + "issue:46697", + "issue:46832", + "issue:46858" + ], + "soft_pairs": [ + "issue:43761|issue:44479", + "issue:43504|issue:45698", + "issue:43540|issue:44514", + "issue:45542|issue:46234", + "issue:43388|issue:43688", + "issue:44792|issue:44977", + "issue:44792|issue:45600", + "issue:43065|issue:44617", + "issue:43572|issue:46178", + "issue:45127|issue:45200", + "issue:43673|issue:44743", + "issue:43352|issue:46640", + "issue:44190|issue:44964", + "issue:41950|issue:43475", + "issue:44534|issue:44913", + "issue:43525|issue:43531", + "issue:44734|issue:45406", + "issue:35141|issue:42490", + "issue:46082|issue:46697", + "issue:45632|issue:46617", + "issue:45245|issue:46858", + "issue:43827|issue:43994", + "issue:44855|issue:45588", + "issue:44869|issue:46832" + ], + "prepared_review_unit_hash": "ea7c6cf7dd2a6e93731466514538c94ca96a56e36ed57e83754f4f826b61970a", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 24, + "serialized_chars": 34531, + "estimated_input_tokens": 8633, + "estimated_eval_tokens": 17522 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is very heterogeneous: the issues span unrelated model families, config/loading regressions, processor/pipeline bugs, cache behavior, and training/runtime edge cases. I don\u2019t see any soft pair that clearly describes the same underlying bug or a PR that would merge into a single fix path.", + "confidence": 0.95, + "canonical_issue_reason": "No clear canonical issue emerges; the cluster appears to be a noisy grouping of unrelated bugs. If a representative must be chosen, issue 45698 is the broadest/openest loading bug, but it does not subsume the others.", + "canonical_pr_reason": null, + "best_issue_reason": "No issue in this set is a strong global duplicate hub. Issue 45698 is the most broadly scoped and best-engaged open issue, but only as a weak representative, not as a true canonical duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "Both are video/vision-regression themed, but they affect different model stacks and failure modes; not the same concrete bug." + }, + { + "left": "issue:43504", + "right": "issue:45698", + "accept": false, + "reason": "Legacy pretrained preset loading for BEiT semantic segmentation is unrelated to custom-module loading after save_pretrained." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Qwen3Omni video-input processing and Qwen2_5_VL chat-template batching are distinct code paths and symptoms." + }, + { + "left": "issue:45542", + "right": "issue:46234", + "accept": false, + "reason": "These are different backend/import-utils issues: tensorboard/TensorFlow detection vs a missing falcon-mamba availability helper." + }, + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "One is gather_for_metrics dropping tuple labels; the other is auxiliary-loss normalization. Different subsystems and fixes." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Janus image-generation test failure and Qwen3.5 flash-attention generation regression are unrelated model-specific problems." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "A failed janus test does not match an auto_mappings CI break from removed Sam3LiteText configs." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder and SAM3 video OOM are different issues, not one fixable code-path bug." + }, + { + "left": "issue:43572", + "right": "issue:46178", + "accept": false, + "reason": "StableLmConfig pad_token_idx regression and a backend-mapping error for tensorflow_text are unrelated configuration/import issues." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "LoRA merge collapse with tie_word_embeddings=True is not the same as Gemma 4 mm_token_type_ids defaulting behavior." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Chunked prefill cache handling and Qwen3.5 recurrent-state reset are both cache-related but on different mechanisms and code paths." + }, + { + "left": "issue:43352", + "right": "issue:46640", + "accept": false, + "reason": "NemotronH Flash Attention 2 unsupported error and a generic Flash Attention 2 issue are not concrete duplicates." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Local dataset loading in an example script is unrelated to loading the Phi-4 multimodal model." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Video-classification processor lookup and a missing fpn_position_embeddings attribute in SAM3 video are different bugs." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer corruption and GPTNeoX rotary_pct reload behavior are separate configuration/runtime problems." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Llama4 pad_token_id missing and Qwen3-MoE sliding_window behavior are unrelated config/model issues." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "Serve KV-cache continuation indexing and Gemma4Processor missing _tokenizer are distinct failures in different components." + }, + { + "left": "issue:35141", + "right": "issue:42490", + "accept": false, + "reason": "Embedding resize/post_init reinitialization and path-with-trailing-slash import warnings do not share a bug mechanism." + }, + { + "left": "issue:46082", + "right": "issue:46697", + "accept": false, + "reason": "LlamaConfig head_dim validation and Swin2SR over-padding are unrelated model-configuration vs image-processing bugs." + }, + { + "left": "issue:45632", + "right": "issue:46617", + "accept": false, + "reason": "Trust-remote-code cache collision by leaf directory name is not the same as FileNotFoundError when loading from cache." + }, + { + "left": "issue:45245", + "right": "issue:46858", + "accept": false, + "reason": "Category-count limit and device-tensor seq_length arithmetic are completely different runtime failures." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs still referencing pipeline() and SigLIP2 producing nonsensical AutoModel/pipeline results are related only by pipelines broadly, not by the same bug." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Python 3.13 indentation/parsing import failure and flash_attention.py crashing on s_aux=None are unrelated code paths." + }, + { + "left": "issue:44869", + "right": "issue:46832", + "accept": false, + "reason": "Whisper timestamp decoding and RT-DETR feature-level indexing are different 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": 24, + "review_unit_index": 38, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:25251", + "issue:27957", + "issue:29870", + "issue:35141", + "issue:39692", + "issue:41720", + "issue:43065", + "issue:43295", + "issue:43335", + "issue:43550", + "issue:43644", + "issue:43673", + "issue:43931", + "issue:44060", + "issue:44077", + "issue:44112", + "issue:44190", + "issue:44242", + "issue:44462", + "issue:44464", + "issue:44479", + "issue:44610", + "issue:44617", + "issue:44734", + "issue:44779", + "issue:44855", + "issue:44877", + "issue:44913", + "issue:44933", + "issue:45072", + "issue:45127", + "issue:45245", + "issue:45276", + "issue:45290", + "issue:45357", + "issue:45362", + "issue:45381", + "issue:45563", + "issue:45584", + "issue:45600", + "issue:45907", + "issue:46369", + "issue:46421", + "issue:46489", + "issue:46655", + "issue:46752", + "issue:46897", + "issue:46932" + ], + "soft_pairs": [ + "issue:45072|issue:46932", + "issue:43673|issue:46421", + "issue:29870|issue:35141", + "issue:44242|issue:46897", + "issue:43931|issue:44060", + "issue:44617|issue:45584", + "issue:43644|issue:44913", + "issue:43335|issue:44855", + "issue:25251|issue:27957", + "issue:45563|issue:45584", + "issue:44933|issue:45362", + "issue:43550|issue:44190", + "issue:44077|issue:44877", + "issue:41720|issue:45245", + "issue:45381|issue:46369", + "issue:44479|issue:44734", + "issue:43335|issue:44190", + "issue:43065|issue:44610", + "issue:39692|issue:44190", + "issue:44464|issue:45563", + "issue:45127|issue:45357", + "issue:45381|issue:46655", + "issue:44779|issue:46489", + "issue:44112|issue:45600", + "issue:43295|issue:45276", + "issue:44617|issue:46369", + "issue:45290|issue:46752", + "issue:44462|issue:45907" + ], + "prepared_review_unit_hash": "1cb8da91e37cc83c813ffd99477578ac99541e38b331fc7e0e712b8d4692c434", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36129, + "estimated_input_tokens": 9033, + "estimated_eval_tokens": 18322 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a bundle of unrelated bug reports that only match at a very broad subsystem level. None of the soft-edge pairs look like the same underlying bug/change, so all should be rejected.", + "confidence": 0.93, + "canonical_issue_reason": "Issue 45276 is the clearest self-contained bug report in the set: it describes a concrete Gemma4 embedding-resize failure with a specific expected/actual mismatch and a narrow fix surface.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45276 is the best anchor candidate because it is specific, reproducible, and actionable; the rest are either broader regressions, unrelated model issues, or follow-on reports rather than true duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45072", + "right": "issue:46932", + "accept": false, + "reason": "Both mention dtype/casting problems, but they affect different model families and different code paths (generic bfloat16 inference vs quantized Gemma4 embedders)." + }, + { + "left": "issue:43673", + "right": "issue:46421", + "accept": false, + "reason": "Both involve chunked prefill/cache behavior, but one is about missing cache state in v5.0.0 while the other is about spurious recompilation from lazy cache init." + }, + { + "left": "issue:29870", + "right": "issue:35141", + "accept": false, + "reason": "These are unrelated failures: BART generate() mask initialization vs token-embedding resize reinitialization." + }, + { + "left": "issue:44242", + "right": "issue:46897", + "accept": false, + "reason": "One is an MoE load-balancing loss bug, the other is a Florence2 training-loss shift bug; they are not the same defect." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Different models and failures: Qwen3-VL weight-shape loading mismatch vs Qwen3-Next tied-weights warning misbinding." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "CUDA OOM in Sam3Video is not the same as Whisper failing on empty transcription after token alignment." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer junk in v5 is unrelated to GPTNeoXConfig rotary_pct reverting on reload." + }, + { + "left": "issue:43335", + "right": "issue:44855", + "accept": false, + "reason": "Sparse-layer construction in SwitchTransformers and a Python 3.13 import indentation failure are different issues." + }, + { + "left": "issue:25251", + "right": "issue:27957", + "accept": false, + "reason": "Pipeline top_k nesting behavior and XLMRoberta Flash Attention 2 are unrelated." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "A stale warning for num_return_sequences is not the same as a Whisper crash on empty transcription." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "A missing import from image_utils and a Qwen3.5 chat crash are unrelated." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "Torch.compile/SDPA failure in Bamba is unrelated to the local dataset loading bug in the example script." + }, + { + "left": "issue:44077", + "right": "issue:44877", + "accept": false, + "reason": "Both touch config validation, but one is about disallowing optional post_init and the other is about a strict config blocking granite_speech loading." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "Qwen3 auto device mapping cudaErrorAssert and the 2^24 category limit error are different failures." + }, + { + "left": "issue:45381", + "right": "issue:46369", + "accept": false, + "reason": "Qwen2.5-VL video-position indexing and Sam3LiteText memory overflow are not the same bug." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Both are generation regressions, but one is a video-input regression in Qwen models and the other is a serve-side KV-cache indexing crash." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "SwitchTransformers sparse-layer config failure and dataset loading in an example script are unrelated." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "A dummy Conv2d in Sam3PixelDecoder is not the same as OmDet-Turbo producing the wrong image resolution." + }, + { + "left": "issue:39692", + "right": "issue:44190", + "accept": false, + "reason": "Documentation example errors for SigLIP2 and a local dataset loading failure do not describe the same bug." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistencies are different from a stale warning about num_return_sequences." + }, + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "LoRA merge collapse with extended vocab and incorrect visual-encoder keys in save_pretrained are unrelated." + }, + { + "left": "issue:45381", + "right": "issue:46655", + "accept": false, + "reason": "Both are video-related, but they describe different model families and different processor bugs (vision_position_ids vs padding calculation)." + }, + { + "left": "issue:44779", + "right": "issue:46489", + "accept": false, + "reason": "These are related by theme, but the reports are not identical enough: generic DeepSeek tokenizer regression vs a DeepSeek-Coder v1 tokenizer gap tied to a specific prior fix." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "A stale GraniteSpeech test and broken auto_mappings references to removed configs are separate CI problems." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Processor.tokenizer regression and Gemma4 embedding-resize behavior affect different APIs and code paths." + }, + { + "left": "issue:44617", + "right": "issue:46369", + "accept": false, + "reason": "Both mention Sam3 models, but one is OOM in Sam3Video while the other is memory overflow in sam3-litetext; not the same concrete bug." + }, + { + "left": "issue:45290", + "right": "issue:46752", + "accept": false, + "reason": "Both crash in apply_chat_template, but one is specifically about assistant tool-call messages with no content and the other is about an empty list input." + }, + { + "left": "issue:44462", + "right": "issue:45907", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json and list_repo_files not forwarding revision are unrelated repository/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": 24, + "review_unit_index": 39, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:25251", + "issue:27957", + "issue:29870", + "issue:32090", + "issue:35141", + "issue:36246", + "issue:38617", + "issue:41628", + "issue:41950", + "issue:42175", + "issue:42796", + "issue:43404", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43643", + "issue:43650", + "issue:43673", + "issue:43854", + "issue:44016", + "issue:44112", + "issue:44220", + "issue:44246", + "issue:44279", + "issue:44295", + "issue:44496", + "issue:44521", + "issue:44734", + "issue:44737", + "issue:44938", + "issue:44964", + "issue:45003", + "issue:45216", + "issue:45276", + "issue:45381", + "issue:45542", + "issue:45584", + "issue:45600", + "issue:45632", + "issue:45803", + "issue:46082", + "issue:46121", + "issue:46234", + "issue:46424", + "issue:46506", + "issue:46697", + "issue:46858" + ], + "soft_pairs": [ + "issue:43526|issue:46697", + "issue:44737|issue:45803", + "issue:45276|issue:45584", + "issue:43540|issue:44734", + "issue:44938|issue:45542", + "issue:44521|issue:45381", + "issue:25251|issue:29870", + "issue:43673|issue:46858", + "issue:43854|issue:44964", + "issue:45632|issue:46234", + "issue:43643|issue:44496", + "issue:44246|issue:45003", + "issue:44964|issue:45600", + "issue:44112|issue:44964", + "issue:27957|issue:35141", + "issue:42175|issue:44279", + "issue:46082|issue:46121", + "issue:42796|issue:44521", + "issue:36246|issue:41628", + "issue:32090|issue:35141", + "issue:43404|issue:45216", + "issue:36246|issue:38617", + "issue:43531|issue:45600", + "issue:41950|issue:44016", + "issue:44295|issue:46506", + "issue:43673|issue:46424", + "issue:43650|issue:44220" + ], + "prepared_review_unit_hash": "5e17d18e0e56ef97e2ef5309b612752a1889764456aa2520b7a6ce571b196170", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35411, + "estimated_input_tokens": 8853, + "estimated_eval_tokens": 17962 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is very noisy and mostly contains unrelated issues that only share superficial transformer/modeling vocabulary. I would not merge any of the suggested soft edges; they are not the same underlying bug or change.", + "confidence": 0.23, + "canonical_issue_reason": "issue:44521 is the most discussion-heavy open issue with inbound references, so it is the best available anchor, but the cluster is heterogeneous and it is not a strong canonical representative.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44521 is the strongest issue candidate by activity and references, though only as a weak representative because the surrounding items do not form a true duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43526", + "right": "issue:46697", + "accept": false, + "reason": "Both are image-processor bugs, but they affect different processors and different failure modes (label reduction vs over-padding)." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "Unrelated bugs: one is XLNet device placement in positional encoding, the other is a broad exception-handling issue in Fuyu batching." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Different subsystems and symptoms: token-embedding resizing/tied weights versus Whisper empty-transcription generation." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Both touch video/generation-related code, but the concrete problems are unrelated: Qwen3OmniMoe video input validation versus serve KV-cache indexing." + }, + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Different causes of import/backend failure: Python 3.14 loading issue versus missing TensorFlow in a torch-only install." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "Both involve multimodal handling, but one is assistant mask computation and the other is vision position IDs for Qwen2.5-VL video input." + }, + { + "left": "issue:25251", + "right": "issue:29870", + "accept": false, + "reason": "No shared bug: pipeline top_k output nesting versus Bart generate decoder_attention_mask initialization." + }, + { + "left": "issue:43673", + "right": "issue:46858", + "accept": false, + "reason": "Both are cache-related, but one is about missing cache during chunked prefill and the other is a device-tensor return breaking slice arithmetic on non-CUDA backends." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Both are model-loading failures, but they concern different models and likely different loader paths." + }, + { + "left": "issue:45632", + "right": "issue:46234", + "accept": false, + "reason": "Different import/runtime issues: trust_remote_code cache-path collision versus a missing import_utils helper symbol." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "Both are config-loading problems, but one drops fields with trust_remote_code and the other rejects a model missing model_type." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Import performance and sys.modules access are related only broadly; they are not the same concrete defect." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "Different failures: loading a multimodal model versus CI breakage from stale auto_mappings references." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "CI test instability in GraniteSpeech is unrelated to multimodal model loading." + }, + { + "left": "issue:27957", + "right": "issue:35141", + "accept": false, + "reason": "Different model features and failures: Flash Attention 2 support versus embedding reinitialization after resize." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "One is a packaging/backend install bug; the other is a vague dependency issue with no same code path." + }, + { + "left": "issue:46082", + "right": "issue:46121", + "accept": false, + "reason": "Distinct config validation bugs: head_dim validation versus rope-parameter list handling." + }, + { + "left": "issue:42796", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer BPE-marker handling is unrelated to multimodal chat template assistant masks." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and different export surfaces." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer GPU broadcast NoneType error is unrelated to output embedding reinitialization." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Both involve model weights/checkpoints, but one is weight tying and the other is save_pretrained regression." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Different missing imports in different modules; no shared underlying bug." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and stale auto_mappings references are unrelated." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "Pipeline image-processor lookup bug versus a notebook syntax error are clearly unrelated." + }, + { + "left": "issue:44295", + "right": "issue:46506", + "accept": false, + "reason": "Position_ids buffer access bug is unrelated to CPU/CUDA result divergence in PPDocLayoutV3." + }, + { + "left": "issue:43673", + "right": "issue:46424", + "accept": false, + "reason": "Both mention cache, but the defects differ: missing cache in one path versus max_cache_length being honored only for quantized cache." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "The titles are too generic; no evidence these are the same code-path or 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": 24, + "review_unit_index": 40, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:39692", + "issue:42907", + "issue:43295", + "issue:43299", + "issue:43381", + "issue:43441", + "issue:43479", + "issue:43526", + "issue:43582", + "issue:43673", + "issue:43824", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44220", + "issue:44279", + "issue:44327", + "issue:44336", + "issue:44496", + "issue:44521", + "issue:44683", + "issue:44779", + "issue:44792", + "issue:44933", + "issue:44987", + "issue:45042", + "issue:45083", + "issue:45216", + "issue:45335", + "issue:45375", + "issue:45405", + "issue:45446", + "issue:45584", + "issue:45588", + "issue:45593", + "issue:45759", + "issue:46082", + "issue:46123", + "issue:46369", + "issue:46399", + "issue:46424", + "issue:46506", + "issue:46515", + "issue:46693", + "issue:46752", + "issue:46932" + ], + "soft_pairs": [ + "issue:43976|issue:44987", + "issue:45083|issue:46506", + "issue:43582|issue:46932", + "issue:43295|issue:45335", + "issue:44016|issue:44220", + "issue:44496|issue:45042", + "issue:43526|issue:43901", + "issue:43381|issue:45593", + "issue:44683|issue:45446", + "issue:42907|issue:43441", + "issue:43901|issue:44496", + "issue:43673|issue:44336", + "issue:43479|issue:45375", + "issue:43976|issue:44279", + "issue:44496|issue:45759", + "issue:44792|issue:46424", + "issue:44016|issue:44062", + "issue:43441|issue:45588", + "issue:43824|issue:44933", + "issue:46123|issue:46369", + "issue:44521|issue:46693", + "issue:44279|issue:46515", + "issue:44327|issue:46752", + "issue:45335|issue:45584", + "issue:43299|issue:45216", + "issue:46082|issue:46399", + "issue:44279|issue:45405", + "issue:43295|issue:44779", + "issue:39692|issue:43901", + "issue:43976|issue:45405" + ], + "prepared_review_unit_hash": "d4aadf20de926eb81e55b1f56a090441c0f40c9bc29accce83ae46bdfdb0336e", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36078, + "estimated_input_tokens": 9020, + "estimated_eval_tokens": 18296 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: most pairs share only a broad subsystem or release context, not the same underlying bug. No soft edge looks safe to merge as a duplicate.", + "confidence": 0.88, + "canonical_issue_reason": "If a single issue must represent the set, #43976 is the cleanest broad regression report, but the cluster is too mixed to have a strong true canonical issue.", + "canonical_pr_reason": null, + "best_issue_reason": "#43976 is the most representative general version-compatibility regression among these disparate reports, though it is not a duplicate umbrella for the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "Both mention 5.1.0, but one is Python-version incompatibility and the other is a model-loading failure; different root causes." + }, + { + "left": "issue:45083", + "right": "issue:46506", + "accept": false, + "reason": "Different models and failure modes: helper length calculation vs CPU/CUDA output mismatch." + }, + { + "left": "issue:43582", + "right": "issue:46932", + "accept": false, + "reason": "Apple Silicon TypeError in allocator warmup is unrelated to Gemma4 quantized embedder dtype casting." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "One is a processor/tokenizer API regression; the other is embedding resize not propagating to decoder weights." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error and fbank feature extraction bug are unrelated." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "Unrecognized model config and torchvision requirement in PIL image processors are different issues." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "One is a BeitImageProcessorFast label-reduction bug; the other is a docs/behavior mismatch for a pipeline flag." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary-loss behavior are not the same bug." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Both touch flex_attention, but one is a compiled torch>=2.9 failure and the other is a version-check/import guard bug." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Failing to save dequantized Ministrals is distinct from a FlashAttention runtime failure in Ministral-3." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "A docs mismatch about return_all_scores is unrelated to model auto-loading/config recognition." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Generation cache regression and ANSI-color logging output are unrelated." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Both are config issues, but for different models and different missing/default fields." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "Python-version breakage and a generic dependency issue are not the same concrete bug." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Both involve model/config loading, but one is an unrecognized model type and the other is failure to unwrap text_config for composite Qwen multimodal configs." + }, + { + "left": "issue:44792", + "right": "issue:46424", + "accept": false, + "reason": "Janus image-generation test failure and cache-length semantics are unrelated." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error versus AddedToken keyword collision; no shared code path." + }, + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Both mention FlashAttention, but they target different model classes and different null/attribute handling paths." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Missing Qwen2.5-VL import and nonexistent image_utils import are separate API/export problems." + }, + { + "left": "issue:46123", + "right": "issue:46369", + "accept": false, + "reason": "Partial-batch result dropping and memory overflow in sam3-litetext are different failures." + }, + { + "left": "issue:44521", + "right": "issue:46693", + "accept": false, + "reason": "All-zero assistant masks in chat templating is unrelated to a long-context FA2 performance regression." + }, + { + "left": "issue:44279", + "right": "issue:46515", + "accept": false, + "reason": "General dependency trouble and torch 2.4.1 incompatibility are too broad to be the same bug." + }, + { + "left": "issue:44327", + "right": "issue:46752", + "accept": false, + "reason": "QA decode_spans kth-out-of-bounds and empty-list chat_template IndexError are unrelated edge cases." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "Embedding resize propagation and Whisper empty-transcription generation failure are different code paths." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Qwen3VL model-loading regression and Qwen3.5 save_pretrained checkpoint regression are distinct." + }, + { + "left": "issue:46082", + "right": "issue:46399", + "accept": false, + "reason": "LlamaConfig head_dim validation and Qwen3.5 rotary-pos-emb registration are unrelated." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "A generic dependency issue is not the same as a PEFT minimum-version bump problem." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Processor/tokenizer API regression and Deepseek tokenizer output corruption do not share a concrete fix." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "SigLIP2 docs example errors and return_all_scores docs mismatch are separate documentation issues." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Python compatibility regression and unreleased PEFT minimum-version bump 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": 24, + "review_unit_index": 41, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:33290", + "issue:41950", + "issue:42796", + "issue:42994", + "issue:43299", + "issue:43317", + "issue:43441", + "issue:43479", + "issue:43525", + "issue:43577", + "issue:43650", + "issue:43653", + "issue:43854", + "issue:44016", + "issue:44075", + "issue:44242", + "issue:44291", + "issue:44336", + "issue:44351", + "issue:44361", + "issue:44368", + "issue:44462", + "issue:44466", + "issue:44623", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:45020", + "issue:45081", + "issue:45103", + "issue:45200", + "issue:45216", + "issue:45276", + "issue:45278", + "issue:45357", + "issue:45372", + "issue:45542", + "issue:45593", + "issue:45656", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:45859", + "issue:45907", + "issue:46082", + "issue:46209", + "issue:46512", + "issue:46851" + ], + "soft_pairs": [ + "issue:44075|issue:44368", + "issue:43577|issue:45216", + "issue:46082|issue:46512", + "issue:44351|issue:45278", + "issue:44779|issue:45812", + "issue:42796|issue:43650", + "issue:41950|issue:45020", + "issue:43299|issue:45357", + "issue:44466|issue:45276", + "issue:43299|issue:43854", + "issue:43653|issue:45593", + "issue:44242|issue:45200", + "issue:41950|issue:45823", + "issue:43650|issue:46209", + "issue:44462|issue:45081", + "issue:43479|issue:43525", + "issue:44016|issue:44291", + "issue:44683|issue:45542", + "issue:44361|issue:45593", + "issue:45081|issue:45907", + "issue:43441|issue:45834", + "issue:33290|issue:45656", + "issue:45103|issue:45372", + "issue:44336|issue:46851", + "issue:42994|issue:43317", + "issue:44623|issue:44749", + "issue:45859|issue:46851" + ], + "prepared_review_unit_hash": "2df0bab1e9ac84a467e65059473a1af99e7d8a80cfe217d0dc538ee72ce22f12", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35931, + "estimated_input_tokens": 8983, + "estimated_eval_tokens": 18222 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The items are mostly distinct Transformers regression reports, not true duplicates. Several are superficially similar around v5-era loading/saving/tokenizer issues, but they hit different models and code paths, so none of the soft pairs should be merged.", + "confidence": 0.93, + "canonical_issue_reason": "No clear canonical issue: these reports span separate bugs across tokenizers, configs, save/load, offload, and optimizer behavior.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45216 is the best single representative if one is needed: it is a clear, versioned save_pretrained regression with a specific model and failure mode.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "SGD optimizer args being ignored is unrelated to a Qwen3.5 tie_word_embeddings warning." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "Both involve model state/dtype concerns, but one is Blip2 dtype propagation and the other is Qwen3.5 checkpoint serialization." + }, + { + "left": "issue:46082", + "right": "issue:46512", + "accept": false, + "reason": "Explicit head_dim validation and stale config.dtype after from_config are different config bugs." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "HybridCache import failure is a specific symbol breakage, while 45278 is a broader upgrade-induced import error report." + }, + { + "left": "issue:44779", + "right": "issue:45812", + "accept": false, + "reason": "Deepseek token IDs and Granite token IDs are separate tokenizer regressions with different model families and likely different root causes." + }, + { + "left": "issue:42796", + "right": "issue:43650", + "accept": false, + "reason": "Tokenizer BPE-marker handling is unrelated to the nonspecific 'ADD THE DATA' issue." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "A video-classification processor lookup bug is unrelated to remote_code model breakages." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Qwen3VL MoE loading and Qwen3.5 save_pretrained visual key serialization are different failure paths." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Inconsistent lm_head serialization and Gemma4 resize_token_embeddings behavior are distinct save vs resize bugs." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Different models and different loading failures; not the same underlying bug." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "BigBirdTokenizer special-token decoding and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "Missing load-balancing loss and mm_token_type_ids default handling affect different parts of model logic." + }, + { + "left": "issue:41950", + "right": "issue:45823", + "accept": false, + "reason": "Pipeline processor selection and Gemma4 device-mismatch forward errors are unrelated." + }, + { + "left": "issue:43650", + "right": "issue:46209", + "accept": false, + "reason": "The first item is not a concrete bug report, and the second is an FP8 inference-on-B200 issue." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "Ignoring tokenizer.json and crashing in fix_mistral_regex are separate tokenizer-loading problems." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Different config classes and different missing-default fields; not the same bug." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Notebook syntax error and init_empty_weights argument mismatch are unrelated." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "Compiled flex_attention on new torch and a TensorBoard-without-TF backend error are unrelated." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "MLukeTokenizer task AttributeError and D-FINE loss handling are unrelated." + }, + { + "left": "issue:45081", + "right": "issue:45907", + "accept": false, + "reason": "Mistral regex patching and list_repo_files revision forwarding are completely different code paths." + }, + { + "left": "issue:43441", + "right": "issue:45834", + "accept": false, + "reason": "Ministral FlashAttention failure and Kosmos2.5 long-OCR index error are different model-specific bugs." + }, + { + "left": "issue:33290", + "right": "issue:45656", + "accept": false, + "reason": "Adafactor OOM under deepspeed and optimizer step called twice under deepspeed are not the same defect." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "Auto-docstring annotation parsing and Gemma4 processor import failure are unrelated." + }, + { + "left": "issue:44336", + "right": "issue:46851", + "accept": false, + "reason": "ANSI escape leakage in loading_report and Qwen3.5 variable-length position_ids are unrelated." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Quantized-model saving and dequantized-model loading with device_map=auto are different failure modes." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "Missing processor.save_pretrained files and a reported slowdown after upgrade are not the same concrete bug." + }, + { + "left": "issue:45859", + "right": "issue:46851", + "accept": false, + "reason": "Both mention Qwen3.5, but CPU offload incompatibility and variable-length position_ids are different code-path problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 42, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:41628", + "issue:42175", + "issue:43441", + "issue:43525", + "issue:43526", + "issue:43643", + "issue:43673", + "issue:43723", + "issue:43761", + "issue:43784", + "issue:43824", + "issue:43881", + "issue:43901", + "issue:43976", + "issue:43994", + "issue:44016", + "issue:44162", + "issue:44188", + "issue:44246", + "issue:44261", + "issue:44263", + "issue:44297", + "issue:44336", + "issue:44373", + "issue:44496", + "issue:44589", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44779", + "issue:45003", + "issue:45083", + "issue:45103", + "issue:45325", + "issue:45362", + "issue:45397", + "issue:45542", + "issue:45561", + "issue:45563", + "issue:45600", + "issue:45712", + "issue:45800", + "issue:45823", + "issue:46291", + "issue:46399", + "issue:46655", + "issue:46726", + "issue:46851" + ], + "soft_pairs": [ + "issue:44261|issue:44263", + "issue:45600|issue:45712", + "issue:42175|issue:43976", + "issue:44671|issue:44779", + "issue:44623|issue:45397", + "issue:43673|issue:43761", + "issue:43901|issue:43994", + "issue:43723|issue:44162", + "issue:43901|issue:44246", + "issue:43824|issue:43976", + "issue:43441|issue:44683", + "issue:43525|issue:43881", + "issue:46291|issue:46399", + "issue:44188|issue:44336", + "issue:45542|issue:45800", + "issue:44016|issue:45823", + "issue:43525|issue:43643", + "issue:44589|issue:45362", + "issue:44336|issue:44373", + "issue:45325|issue:46851", + "issue:46655|issue:46726", + "issue:41628|issue:43976", + "issue:45083|issue:46851", + "issue:44246|issue:44496", + "issue:43784|issue:45542", + "issue:44297|issue:45397", + "issue:46399|issue:46851", + "issue:45561|issue:45563", + "issue:45103|issue:46399", + "issue:43526|issue:45003" + ], + "prepared_review_unit_hash": "2435a7998784a022af76eb97ac80dd26fbdcb40ceae9394e54f74792e594afd4", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36637, + "estimated_input_tokens": 9160, + "estimated_eval_tokens": 18576 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No true duplicate cluster emerges here: the items cover unrelated bugs, packaging issues, docs, and model-specific regressions. The soft edges are mostly superficial title similarity or shared subsystem vocabulary, so all should be rejected.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:45561 is the strongest standalone bug report here: it is open, well-scoped, has discussion and inbound references, and describes a concrete race in testing utils. It is not, however, representative of the rest of the heterogeneous set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "Different bugs: one is a MLA layernorm precision/config issue, the other is a torch.split return-value handling bug in GlmMoeDsaIndexer." + }, + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Both are repo-maintenance problems, but one is auto_mappings CI breakage while the other is leftover dummy classes leaking into dir(transformers); not the same underlying change." + }, + { + "left": "issue:42175", + "right": "issue:43976", + "accept": false, + "reason": "Packaging/backend dependency issue versus Python version compatibility regression; different failure modes and fixes." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "Different model regressions: CamemBERT masked LM outputs vs Deepseek tokenizer behavior." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "Processor save_pretrained file omission is unrelated to gemma-4 zero3 from_pretrained behavior." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Generation cache/chunked_prefill bug versus CLIPVisionModel hidden_states regression; separate code paths." + }, + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "Docs mismatch for text classification pipeline is not the same as SIGLIP2 model inference producing nonsense results." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "Tokenizer loading regression and ESM2 model breakage are different layers of the stack." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "Pipeline docs/behavior mismatch does not match intermittent import-time slowness." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Qwen2.5-VL missing importable class and Python version support are unrelated issues." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "Ministral FlashAttention failure and compiled flex_attention failure are different model/kernel regressions." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Llama4Config missing attribute and glm-4v-9b loading failure are separate model configuration problems." + }, + { + "left": "issue:46291", + "right": "issue:46399", + "accept": false, + "reason": "Importing pipelines due to kernels package requirements is not the same as missing decorator registration for apply_rotary_pos_emb." + }, + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "torch.compile attention-kernel divergence is unrelated to ANSI code leakage in loading_report." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "Undefined tf backend from optional installs is not the same as a torch/transformers version incompatibility." + }, + { + "left": "issue:44016", + "right": "issue:45823", + "accept": false, + "reason": "Notebook syntax error and Gemma4 PLE device mismatch are unrelated." + }, + { + "left": "issue:43525", + "right": "issue:43643", + "accept": false, + "reason": "Missing pad_token_id on Llama4Config is not the same as trust_remote_code causing missing fields in AutoConfig." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Storage backend/type error and Qwen3.5 chat crash are different problems." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "Terminal ANSI escaping bug versus incorrect docstring text; no shared code defect." + }, + { + "left": "issue:45325", + "right": "issue:46851", + "accept": false, + "reason": "Both involve Qwen3.5/Qwen2.5-VL positional handling, but these are distinct model-specific bugs with different symptoms and fixes." + }, + { + "left": "issue:46655", + "right": "issue:46726", + "accept": false, + "reason": "Incorrect video padding calculation in SmolVLM is not the same as AutoProcessor erroring when vision backend is missing." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "AutoImageProcessor import failure and Python version compatibility are unrelated." + }, + { + "left": "issue:45083", + "right": "issue:46851", + "accept": false, + "reason": "Helper output-length computation bug in qwen3_omni_moe is different from variable-length position_ids handling in Qwen3.5." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Import slowness versus unrecognized model/config loading failure are different issues." + }, + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "Sentence-transformers import NameError and missing TensorFlow backend are unrelated dependency issues." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch is not the same as gemma-4 zero3 loading failure." + }, + { + "left": "issue:46399", + "right": "issue:46851", + "accept": false, + "reason": "Missing decorator for a rotary-pos-emb registration error is separate from variable-length input position_ids bugs." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "Race on captured_info.txt in testing utils and a stale num_return_sequences warning in paged generate() are different behaviors." + }, + { + "left": "issue:45103", + "right": "issue:46399", + "accept": false, + "reason": "auto_docstring annotation handling crash is unrelated to kernel registration/decorator errors." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug and unsafe sys.modules access in modeling_utils are not the same defect." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 43, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:36246", + "issue:41950", + "issue:42491", + "issue:42757", + "issue:42886", + "issue:43257", + "issue:43262", + "issue:43299", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43901", + "issue:44162", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44336", + "issue:44351", + "issue:44451", + "issue:44479", + "issue:44485", + "issue:44496", + "issue:44779", + "issue:44871", + "issue:45003", + "issue:45072", + "issue:45137", + "issue:45200", + "issue:45292", + "issue:45362", + "issue:45375", + "issue:45397", + "issue:45561", + "issue:45588", + "issue:45593", + "issue:45759", + "issue:45859", + "issue:46132", + "issue:46399", + "issue:46506", + "issue:46614", + "issue:46655", + "issue:46665", + "issue:46752", + "issue:46821", + "issue:46829", + "issue:46851" + ], + "soft_pairs": [ + "issue:43540|issue:46752", + "issue:45072|issue:46829", + "issue:46614|issue:46665", + "issue:42491|issue:43299", + "issue:43531|issue:45362", + "issue:44246|issue:44871", + "issue:45588|issue:46851", + "issue:44295|issue:45397", + "issue:44242|issue:45593", + "issue:44336|issue:44485", + "issue:45003|issue:46821", + "issue:42886|issue:46132", + "issue:44451|issue:46752", + "issue:43550|issue:46506", + "issue:44779|issue:45292", + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:43901|issue:45003", + "issue:43257|issue:45137", + "issue:41950|issue:44220", + "issue:45003|issue:45561", + "issue:44496|issue:45003", + "issue:43540|issue:46655", + "issue:45375|issue:45759", + "issue:44162|issue:45200", + "issue:43262|issue:44246", + "issue:44479|issue:46655", + "issue:45859|issue:46399" + ], + "prepared_review_unit_hash": "b7e2d938e808d3b420a1eb3429323678dc390a3c69bc70b5f788a915d1500868", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36046, + "estimated_input_tokens": 9012, + "estimated_eval_tokens": 18280 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose similarity set of unrelated Transformers issues, with only broad thematic overlap around model loading, video/audio processors, and Qwen variants. None of the soft pairs look like the same concrete bug/change, so all are rejected.", + "confidence": 0.93, + "canonical_issue_reason": "Best representative is issue 45003 because it is a core internal bug in modeling_utils with broad potential impact across many loading paths, unlike the rest which are mostly model-specific regressions.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45003 is the most cross-cutting and infrastructure-level item in the set; it is more suitable as a cluster representative than the many narrowly scoped model bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43540", + "right": "issue:46752", + "accept": false, + "reason": "Different code paths: Qwen3OmniMoe video input processing vs empty-chat-template handling. Same broad area, not the same bug." + }, + { + "left": "issue:45072", + "right": "issue:46829", + "accept": false, + "reason": "Unrelated failures: bfloat16 dtype mismatch in two models vs hidden_states field wiring in video classification output." + }, + { + "left": "issue:46614", + "right": "issue:46665", + "accept": false, + "reason": "Stale phonemizer backend state is unrelated to from_pretrained behavior inside Python classes." + }, + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both mention Qwen3/MoE loading, but one is a LoRA compatibility problem and the other is a model-loading breakage in v5; not the same concrete defect." + }, + { + "left": "issue:43531", + "right": "issue:45362", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to Qwen3.5 chat crashes." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Import-time performance and eos_token_id config mismatch are unrelated." + }, + { + "left": "issue:45588", + "right": "issue:46851", + "accept": false, + "reason": "Flash-attention s_aux=None crash is unrelated to variable-length position_ids handling." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "Buffer access after registering position_ids is unrelated to Gemma-4 ZeRO3 from_pretrained loading." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "Both concern losses not being added under a condition, but for different models and loss mechanisms; not mergeable as one bug." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "ANSI code output in loading_report is unrelated to GLM-5 RoPE implementation." + }, + { + "left": "issue:45003", + "right": "issue:46821", + "accept": false, + "reason": "Unsafe sys.modules access is unrelated to TimesFM window_size attribute errors." + }, + { + "left": "issue:42886", + "right": "issue:46132", + "accept": false, + "reason": "Offline tokenizer cache loading is unrelated to AttentionInterface.register behavior changes." + }, + { + "left": "issue:44451", + "right": "issue:46752", + "accept": false, + "reason": "ScandiBERT loading failure is unrelated to empty-list apply_chat_template crashes." + }, + { + "left": "issue:43550", + "right": "issue:46506", + "accept": false, + "reason": "torch.compile+SDPA failure in Bamba is unrelated to CPU/CUDA result divergence in PPDocLayoutV3." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "DeepSeek tokenizer regression is unrelated to resize_token_embeddings not updating output embeddings." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols and different packages; not the same underlying change." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Missing Qwen2_5_VLImageProcessor export and missing HybridCache export are distinct API breakages." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Documentation mismatch for return_all_scores is unrelated to sys.modules access in modeling_utils." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed and Zero3 deque underflow are separate DeepSpeed-related failures." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification image-processor lookup and _torch_extract_fbank_features are different subsystems." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "Core loading bug is unrelated to pytest-xdist file race in testing utils." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model_type in config loading is a model registry issue, not the sys.modules access bug." + }, + { + "left": "issue:43540", + "right": "issue:46655", + "accept": false, + "reason": "Both are video-related, but Qwen3OmniMoe video-input ValueError and SmolVLM padding calculation are model-specific and not evidently the same bug." + }, + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Both involve Qwen 3.5 multimodal config handling, but one is a missing vision config field and the other is from_config text_config unwrapping; different fixes." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "ESM2 breakage and Gemma-4 mm_token_type_ids defaulting are unrelated model-specific issues." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio chat-template sampling-rate default and import performance are unrelated." + }, + { + "left": "issue:44479", + "right": "issue:46655", + "accept": false, + "reason": "Both concern video inputs, but the reported regressions target different model families and different failure modes." + }, + { + "left": "issue:45859", + "right": "issue:46399", + "accept": false, + "reason": "Qwen3.5 CPU offload rotary embedding incompatibility and missing decorator on apply_rotary_pos_emb registration are distinct rotary/kernel 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": 24, + "review_unit_index": 44, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:29870", + "issue:30333", + "issue:36246", + "issue:42199", + "issue:42886", + "issue:43295", + "issue:43352", + "issue:43475", + "issue:43526", + "issue:43540", + "issue:43596", + "issue:43653", + "issue:44038", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44242", + "issue:44295", + "issue:44368", + "issue:44466", + "issue:44479", + "issue:44623", + "issue:44749", + "issue:44857", + "issue:44877", + "issue:44977", + "issue:44987", + "issue:45137", + "issue:45229", + "issue:45250", + "issue:45292", + "issue:45335", + "issue:45375", + "issue:45397", + "issue:45412", + "issue:45459", + "issue:45584", + "issue:45600", + "issue:45656", + "issue:46123", + "issue:46170", + "issue:46369", + "issue:46506", + "issue:46726", + "issue:46832", + "issue:46851", + "issue:46894", + "issue:46897" + ], + "soft_pairs": [ + "issue:43475|issue:46369", + "issue:44295|issue:44749", + "issue:45137|issue:45600", + "issue:45292|issue:45584", + "issue:43295|issue:45292", + "issue:44877|issue:45375", + "issue:44466|issue:45335", + "issue:36246|issue:43596", + "issue:44242|issue:44857", + "issue:44623|issue:46726", + "issue:43526|issue:46897", + "issue:44987|issue:45459", + "issue:44038|issue:46506", + "issue:29870|issue:30333", + "issue:42199|issue:46851", + "issue:43352|issue:45250", + "issue:43526|issue:46894", + "issue:45412|issue:46123", + "issue:43653|issue:44162", + "issue:44479|issue:46170", + "issue:44060|issue:44077", + "issue:44368|issue:44977", + "issue:42886|issue:45229", + "issue:45397|issue:45656", + "issue:44877|issue:46170", + "issue:43475|issue:43540", + "issue:46170|issue:46832" + ], + "prepared_review_unit_hash": "72de5f4c9e64dbed4636a6426c0775e2c0000b5e7e726d9f77f875c932b49390", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35969, + "estimated_input_tokens": 8993, + "estimated_eval_tokens": 18242 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is highly heterogeneous; the soft pairs are topical matches at best, not true duplicates. I would not merge any of the proposed pairs.", + "confidence": 0.96, + "canonical_issue_reason": "No clear canonical issue: the items span unrelated bugs across models, tokenizers, configs, deepspeed, and pipelines.", + "canonical_pr_reason": "No PRs are present in this cluster.", + "best_issue_reason": "No single issue clearly subsumes the rest, so there is no defensible global best issue for duplicate triage.", + "best_pr_reason": "No PRs are present in this cluster.", + "soft_edge_verdicts": [ + { + "left": "issue:43475", + "right": "issue:46369", + "accept": false, + "reason": "Both mention SAM 3, but one is an attribute error in vision encoder output and the other is a memory overflow report; different failures and code paths." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "Unrelated symptoms: position_ids buffer access error vs a performance regression after upgrading Transformers." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque IndexError and broken auto_mappings CI are unrelated issues." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "resize_token_embeddings behavior is unrelated to Whisper empty-transcription generation after align_special_tokens." + }, + { + "left": "issue:43295", + "right": "issue:45292", + "accept": false, + "reason": "Processor/tokenizer regression in v4.57.5 is not the same underlying bug as output_embeddings not updating on resize." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Both involve strict config handling, but they affect different configs and different missing fields/load failures." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Serialization of tied weights on device vs t5gemma resize_token_embeddings not updating decoder.embed_tokens are different code paths." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "Qwen2.5-VL import error and BertModel zero3 IndexError are unrelated." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Load-balancing loss omission in MoE routing is unrelated to an AMP/CUDA crash in LwDetrImageLoss." + }, + { + "left": "issue:44623", + "right": "issue:46726", + "accept": false, + "reason": "Missing files in processor.save_pretrained is different from AutoProcessor failing when a vision backend is absent." + }, + { + "left": "issue:43526", + "right": "issue:46897", + "accept": false, + "reason": "BeitImageProcessorFast reducing labels incorrectly is not the same bug as Florence2 training-loss double shifting." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "Loading failure in transformers>=5.1.0 and protobuf-related tokenizer error masking are different failures." + }, + { + "left": "issue:44038", + "right": "issue:46506", + "accept": false, + "reason": "Qwen3-VL-Moe compatibility issue and CPU/CUDA output mismatch in PPDocLayoutV3 are unrelated." + }, + { + "left": "issue:29870", + "right": "issue:30333", + "accept": false, + "reason": "Bart generate() UnboundLocalError and MLFlow job-state reporting are unrelated." + }, + { + "left": "issue:42199", + "right": "issue:46851", + "accept": false, + "reason": "DETR cardinality error and Qwen3.5 variable-length position_ids issue affect different models and logic." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "A specific model not supporting Flash Attention 2.0 yet is not the same as a generic Flash Attention 2.0 issue." + }, + { + "left": "issue:43526", + "right": "issue:46894", + "accept": false, + "reason": "Label reduction in BeitImageProcessorFast is unrelated to the ASR training-loss double-shift bug." + }, + { + "left": "issue:45412", + "right": "issue:46123", + "accept": false, + "reason": "RT-DETR memory leak and MaskGenerationPipeline partial-batch dropping are different bugs." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "BigBirdTokenizer special-token decode bug and ESM2 breakage are not the same underlying defect." + }, + { + "left": "issue:44479", + "right": "issue:46170", + "accept": false, + "reason": "Qwen video-input regression and Windows ROCm/distributed import failure are unrelated." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "Tied-weights warning in Qwen3-Next and patchtsmixer post_init policy are different model/config issues." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "A warning about tie_word_embeddings and a flash-attention generation failure are not the same bug." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Offline tokenizer cache loading and Gemma4 multi-GPU OOM are unrelated." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Gemma-4 ZeRO3 from_pretrained failure and deepspeed optimizer double-step are different code paths." + }, + { + "left": "issue:44877", + "right": "issue:46170", + "accept": false, + "reason": "Strict config loading and Windows ROCm torch.distributed dependency errors are unrelated." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "SAM 3 video encoder attribute error and Qwen3OmniMoe video ValueError are different model families and failures." + }, + { + "left": "issue:46170", + "right": "issue:46832", + "accept": false, + "reason": "Windows ROCm import/type error and RT-DETR feature-level TypeError are unrelated 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": 24, + "review_unit_index": 45, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:42994", + "issue:43065", + "issue:43232", + "issue:43408", + "issue:43441", + "issue:43475", + "issue:43482", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43701", + "issue:43723", + "issue:43906", + "issue:43931", + "issue:43994", + "issue:44162", + "issue:44188", + "issue:44297", + "issue:44361", + "issue:44368", + "issue:44483", + "issue:44623", + "issue:44749", + "issue:44871", + "issue:44995", + "issue:45103", + "issue:45200", + "issue:45375", + "issue:45478", + "issue:45563", + "issue:45588", + "issue:45656", + "issue:45685", + "issue:45987", + "issue:46082", + "issue:46123", + "issue:46132", + "issue:46202", + "issue:46421", + "issue:46431", + "issue:46710", + "issue:46726", + "issue:46752", + "issue:46821", + "issue:46851", + "issue:46858" + ], + "soft_pairs": [ + "issue:43065|issue:43408", + "issue:43723|issue:46710", + "issue:44368|issue:45563", + "issue:44995|issue:45200", + "issue:43931|issue:44368", + "issue:45987|issue:46082", + "issue:44188|issue:46202", + "issue:44483|issue:46202", + "issue:43906|issue:46202", + "issue:44188|issue:46421", + "issue:44995|issue:45588", + "issue:45685|issue:46431", + "issue:45103|issue:46858", + "issue:43065|issue:46123", + "issue:44361|issue:44749", + "issue:44297|issue:44623", + "issue:43994|issue:46726", + "issue:44749|issue:45656", + "issue:43701|issue:44749", + "issue:44871|issue:46132", + "issue:43646|issue:44162", + "issue:44623|issue:45656", + "issue:43441|issue:46752", + "issue:43475|issue:43643", + "issue:42994|issue:43482", + "issue:28282|issue:30990", + "issue:45375|issue:45478", + "issue:44188|issue:46851", + "issue:43475|issue:46821", + "issue:43232|issue:43673" + ], + "prepared_review_unit_hash": "21e0208a65cc23330133d4b9ce07645c9d371230ccfbdb8a164e668e0c5723bf", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36783, + "estimated_input_tokens": 9196, + "estimated_eval_tokens": 18648 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a loose collection of unrelated Transformers issues spanning model loading, generation/cache, serialization, and model-specific regressions. No pull requests are present. I found no soft pair that is clearly the same underlying bug with enough specificity to merge, so I kept all soft edges rejected.", + "confidence": 0.91, + "canonical_issue_reason": "If forced to pick one representative issue, 46858 is the clearest concrete core-runtime regression: it names a specific failing path (_prefill/cache.get_seq_length on static cache with torch.compile on non-CUDA backends) rather than a broad symptom.", + "canonical_pr_reason": null, + "best_issue_reason": "46858 is the best global representative because it is precise, actionable, and tied to a core cache/generation bug that is easy to understand and likely to have high impact.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "Both are SAM3-related, but one is a dummy Conv2d implementation choice and the other is a config/model-type mismatch warning; different bug classes." + }, + { + "left": "issue:43723", + "right": "issue:46710", + "accept": false, + "reason": "Tokenizer loading failure and an accuracy regression after upgrade are unrelated failure modes." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "Both mention warnings, but one is about tie_word_embeddings and the other is a stale num_return_sequences warning; not the same issue." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "Stale indexer cache on second forward pass and missing mm_token_type_ids defaults are different problems." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Weight-shape mismatch for Qwen3-VL is unrelated to a tie_word_embeddings warning." + }, + { + "left": "issue:45987", + "right": "issue:46082", + "accept": false, + "reason": "StaticCache seq_length return-type bug and LlamaConfig head_dim validation are separate code paths." + }, + { + "left": "issue:44188", + "right": "issue:46202", + "accept": false, + "reason": "torch.compile attention-kernel branching and Gemma pipeline roundtrip behavior are unrelated." + }, + { + "left": "issue:44483", + "right": "issue:46202", + "accept": false, + "reason": "Chat-completions API acceptance failure is not the same as a Gemma pipeline token roundtrip bug." + }, + { + "left": "issue:43906", + "right": "issue:46202", + "accept": false, + "reason": "An isolated reproduction of another issue is not the same as a Gemma pipeline failure." + }, + { + "left": "issue:44188", + "right": "issue:46421", + "accept": false, + "reason": "Both involve cache/compile themes, but one is divergent attention kernels from branching and the other is spurious recompilation from lazy cache init; not the same bug." + }, + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "Second-forward stale cache and flash_attention s_aux=None crash are different failures." + }, + { + "left": "issue:45685", + "right": "issue:46431", + "accept": false, + "reason": "MPS histogram dtype support and missing chat template on a processor are unrelated." + }, + { + "left": "issue:45103", + "right": "issue:46858", + "accept": false, + "reason": "Docstring helper crash under future annotations and static-cache seq_length/device-tensor arithmetic are unrelated." + }, + { + "left": "issue:43065", + "right": "issue:46123", + "accept": false, + "reason": "Dummy Conv2d in SAM3 and MaskGenerationPipeline final-batch handling are different issues." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "Tokenizer AttributeError and a performance regression during filtering are unrelated." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "Tokenizer config mismatch on save and missing processor files on save are both serialization-related, but they are different APIs and symptoms; not enough to call them the same bug." + }, + { + "left": "issue:43994", + "right": "issue:46726", + "accept": false, + "reason": "Nonsensical SigLIP2 model outputs and missing vision-backend image processor handling are different problems." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "Data filtering slowdown and duplicated optimizer steps under deepspeed are unrelated." + }, + { + "left": "issue:43701", + "right": "issue:44749", + "accept": false, + "reason": "Checkpoint key mismatch and a performance regression do not describe the same failure." + }, + { + "left": "issue:44871", + "right": "issue:46132", + "accept": false, + "reason": "Gemma eos_token_id inconsistency and AttentionInterface.register behavior change are unrelated." + }, + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "A general v5 custom-model-init breakage and ESM2-specific breakage are not the same underlying bug." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "Missing processor save files and doubled deepspeed optimizer steps are different subsystems and symptoms." + }, + { + "left": "issue:43441", + "right": "issue:46752", + "accept": false, + "reason": "FlashAttention failure for Ministral-3 and empty-list apply_chat_template IndexError are unrelated." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "Missing fpn_position_embeddings on Sam3 output and missing config fields with trust_remote_code are different issues." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Quantized model saving failure and GGUF loading failure are opposite directions and distinct bugs." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "ImportError for missing PyTorch and Sentence Transformers hanging on load are unrelated." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "These are related in model family and loading path, but one is a specific config field being dropped while the other is a broad from_pretrained error; not enough evidence they are the same concrete bug." + }, + { + "left": "issue:44188", + "right": "issue:46851", + "accept": false, + "reason": "Attention-kernel branching under torch.compile and Qwen3.5 variable-length position_ids handling are different failures." + }, + { + "left": "issue:43475", + "right": "issue:46821", + "accept": false, + "reason": "Sam3 vision-encoder missing field and TimesFM window_size AttributeError are unrelated." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "A generation kwargs update bug after sync_gpus and a missing cache during chunked_prefill are both generation-related but clearly different code paths." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 46, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:36010", + "issue:39692", + "issue:41628", + "issue:42371", + "issue:42673", + "issue:42831", + "issue:42994", + "issue:43278", + "issue:43504", + "issue:43525", + "issue:43704", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43873", + "issue:43986", + "issue:44038", + "issue:44222", + "issue:44351", + "issue:44373", + "issue:44623", + "issue:44857", + "issue:44869", + "issue:44933", + "issue:44945", + "issue:44964", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45230", + "issue:45310", + "issue:45356", + "issue:45375", + "issue:45397", + "issue:45406", + "issue:45561", + "issue:45593", + "issue:45753", + "issue:46202", + "issue:46326", + "issue:46399", + "issue:46463", + "issue:46701", + "issue:46726", + "issue:46851", + "issue:46877" + ], + "soft_pairs": [ + "issue:44964|issue:45137", + "issue:44869|issue:45356", + "issue:43986|issue:46726", + "issue:45310|issue:45375", + "issue:41628|issue:44933", + "issue:46463|issue:46877", + "issue:39692|issue:46726", + "issue:45230|issue:45397", + "issue:44038|issue:44623", + "issue:44373|issue:46851", + "issue:43723|issue:46326", + "issue:42371|issue:45092", + "issue:42994|issue:43525", + "issue:43873|issue:44945", + "issue:45406|issue:46202", + "issue:43784|issue:46399", + "issue:44945|issue:45753", + "issue:44857|issue:46877", + "issue:45593|issue:46877", + "issue:41628|issue:44351", + "issue:36010|issue:41628", + "issue:43278|issue:45137", + "issue:45081|issue:45103", + "issue:46202|issue:46701", + "issue:44038|issue:45310", + "issue:42673|issue:43704", + "issue:44222|issue:45310", + "issue:42994|issue:43824", + "issue:42831|issue:44222", + "issue:43504|issue:45561" + ], + "prepared_review_unit_hash": "81b1ab48c68a08b796c4a8e08b4cef0ffc0063c2a26bc3152ce3fa35570df57b", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 35841, + "estimated_input_tokens": 8961, + "estimated_eval_tokens": 18178 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous and does not form a broad duplicate set. The only clear duplicate is the repeated Qwen3ForCausalLM VRAM leak report (issue:42673 and issue:43704); the rest are separate bugs spanning imports, tokenizers, quantization, processors, and model-specific regressions.", + "confidence": 0.92, + "canonical_issue_reason": "issue:42673 is the strongest anchor because it has a true duplicate in issue:43704 with the exact same title and bug description, while also having much richer discussion context.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42673 is the best representative issue for the only confirmed duplicate pair in this cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Different failures: Phi-4 multimodal model loading vs DeepSpeed ZeRO3 deque underflow." + }, + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Both involve tokenization, but one is a Whisper timestamp decode crash and the other is a Kimi tokenizer regression with mistral regex handling." + }, + { + "left": "issue:43986", + "right": "issue:46726", + "accept": false, + "reason": "Both mention processor/backend availability, but they are different code paths and errors (torchvision-missing video model vs unrecognized image processor)." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "Same model family, but one is a from_pretrained error and the other is a missing config field silently dropped; not enough to treat as the same bug." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Both are import errors, but they concern different symbols and modules; no evidence of one underlying fix." + }, + { + "left": "issue:46463", + "right": "issue:46877", + "accept": false, + "reason": "Unrelated: T5 decoder-input validation message vs auto_docstring import-time diagnostics output." + }, + { + "left": "issue:39692", + "right": "issue:46726", + "accept": false, + "reason": "SigLIP2 documentation/example errors are not the same bug as AutoProcessor backend-missing runtime handling." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "Generic bug report vs Gemma-4 ZeRO3 from_pretrained failure; no shared underlying code-path." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe load incompatibility and processor.save_pretrained missing files are separate issues." + }, + { + "left": "issue:44373", + "right": "issue:46851", + "accept": false, + "reason": "Docstring mismatch for position_ids is unrelated to a runtime variable-length input bug." + }, + { + "left": "issue:43723", + "right": "issue:46326", + "accept": false, + "reason": "Tokenizer loading in v5 and continue_final_message for mistral_common are different bugs despite both touching tokenization." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 API settings guidance and InternVL2 meta-initialization incompatibility are unrelated." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized model saving failure and missing pad_token_id on Llama4Config are different defects." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Offloading/quantization behavior and pipeline-parallel incorrect output are separate issues." + }, + { + "left": "issue:45406", + "right": "issue:46202", + "accept": false, + "reason": "Gemma processor missing _tokenizer and Gemma roundtrip on \u2581 are different failure modes." + }, + { + "left": "issue:43784", + "right": "issue:46399", + "accept": false, + "reason": "Sentence-transformers import NameError and Qwen3.5 kernel registration crash are unrelated." + }, + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Pipeline parallelism output corruption and Qwen3.5 infinite-loop-on-image are different bugs." + }, + { + "left": "issue:44857", + "right": "issue:46877", + "accept": false, + "reason": "LwDetrImageLoss AMP crash and auto_docstring stdout diagnostics are unrelated." + }, + { + "left": "issue:45593", + "right": "issue:46877", + "accept": false, + "reason": "D-FINE auxiliary-loss behavior is unrelated to auto_docstring import-time logging." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Both are import/export regressions, but they involve different missing symbols and different modules." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "GenerationMixin import failure and AutoImageProcessor import failure are different missing-export regressions." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Embedding dtype drift in evaluation is unrelated to a ZeRO3 deque error." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "Tokenizer regex patch crash and auto_docstring future-annotations crash are distinct code paths." + }, + { + "left": "issue:46202", + "right": "issue:46701", + "accept": false, + "reason": "Both involve Gemma, but the pipeline roundtrip bug and the image-text-to-text pipeline failure are not the same underlying issue." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Exact same VRAM leak report for Qwen3ForCausalLM in multiple dataloader threads; clearly the same bug." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "FP8 save_pretrained moe and Qwen3.5 from_pretrained error are different operations and failures." + }, + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Quantized model saving failure is unrelated to Qwen2_5_VL import failure." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "Both mention FP8, but one is an accuracy problem and the other is a saving bug; not the same fix." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Legacy-field model loading bug and pytest-xdist file race are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: the only clearly duplicate pair is issue:42673 and issue:43704, which share the exact same title, and the discussion-count contrast supports the canonical anchor choice. The other soft-edge rejections are appropriately conservative and do not overclaim equivalence from weak title similarity alone." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 47, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:27957", + "issue:30333", + "issue:30990", + "issue:33290", + "issue:34689", + "issue:35141", + "issue:39692", + "issue:42175", + "issue:42222", + "issue:42673", + "issue:42898", + "issue:43232", + "issue:43295", + "issue:43479", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43976", + "issue:44016", + "issue:44079", + "issue:44496", + "issue:44568", + "issue:44623", + "issue:44625", + "issue:44734", + "issue:44779", + "issue:44792", + "issue:44869", + "issue:44933", + "issue:45092", + "issue:45103", + "issue:45200", + "issue:45310", + "issue:45356", + "issue:45405", + "issue:45561", + "issue:45600", + "issue:45712", + "issue:45812", + "issue:46077", + "issue:46123", + "issue:46231", + "issue:46693", + "issue:46829", + "issue:46877", + "issue:46909" + ], + "soft_pairs": [ + "issue:43232|issue:44792", + "issue:45561|issue:45600", + "issue:42175|issue:44933", + "issue:44496|issue:45561", + "issue:45405|issue:46877", + "issue:43976|issue:44933", + "issue:30333|issue:33290", + "issue:30990|issue:35141", + "issue:42898|issue:44568", + "issue:42175|issue:45712", + "issue:43479|issue:46829", + "issue:46077|issue:46123", + "issue:44623|issue:45310", + "issue:42673|issue:46693", + "issue:43479|issue:44869", + "issue:42898|issue:45812", + "issue:43723|issue:44568", + "issue:44016|issue:46231", + "issue:44625|issue:46877", + "issue:39692|issue:46909", + "issue:42898|issue:44779", + "issue:43479|issue:44079", + "issue:43784|issue:43824", + "issue:44734|issue:45103", + "issue:43874|issue:45092", + "issue:27957|issue:30333", + "issue:42898|issue:43295", + "issue:30333|issue:34689", + "issue:42222|issue:45200", + "issue:42898|issue:45356" + ], + "prepared_review_unit_hash": "fc90859cc6d22c238cdeda93e0f9c6dabb80094b3d5c808020df59e56683eb80", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36369, + "estimated_input_tokens": 9093, + "estimated_eval_tokens": 18442 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous; none of the soft pairs look like the same underlying bug/change. I\u2019d keep issue 45812 as the best representative standalone issue because it describes a broad, high-impact silent tokenizer regression.", + "confidence": 0.81, + "canonical_issue_reason": "Issue 45812 is the strongest canonical issue here: it reports a broad, concrete regression affecting all Granite models, with clear user impact and a precise failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45812 is also the best standalone issue in the cluster: it is actionable, high-impact, and more central than the narrower infra/docs/config reports around it.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Different problems: generation kwargs/sync_gpus handling vs janus image-generation test failure." + }, + { + "left": "issue:45561", + "right": "issue:45600", + "accept": false, + "reason": "Unrelated: xdist race on captured_info.txt vs broken auto_mappings references to removed configs." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Different code paths: missing TensorFlow backend in install extras vs nonexistent image_utils import." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Unrelated model loading/config error vs testing-utils file race." + }, + { + "left": "issue:45405", + "right": "issue:46877", + "accept": false, + "reason": "Package version bump issue vs auto_docstring stdout noise; not the same bug." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python version compatibility problem vs image_utils import error." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "MLflow reporting on training failure vs deepspeed adafactor OOM; different symptoms and fixes." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "Loading hang in Sentence Transformers vs embedding reinitialization after resize; not the same failure." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Both are tokenizer regressions in v5, but they are different behaviors and model-specific bugs." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Install/backend dependency issue vs leftover dummy classes leaking into namespace/check_repo." + }, + { + "left": "issue:43479", + "right": "issue:46829", + "accept": false, + "reason": "Multimodal config initialization bug vs classifier output field misassignment." + }, + { + "left": "issue:46077", + "right": "issue:46123", + "accept": false, + "reason": "Causal mask construction bug vs batch-finalization bug in MaskGenerationPipeline; unrelated." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Saving processor artifacts vs Qwen3.5 MoE from_pretrained load failure; different paths." + }, + { + "left": "issue:42673", + "right": "issue:46693", + "accept": false, + "reason": "VRAM leak with multiple dataloader threads vs FA2 long-context performance regression; not the same defect." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Different multimodal/model bugs: config defaults vs Whisper word-timestamp decode crash." + }, + { + "left": "issue:42898", + "right": "issue:45812", + "accept": false, + "reason": "Both concern v5 tokenizer behavior, but one is a generic cleanup-space change and the other is a Granite-specific token-id regression." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "AutoTokenizer loading failure vs missing BOS/EOS insertion for one tokenizer model." + }, + { + "left": "issue:44016", + "right": "issue:46231", + "accept": false, + "reason": "Notebook syntax error vs docs parameter-name mismatch." + }, + { + "left": "issue:44625", + "right": "issue:46877", + "accept": false, + "reason": "Qwen3.5 num_labels propagation bug vs docstring import-time stdout diagnostics." + }, + { + "left": "issue:39692", + "right": "issue:46909", + "accept": false, + "reason": "SigLIP2 example errors vs DIA documentation example breakage; different docs and failures." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Tokenizer behavior change in v5 vs Deepseek tokenizer producing incorrect results; related area but not the same bug." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Multimodal config defaults vs ModelOutput key assignment logic; unrelated." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Different import/loading issues: sentence-transformers NameError vs missing Qwen2_5_VL export." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Generation/KV-cache indexing bug vs auto_docstring crash with future annotations; unrelated." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Image processor API missing method vs old InternVL2 remote-code checkpoint/meta-init incompatibility." + }, + { + "left": "issue:27957", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention 2 model support vs MLflow job-reporting bug; no shared underlying change." + }, + { + "left": "issue:42898", + "right": "issue:43295", + "accept": false, + "reason": "Tokenizer cleanup behavior in v5 vs regression in custom code accessing processor.tokenizer/passing images; different breakages." + }, + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "MLflow reporting bug vs Llama 3.2 Vision model-loading regression; unrelated." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "Broken VitPose model files vs Gemma 4 multimodal token-type-id defaulting." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Generic v5 tokenizer cleanup change vs Kimi-K2.5 codec/regex regression; same subsystem but different concrete bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 48, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:22355", + "issue:28282", + "issue:30333", + "issue:36331", + "issue:41628", + "issue:42757", + "issue:42796", + "issue:42831", + "issue:43116", + "issue:43122", + "issue:43232", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:44060", + "issue:44075", + "issue:44190", + "issue:44222", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44448", + "issue:44484", + "issue:44485", + "issue:44623", + "issue:44933", + "issue:44977", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45245", + "issue:45310", + "issue:45405", + "issue:45656", + "issue:45712", + "issue:45834", + "issue:46082", + "issue:46132", + "issue:46209", + "issue:46231", + "issue:46424", + "issue:46614", + "issue:46710", + "issue:46762", + "issue:46932" + ], + "soft_pairs": [ + "issue:44295|issue:45656", + "issue:43116|issue:44190", + "issue:30333|issue:36331", + "issue:45405|issue:45712", + "issue:22355|issue:28282", + "issue:44222|issue:44297", + "issue:44060|issue:45125", + "issue:44075|issue:46231", + "issue:44448|issue:46132", + "issue:44933|issue:45712", + "issue:42831|issue:44623", + "issue:44360|issue:46762", + "issue:42796|issue:43723", + "issue:44297|issue:45656", + "issue:45245|issue:46132", + "issue:44222|issue:44623", + "issue:43479|issue:43644", + "issue:46132|issue:46710", + "issue:41628|issue:42757", + "issue:44373|issue:44485", + "issue:44297|issue:45310", + "issue:45071|issue:46082", + "issue:43723|issue:44977", + "issue:46614|issue:46932", + "issue:44222|issue:44484", + "issue:43122|issue:46209", + "issue:44190|issue:46082", + "issue:44263|issue:45092", + "issue:43232|issue:46424", + "issue:43232|issue:45834" + ], + "prepared_review_unit_hash": "f6410a1fa95f9236a1de11a57a86ed1b41f7942735395b3bd9e91575a671e691", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 35760, + "estimated_input_tokens": 8940, + "estimated_eval_tokens": 18136 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: most items are unrelated import, training, tokenizer, config, and serialization bugs. None of the soft pairs look like the same underlying change or defect, so all soft edges are rejected.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue stands out; the issues span multiple unrelated subsystems and failure modes, so there is no clear duplicate representative.", + "canonical_pr_reason": null, + "best_issue_reason": "No issue is a good global representative because the cluster is not a duplicate set; the closest-looking items still differ in concrete code path and symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "Position_ids buffer read error is unrelated to DeepSpeed calling optimizer.step twice." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "Empty multi-label results in an example script is not the same bug as failing to load a local dataset." + }, + { + "left": "issue:30333", + "right": "issue:36331", + "accept": false, + "reason": "MLflow reporting on failed training and CustomTrainer.compute_loss signature mismatch are different problems." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "A PEFT version bump and leftover dummy classes failing repo checks are unrelated." + }, + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Missing transformers.onnx and missing PyTorch for AutoModel are distinct import errors." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "Both involve save_pretrained, but one is FP8 MoE saving and the other is tokenizer_class metadata mismatch." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Qwen3-Next tied-weights warning and missing _tp_plan for Qwen3_5MoE are different model bugs." + }, + { + "left": "issue:44075", + "right": "issue:46231", + "accept": false, + "reason": "Unused SGD args and a docstring parameter name mismatch are not the same defect." + }, + { + "left": "issue:44448", + "right": "issue:46132", + "accept": false, + "reason": "Pegasus v4/v5 output regression and AttentionInterface.register behavior change are separate code paths." + }, + { + "left": "issue:44933", + "right": "issue:45712", + "accept": false, + "reason": "A missing import from image_utils is unrelated to dummy classes leaking into dir(transformers)." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FineGrainedFP8 accuracy issues are not the same as processor.save_pretrained missing files." + }, + { + "left": "issue:44360", + "right": "issue:46762", + "accept": false, + "reason": "A missing ReLU in the DSA indexer is not the same as Minimax MSA numerical drift." + }, + { + "left": "issue:42796", + "right": "issue:43723", + "accept": false, + "reason": "Tokenizer BPE marker handling and AutoTokenizer.from_pretrained loading failures are different issues." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch is unrelated to DeepSpeed optimizer stepping." + }, + { + "left": "issue:45245", + "right": "issue:46132", + "accept": false, + "reason": "Category-count limits and AttentionInterface.register behavior are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "FP8 MoE save_pretrained and processor.save_pretrained missing files are different serialization bugs." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Phi4MultimodalConfig None-handling and non-persistent buffer junk are unrelated initialization problems." + }, + { + "left": "issue:46132", + "right": "issue:46710", + "accept": false, + "reason": "AttentionInterface.register behavior change does not match the DeepSeek accuracy regression." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Missing AutoImageProcessor from transformers and missing is_offline_mode from huggingface_hub are distinct import issues." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "A wrong docstring for position_ids is not the same as a GLM-5 RoPE implementation bug." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch and Qwen3.5 MoE from_pretrained error are unrelated." + }, + { + "left": "issue:45071", + "right": "issue:46082", + "accept": false, + "reason": "PretrainedConfig type checking breakage and LlamaConfig head_dim validation are different config bugs." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer loading failure and Qwen3.5 flash-attention generation regression are not the same bug." + }, + { + "left": "issue:46614", + "right": "issue:46932", + "accept": false, + "reason": "Stale phonemizer backend state and Gemma4 quantized vision/audio dtype casting are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "Both mention save_pretrained, but one is FP8 MoE saving and the other is the max_shard_size default." + }, + { + "left": "issue:43122", + "right": "issue:46209", + "accept": false, + "reason": "Tokenization changes across versions and FP8 inference failure on B200 are unrelated." + }, + { + "left": "issue:44190", + "right": "issue:46082", + "accept": false, + "reason": "Local dataset loading failure is not the same as LlamaConfig head_dim validation." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "GlmMoeDsaIndexer torch.split output handling and remote-code meta init incompatibility are different issues." + }, + { + "left": "issue:43232", + "right": "issue:46424", + "accept": false, + "reason": "Generation kwargs update after sync_gpus and cache max length handling are unrelated cache/generation behaviors." + }, + { + "left": "issue:43232", + "right": "issue:45834", + "accept": false, + "reason": "Generation kwargs update and Kosmos2.5 OCR index errors are not the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 49, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:28218", + "issue:30990", + "issue:38175", + "issue:39401", + "issue:41628", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43066", + "issue:43232", + "issue:43408", + "issue:43475", + "issue:43644", + "issue:43824", + "issue:44038", + "issue:44188", + "issue:44261", + "issue:44263", + "issue:44279", + "issue:44295", + "issue:44360", + "issue:44373", + "issue:44483", + "issue:44568", + "issue:44623", + "issue:44829", + "issue:45092", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45381", + "issue:45406", + "issue:45593", + "issue:45753", + "issue:46002", + "issue:46050", + "issue:46209", + "issue:46305", + "issue:46328", + "issue:46369", + "issue:46402", + "issue:46431", + "issue:46655", + "issue:46909" + ], + "soft_pairs": [ + "issue:43475|issue:46305", + "issue:46209|issue:46431", + "issue:44295|issue:44623", + "issue:46305|issue:46655", + "issue:30990|issue:38175", + "issue:44263|issue:44360", + "issue:44263|issue:44373", + "issue:45092|issue:45357", + "issue:42907|issue:46050", + "issue:42831|issue:45310", + "issue:44279|issue:46328", + "issue:46369|issue:46909", + "issue:45381|issue:46305", + "issue:44483|issue:45406", + "issue:22355|issue:27957", + "issue:43644|issue:46305", + "issue:44261|issue:45092", + "issue:44829|issue:46002", + "issue:42831|issue:44038", + "issue:41628|issue:43824", + "issue:43408|issue:46402", + "issue:42831|issue:45753", + "issue:22355|issue:28218", + "issue:39401|issue:44568", + "issue:42886|issue:43066", + "issue:27957|issue:30990", + "issue:44188|issue:44373", + "issue:42886|issue:42994", + "issue:42757|issue:45278", + "issue:43408|issue:46369", + "issue:43232|issue:45593" + ], + "prepared_review_unit_hash": "40b4395056d22cefa8ec528b1e92fd03f3731fb13b83be3a6a56ec9c579d887c", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36516, + "estimated_input_tokens": 9129, + "estimated_eval_tokens": 18514 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft pair looks like the same underlying bug/change closely enough to merge. The set is mostly unrelated issues within Transformers, often sharing only broad themes like tokenizers, imports, model loading, or serialization.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:46305 is the most specific, concrete, and actionable bug report in the set (a precise off-by-one in sliding_window_mask_function with clear expected vs actual behavior).", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43475", + "right": "issue:46305", + "accept": false, + "reason": "Different SAM3 memory/attribute failure vs MoonshineStreaming mask off-by-one; unrelated code paths." + }, + { + "left": "issue:46209", + "right": "issue:46431", + "accept": false, + "reason": "FP8 inference on B200 vs missing chat template error; no shared bug mechanism." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "Buffer registration/readback issue vs processor save_pretrained file omission; both serialization-adjacent but not the same defect." + }, + { + "left": "issue:46305", + "right": "issue:46655", + "accept": false, + "reason": "Different streaming/video processor logic and different off-by-one/padding behavior." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "Loading stall vs zero probabilities; different runtime symptoms and model paths." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "DSA indexer split behavior vs lack of ReLU in a different indexer; not the same bug." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "Indexer torch.split values vs docstring typo for position_ids; one is functional, one documentation-related." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "Meta initialization incompatibility vs incorrect visual encoder keys on save_pretrained; distinct save/load issues." + }, + { + "left": "issue:42907", + "right": "issue:46050", + "accept": false, + "reason": "Saving dequantized Ministral/Devstral models vs sinq quantized model round-trip breakage; different quantization pipelines." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "FineGrainedFP8 accuracy regression vs Qwen3.5 from_pretrained error; different failure modes." + }, + { + "left": "issue:44279", + "right": "issue:46328", + "accept": false, + "reason": "Generic dependency issue vs pad performance inefficiency; unrelated." + }, + { + "left": "issue:46369", + "right": "issue:46909", + "accept": false, + "reason": "SAM3-LiteText memory overflow vs documentation example bug; unrelated." + }, + { + "left": "issue:45381", + "right": "issue:46305", + "accept": false, + "reason": "Qwen2.5-VL video position_ids bug vs MoonshineStreaming mask off-by-one; different models and code paths." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "Chat-completions request rejection vs Gemma4Processor missing _tokenizer; unrelated API failures." + }, + { + "left": "issue:22355", + "right": "issue:27957", + "accept": false, + "reason": "onnx import failure vs XLMRoberta Flash Attention 2 behavior; unrelated." + }, + { + "left": "issue:43644", + "right": "issue:46305", + "accept": false, + "reason": "Non-persistent buffer junk in v5.0 vs streaming mask off-by-one; different subsystems." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "MLA layernorm epsilon precision issue vs remote-code/meta-init incompatibility; not the same defect." + }, + { + "left": "issue:44829", + "right": "issue:46002", + "accept": false, + "reason": "Flash Attention 3 degenerate training vs NaNs after checkout+init; different training-path problems." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "FP8 accuracy issue vs Qwen3-VL-Moe from_pretrained error; not the same bug." + }, + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Two different import errors for different symbols; same broad area but not one underlying fix." + }, + { + "left": "issue:43408", + "right": "issue:46402", + "accept": false, + "reason": "sam3_tracker class mismatch warning vs sam3 model text encoder weights not loaded; different SAM3 issues." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "FP8 accuracy regression vs infinite loop on a specific image; unrelated symptoms and causes." + }, + { + "left": "issue:22355", + "right": "issue:28218", + "accept": false, + "reason": "onnx import failure vs tokenizer added-space behavior; unrelated tokenizer/import topics." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "offset_mapping bug vs add_special_tokens not adding BOS/EOS; both tokenizer issues but different mechanics." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Offline cache load failure vs wrong tokenizer decoder type in v5; different tokenizer regressions." + }, + { + "left": "issue:27957", + "right": "issue:30990", + "accept": false, + "reason": "Flash Attention 2 model support vs Sentence Transformers loading stall; unrelated." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "torch.compile attention-kernel divergence vs docstring error; not the same kind of bug." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Offline tokenizer cache loading vs quantized model saving failure; unrelated save/load paths." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "Single missing huggingface_hub symbol vs many import errors after upgrade; same ecosystem, not same underlying issue." + }, + { + "left": "issue:43408", + "right": "issue:46369", + "accept": false, + "reason": "SAM3 model-type warning vs SAM3-LiteText memory overflow; different failures." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation kwargs update after sync_gpus vs D-FINE auxiliary loss logic; 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": 24, + "review_unit_index": 50, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:22355", + "issue:29159", + "issue:30990", + "issue:33357", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42796", + "issue:42907", + "issue:43122", + "issue:43408", + "issue:43452", + "issue:43475", + "issue:43540", + "issue:43576", + "issue:43646", + "issue:43756", + "issue:43784", + "issue:43937", + "issue:43976", + "issue:44038", + "issue:44060", + "issue:44117", + "issue:44222", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44485", + "issue:44488", + "issue:44521", + "issue:44625", + "issue:44704", + "issue:44829", + "issue:44945", + "issue:44964", + "issue:45216", + "issue:45278", + "issue:45290", + "issue:45381", + "issue:45478", + "issue:45600", + "issue:46002", + "issue:46172", + "issue:46326", + "issue:46614", + "issue:46762", + "issue:46909" + ], + "soft_pairs": [ + "issue:42796|issue:43122", + "issue:43646|issue:46002", + "issue:45381|issue:46909", + "issue:43540|issue:46909", + "issue:30990|issue:33357", + "issue:42907|issue:44945", + "issue:42907|issue:44964", + "issue:39401|issue:44488", + "issue:44295|issue:44297", + "issue:43976|issue:44945", + "issue:43576|issue:43784", + "issue:42371|issue:44263", + "issue:43756|issue:44060", + "issue:46172|issue:46762", + "issue:42907|issue:43937", + "issue:41628|issue:45278", + "issue:43784|issue:45600", + "issue:44117|issue:44704", + "issue:39401|issue:44521", + "issue:44625|issue:44829", + "issue:44704|issue:45290", + "issue:43408|issue:43475", + "issue:43122|issue:44521", + "issue:44360|issue:44485", + "issue:43122|issue:46326", + "issue:42907|issue:43452", + "issue:44038|issue:44297", + "issue:22355|issue:29159", + "issue:44038|issue:44295", + "issue:45216|issue:45478", + "issue:39401|issue:46614", + "issue:44038|issue:44222" + ], + "prepared_review_unit_hash": "7fdd2f55d0218527b06b30d09e858ff09e2c0530c2e3d29cb331a2a1d5582c0d", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 36833, + "estimated_input_tokens": 9209, + "estimated_eval_tokens": 18674 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: it mixes unrelated tokenizer, model-loading, multimodal, and API/regression bugs. The soft pairs mostly share only broad Transformers similarity, not the same concrete underlying defect, so none should be merged as duplicates.", + "confidence": 0.34, + "canonical_issue_reason": "issue:45278 is the broadest regression report in the set and aligns with several import/breakage complaints, making it the least specific anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "Best representative of the cluster\u2019s shared theme of Transformers v5 breakages; broad user impact and generic wording make it the most usable single issue anchor.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42796", + "right": "issue:43122", + "accept": false, + "reason": "Both are tokenizer-related, but one is a BPE-marker handling bug and the other is an offset_mapping regression; different failure modes." + }, + { + "left": "issue:43646", + "right": "issue:46002", + "accept": false, + "reason": "Custom model initialization breakage and NaNs in classification heads are unrelated bugs." + }, + { + "left": "issue:45381", + "right": "issue:46909", + "accept": false, + "reason": "Video position-id regression vs a broken documentation example; not the same issue." + }, + { + "left": "issue:43540", + "right": "issue:46909", + "accept": false, + "reason": "Qwen3Omni video-input ValueError and a docs-example issue are unrelated." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Loading hang for Sentence Transformers is not the same as a MacOS bus error on a CLIP model." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Dequantized save failure and incorrect pipeline-parallel LLM output are different code paths." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Saving dequantized Ministrals/Devstrals and loading Phi-4 multimodal are separate failures." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Qwen3 tokenizer offset mapping and a model-load failure for sleng-bert are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "Buffer access on position_ids and tokenizer_class mismatch on save_pretrained are different bugs." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python version compatibility and pipeline parallelism output corruption are not the same defect." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "Broken env command and a NameError during sentence-transformers import are distinct import-time issues." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 settings guidance and a torch.split return-value bug have no concrete overlap." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Smollm3 RoPE-layer mismatch and Qwen3-Next tied-weights warning are unrelated model-specific problems." + }, + { + "left": "issue:46172", + "right": "issue:46762", + "accept": false, + "reason": "DeepSeek-v4 transpose logic and Minimax MSA numerical differences are different implementations and symptoms." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Saving dequantized models and invalid GenerationConfig handling are separate issues." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "Both involve imports, but one is a specific AutoImageProcessor import error and the other is a broad post-upgrade import regression." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "Sentence-transformers import NameError and broken auto_mappings references are unrelated import problems." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "Tokenizer mapping None handling and AutoProcessor kwarg forwarding are different loader bugs." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer offset_mapping and multimodal assistant_masks are not the same underlying bug." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and FA3-induced degenerate training are different failures." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "cached_file kwarg forwarding and chat-template crashing on tool-call messages are unrelated." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Model-type warning and missing fpn_position_embeddings are separate SAM 3 issues." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Tokenization differences across versions and multimodal assistant_masks zeroing are different bugs." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer missing ReLU and GLM-5 RoPE implementation are not the same defect." + }, + { + "left": "issue:43122", + "right": "issue:46326", + "accept": false, + "reason": "General tokenization changes and mistral_common continue_final_message behavior are different code paths." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Dequantized save failure and gguf_file loading breakage are unrelated model I/O bugs." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe model bug and tokenizer_class mismatch on save_pretrained are different issues." + }, + { + "left": "issue:22355", + "right": "issue:29159", + "accept": false, + "reason": "Missing transformers.onnx module and slow/fast tokenizer inconsistency are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Qwen3-VL-Moe behavior and position_ids buffer access are different bugs." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Both touch Qwen3.5 save/load flows, but one is a save_pretrained regression and the other is a from_pretrained error; not the same concrete bug." + }, + { + "left": "issue:39401", + "right": "issue:46614", + "accept": false, + "reason": "Qwen3 tokenizer offset mapping and stale phonemizer language state are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Qwen3-VL-Moe loading/processing and FP8 save_pretrained for moe are different failure modes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 51, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:36296", + "issue:38617", + "issue:42175", + "issue:42371", + "issue:42673", + "issue:42757", + "issue:42898", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43278", + "issue:43493", + "issue:43611", + "issue:43618", + "issue:43824", + "issue:43825", + "issue:43874", + "issue:43937", + "issue:43976", + "issue:44038", + "issue:44075", + "issue:44188", + "issue:44351", + "issue:44355", + "issue:44485", + "issue:44610", + "issue:44625", + "issue:44779", + "issue:44936", + "issue:44938", + "issue:44945", + "issue:45072", + "issue:45083", + "issue:45200", + "issue:45278", + "issue:45357", + "issue:45446", + "issue:45478", + "issue:45753", + "issue:46002", + "issue:46231", + "issue:46328", + "issue:46614", + "issue:46829", + "issue:46877", + "issue:46909" + ], + "soft_pairs": [ + "issue:43824|issue:44351", + "issue:43116|issue:45200", + "issue:43122|issue:44779", + "issue:43278|issue:44945", + "issue:36296|issue:43116", + "issue:42757|issue:44351", + "issue:22355|issue:29127", + "issue:42175|issue:44938", + "issue:43493|issue:45753", + "issue:44188|issue:46328", + "issue:46231|issue:46877", + "issue:43937|issue:44485", + "issue:42898|issue:46614", + "issue:38617|issue:45278", + "issue:43066|issue:44779", + "issue:43611|issue:46002", + "issue:43976|issue:45446", + "issue:44038|issue:45478", + "issue:43116|issue:43493", + "issue:42673|issue:45072", + "issue:45357|issue:45478", + "issue:44355|issue:46231", + "issue:44625|issue:46002", + "issue:42371|issue:43874", + "issue:45083|issue:45446", + "issue:44610|issue:46909", + "issue:44075|issue:44936", + "issue:43618|issue:46829", + "issue:43116|issue:43825" + ], + "prepared_review_unit_hash": "ef23e0b9ed7bcba275751fe665d3856a538fd429956baad05054371a3b7889e9", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 35979, + "estimated_input_tokens": 8995, + "estimated_eval_tokens": 18246 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a loose similarity cluster around Transformers v5 regressions, import/API breakages, tokenizer/model-specific bugs, and assorted unrelated issues. I did not find any soft pair that clearly describes the same underlying bug/change.", + "confidence": 0.83, + "canonical_issue_reason": "issue:45278 is the broadest representative of the cluster because it explicitly summarizes widespread import breakage after upgrading to 5.5.0; most other items are narrower, model-specific, or unrelated.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best single issue to anchor the cluster since it captures the general upgrade/regression theme better than the more specific one-off bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors in Transformers, but they involve different missing symbols (`Qwen2_5_VLForConditionalGeneration` vs `HybridCache`) and likely different breakages." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "One is a multi-label example-script bug; the other is a Gemma 4 multimodal config default issue. Different code paths and symptoms." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions in v5, but for different models and outputs. Too broad to treat as the same bug." + }, + { + "left": "issue:43278", + "right": "issue:44945", + "accept": false, + "reason": "Different failures: dtype mismatch between train/eval versus incorrect outputs under pipeline parallelism." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor parallel training and a classification example script issue are unrelated." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Both are import-related, but they fail on different missing exports from different dependency surfaces; not the same underlying defect." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "A missing module import and an error-message clarity issue for LayoutLMv3 are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Different environment/version problems: backend extras not including TensorFlow versus Python 3.14 load failure." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "Both involve Qwen models, but one is a SigLIP2 implementation discrepancy and the other is an infinite-loop image-specific generation bug." + }, + { + "left": "issue:44188", + "right": "issue:46328", + "accept": false, + "reason": "One is a torch.compile attention-kernel divergence; the other is a performance issue in feature padding. Not the same bug." + }, + { + "left": "issue:46231", + "right": "issue:46877", + "accept": false, + "reason": "Documentation parameter mismatch versus import-time doc-lint stdout noise; both are doc-related, but not the same change or defect." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "Both concern GLM-5, but one is invalid GenerationConfig handling and the other is RoPE implementation discussion. Different concrete issues." + }, + { + "left": "issue:42898", + "right": "issue:46614", + "accept": false, + "reason": "Tokenizer cleanup behavior in v5 is unrelated to stale phonemizer language state in Wav2Vec2 phonemization." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "The latter is a broad import-error umbrella, but these are not clearly the same defect; one specific missing import is not enough to merge them." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions in v5, but they describe different model/tokenization problems and likely different root causes." + }, + { + "left": "issue:43611", + "right": "issue:46002", + "accept": false, + "reason": "Model loading via `base_model_prefix` and NaNs in classification heads are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python-version compatibility and a PyTorch version check for AuxRequest are separate version-guard bugs." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Both mention Qwen3.5/Moe, but one is a 5.0 bug and the other is a 5.4+ from_pretrained error; insufficient evidence they are the same underlying issue." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Classification example-script output bug versus SigLIP2 implementation discrepancy are unrelated." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "VRAM leak in multi-threaded loading and dtype mismatches in bfloat16 inference are different code paths." + }, + { + "left": "issue:45357", + "right": "issue:45478", + "accept": false, + "reason": "Both are Qwen3.5 issues, but one is save_pretrained key serialization and the other is a from_pretrained loading error; not the same bug." + }, + { + "left": "issue:44355", + "right": "issue:46231", + "accept": false, + "reason": "Compiled Python-file loading errors and a documentation parameter-name mismatch are unrelated." + }, + { + "left": "issue:44625", + "right": "issue:46002", + "accept": false, + "reason": "A missing num_labels propagation bug and NaNs in classification heads are different defects." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 behavior settings and a missing image-patch-count method are unrelated." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "Unexpected helper output lengths in qwen3_omni_moe and an AuxRequest import version check are not the same issue." + }, + { + "left": "issue:44610", + "right": "issue:46909", + "accept": false, + "reason": "Processor/model input-size mismatch versus a broken documentation example are unrelated." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "SGD argument handling and trainer evaluate-after-train failure are different training issues." + }, + { + "left": "issue:43618", + "right": "issue:46829", + "accept": false, + "reason": "Both touch output dataclasses, but they concern different models and different missing/wrong fields." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "A multi-label example script bug and an incorrect pipeline translation-task error message 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": 24, + "review_unit_index": 52, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28348", + "issue:33453", + "issue:39401", + "issue:42222", + "issue:42673", + "issue:42796", + "issue:43064", + "issue:43066", + "issue:43441", + "issue:43450", + "issue:43673", + "issue:43827", + "issue:43937", + "issue:43986", + "issue:43992", + "issue:44117", + "issue:44263", + "issue:44291", + "issue:44380", + "issue:44410", + "issue:44466", + "issue:44704", + "issue:44779", + "issue:44829", + "issue:44977", + "issue:44995", + "issue:45071", + "issue:45081", + "issue:45092", + "issue:45216", + "issue:45310", + "issue:45341", + "issue:45405", + "issue:45446", + "issue:45561", + "issue:45563", + "issue:45588", + "issue:46002", + "issue:46077", + "issue:46328", + "issue:46512", + "issue:46614", + "issue:46640", + "issue:46655", + "issue:46697", + "issue:46710", + "issue:46752", + "issue:46762" + ], + "soft_pairs": [ + "issue:43992|issue:44704", + "issue:43066|issue:46710", + "issue:46614|issue:46752", + "issue:28348|issue:46640", + "issue:44995|issue:46697", + "issue:44263|issue:46762", + "issue:45092|issue:46762", + "issue:43827|issue:45446", + "issue:42222|issue:43441", + "issue:33453|issue:39401", + "issue:43064|issue:43673", + "issue:42673|issue:42796", + "issue:43937|issue:45563", + "issue:44380|issue:46077", + "issue:43673|issue:46328", + "issue:44291|issue:45071", + "issue:42673|issue:46655", + "issue:42222|issue:46752", + "issue:45405|issue:45446", + "issue:44117|issue:45588", + "issue:44466|issue:46002", + "issue:39401|issue:44779", + "issue:43937|issue:44977", + "issue:44410|issue:45310", + "issue:43450|issue:43986", + "issue:45341|issue:45561", + "issue:45071|issue:46512", + "issue:45341|issue:46328", + "issue:44829|issue:45216", + "issue:44117|issue:45081" + ], + "prepared_review_unit_hash": "06ebe3068c4a0c9edbc583b5f8ab298bd46158aa64d29b902bf0715da3ed6a2a", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36673, + "estimated_input_tokens": 9169, + "estimated_eval_tokens": 18594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is a heterogeneous mix of Transformers v5 tokenizer, loading, attention, generation, and model-specific regressions. None of the suggested soft pairs look like the same underlying bug or change; they mostly share only broad subsystem keywords.", + "confidence": 0.89, + "canonical_issue_reason": "issue:33453 is the broadest and most umbrella-like report in the set, centered on tokenizer loading regressions rather than a narrow model-specific symptom.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:33453 has the clearest general loading-regression framing and the highest discussion activity, making it the best representative issue for this heterogeneous cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Different code paths: missing model weight handling vs AutoProcessor kwarg propagation." + }, + { + "left": "issue:43066", + "right": "issue:46710", + "accept": false, + "reason": "Tokenizer decoder typing issue vs a DeepSeek accuracy regression; not the same bug." + }, + { + "left": "issue:46614", + "right": "issue:46752", + "accept": false, + "reason": "Stale phonemizer backend state vs empty chat template crash; unrelated symptoms." + }, + { + "left": "issue:28348", + "right": "issue:46640", + "accept": false, + "reason": "Feature request for GPT-2 FlashAttention support vs a vague Flash Attention 2 issue; not clearly one defect." + }, + { + "left": "issue:44995", + "right": "issue:46697", + "accept": false, + "reason": "GLM-MoE stale indexer cache crash vs Swin2SR image over-padding; unrelated." + }, + { + "left": "issue:44263", + "right": "issue:46762", + "accept": false, + "reason": "torch.split return semantics in GlmMoeDsaIndexer vs numerical mismatch in Minimax attention; different problems." + }, + { + "left": "issue:45092", + "right": "issue:46762", + "accept": false, + "reason": "Meta-init incompatibility for InternVL2 vs Minimax attention numeric drift; no shared failure mode." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs referencing removed pipeline() vs PyTorch version check bug in flex_attention import." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "VitPose model breakage vs Ministral FlashAttention failure; only broad model-loading similarity." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Generic tokenizer-loading regression vs Qwen3 offset_mapping bug; different concrete defects." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "FSDP2/PEFT optimizer-state corruption vs missing generation cache during chunked prefill." + }, + { + "left": "issue:42673", + "right": "issue:42796", + "accept": false, + "reason": "VRAM leak in multi-threaded Qwen3 inference vs tokenizer BPE marker handling." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "Invalid GenerationConfig vs stale warning in paged generate(); unrelated." + }, + { + "left": "issue:44380", + "right": "issue:46077", + "accept": false, + "reason": "Ignored GPT-2 attention scaling under SDPA/FlashAttention vs Gemma4 mask construction bug." + }, + { + "left": "issue:43673", + "right": "issue:46328", + "accept": false, + "reason": "Generation cache missing vs SequenceFeatureExtractor padding inefficiency; unrelated." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "init_empty_weights unexpected argument vs PretrainedConfig type checking regression; not the same issue." + }, + { + "left": "issue:42673", + "right": "issue:46655", + "accept": false, + "reason": "VRAM leak in Qwen3 threading vs SmolVLM video padding miscalculation." + }, + { + "left": "issue:42222", + "right": "issue:46752", + "accept": false, + "reason": "VitPose model breakage vs empty-list chat template crash; unrelated." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "Dependency version bump problem vs flex_attention import version check bug." + }, + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "Tokenizer mapping fallback bug vs flash_attention.py crashing on s_aux=None; different layers." + }, + { + "left": "issue:44466", + "right": "issue:46002", + "accept": false, + "reason": "Serialization of tied lm_head weights vs NaNs in classification heads; not the same defect." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Qwen3 offset_mapping bug vs DeepSeek tokenizer regression; similar area, different model and behavior." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "GenerationConfig validation failure vs FlashAttention generation failure in Qwen3.5." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Missing qkv/gate projections in qwen3next vs Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Video processor output shape bug vs confusing missing-torchvision crash through AutoProcessor." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "A testing_utils bug vs pytest-xdist race on captured_info.txt; not enough to treat as duplicate." + }, + { + "left": "issue:45071", + "right": "issue:46512", + "accept": false, + "reason": "Type checking regression in PretrainedConfig vs stale config.dtype after from_config()." + }, + { + "left": "issue:45341", + "right": "issue:46328", + "accept": false, + "reason": "testing_utils issue vs feature extractor padding performance regression; unrelated." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "FlashAttention-3 causes degenerate training vs Qwen3.5 checkpoint save regression." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer mapping None-handling vs Mistral regex patch crash; same broad subsystem, different concrete bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 53, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:27957", + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42222", + "issue:42371", + "issue:42617", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43208", + "issue:43352", + "issue:43408", + "issue:43673", + "issue:43742", + "issue:43874", + "issue:43931", + "issue:43937", + "issue:44075", + "issue:44220", + "issue:44261", + "issue:44263", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44493", + "issue:44496", + "issue:44877", + "issue:44933", + "issue:45071", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45362", + "issue:45464", + "issue:45478", + "issue:45542", + "issue:45656", + "issue:46172", + "issue:46178", + "issue:46209", + "issue:46231", + "issue:46298", + "issue:46328", + "issue:46617", + "issue:46762", + "issue:46894" + ], + "soft_pairs": [ + "issue:43408|issue:43742", + "issue:42617|issue:45341", + "issue:43122|issue:46298", + "issue:38617|issue:41628", + "issue:45362|issue:46328", + "issue:42222|issue:46178", + "issue:44877|issue:46231", + "issue:44484|issue:45310", + "issue:43937|issue:44496", + "issue:42371|issue:44261", + "issue:39401|issue:46209", + "issue:44261|issue:46762", + "issue:42371|issue:46617", + "issue:44410|issue:45478", + "issue:44933|issue:45278", + "issue:38617|issue:44220", + "issue:44075|issue:45656", + "issue:43208|issue:46894", + "issue:43066|issue:43122", + "issue:44263|issue:44485", + "issue:44483|issue:45464", + "issue:43097|issue:44493", + "issue:43673|issue:45341", + "issue:43931|issue:44410", + "issue:42222|issue:45542", + "issue:41628|issue:45071", + "issue:45071|issue:45542", + "issue:44484|issue:45478", + "issue:43874|issue:44261", + "issue:44263|issue:46172", + "issue:27957|issue:43352" + ], + "prepared_review_unit_hash": "9579dd932e16b800e1873277670db8e529f06b52e72a39402d6d0ae86f72edc0", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36277, + "estimated_input_tokens": 9070, + "estimated_eval_tokens": 18396 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is not a duplicate set: it mixes unrelated import errors, tokenizer regressions, model-loading/config issues, API behavior changes, and performance/logging bugs. The soft pairs mostly share only broad subsystem names, not the same underlying defect.", + "confidence": 0.97, + "canonical_issue_reason": "No canonical issue: the items are heterogeneous and do not appear to describe one shared bug or change.", + "canonical_pr_reason": "No canonical PR candidate is present.", + "best_issue_reason": "No single issue is a strong representative because there is no clear duplicate cluster here.", + "best_pr_reason": "No PR candidate is present.", + "soft_edge_verdicts": [ + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "Different model-loading problems: SAM3 tracker/type warning vs MobileLLM key error." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "Unrelated: a runtime failure in 3d_parallel.py vs a small testing_utils.py bug." + }, + { + "left": "issue:43122", + "right": "issue:46298", + "accept": false, + "reason": "Tokenizer regression vs noisy logging; no shared underlying defect." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but for different symbols and different root causes." + }, + { + "left": "issue:45362", + "right": "issue:46328", + "accept": false, + "reason": "Chat model crash vs SequenceFeatureExtractor padding inefficiency; unrelated." + }, + { + "left": "issue:42222", + "right": "issue:46178", + "accept": false, + "reason": "VitPose model breakage vs tensorflow_text backend mapping error; different bugs." + }, + { + "left": "issue:44877", + "right": "issue:46231", + "accept": false, + "reason": "Strict config loading issue vs documentation parameter mismatch; not the same bug." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "Different topics: save_pretrained shard size behavior vs Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "GenerationConfig validation issue vs unrecognized model/config loading error." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 API guidance vs MLA layernorm precision/config bug; unrelated." + }, + { + "left": "issue:39401", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer offset_mapping bug vs FP8 inference failure on B200; no match." + }, + { + "left": "issue:44261", + "right": "issue:46762", + "accept": false, + "reason": "Both mention numerical differences, but one is a missing config epsilon and the other is an attention implementation mismatch." + }, + { + "left": "issue:42371", + "right": "issue:46617", + "accept": false, + "reason": "TF32 settings guidance vs cache FileNotFoundError for trust_remote_code; unrelated." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3Next missing projections vs Qwen3.5 MoE from_pretrained error; different code paths." + }, + { + "left": "issue:44933", + "right": "issue:45278", + "accept": false, + "reason": "A missing image_utils import vs broad import errors after upgrade; too general to be the same bug." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "ImportError for layer_type_validation vs audio feature extraction issue; unrelated." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "SGD args not used vs optimizer step called twice with deepspeed; different optimizer bugs." + }, + { + "left": "issue:43208", + "right": "issue:46894", + "accept": false, + "reason": "xLSTM training blockers vs CohereASR double-shift loss bug; no shared defect." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Both are tokenizer-related regressions, but one is decoder type mismatch and the other is changed tokenization output; not the same bug." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "torch.split return issue in GLM-MoE indexer vs GLM-5 RoPE implementation; distinct code paths." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Both hit chat/completions, but one is request acceptance failure and the other is streaming inference failure on a specific model." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "Removed tie_embeddings flag vs unexpected position-id keys; unrelated model/config regressions." + }, + { + "left": "issue:43673", + "right": "issue:45341", + "accept": false, + "reason": "Cache missing in chunked_prefill vs a testing utility bug; no overlap." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Qwen3-VL weight shape mismatch vs Qwen3Next missing projections; different model-loading problems." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "VitPose model breakage vs TensorFlow backend detection error; unrelated." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "ImportError for AutoImageProcessor vs PreTrainedConfig type checking regression; different issues." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Config type-checking regression vs undefined tf backend error; unrelated." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "save_pretrained shard size behavior vs Qwen3.5 MoE loading error; different bugs." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Missing image patch helper vs missing rms_norm_eps precision bug; different multimodal/model issues." + }, + { + "left": "issue:44263", + "right": "issue:46172", + "accept": false, + "reason": "GLM-MoE indexing bug vs deepseek-v4 tensor transpose issue; unrelated." + }, + { + "left": "issue:27957", + "right": "issue:43352", + "accept": false, + "reason": "Both mention Flash Attention 2 support, but for different models and likely separate model-specific support gaps." + } + ] + }, + "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": 24, + "review_unit_index": 54, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:38617", + "issue:42757", + "issue:42890", + "issue:43012", + "issue:43064", + "issue:43097", + "issue:43208", + "issue:43441", + "issue:43452", + "issue:43531", + "issue:43577", + "issue:43673", + "issue:43873", + "issue:43874", + "issue:43931", + "issue:43986", + "issue:44038", + "issue:44062", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44410", + "issue:44484", + "issue:44485", + "issue:44829", + "issue:44987", + "issue:45003", + "issue:45341", + "issue:45372", + "issue:45375", + "issue:45440", + "issue:45588", + "issue:45753", + "issue:46050", + "issue:46123", + "issue:46172", + "issue:46298", + "issue:46369", + "issue:46655", + "issue:46664", + "issue:46701", + "issue:46710", + "issue:46752", + "issue:46821", + "issue:46897" + ], + "soft_pairs": [ + "issue:44360|issue:46172", + "issue:43931|issue:44038", + "issue:43441|issue:44829", + "issue:43012|issue:43577", + "issue:43097|issue:43452", + "issue:45003|issue:45341", + "issue:43986|issue:45341", + "issue:44261|issue:44485", + "issue:45753|issue:46369", + "issue:44117|issue:44987", + "issue:43531|issue:45341", + "issue:43874|issue:44263", + "issue:45440|issue:46172", + "issue:43097|issue:45588", + "issue:43673|issue:46123", + "issue:44410|issue:45375", + "issue:43986|issue:45372", + "issue:44038|issue:44410", + "issue:38617|issue:42757", + "issue:43097|issue:46050", + "issue:43208|issue:46897", + "issue:43064|issue:44315", + "issue:43986|issue:44484", + "issue:46655|issue:46664", + "issue:46298|issue:46752", + "issue:42890|issue:46298", + "issue:46172|issue:46710", + "issue:44062|issue:44117", + "issue:38617|issue:44351", + "issue:43097|issue:43873", + "issue:43012|issue:46821", + "issue:44117|issue:46701" + ], + "prepared_review_unit_hash": "0cd39a55a124aaec4a5ba26bb1d8b7212055ab271ccf538ee7126c938dabf2b2", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37183, + "estimated_input_tokens": 9296, + "estimated_eval_tokens": 18848 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated Transformers bug reports. The soft-similarity links are largely false positives across different models, APIs, or failure modes; none look like the same underlying bug or change.", + "confidence": 0.78, + "canonical_issue_reason": "issue:43873 is the broadest and most triage-relevant item here (open, high discussion, inbound refs), so it is the closest thing to an umbrella anchor\u2014but the cluster is still too heterogeneous to treat it as a true duplicate canonical.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best representative issue because it has the strongest evidence of active impact and ongoing discussion, even though it does not subsume the other reports as duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44360", + "right": "issue:46172", + "accept": false, + "reason": "Different models and different bugs: DSA indexer math vs DeepSeek-v4 dtype/transpose behavior." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "Both are Qwen3-VL-related, but one is a weight-shape mismatch and the other is a Qwen3-VL-Moe bug; not the same failure." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "Both mention flash attention, but one is a Ministral-3 load failure and the other is a training regression in sequence classification." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "Both involve dtype/precision, but one is a PyTorch warning during compile and the other is incorrect dtype propagation when loading Blip2." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "A v5 API removal issue and a gguf tokenizer/model loading breakage are different code paths and symptoms." + }, + { + "left": "issue:45003", + "right": "issue:45341", + "accept": false, + "reason": "sys.modules access in modeling_utils is unrelated to a testing_utils bug." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "Video-model processor import crash without torchvision is unrelated to a testing utility issue." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "Different model internals: MLA q_a_layernorm eps handling vs GLM-5 RoPE implementation." + }, + { + "left": "issue:45753", + "right": "issue:46369", + "accept": false, + "reason": "Different symptoms and subsystems: Qwen3_5 infinite loop on an image vs SAM memory overflow." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer mapping None handling is a different bug from loading physical-intelligence/fast." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to a testing_utils bug." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Image patch-count helper missing in GLM46V is unrelated to torch.split output handling in GlmMoeDsaIndexer." + }, + { + "left": "issue:45440", + "right": "issue:46172", + "accept": false, + "reason": "DeepSeekV3MoE divergence and DeepSeek-v4 tensor conversion/transposition are different bugs." + }, + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "One is a removed config attribute; the other is a flash_attention crash on s_aux=None." + }, + { + "left": "issue:43673", + "right": "issue:46123", + "accept": false, + "reason": "Generation cache missing during chunked prefill is unrelated to MaskGenerationPipeline batch-finalization logic." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Missing attn projections in qwen3next is unrelated to a missing config field being dropped by strict loading." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "Different import failures in different processor stacks: missing torchvision vs missing ReasoningEffort." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Qwen3-VL-Moe bug and qwen3next projection omission are distinct model families and failures." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they are different missing symbols from different modules and likely different fixes." + }, + { + "left": "issue:43097", + "right": "issue:46050", + "accept": false, + "reason": "Config API removal is unrelated to save/load breakage for sinq-quantized models." + }, + { + "left": "issue:43208", + "right": "issue:46897", + "accept": false, + "reason": "xLSTM training blockers and Florence2 loss-shift bug are different model-specific training issues." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "Wrong optimizer states under FSDP2+PEFT is a different code path from Liger Kernel not being applied with model_init." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "A missing torchvision dependency crash is unrelated to max_shard_size behavior." + }, + { + "left": "issue:46655", + "right": "issue:46664", + "accept": false, + "reason": "Both are SmolVLM video-processor issues, but one is padding calculation and the other is the meaning of num_frames." + }, + { + "left": "issue:46298", + "right": "issue:46752", + "accept": false, + "reason": "Noisy logging and empty-list chat-template IndexError are unrelated." + }, + { + "left": "issue:42890", + "right": "issue:46298", + "accept": false, + "reason": "Missing set_seed in tests is unrelated to a logging-message issue." + }, + { + "left": "issue:46172", + "right": "issue:46710", + "accept": false, + "reason": "DeepSeek-v4 tensor conversion bug is unrelated to an accuracy regression after upgrading Transformers." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "AddedToken special-argument conflict is unrelated to TOKENIZER_MAPPING_NAMES returning None." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Different missing imports from different surfaces; same general class, but not the same bug." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "A removed config field and a quantization offloading problem are different issues." + }, + { + "left": "issue:43012", + "right": "issue:46821", + "accept": false, + "reason": "Precision-warning behavior is unrelated to a window_size AttributeError in TimesFM 2.5." + }, + { + "left": "issue:44117", + "right": "issue:46701", + "accept": false, + "reason": "Tokenizer mapping None handling is unrelated to the DiffusionGemma generation-output pipeline failure." + } + ] + }, + "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": 24, + "review_unit_index": 55, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:36683", + "issue:38617", + "issue:42175", + "issue:42490", + "issue:42757", + "issue:42907", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43352", + "issue:43502", + "issue:43526", + "issue:43550", + "issue:43582", + "issue:43618", + "issue:43824", + "issue:43825", + "issue:43906", + "issue:43976", + "issue:43986", + "issue:43992", + "issue:44016", + "issue:44077", + "issue:44162", + "issue:44222", + "issue:44368", + "issue:44380", + "issue:44483", + "issue:44484", + "issue:44514", + "issue:44655", + "issue:44704", + "issue:44936", + "issue:44938", + "issue:45103", + "issue:45478", + "issue:45479", + "issue:45584", + "issue:45712", + "issue:45753", + "issue:46132", + "issue:46172", + "issue:46178", + "issue:46298", + "issue:46365", + "issue:46421" + ], + "soft_pairs": [ + "issue:43502|issue:43992", + "issue:44936|issue:46365", + "issue:44704|issue:45103", + "issue:44380|issue:46421", + "issue:44938|issue:46178", + "issue:43986|issue:44016", + "issue:42907|issue:45753", + "issue:42907|issue:44655", + "issue:43825|issue:44655", + "issue:43012|issue:43550", + "issue:45712|issue:46298", + "issue:43208|issue:43550", + "issue:43012|issue:44368", + "issue:43992|issue:44484", + "issue:43352|issue:43976", + "issue:43352|issue:44380", + "issue:43066|issue:43906", + "issue:43097|issue:46298", + "issue:44162|issue:46172", + "issue:45584|issue:46365", + "issue:44380|issue:46132", + "issue:44077|issue:46365", + "issue:44222|issue:45478", + "issue:36683|issue:43582", + "issue:44380|issue:45479", + "issue:42175|issue:46365", + "issue:42757|issue:43824", + "issue:43064|issue:44380", + "issue:42490|issue:43012", + "issue:44483|issue:44514", + "issue:43618|issue:44380", + "issue:43208|issue:43526", + "issue:38617|issue:43502" + ], + "prepared_review_unit_hash": "469f5493a9ce934df1cf2b8dad579db0bd2fe57e3a62eb456b208888c51f1af4", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37438, + "estimated_input_tokens": 9360, + "estimated_eval_tokens": 18976 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous; none of the soft pairs look like the same underlying bug or change, so all are rejected. No PRs are present, and there is no clear canonical duplicate center.", + "confidence": 0.92, + "canonical_issue_reason": "No strong duplicate center exists in this set; if one issue must anchor triage, #46365 is the clearest user-facing regression with broad impact and active discussion.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #46365 is the best standalone representative because it is a straightforward import regression, clearly actionable, and likely to matter to many users.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "Different failures: unexpected network access with local_files_only vs missing encoder weights during from_pretrained." + }, + { + "left": "issue:44936", + "right": "issue:46365", + "accept": false, + "reason": "Trainer evaluation/runtime behavior is unrelated to the separate import error involving datasets." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "Kwargs propagation in AutoProcessor is unrelated to the docstring/annotations AttributeError." + }, + { + "left": "issue:44380", + "right": "issue:46421", + "accept": false, + "reason": "GPT2 attention scaling backend bug is unrelated to cache lazy initialization and recompilation." + }, + { + "left": "issue:44938", + "right": "issue:46178", + "accept": false, + "reason": "Python 3.14 import/load failure is not the same as an invalid BACKENDS_MAPPING backend name." + }, + { + "left": "issue:43986", + "right": "issue:44016", + "accept": false, + "reason": "AutoProcessor/torchvision crash and notebook syntax error are completely different issues." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Saving dequantized models is unrelated to a Qwen3_5 image generation loop." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Pipeline save_pretrained failure is a different save-path than dequantized model saving." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "Error-message wording for translation tasks is unrelated to Pipeline serialization." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "A PyTorch precision warning is unrelated to a torch.compile SDPA model failure." + }, + { + "left": "issue:45712", + "right": "issue:46298", + "accept": false, + "reason": "Dummy class cleanup and noisy logging are unrelated repository maintenance problems." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "xLSTM training blockers are unrelated to Bamba SDPA compile failures." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "Float32 warning on bfloat16 compile is unrelated to tie_word_embeddings warnings." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Missing embedding weights on load is unrelated to save_pretrained shard-size defaults." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "Flash Attention support in a model class is unrelated to the Python version compatibility issue." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Different model/backends and different symptoms; these are not the same code-path bug." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Tokenizer decoder type regression is not the same as a reproduction of another issue." + }, + { + "left": "issue:43097", + "right": "issue:46298", + "accept": false, + "reason": "A removed config field is unrelated to a noisy logging message." + }, + { + "left": "issue:44162", + "right": "issue:46172", + "accept": false, + "reason": "ESM2 breakage and a DeepSeek tensor transpose bug are unrelated." + }, + { + "left": "issue:45584", + "right": "issue:46365", + "accept": false, + "reason": "Whisper empty-transcription failure is unrelated to importing Trainer with datasets installed." + }, + { + "left": "issue:44077", + "right": "issue:46365", + "accept": false, + "reason": "PatchTSMixer post_init validation is unrelated to the Trainer import dependency issue." + }, + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "FP8 save_pretrained and Qwen3.5 MoE from_pretrained errors involve different directions and different failures." + }, + { + "left": "issue:36683", + "right": "issue:43582", + "accept": false, + "reason": "Gemma3Config vocab_size and AppleSilicon allocator warmup are unrelated." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Attention scaling backend behavior is unrelated to degenerate zero-loss classification." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "CLIPOutput attentions assignment and GPT2 attention scaling are separate model bugs." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Two different import errors for different symbols; not the same missing-export bug." + }, + { + "left": "issue:42175", + "right": "issue:46365", + "accept": false, + "reason": "Backend packaging for TensorFlow is unrelated to importing Trainer with datasets present." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "FSDP2/PEFT optimizer-state corruption is a different training bug than GPT2 attention scaling." + }, + { + "left": "issue:42490", + "right": "issue:43012", + "accept": false, + "reason": "Path-trailing-slash import warning is unrelated to a bfloat16 precision warning." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "Chat-completions API request rejection is unrelated to Qwen2_5_VLProcessor batching crash." + }, + { + "left": "issue:44655", + "right": "issue:45103", + "accept": false, + "reason": "Pipeline serialization and docstring annotation parsing are unrelated." + }, + { + "left": "issue:45478", + "right": "issue:45479", + "accept": false, + "reason": "MoE from_pretrained regression and zero-loss classification are different model-path problems." + }, + { + "left": "issue:45584", + "right": "issue:46365", + "accept": false, + "reason": "Whisper generation edge case is unrelated to Trainer import dependency behavior." + }, + { + "left": "issue:45712", + "right": "issue:46298", + "accept": false, + "reason": "Leftover dummy classes and noisy logging do not describe the same fix." + }, + { + "left": "issue:46132", + "right": "issue:46172", + "accept": false, + "reason": "AttentionInterface registration semantics and a tensor transpose bug are not the same issue." + }, + { + "left": "issue:46178", + "right": "issue:44938", + "accept": false, + "reason": "Backend mapping validation and Python 3.14 loading failure are different import/runtime problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 56, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:29870", + "issue:33453", + "issue:36010", + "issue:36296", + "issue:36683", + "issue:42757", + "issue:42994", + "issue:43010", + "issue:43064", + "issue:43208", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43525", + "issue:43611", + "issue:43618", + "issue:43756", + "issue:43761", + "issue:43867", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44075", + "issue:44077", + "issue:44230", + "issue:44355", + "issue:44483", + "issue:44485", + "issue:44492", + "issue:44625", + "issue:44655", + "issue:44855", + "issue:44908", + "issue:44938", + "issue:45071", + "issue:45216", + "issue:45276", + "issue:45278", + "issue:45310", + "issue:46050", + "issue:46077", + "issue:46082", + "issue:46365", + "issue:46369", + "issue:46829", + "issue:46851", + "issue:46877" + ], + "soft_pairs": [ + "issue:43756|issue:44485", + "issue:43618|issue:43761", + "issue:44038|issue:44230", + "issue:42994|issue:44655", + "issue:43299|issue:43867", + "issue:36010|issue:42757", + "issue:44625|issue:46365", + "issue:45276|issue:46077", + "issue:43611|issue:44655", + "issue:43352|issue:43937", + "issue:44483|issue:46369", + "issue:44355|issue:46877", + "issue:36683|issue:43475", + "issue:44655|issue:46050", + "issue:36683|issue:44077", + "issue:43976|issue:44938", + "issue:44625|issue:46077", + "issue:43867|issue:45216", + "issue:46077|issue:46082", + "issue:43618|issue:45216", + "issue:43208|issue:46829", + "issue:36683|issue:43525", + "issue:36683|issue:44625", + "issue:29870|issue:36296", + "issue:43010|issue:45278", + "issue:43992|issue:45310", + "issue:43064|issue:44492", + "issue:44075|issue:46365", + "issue:44908|issue:46851", + "issue:29870|issue:33453", + "issue:44938|issue:45071", + "issue:44355|issue:44855" + ], + "prepared_review_unit_hash": "123236c6e6bd81a591baa1c4477f7f0bd7bff8be9b942b250fb9978e38a82c1f", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 36860, + "estimated_input_tokens": 9215, + "estimated_eval_tokens": 18686 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is a loose grab bag of unrelated Transformers issues (imports, model config regressions, save/load bugs, training bugs). No PRs are present, and none of the soft-edge pairs are strong enough to merge as duplicates or the same fix.", + "confidence": 0.94, + "canonical_issue_reason": "issue:45216 is the cleanest representative: it is a concrete, versioned regression (`save_pretrained` on Qwen3.5 since 5.4.0) with a specific failure mode and likely actionable reproduction.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45216 has the strongest issue quality overall: clear regression boundary, specific model family, and a focused save-path corruption problem.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Both mention RoPE, but they are different models and different problems (Smollm3 layer dropping vs GLM-5 RoPE implementation)." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": false, + "reason": "Both are CLIP v5 regressions, but one is missing attentions and the other is missing hidden_states; likely separate bugs in the same area, not clearly one fix." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Different Qwen3-VL-Moe topics: general breakage/loading regression vs fp8 inference support." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Both involve saving, but one is quantized model persistence and the other is Pipeline.save_pretrained; different code paths and artifacts." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Both are loading failures, but the symptoms and likely root causes differ (Qwen3.5 model loading vs sorted state_dict ordering)." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols and different dependency breakages." + }, + { + "left": "issue:44625", + "right": "issue:46365", + "accept": false, + "reason": "Unrelated: Qwen3.5 num_labels propagation vs Trainer import affected by datasets installation." + }, + { + "left": "issue:45276", + "right": "issue:46077", + "accept": false, + "reason": "Same model family, but one is token embedding resize propagation and the other is attention mask construction; different bugs." + }, + { + "left": "issue:43611", + "right": "issue:44655", + "accept": false, + "reason": "Different problem areas: base_model_prefix loading regression vs pipeline serialization." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "Different failures: Flash Attention unsupported for a model vs invalid GenerationConfig for GLM-5." + }, + { + "left": "issue:44483", + "right": "issue:46369", + "accept": false, + "reason": "Completely different symptoms and subsystems: chat completions request rejection vs sam3-litetext memory overflow." + }, + { + "left": "issue:44355", + "right": "issue:46877", + "accept": false, + "reason": "Both are import-time issues, but one concerns compiled Python files and the other noisy docstring lint output." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Different model families and attributes: Gemma3 vocab_size vs SAM3 vision output fields." + }, + { + "left": "issue:44655", + "right": "issue:46050", + "accept": false, + "reason": "Both are save/load related, but one is Pipeline.save_pretrained and the other is quantized sinq round-trip behavior." + }, + { + "left": "issue:36683", + "right": "issue:44077", + "accept": false, + "reason": "No relation beyond being model-related issues; one is missing config attribute, the other is a post_init API constraint." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both are Python-version compatibility reports, but for different interpreter versions and likely unrelated causes." + }, + { + "left": "issue:44625", + "right": "issue:46077", + "accept": false, + "reason": "Different Gemma4/Qwen3.5 configuration bugs with no shared failing path." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "Both involve load/save behavior, but one is state_dict ordering during load and the other is Qwen3.5 checkpoint corruption on save." + }, + { + "left": "issue:46077", + "right": "issue:46082", + "accept": false, + "reason": "Different configuration/validation bugs in different models (Gemma4 attention mode vs Llama head_dim validation)." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "Different model families and different outputs; CLIP output fields vs Qwen3.5 checkpoint saving." + }, + { + "left": "issue:43208", + "right": "issue:46829", + "accept": false, + "reason": "Different models and domains: xLSTM training blockers vs VideoPrism output field mixup." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Both are missing config attributes, but for unrelated models and different fields." + }, + { + "left": "issue:36683", + "right": "issue:44625", + "accept": false, + "reason": "Different missing-attribute/config propagation bugs in different model families." + }, + { + "left": "issue:29870", + "right": "issue:36296", + "accept": false, + "reason": "Different training/runtime failures: BART generate decoder mask bug vs tensor-parallel training issue." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "Unrelated: cache update no_grad decoration vs broad import errors after version upgrade." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "Different Qwen3.5 MoE issues: missing embed token weights vs from_pretrained error on newer transformers." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Completely unrelated: optimizer state corruption under FSDP/PEFT vs a cache strategy typo." + }, + { + "left": "issue:44075", + "right": "issue:46365", + "accept": false, + "reason": "Different subsystems: SGD argument handling vs Trainer import failure due to datasets." + }, + { + "left": "issue:44908", + "right": "issue:46851", + "accept": false, + "reason": "Different features: scheduler kwargs ignored vs variable-length position_ids in Qwen3.5." + }, + { + "left": "issue:29870", + "right": "issue:33453", + "accept": false, + "reason": "Different issues: BART generation mask bug vs tokenizer loading regression." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Both are version/regression complaints, but one is Python 3.14 import failure and the other is PretrainedConfig type checking." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Different failure modes: compiled files import issues vs Python 3.13 IndentationError from TorchScript parsing." + } + ] + }, + "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": 24, + "review_unit_index": 57, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28282", + "issue:30333", + "issue:33453", + "issue:36296", + "issue:36683", + "issue:41669", + "issue:42890", + "issue:42898", + "issue:43010", + "issue:43064", + "issue:43065", + "issue:43502", + "issue:43575", + "issue:43576", + "issue:43618", + "issue:43723", + "issue:43824", + "issue:43867", + "issue:43874", + "issue:43976", + "issue:43992", + "issue:44062", + "issue:44077", + "issue:44112", + "issue:44222", + "issue:44246", + "issue:44355", + "issue:44360", + "issue:44387", + "issue:44485", + "issue:44556", + "issue:44734", + "issue:44749", + "issue:44964", + "issue:45125", + "issue:45357", + "issue:45375", + "issue:45440", + "issue:45478", + "issue:45561", + "issue:45600", + "issue:45850", + "issue:46132", + "issue:46202", + "issue:46285", + "issue:46617", + "issue:46664" + ], + "soft_pairs": [ + "issue:43867|issue:46202", + "issue:36683|issue:43064", + "issue:42890|issue:45561", + "issue:43867|issue:45375", + "issue:43010|issue:45850", + "issue:41669|issue:44246", + "issue:36296|issue:43575", + "issue:43010|issue:43976", + "issue:43618|issue:44077", + "issue:43867|issue:45357", + "issue:43992|issue:45478", + "issue:28282|issue:33453", + "issue:43618|issue:46285", + "issue:44360|issue:45440", + "issue:42890|issue:44964", + "issue:42890|issue:44112", + "issue:42890|issue:45600", + "issue:42898|issue:46664", + "issue:43502|issue:44062", + "issue:44734|issue:46617", + "issue:43723|issue:44556", + "issue:44222|issue:45125", + "issue:43824|issue:46617", + "issue:44387|issue:44556", + "issue:43502|issue:44485", + "issue:43874|issue:46617", + "issue:43576|issue:46132", + "issue:43010|issue:43065", + "issue:45125|issue:45478", + "issue:30333|issue:33453", + "issue:44355|issue:44749", + "issue:45125|issue:45375" + ], + "prepared_review_unit_hash": "3afccda443ab027e16ff2b2db3decdcc40519b8045d48346a3df977f928b53e3", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36428, + "estimated_input_tokens": 9107, + "estimated_eval_tokens": 18470 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a semantic neighborhood of Transformers regressions, but the soft pairs do not look like true duplicates of the same underlying bug/change. No PRs are present. The best representative issue is a generic loading-failure report, but it does not meaningfully subsume the rest.", + "confidence": 0.18, + "canonical_issue_reason": "Issue 43867 is the broadest and most generic loading-error report in the set, but the cluster is too heterogeneous for it to be a true canonical duplicate.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43867 is the most representative umbrella-style issue by title, but it should only be treated as a loose thematic anchor, not a duplicate target for the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43867", + "right": "issue:46202", + "accept": false, + "reason": "Both mention model/pipeline failure, but one is a sorted state_dict loading bug and the other is a Gemma roundtrip issue; different code paths." + }, + { + "left": "issue:36683", + "right": "issue:43064", + "accept": false, + "reason": "Gemma3Config vocab_size missing and FSDP2/PEFT optimizer-state corruption are unrelated bugs." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "One is flaky SamHQ integration tests from missing set_seed; the other is an xdist file race in testing utils." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "Sorted state_dict loading error and missing Qwen3.5 vision config field are distinct failures." + }, + { + "left": "issue:43010", + "right": "issue:45850", + "accept": false, + "reason": "torch.no_grad on cache updates and a broad 'many models break' complaint are not the same bug." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": false, + "reason": "Both concern import time, but one attributes a specific import-star slowdown while the other is a general intermittent slow import report." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Tensor-parallel training bug and tp-related OOM while loading Qwen2-57B are different symptoms and likely different paths." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Cache update no_grad and Python version compatibility are unrelated." + }, + { + "left": "issue:43618", + "right": "issue:44077", + "accept": false, + "reason": "CLIPOutput attentions regression and patchtsmixer post_init validation are unrelated model issues." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "Loading sorted state_dict and incorrect visual encoder keys in save_pretrained are different bugs." + }, + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Missing embed_tokens.weight in UMT5Encoder and Qwen3.5 Moe from_pretrained failure are different model-loading issues." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "PyTorch-missing import error and tokenizer-loading regression are unrelated." + }, + { + "left": "issue:43618", + "right": "issue:46285", + "accept": false, + "reason": "These are both CLIP-related, but one is attentions assignment and the other is class-structure inconsistency." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "DSA indexer ReLU discussion and DeepSeekV3MoE implementation divergence are unrelated." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "Test seeding flake and Phi-4 multimodal loading failure are unrelated." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "Different CI/test failures: missing seed versus stale device override." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "Test flakiness and auto_mappings references to removed configs are not the same issue." + }, + { + "left": "issue:42898", + "right": "issue:46664", + "accept": false, + "reason": "Tokenizer clean_up_tokenization_spaces behavior and SmolVLM num_frames semantics are unrelated." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "local_files_only still triggering API calls and AddedToken keyword duplication are different subsystems." + }, + { + "left": "issue:44734", + "right": "issue:46617", + "accept": false, + "reason": "Serve KV-cache continuation indexing bug and trust_remote_code cache FileNotFoundError are different failure modes." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "AutoTokenizer.from_pretrained regression and checkpoint reload incompatibility after upgrade are distinct load-path problems." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 save_pretrained on MoE and missing _tp_plan for tensor parallelism are different issues." + }, + { + "left": "issue:43824", + "right": "issue:46617", + "accept": false, + "reason": "ImportError for Qwen2_5_VLForConditionalGeneration and cache FileNotFoundError are unrelated." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Int4 quantization memory regression and model reload failure after version upgrade are not the same bug." + }, + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "Remote-access leakage with local_files_only and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:43874", + "right": "issue:46617", + "accept": false, + "reason": "Missing image-patch helper in GLM46V and trust_remote_code cache loading failure are different code paths." + }, + { + "left": "issue:43576", + "right": "issue:46132", + "accept": false, + "reason": "Broken transformers env command and AttentionInterface.register behavior change are unrelated." + }, + { + "left": "issue:43010", + "right": "issue:43065", + "accept": false, + "reason": "Cache update no_grad and a dummy Conv2d in Sam3PixelDecoder are unrelated." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5 Moe, but one is missing _tp_plan and the other is a broader from_pretrained error; not enough to treat as the same bug." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLFlow reporting on failed training jobs and tokenizer loading regression are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Compiled Python file errors and post-upgrade filtering slowdown are different issues." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Missing _tp_plan for tensor parallelism and missing deepstack_visual_indexes in vision config are different config-field 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": 24, + "review_unit_index": 58, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:29870", + "issue:30333", + "issue:30990", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:38617", + "issue:41628", + "issue:42548", + "issue:42757", + "issue:43010", + "issue:43232", + "issue:43299", + "issue:43475", + "issue:43502", + "issue:43576", + "issue:43673", + "issue:43824", + "issue:43827", + "issue:43901", + "issue:44222", + "issue:44230", + "issue:44273", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44485", + "issue:44492", + "issue:44556", + "issue:44623", + "issue:44908", + "issue:44938", + "issue:44961", + "issue:45003", + "issue:45071", + "issue:45230", + "issue:45362", + "issue:45468", + "issue:45542", + "issue:45850", + "issue:46143", + "issue:46144", + "issue:46209", + "issue:46617", + "issue:46664", + "issue:46726" + ], + "soft_pairs": [ + "issue:30990|issue:31795", + "issue:43010|issue:43232", + "issue:46664|issue:46726", + "issue:42548|issue:44351", + "issue:43502|issue:44961", + "issue:43502|issue:44222", + "issue:43827|issue:44961", + "issue:30333|issue:36296", + "issue:43299|issue:44556", + "issue:29870|issue:31795", + "issue:44355|issue:45071", + "issue:33453|issue:34689", + "issue:43576|issue:45071", + "issue:42757|issue:43576", + "issue:44315|issue:44360", + "issue:31795|issue:38617", + "issue:43576|issue:43824", + "issue:44961|issue:45468", + "issue:44485|issue:44961", + "issue:44351|issue:46617", + "issue:36296|issue:43901", + "issue:44908|issue:44961", + "issue:43576|issue:44938", + "issue:43475|issue:46617", + "issue:46143|issue:46144", + "issue:44222|issue:44230", + "issue:44315|issue:46209", + "issue:41628|issue:42548", + "issue:44485|issue:46664", + "issue:43673|issue:44492", + "issue:45362|issue:45850", + "issue:44273|issue:45542", + "issue:44623|issue:45230", + "issue:44360|issue:46664", + "issue:44273|issue:45003" + ], + "prepared_review_unit_hash": "a52de9ea654dac04b21cd4e8ae507eec19b5be6c0cec18c61b9471051f141186", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 37295, + "estimated_input_tokens": 9324, + "estimated_eval_tokens": 18904 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is very noisy and mostly contains distinct bugs rather than true duplicates; the closest shared theme is v5-era loading/import regressions, but there is no single strong duplicate cluster.", + "confidence": 0.74, + "canonical_issue_reason": "issue:45850 is the broadest umbrella report in the set and best matches the common v5 breakage/loading-regression theme, though it is still only a loose representative because the cluster is heterogeneous.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45850 is the most representative issue for the overall cluster because it is the most general user-facing report of widespread v5 regressions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "SentenceTransformers loading gets stuck is a runtime hang; the other is a documentation clarity issue about forward arguments." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "Both mention generation/cache internals, but one is about no_grad decoration and the other about sync_gpus/model kwargs handling; different bugs." + }, + { + "left": "issue:46664", + "right": "issue:46726", + "accept": false, + "reason": "Different failures in different processor paths: video num_frames semantics vs missing vision-backend handling in AutoProcessor." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they concern different missing exports and likely different breakages." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "API requests despite local_files_only is unrelated to the unrelated one-word issue 'racoon'." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "One is about unwanted network access during loading; the other is an FP8 save_pretrained issue for moe models." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs referencing pipeline() after removals is unrelated to the unrelated 'racoon' issue." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLflow reporting on failed jobs is unrelated to a tensor-parallel training bug." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Both are version-related loading regressions, but they affect different model families and code paths; not the same concrete bug." + }, + { + "left": "issue:29870", + "right": "issue:31795", + "accept": false, + "reason": "Generation UnboundLocalError is a runtime bug; the other is documentation confusion." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled Python file loading errors are unrelated to PreTrainedModel type-checking regressions." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Both are loading regressions, but one is tokenizer loading and the other is a specific Llama 3.2 vision model-loading failure." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "Broken env command in v5 is unrelated to PreTrainedModel type checking." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Missing is_offline_mode from huggingface_hub is a dependency/export issue, not the env command bug." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger kernel not applied in model_init and the DSA indexer ReLU discussion are unrelated code paths." + }, + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Documentation clarification is unrelated to the layer_type_validation import error." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "Broken env command and missing Qwen2_5_VLForConditionalGeneration import are different failures." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "Unrelated issues with no shared code path or symptom." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "GLM-5 RoPE implementation is unrelated to the unrelated 'racoon' report." + }, + { + "left": "issue:44351", + "right": "issue:46617", + "accept": false, + "reason": "Importing HybridCache and trust_remote_code cache FileNotFoundError are different loading failures with different causes." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor-parallel training and pipeline doc behavior are unrelated." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs handling is unrelated to the unrelated 'racoon' issue." + }, + { + "left": "issue:43576", + "right": "issue:44938", + "accept": false, + "reason": "Broken env command and Python 3.14 load failure are different environment-related problems." + }, + { + "left": "issue:43475", + "right": "issue:46617", + "accept": false, + "reason": "SAM3 video encoder attribute error is unrelated to trust_remote_code cache loading." + }, + { + "left": "issue:46143", + "right": "issue:46144", + "accept": false, + "reason": "Both concern RoFormer, but one is kwargs forwarding and the other is attention-interface usage; not the same concrete bug." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "FP8 save_pretrained for moe and FP8 inference support are related but distinct save vs inference issues." + }, + { + "left": "issue:44315", + "right": "issue:46209", + "accept": false, + "reason": "Liger kernel application and FP8 inference on B200 are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Different missing imports from transformers; no evidence they stem from one exact export regression." + }, + { + "left": "issue:44485", + "right": "issue:46664", + "accept": false, + "reason": "GLM-5 RoPE implementation and SmolVLM video frame semantics are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Generation cache missing in chunked_prefill is unrelated to a cache-strategy typo." + }, + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "Qwen3.5 chat crash is a specific model/chat-path bug; 45850 is a broad meta complaint about many v5 regressions." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading behavior and missing TensorFlow backend handling are different problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 24, + "review_unit_index": 59, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:31795", + "issue:33453", + "issue:41669", + "issue:42371", + "issue:42890", + "issue:43352", + "issue:43408", + "issue:43519", + "issue:43525", + "issue:43572", + "issue:43576", + "issue:43673", + "issue:43976", + "issue:44038", + "issue:44077", + "issue:44230", + "issue:44246", + "issue:44273", + "issue:44295", + "issue:44297", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44393", + "issue:44485", + "issue:44492", + "issue:44556", + "issue:44655", + "issue:44704", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45468", + "issue:45478", + "issue:45753", + "issue:45850", + "issue:46095", + "issue:46218", + "issue:46328", + "issue:46655", + "issue:46894" + ], + "soft_pairs": [ + "issue:44038|issue:45125", + "issue:44393|issue:45071", + "issue:44393|issue:45753", + "issue:41669|issue:43673", + "issue:43576|issue:44556", + "issue:43525|issue:45850", + "issue:43572|issue:45850", + "issue:44556|issue:45092", + "issue:44492|issue:45850", + "issue:45278|issue:45850", + "issue:42371|issue:44492", + "issue:44230|issue:46095", + "issue:43408|issue:44315", + "issue:41669|issue:46218", + "issue:44246|issue:44273", + "issue:42890|issue:44908", + "issue:43576|issue:44908", + "issue:43352|issue:44393", + "issue:43976|issue:44393", + "issue:44908|issue:45071", + "issue:44295|issue:45230", + "issue:44704|issue:44908", + "issue:44297|issue:45230", + "issue:44492|issue:46328", + "issue:41669|issue:44351", + "issue:41669|issue:45341", + "issue:43976|issue:45850", + "issue:43976|issue:46218", + "issue:44393|issue:45478", + "issue:44077|issue:44861", + "issue:31795|issue:33453", + "issue:43576|issue:44273", + "issue:44315|issue:44829", + "issue:43519|issue:46655", + "issue:44360|issue:45468", + "issue:44273|issue:44908", + "issue:45468|issue:46894", + "issue:45230|issue:45310", + "issue:43576|issue:46218", + "issue:44655|issue:44861", + "issue:43519|issue:44485", + "issue:43519|issue:44908" + ], + "prepared_review_unit_hash": "a3ff16ac0ae5b78f6a0dbfb83c29a086855dc70504176ef83f31885a78d02404", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 42, + "serialized_chars": 39666, + "estimated_input_tokens": 9917, + "estimated_eval_tokens": 20090 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: docs, tokenizer/loading, config-attribute, Qwen/vision, caching, and training issues mixed together. The only loose common thread is broad v5 regression complaints, so there is no strong duplicate set; issue:45850 is just the closest umbrella among them.", + "confidence": 0.92, + "canonical_issue_reason": "issue:45850 is the closest umbrella for the v5-breakage complaints, but the cluster is too heterogeneous for a true duplicate canonical.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45850 is the broadest and most representative issue here, but it is still too vague to act as a precise duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Different bugs: Qwen3-VL FP8/inference support vs missing `_tp_plan` for Qwen3.5 MoE tensor parallelism." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "Qwen3-VL 2D bbox hallucination is unrelated to `PretrainedConfig` type-checking regressions." + }, + { + "left": "issue:44393", + "right": "issue:45753", + "accept": false, + "reason": "Same family (Qwen3-VL/Qwen3.5) but different failures: bbox output issue vs infinite-loop on one image." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import-star cleanup/performance issue vs missing GenerationMixin cache in chunked prefill; different code paths." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Both are v5 complaints, but one is the `transformers env` command and the other is checkpoint reload compatibility." + }, + { + "left": "issue:43525", + "right": "issue:45850", + "accept": false, + "reason": "Specific `pad_token_id` attribute bug vs a vague broad v5-breakage report; not the same defect." + }, + { + "left": "issue:43572", + "right": "issue:45850", + "accept": false, + "reason": "Concrete StableLm `pad_token_idx` regression vs generic v5 breakage; too broad to be the same issue." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Both involve v5 migration, but one is checkpoint reload failure and the other is old remote-code/meta-init incompatibility." + }, + { + "left": "issue:44492", + "right": "issue:45850", + "accept": false, + "reason": "Cache-strategy typo is unrelated to broad model breakage reports." + }, + { + "left": "issue:45278", + "right": "issue:45850", + "accept": false, + "reason": "General import errors after upgrade is not the same as the broad 'too many models break' complaint." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 API warning is unrelated to a cache-strategy typo." + }, + { + "left": "issue:44230", + "right": "issue:46095", + "accept": false, + "reason": "FP8 Qwen3-VL/Qwen3.5 inference support vs DeepSeek weight-conversion tooling; different change requests." + }, + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "Model-type warning for sam3_tracker vs Liger Kernel application with `model_init`; unrelated." + }, + { + "left": "issue:41669", + "right": "issue:46218", + "accept": false, + "reason": "Import-star performance cleanup vs missing `Qwen3_5ForCausalLM` import/export; different problems." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Both mention import/lazy loading, but one is slow import sometimes and the other is a specific lazy-loading bug; not enough to unify." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "Integration-test seeding failures are unrelated to inverse_sqrt scheduler kwargs handling." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "Broken env CLI and scheduler kwargs are separate features with no shared code-path bug." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Flash Attention 2 support for Nemotron is unrelated to Qwen3-VL bounding-box output errors." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Python-version compatibility regression vs Qwen3-VL output bug; different scope and cause." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwargs ignored is unrelated to `PretrainedConfig` type-checking breaks." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "Position_ids buffer access bug is too specific to match a generic 'Bug report' entry." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "AutoProcessor kwargs forwarding issue vs scheduler kwargs bug; unrelated components." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "Tokenizer save metadata mismatch is not the same as an unspecified generic bug report." + }, + { + "left": "issue:44492", + "right": "issue:46328", + "accept": false, + "reason": "Cache-strategy typo and feature-extractor padding performance are different areas." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-star removal/performance vs missing `HybridCache` symbol export; different root causes." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Import hygiene/performance issue vs a testing utility bug; no shared defect." + }, + { + "left": "issue:43976", + "right": "issue:45850", + "accept": false, + "reason": "Python 3.9/3.10 compatibility is a platform issue, not the same as generic v5 model breakage." + }, + { + "left": "issue:43976", + "right": "issue:46218", + "accept": false, + "reason": "Platform compatibility regression and missing model import are unrelated." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Different Qwen issues: bbox hallucination in Qwen3-VL vs Qwen3.5 MoE `from_pretrained` error." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "Model `post_init` policy validation vs tied-weights helper crashing on list input; different code paths." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Documentation confusion about forward args is unrelated to tokenizer loading regression." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Broken CLI command and lazy-loading failure are not the same bug." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel not applied in `model_init` is unrelated to flash_attention_3 causing degenerate training." + }, + { + "left": "issue:43519", + "right": "issue:46655", + "accept": false, + "reason": "Both are processor math bugs, but timestamps in Qwen3VL and padding in SmolVLM are different underlying defects." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "DSA indexer ReLU discussion is unrelated to Gemma-4 audio positional encoding." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading/import slowness is unrelated to scheduler keyword forwarding." + }, + { + "left": "issue:45468", + "right": "issue:46894", + "accept": false, + "reason": "Different model families and failures: Gemma-4 audio positional encoding vs CohereASR training-loss shift bug." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "A generic bug placeholder is not enough to match the specific Qwen3.5 MoE `from_pretrained` failure." + } + ] + }, + "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": 24, + "review_unit_index": 60, + "review_unit_count": 60, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-502", + "nodes": [ + "issue:28348", + "issue:29870", + "issue:30333", + "issue:31515", + "issue:31795", + "issue:34689", + "issue:36296", + "issue:37934", + "issue:38617", + "issue:41669", + "issue:42371", + "issue:42913", + "issue:42936", + "issue:43066", + "issue:43240", + "issue:43298", + "issue:43352", + "issue:43502", + "issue:43519", + "issue:43525", + "issue:43572", + "issue:43606", + "issue:43618", + "issue:43668", + "issue:43704", + "issue:43717", + "issue:43813", + "issue:44031", + "issue:44230", + "issue:44263", + "issue:44351", + "issue:44485", + "issue:44492", + "issue:44623", + "issue:44861", + "issue:44908", + "issue:44938", + "issue:44998", + "issue:45250", + "issue:45290", + "issue:45310", + "issue:45468", + "issue:45850", + "issue:46018", + "issue:46095", + "issue:46144", + "issue:46172", + "issue:46218" + ], + "soft_pairs": [ + "issue:41669|issue:43704", + "issue:31795|issue:34689", + "issue:43704|issue:44485", + "issue:46018|issue:46172", + "issue:44485|issue:45468", + "issue:42371|issue:43704", + "issue:43502|issue:43519", + "issue:45850|issue:46218", + "issue:43519|issue:44623", + "issue:44263|issue:45468", + "issue:43606|issue:45468", + "issue:43298|issue:46095", + "issue:43519|issue:43704", + "issue:43352|issue:46218", + "issue:44351|issue:46218", + "issue:42913|issue:43066", + "issue:28348|issue:45250", + "issue:44908|issue:44998", + "issue:44492|issue:44998", + "issue:43618|issue:44998", + "issue:44998|issue:45850", + "issue:44938|issue:44998", + "issue:44861|issue:45290", + "issue:31515|issue:38617", + "issue:43572|issue:44861", + "issue:43525|issue:44861", + "issue:44230|issue:44623", + "issue:28348|issue:43352", + "issue:29870|issue:31515", + "issue:30333|issue:31515", + "issue:31515|issue:36296", + "issue:31515|issue:31795", + "issue:44230|issue:45310", + "issue:43298|issue:44230", + "issue:43717|issue:46144", + "issue:42936|issue:46018", + "issue:43668|issue:44031", + "issue:42936|issue:43240", + "issue:43240|issue:46144", + "issue:42936|issue:44031", + "issue:42936|issue:46144", + "issue:42936|issue:43813", + "issue:43240|issue:43668", + "issue:43668|issue:43813", + "issue:37934|issue:43668", + "issue:42936|issue:43668" + ], + "prepared_review_unit_hash": "b78e3a01bc88e9332a776708d7c312924f9bb366e58d74d968646dc734717f3c", + "component_budget": { + "node_count": 502, + "item_count": 502, + "soft_pair_count": 1844, + "serialized_chars": 956908, + "estimated_input_tokens": 239227, + "estimated_eval_tokens": 478710 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 46, + "serialized_chars": 40600, + "estimated_input_tokens": 10150, + "estimated_eval_tokens": 20556 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous regression/feature issues, not true duplicates. None of the soft edges look like the same underlying bug or change, so I reject all pairwise merges. If an umbrella issue is needed, the broad Transformers 5.0 breakage report is the best anchor.", + "confidence": 0.26, + "canonical_issue_reason": "issue:45850 is the broadest umbrella-style report in the set (\u201cSince 5.0 version it breaks too many models\u201d), so it is the most suitable canonical anchor among these otherwise unrelated issues.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45850 best fits as the global representative because it is the only item that broadly describes the overall regression theme rather than a narrow model-specific or unrelated bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import-star/model import performance is unrelated to a VRAM leak in Qwen3ForCausalLM." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Documentation confusion about input_ids/past_key_values is not the same as a Llama 3.2 Vision model loading regression." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "VRAM leak in Qwen3ForCausalLM and GLM-5 RoPE behavior are different code paths and symptoms." + }, + { + "left": "issue:46018", + "right": "issue:46172", + "accept": false, + "reason": "DeepSeek-V4 shared expert gating and a DeepSeek-V4 tensor transpose issue are distinct bugs." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "GLM-5 RoPE implementation and Gemma-4 audio positional encoding are unrelated model-specific issues." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 settings behavior is unrelated to the VRAM leak report." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "Local-files-only network access and Qwen3VL timestamp calculation are different problems." + }, + { + "left": "issue:45850", + "right": "issue:46218", + "accept": false, + "reason": "A broad \u201cmany models break in 5.0\u201d complaint is not the same as a specific Qwen3_5 import failure." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Timestamp calculation in a processor and missing save_pretrained files are different processor bugs." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "A torch.split return-value issue in GlmMoeDsaIndexer is unrelated to Gemma4AudioRelPositionalEncoding." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch in bark-small is not the same as a Gemma-4 audio encoding bug." + }, + { + "left": "issue:43298", + "right": "issue:46095", + "accept": false, + "reason": "Audio-model support for continuous batching and a DeepSeek-V4 weight-conversion question are unrelated." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Timestamp calculation in Qwen3VL and a Qwen3ForCausalLM VRAM leak are different failures." + }, + { + "left": "issue:43352", + "right": "issue:46218", + "accept": false, + "reason": "Flash Attention 2.0 support for NemotronH and a Qwen3_5 import error are unrelated." + }, + { + "left": "issue:44351", + "right": "issue:46218", + "accept": false, + "reason": "HybridCache import failure and Qwen3_5 module import failure are different import regressions." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Tokenizer behavior differences across versions and a wrong tokenizer decoder type are related in theme but not the same concrete bug." + }, + { + "left": "issue:28348", + "right": "issue:45250", + "accept": false, + "reason": "GPT2-specific Flash Attention 2.0 support and a generic Flash Attention 2.0 issue are not the same concrete change." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwargs handling and the unrelated spam issue are not duplicates." + }, + { + "left": "issue:44492", + "right": "issue:44998", + "accept": false, + "reason": "Cache-strategy typo and the unrelated spam issue are not duplicates." + }, + { + "left": "issue:43618", + "right": "issue:44998", + "accept": false, + "reason": "CLIPOutput attentions regression and the unrelated spam issue are not duplicates." + }, + { + "left": "issue:44998", + "right": "issue:45850", + "accept": false, + "reason": "An unrelated spam issue is not a duplicate of a broad Transformers 5.0 regression report." + }, + { + "left": "issue:44938", + "right": "issue:44998", + "accept": false, + "reason": "Python 3.14 import failure and the unrelated spam issue are not duplicates." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Tied-weight-key handling and apply_chat_template tool-call crashes affect different code paths." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "from_pretrained slowness and a missing layer_type_validation import are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig is unrelated to tied-weight-key attribute handling." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_id in Llama4Config is unrelated to tied-weight-key attribute handling." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 support for qwen3-vl/qwen3.5 and processor.save_pretrained missing files are different issues." + }, + { + "left": "issue:28348", + "right": "issue:43352", + "accept": false, + "reason": "Both mention Flash Attention 2.0, but they concern different models and separate support requests." + }, + { + "left": "issue:29870", + "right": "issue:31515", + "accept": false, + "reason": "BART generate() decoder_attention_mask bug and checkpoint loading slowness are unrelated." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLflow job-reporting behavior and checkpoint loading slowness are unrelated." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "from_pretrained slowness and tensor-parallel training bugs are different subsystems." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Loading slowness and documentation confusion are not the same bug." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 support request for Qwen3-VL/Qwen3.5 and a Qwen3.5 from_pretrained error are not the same concrete issue." + }, + { + "left": "issue:43717", + "right": "issue:46144", + "accept": false, + "reason": "Mamba init_weights behavior and RoFormer attention-interface usage are unrelated model implementations." + }, + { + "left": "issue:42936", + "right": "issue:46018", + "accept": false, + "reason": "Mask2former ignore_value handling and DeepSeek-V4 shared expert gating are unrelated." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "ModernBERTConfig type-hint correctness and tokenizer regex warnings are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Mask2former ignore_value handling and loss kwargs forwarding are different bugs." + }, + { + "left": "issue:43240", + "right": "issue:46144", + "accept": false, + "reason": "Loss kwargs forwarding and RoFormer attention-interface usage are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "Mask2former ignore_value handling and tokenizer regex warnings are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:46144", + "accept": false, + "reason": "Mask2former ignore_value handling and RoFormer attention-interface usage are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "Mask2former ignore_value handling and a typo in quantization_operation are unrelated." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "ModernBERTConfig type-hint correctness and a quantization typo are unrelated." + }, + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "Llama4TextL2Norm semantics and a ModernBERTConfig type hint are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "Mask2former ignore_value handling and ModernBERTConfig type-hint correctness are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 24, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10519-3", + "nodes": [ + "issue:10519", + "issue:11005", + "issue:11307" + ], + "soft_pairs": [ + "issue:10519|issue:11307", + "issue:10519|issue:11005", + "issue:11005|issue:11307" + ], + "prepared_review_unit_hash": "1e109591f1c0208abf877dc1e810dbcaf685bc8d8f9ddcfe33c2b8e6f3c42958", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2875, + "estimated_input_tokens": 719, + "estimated_eval_tokens": 1694 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2875, + "estimated_input_tokens": 719, + "estimated_eval_tokens": 1694 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These three issues are unrelated: token truncation direction, ReduceLROnPlateau-like scheduler behavior, and Wav2Vec2 word timestamp offsets. None should be clustered as duplicates.", + "confidence": 0.99, + "canonical_issue_reason": "No single canonical issue fits the cluster because the topics are different and do not describe the same underlying bug or feature request.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 10519 is the most self-contained and specific, but it is not a true representative for this cluster since the other issues are on unrelated topics.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:10519", + "right": "issue:11307", + "accept": false, + "reason": "Truncation-direction behavior and Wav2Vec2 word offset timing are unrelated features." + }, + { + "left": "issue:10519", + "right": "issue:11005", + "accept": false, + "reason": "Token truncation options and ReduceLROnPlateau-like scheduling address different code paths and problems." + }, + { + "left": "issue:11005", + "right": "issue:11307", + "accept": false, + "reason": "A learning-rate scheduler feature request and word timestamp offsets 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": 3, + "component_count": 24, + "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": "cc06f0792774d6a7e57bd662b0a3ba556a92acd5496225a2677d46bdab9c9ef7", + "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 are not duplicates: one is about Tapas tokenization differing from TensorFlow code, one is about an incorrect error message when downloading a private model while logged out, and one is a generic inability to load bert-base-uncased. The cluster is heterogeneous, with no safe canonical duplicate.", + "confidence": 0.88, + "canonical_issue_reason": "No single canonical issue: the titles describe different bug classes and subsystems, so there is no clear duplicate target.", + "canonical_pr_reason": null, + "best_issue_reason": "If a representative must be chosen, issue 16618 is the broadest model-loading report, but it still does not match the other issues closely enough to serve as a true duplicate canonical.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Different problems: Tapas tokenization mismatch vs a model loading failure for bert-base-uncased; not the same underlying bug." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "Related to model access, but one is a bad private-model download/login error and the other is a generic load failure; too different to merge as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 4, + "component_count": 24, + "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": "f4528b4bb127de021bed169922a2fed1586c7b537aa1ef22530e6da4a14531e2", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are both architecture-support community-contribution issues, but they target different features: Flash Attention 2 vs GGUF. They do not describe the same underlying bug or change, so they should not be clustered as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Same broad theme of expanding architecture support, but the concrete changes are unrelated (Flash Attention 2 support vs GGUF support) and would not be merged as one issue/change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 5, + "component_count": 24, + "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-27088-9", + "nodes": [ + "issue:27088", + "issue:28936", + "issue:38435", + "issue:43834", + "issue:44365", + "issue:44370", + "issue:46222", + "issue:46226", + "issue:46931" + ], + "soft_pairs": [ + "issue:27088|issue:44370", + "issue:27088|issue:44365", + "issue:27088|issue:43834", + "issue:27088|issue:38435", + "issue:27088|issue:28936", + "issue:28936|issue:43834", + "issue:28936|issue:44365", + "issue:28936|issue:38435", + "issue:28936|issue:44370", + "issue:38435|issue:44370", + "issue:38435|issue:44365", + "issue:38435|issue:43834", + "issue:38435|issue:46222", + "issue:38435|issue:46226", + "issue:43834|issue:44370", + "issue:43834|issue:44365", + "issue:43834|issue:46222", + "issue:43834|issue:46226", + "issue:43834|issue:46931", + "issue:44365|issue:44370", + "issue:44365|issue:46222", + "issue:44365|issue:46226", + "issue:44365|issue:46931", + "issue:44370|issue:46222", + "issue:44370|issue:46226", + "issue:44370|issue:46931", + "issue:46222|issue:46226", + "issue:46222|issue:46931", + "issue:46226|issue:46931" + ], + "prepared_review_unit_hash": "55aedb2f7895a058ebef72d32a7afbdd53f862064b554f6fc15c2094ebb988d0", + "component_budget": { + "node_count": 9, + "item_count": 9, + "soft_pair_count": 29, + "serialized_chars": 15977, + "estimated_input_tokens": 3995, + "estimated_eval_tokens": 8246 + }, + "budget": { + "node_count": 9, + "item_count": 9, + "soft_pair_count": 29, + "serialized_chars": 15977, + "estimated_input_tokens": 3995, + "estimated_eval_tokens": 8246 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a mix of separate i18n translation requests by language, so most cross-language pairs are not duplicates. The only convincing duplicates are the identical placeholder-title issues among the English-template entries, plus the exact same Thai-title pair.", + "confidence": 0.81, + "canonical_issue_reason": "issue:43834 is the oldest clean representative of the exact-title placeholder duplicate subgroup and is the best single issue to anchor that duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43834 is the strongest representative issue in the cluster because it sits in the clearest duplicate subgroup and predates the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Different language-specific translation request vs. generic placeholder entry; not the same issue." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "Turkish docs translation request is distinct from the placeholder entry." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "Different language request; no evidence of the same translation task." + }, + { + "left": "issue:27088", + "right": "issue:38435", + "accept": false, + "reason": "Turkish and Romanian translation requests are separate issues." + }, + { + "left": "issue:27088", + "right": "issue:28936", + "accept": false, + "reason": "Turkish and Spanish translation requests are different tasks." + }, + { + "left": "issue:28936", + "right": "issue:43834", + "accept": false, + "reason": "Spanish docs request does not match the generic placeholder issue." + }, + { + "left": "issue:28936", + "right": "issue:44365", + "accept": false, + "reason": "Different language request; not a duplicate." + }, + { + "left": "issue:28936", + "right": "issue:38435", + "accept": false, + "reason": "Spanish and Romanian translation requests are separate." + }, + { + "left": "issue:28936", + "right": "issue:44370", + "accept": false, + "reason": "Spanish docs request is unrelated to the placeholder entry." + }, + { + "left": "issue:38435", + "right": "issue:44370", + "accept": false, + "reason": "Romanian request vs. generic placeholder; different underlying issue." + }, + { + "left": "issue:38435", + "right": "issue:44365", + "accept": false, + "reason": "Different translation request; no concrete duplicate evidence." + }, + { + "left": "issue:38435", + "right": "issue:43834", + "accept": false, + "reason": "Romanian docs issue is distinct from the placeholder entry." + }, + { + "left": "issue:38435", + "right": "issue:46222", + "accept": false, + "reason": "Romanian request and Thai-titled request are different issues." + }, + { + "left": "issue:38435", + "right": "issue:46226", + "accept": false, + "reason": "Different language/title; not the same change." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": true, + "reason": "Same placeholder title and nearly identical body; looks like a duplicate of the same translation request." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": true, + "reason": "Same placeholder title and near-identical issue text; likely duplicate." + }, + { + "left": "issue:43834", + "right": "issue:46222", + "accept": false, + "reason": "Different language/title content; no evidence these are the same issue." + }, + { + "left": "issue:43834", + "right": "issue:46226", + "accept": false, + "reason": "Placeholder entry and Thai-titled entry are not the same request." + }, + { + "left": "issue:43834", + "right": "issue:46931", + "accept": false, + "reason": "Different titles and different underlying translation task." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": true, + "reason": "Same placeholder title with very similar body text; likely duplicate." + }, + { + "left": "issue:44365", + "right": "issue:46222", + "accept": false, + "reason": "Different language/title; not a duplicate." + }, + { + "left": "issue:44365", + "right": "issue:46226", + "accept": false, + "reason": "Different request; no same-issue evidence." + }, + { + "left": "issue:44365", + "right": "issue:46931", + "accept": false, + "reason": "Different translation issue; not the same underlying task." + }, + { + "left": "issue:44370", + "right": "issue:46222", + "accept": false, + "reason": "Different language/title and no concrete overlap beyond broad i18n similarity." + }, + { + "left": "issue:44370", + "right": "issue:46226", + "accept": false, + "reason": "Different request; not the same change." + }, + { + "left": "issue:44370", + "right": "issue:46931", + "accept": false, + "reason": "Different underlying translation task." + }, + { + "left": "issue:46222", + "right": "issue:46226", + "accept": true, + "reason": "Exact same Thai title and identical body size; these look like duplicates." + }, + { + "left": "issue:46222", + "right": "issue:46931", + "accept": false, + "reason": "Similar language family is not enough; the Thai titles describe different requests." + }, + { + "left": "issue:46226", + "right": "issue:46931", + "accept": false, + "reason": "Different Thai-titled request; no evidence it is the same issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: it distinguishes the generic English-template placeholder cluster from the separate language-specific translation requests, and it only elevates the exact-title pairs (43834/44365/44370 and 46222/46226) as duplicates. The canonical issue choice is also conservative and matches the oldest representative in the placeholder subgroup." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 6, + "component_count": 24, + "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-36718-19", + "nodes": [ + "issue:36718", + "issue:36817", + "issue:36879", + "issue:37015", + "issue:41115", + "issue:41211", + "issue:42503", + "issue:42738", + "issue:43671", + "issue:43837", + "issue:43909", + "issue:44121", + "issue:44205", + "issue:45027", + "issue:45177", + "issue:45322", + "issue:46133", + "issue:46645", + "issue:46653" + ], + "soft_pairs": [ + "issue:36718|issue:41115", + "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:42738|issue:46653", + "issue:43671|issue:43837", + "issue:43671|issue:45027", + "issue:43671|issue:45177", + "issue:43671|issue:44121", + "issue:43837|issue:43909", + "issue:43837|issue:45322", + "issue:43837|issue:44121", + "issue:43909|issue:44205", + "issue:43909|issue:46645", + "issue:44121|issue:45322", + "issue:43837|issue:45027", + "issue:37015|issue:45027", + "issue:43837|issue:45177", + "issue:45322|issue:46133", + "issue:42503|issue:46653", + "issue:45027|issue:46653" + ], + "prepared_review_unit_hash": "2770cd8f81b19707a4a3b8701b8cce596b92ca718dddfa807b892ecea7567981", + "component_budget": { + "node_count": 19, + "item_count": 19, + "soft_pair_count": 40, + "serialized_chars": 24676, + "estimated_input_tokens": 6169, + "estimated_eval_tokens": 12594 + }, + "budget": { + "node_count": 19, + "item_count": 19, + "soft_pair_count": 40, + "serialized_chars": 24676, + "estimated_input_tokens": 6169, + "estimated_eval_tokens": 12594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are mostly unrelated model-request issues connected only by generic \"add model/support\" wording; I see no true duplicate groups or mergeable PRs in the provided set.", + "confidence": 0.98, + "canonical_issue_reason": "No true duplicate anchor exists; the cluster is overbroad. If one issue must stand in, 36879 is the most generic and central model-add request.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 36879 is the best overall representative only as a broad model-addition request, but it should not absorb the other issues as duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36718", + "right": "issue:41115", + "accept": false, + "reason": "Different model domains: keypoint matching vs MiniCPM3 model support; shared phrasing only." + }, + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "Both are model-add requests, but EuroBert config and DEIMv2 are unrelated changes." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "Different models and modalities; BERT config vs Qwen3-TTS support." + }, + { + "left": "issue:36817", + "right": "issue:36879", + "accept": false, + "reason": "Both are generic model additions, but EuroBert and RF-DETR are distinct models with separate integration work." + }, + { + "left": "issue:36879", + "right": "issue:41211", + "accept": false, + "reason": "RF-DETR and DEIMv2 are different vision models; not the same underlying request." + }, + { + "left": "issue:36879", + "right": "issue:43909", + "accept": false, + "reason": "RF-DETR model addition is unrelated to adding LFM2.5 Audio 1.5B support." + }, + { + "left": "issue:36879", + "right": "issue:44121", + "accept": false, + "reason": "Different architectures and codepaths; RF-DETR vs weight-sparse transformer request." + }, + { + "left": "issue:36879", + "right": "issue:43837", + "accept": false, + "reason": "Different models and tasks; RF-DETR is not the same as Qwen3-ASR support." + }, + { + "left": "issue:36879", + "right": "issue:43671", + "accept": false, + "reason": "Different model families and modalities; RF-DETR vs Qwen3-TTS." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT and Qwen3-ASR are unrelated model requests." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT vs weight-sparse transformer are distinct model additions, not one bug/change." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "Different requests: NeoBERT and LFM2.5 Audio 1.5B." + }, + { + "left": "issue:36879", + "right": "issue:37015", + "accept": false, + "reason": "RF-DETR and NeoBERT are separate model integrations with no shared concrete change." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT and ModernVBERT are different models; only broad similarity in request type." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 model architecture request and Qwen3-ASR support are distinct." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 support is unrelated to adding a weight-sparse transformer implementation." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "MiniCPM3 and ModernVBERT are separate model requests." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "Different audio/text model support requests; no shared code-path problem." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 and Qwen3-TTS are unrelated model additions." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT models and Qwen3-ASR support are different integration tasks." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT and Qwen3-TTS are unrelated model requests." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "ModernVBERT models and generic BERT-like RoPE support are not the same concrete change." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "BERT-like models with RoPE and Voxtral TTS support target different model families." + }, + { + "left": "issue:42738", + "right": "issue:46653", + "accept": false, + "reason": "Generic model-architecture issue vs an unrelated personal-name placeholder issue." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "Same Qwen3 family, but TTS vs ASR are different tasks and likely separate changes." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-TTS and Voxtral-4B-TTS are separate model-support requests." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-TTS support is unrelated to DeepSeek-OCR2." + }, + { + "left": "issue:43671", + "right": "issue:44121", + "accept": false, + "reason": "Audio model support and a sparse-transformer proposal are different codepaths." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-ASR support and LFM2.5 Audio 1.5B are different model requests." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "Qwen3-ASR and EUPE are unrelated model integrations." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "ASR model support is not the same change as adding a weight-sparse transformer." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "LFM2.5 Audio 1.5B and SAM3-LiteText are different model families." + }, + { + "left": "issue:43909", + "right": "issue:46645", + "accept": false, + "reason": "Audio model support and SWA+sink for Granite are unrelated changes." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Weight-sparse transformer proposal and EUPE model request are different." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral-4B-TTS are distinct audio-model support requests." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT and Voxtral-4B-TTS are unrelated model additions." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-ASR support and DeepSeek-OCR2 are different model requests." + }, + { + "left": "issue:45322", + "right": "issue:46133", + "accept": false, + "reason": "EUPE and TIPSv2 are different vision-language encoder requests from different projects." + }, + { + "left": "issue:42503", + "right": "issue:46653", + "accept": false, + "reason": "ModernVBERT models and the placeholder-name issue have no substantive overlap." + }, + { + "left": "issue:45027", + "right": "issue:46653", + "accept": false, + "reason": "Voxtral TTS support is unrelated to the placeholder-name issue." + } + ] + }, + "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": 24, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-38740-2", + "nodes": [ + "issue:38740", + "issue:46443" + ], + "soft_pairs": [ + "issue:38740|issue:46443" + ], + "prepared_review_unit_hash": "9d3a1598745feb6ac90598e1963f56971048b875f94ed89685f01ec63cf6ca54", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1506, + "estimated_input_tokens": 377, + "estimated_eval_tokens": 1010 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1506, + "estimated_input_tokens": 377, + "estimated_eval_tokens": 1010 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are unrelated: one is a docs request to add Pruna as an optimization framework, the other is a technical Vision RoPE refactor. The soft similarity signal is not enough to treat them as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": "Prefer issue 46443 as the cluster representative: it is the substantive technical issue, while 38740 is a separate docs-oriented request with no overlap in problem statement.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 46443 is the best issue in this set to surface as the representative artifact because it is a concrete engineering task and the only one plausibly tied to a code change.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:38740", + "right": "issue:46443", + "accept": false, + "reason": "Different topics and artifact types: a documentation/addition request for Pruna vs. a Vision RoPE refactor. No shared targets, filenames, or concrete code-path overlap." + } + ] + }, + "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": 24, + "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": "c071042d5dc526504e268c89c3cf1b22f83f9191ff5b4dbc1af5fa1d00a8ab64", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are unrelated: one is about block decoding, the other requests DeepSeek v3.2 support. They do not describe the same bug or change, so the soft edge should not be merged.", + "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": "Different topics and scopes: block decoding versus model/version support. No shared code path or concrete duplicate behavior 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": 9, + "component_count": 24, + "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-41306-2", + "nodes": [ + "issue:41306", + "issue:41826" + ], + "soft_pairs": [ + "issue:41306|issue:41826" + ], + "prepared_review_unit_hash": "6de6e7015ee51e418111acc4051ac3f475e2f7a2fb1855af5e49174be5ba4b6d", + "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": "The two issues are related only by a broad theme of tiled/memory-efficient layers, but they describe different components and different asks: one is about a `from_pretrained()` flag for TiledFusedLogitsLoss / LigerFusedLinearCrossEntropy, the other is about integrating TiledMLP for lower memory usage.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 41306 is the more concrete, self-contained request: it names the exact feature (`from_pretrained()` flag) and the specific fused loss path affected.", + "canonical_pr_reason": null, + "best_issue_reason": "41306 is the best representative because it is narrowly scoped and likely resolvable without needing broader architectural changes.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41306", + "right": "issue:41826", + "accept": false, + "reason": "Different underlying changes: one concerns pretrained loading behavior for a fused logits/loss implementation, the other concerns adding TiledMLP for memory reduction. Shared tiling vocabulary is too broad to treat 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": 10, + "component_count": 24, + "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": "75cfdbcf387dd81bea5a2a892eb599947140ea57e95fee6eb318487f106a1dee", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are only loosely related through custom trainers, but they describe different problems: one is about `remove_unused_column` behavior, the other about FLOPS estimation. They should not be treated as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": "No canonical issue here: 43598 concerns Trainer column pruning/customizability, while 43600 concerns FLOPS counting for custom trainers.", + "canonical_pr_reason": null, + "best_issue_reason": "If forced to pick a representative issue, 43598 is the more concrete Trainer behavior request, but it still does not subsume 43600.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "They share the broad custom-trainer context, but the underlying changes are different: one targets unused-column removal, the other FLOPS estimation. Not the same bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 11, + "component_count": 24, + "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": "65c53f4a0574159181a949dd2afbc64771282b466de2c2db42b687f7a575c077", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related Qwen-model support requests, but they target different model families and different auto-class APIs: Qwen3-Omni registration vs Qwen3.5 sequence-classification support. They are not the same underlying change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43684 is the broader foundational registration request, but it does not cover the separate AutoModelForSequenceClassification support requested in 44405.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Different model/config and different API surface: Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration versus Qwen3.5 AutoModelForSequenceClassification support. Not the same bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 12, + "component_count": 24, + "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": "e40aa6528e1a0ca4e4097ad9358492bf81352b0d6e165ac56aaead653abb29b7", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4264, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4264, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The three \"huggingface \" issues are very similar, but the packet only provides soft_similarity and no deterministic evidence, so none of the pairings should be accepted as duplicates. \"Fett\" still looks unrelated. No PRs are present.", + "confidence": 0.89, + "canonical_issue_reason": "issue:43845 is the earliest of the three repeated \"huggingface \" issues, so it is the best canonical issue candidate.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43845 is the safest cluster representative because it is the earliest of the repeated-title issues, even though the duplicate evidence is limited.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "Only weak similarity is present, with different titles and no shared targets, filenames, or content evidence tying them to the same bug." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "\"Fett\" does not match the repeated \"huggingface \" reports, and there is no concrete evidence of the same underlying issue." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "The title is unrelated to the other reports and the packet contains no shared issue-specific evidence." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": false, + "reason": "These are the strongest match because they share the same title, but the evidence is still only soft similarity with no deterministic support." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": false, + "reason": "Same title, but the packet provides no concrete details beyond sparse metadata, so this is not enough to merge them confidently." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": false, + "reason": "Same-title similarity exists, but there is no additional evidence showing they describe the same underlying report." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster summary is mostly grounded, but the soft-edge judgments are too strong for the evidence in the packet. All pairings are supported only by soft_similarity, and the packet\u2019s deterministic_accept is false for every pair. In particular, the acceptance of 43845\u219443859 and 43846\u219443859 overstates what can be inferred from shared title plus sparse metadata; 43845\u219443846 is the strongest match, but the evidence is still limited. The canonical issue choice (43845 as earliest of the three 'huggingface ' issues) is fine." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 24, + "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": "fe46d237ac961ca84d9955a2ae3b7e6da0a7a6f87bedbc1f2078ea5d59274fb9", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These look like unrelated low-information spam/vague issues, not a real duplicate cluster. There isn\u2019t enough evidence to merge any pair as the same underlying report.", + "confidence": 0.69, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:44418 is the only item with a somewhat descriptive, non-spam title, but it still lacks body detail and does not clearly match the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44371", + "right": "issue:44572", + "accept": false, + "reason": "Both are titled \"\" with no body or targets; that is not enough to prove the same underlying issue." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Both are titled \"\" with no substantive details, so they could be unrelated spam posts." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "One is spam-like, the other is a generic documentation suggestion; no evidence they report the same problem." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder vs. \"Check out 'Google LLC'\" are not the same concrete issue." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "A documentation suggestion and a spam placeholder do not indicate the same underlying bug/change." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "A documentation suggestion and a spam placeholder are too different to treat as duplicates." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Same placeholder title, but no body or contextual evidence tying them to the same report." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder vs. unrelated \"Google LLC\" title; no shared concrete issue is visible." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "No sign these are the same underlying issue; the titles are unrelated and the bodies are empty." + } + ] + }, + "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": 24, + "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": "1bb9be3a43e20797a8835bc8d3cff7c03959b505968200314f6a2ff02ca619bd", + "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 closed issues have the same title, highly similar body size, and strong textual overlap, so they appear to be duplicate reports of the same underlying energy-efficiency proposal. Issue #45145 is the better canonical pick because it has an inbound reference; no PRs are present.", + "confidence": 0.98, + "canonical_issue_reason": "#45145 and #45151 share the exact title and appear to describe the same proposal. #45145 is preferable as canonical because it has an inbound reference, suggesting it is the more-referenced duplicate target.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the best issue candidate: same subject as #45151, but with at least one inbound reference and no evidence that #45151 is more central.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": true, + "reason": "Same title and near-identical topic/body footprint; these look like duplicate issues for the same underlying request." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The analyst stays grounded in the packet: the two issues have the exact same title, similar body lengths, and a soft-similarity signal, so treating them as likely duplicates is conservative. Choosing #45145 as canonical because it has the only inbound reference is also supported. No PR claims were made." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 15, + "component_count": 24, + "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": 16, + "component_count": 24, + "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": "1fd044e2d53b4cb29e95bdcd307c5aa7e64ed8653376779ecea2cad58a34a755", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1476, + "estimated_input_tokens": 369, + "estimated_eval_tokens": 994 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1476, + "estimated_input_tokens": 369, + "estimated_eval_tokens": 994 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two closed issues are just person-name/date entries and do not indicate the same underlying bug or request; this looks like unrelated content, not a duplicate pair.", + "confidence": 0.97, + "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 names and birthdates with no shared issue target, body, or code context; they do not appear to describe the same problem 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": 24, + "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": "e412a072a723cc057635f78755c179f4d30711a8e73f5ac620baa0b3b85824b2", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1479, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1479, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two closed issues do not appear to describe the same bug or request. They are just short name/date titles with no body or shared technical context, so the soft similarity signal is not enough to treat them as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": "No clear canonical issue: both issues lack descriptions and appear to be unrelated name/date entries rather than a shared problem report.", + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a good canonical choice; both are equally non-technical and provide no actionable context.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45764", + "right": "issue:45765", + "accept": false, + "reason": "The titles only resemble personal names with dates and provide no evidence of the same underlying bug, task, or code path." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 18, + "component_count": 24, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45810-2", + "nodes": [ + "issue:45810", + "issue:46295" + ], + "soft_pairs": [ + "issue:45810|issue:46295" + ], + "prepared_review_unit_hash": "eff1fda2d23b228a1b250da814dd1884374379aacdc19a65a952557bc302bf38", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1541, + "estimated_input_tokens": 386, + "estimated_eval_tokens": 1028 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1541, + "estimated_input_tokens": 386, + "estimated_eval_tokens": 1028 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster does not look like a duplicate set: one issue asks for a new Qwen3.5 token-classification model head, while the other requests FP8 training support. They are different feature areas and code paths.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue 46295 is the better cluster representative only because it appears more broadly discussed and referenced, but it is not a true duplicate of 45810.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45810", + "right": "issue:46295", + "accept": false, + "reason": "Different requests: model-classification support vs FP8 training support. No shared target, filename, or underlying bug/change path." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 19, + "component_count": 24, + "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": "4af850637523eaf55e5d3dc0940b980d3753fcfbde552c8ab3d571f1aa4a8618", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5727, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5727, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The set is not one coherent duplicate cluster: it contains two separate exact-duplicate issue pairs (45870/45871 and 45872/45873), while 45869 is only loosely related by topic and should not be merged with either pair.", + "confidence": 0.98, + "canonical_issue_reason": "45870 is the cleanest canonical candidate for its subcluster: 45871 is an exact later duplicate, with identical title/body shape, and 45870 is the earliest instance. The other exact-duplicate pair (45872/45873) is a separate issue theme, so there is no single canonical issue for all five items.", + "canonical_pr_reason": null, + "best_issue_reason": "45870 is the best representative within the 'Example: Financial time series prediction with DataForge' duplicate pair because it is an exact duplicate target and the earliest submission. That said, the overall set splits into two unrelated duplicate pairs, so no one issue cleanly represents the entire cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45869", + "right": "issue:45870", + "accept": false, + "reason": "Related financial-data topic, but one is a dataset promo and the other is an example issue; not the same underlying bug/change." + }, + { + "left": "issue:45869", + "right": "issue:45871", + "accept": false, + "reason": "Same as above: similar theme, but different artifact and intent." + }, + { + "left": "issue:45870", + "right": "issue:45871", + "accept": true, + "reason": "Exact duplicate pair: same title and matching body size, with no evidence of a different underlying issue." + }, + { + "left": "issue:45870", + "right": "issue:45872", + "accept": false, + "reason": "Both mention financial time series/DataForge, but one is an example and the other adds a scraping service; these are different issue requests." + }, + { + "left": "issue:45870", + "right": "issue:45873", + "accept": false, + "reason": "Different underlying content and scope; not the same issue." + }, + { + "left": "issue:45871", + "right": "issue:45872", + "accept": false, + "reason": "Different issue themes despite shared finance/DataForge keywords." + }, + { + "left": "issue:45871", + "right": "issue:45873", + "accept": false, + "reason": "Different issue themes despite shared finance/DataForge keywords." + }, + { + "left": "issue:45872", + "right": "issue:45873", + "accept": true, + "reason": "Exact duplicate pair: same title and matching body size, indicating the same underlying issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The conclusions are grounded in the packet: 45870/45871 and 45872/45873 are the only clearly duplicate-looking pairs by title/body length and creation order, while 45869 is only loosely related. The soft-edge rejections are conservative, and the note that there is no single canonical issue for all five items is supported by the split into two separate pairs." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 20, + "component_count": 24, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46439-2", + "nodes": [ + "issue:46439", + "issue:46441" + ], + "soft_pairs": [ + "issue:46439|issue:46441" + ], + "prepared_review_unit_hash": "24a1ba1b9ca4b5c680683fcdad2897c2e26be82c44d689e5962a1cb7435368d3", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1636, + "estimated_input_tokens": 409, + "estimated_eval_tokens": 1074 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1636, + "estimated_input_tokens": 409, + "estimated_eval_tokens": 1074 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both issues point at static-cache problems around linear-attention models, but they describe different concrete failures: one is shape corruption leading to a RuntimeError, the other is a KeyError for a missing `linear_attention` cache entry in Qwen3.5. They are related in subsystem, but not clearly the same underlying bug.", + "confidence": 0.67, + "canonical_issue_reason": "Issue 46439 is the better canonical issue because it describes the broader, more actionable failure mode in `early_initialization` and names the affected component (`linear-attention` layers) rather than a single model symptom.", + "canonical_pr_reason": null, + "best_issue_reason": "46439 is the strongest representative of the cluster: it captures the likely root-cause area and the generic breakage pattern, while 46441 reads as a specific manifestation for Qwen3.5.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46439", + "right": "issue:46441", + "accept": false, + "reason": "Related static-cache/linear-attention symptoms, but one is a shape-corruption RuntimeError and the other is a KeyError for `linear_attention`; not enough evidence they are the same exact bug or would be fixed by one change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 21, + "component_count": 24, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46551-2", + "nodes": [ + "issue:46551", + "issue:46552" + ], + "soft_pairs": [ + "issue:46551|issue:46552" + ], + "prepared_review_unit_hash": "ea98469aa76957dc679bcc105e9236fe0bb37b9e7c3556252bac4a330f777477", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1707, + "estimated_input_tokens": 427, + "estimated_eval_tokens": 1110 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1707, + "estimated_input_tokens": 427, + "estimated_eval_tokens": 1110 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are duplicates of the same Wav2Vec2CTCTokenizer v5 bug report; they have identical titles and bodies, differing only in issue number and activity state.", + "confidence": 0.99, + "canonical_issue_reason": "issue:46552 is the better canonical issue because it is the open, more-referenced copy of the same report and is more likely to be the active tracking item.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:46552 is the best representative of the cluster since it is the same problem statement and has inbound references, making it the likelier primary tracker.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46551", + "right": "issue:46552", + "accept": true, + "reason": "Same title, same body, and same underlying tokenizer bug; these are duplicate reports of the exact same issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The conclusion is well grounded: the two issues have identical titles and bodies, and the packet\u2019s soft-similarity signal is maximal (Jaccard 1.0). The canonical-issue preference for issue:46552 is also supported by its open state and higher inbound references. The soft-edge verdict is conservative enough." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 22, + "component_count": 24, + "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-46657-2", + "nodes": [ + "issue:46657", + "issue:46692" + ], + "soft_pairs": [ + "issue:46657|issue:46692" + ], + "prepared_review_unit_hash": "853dc8c1addcfacb8462cbcd006c8a286402c58ac097c0c3daeab5219dfee814", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1569, + "estimated_input_tokens": 393, + "estimated_eval_tokens": 1042 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1569, + "estimated_input_tokens": 393, + "estimated_eval_tokens": 1042 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are both about training-time failures, but they describe different underlying problems: a tensor size mismatch vs. out-of-memory during RT-DETR v2 training. They should not be merged as duplicates.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 46657 is the cleaner representative bug report: it names a concrete tensor-shape mismatch during object detection training. Issue 46692 is a different failure mode entirely (OOM after several epochs).", + "canonical_pr_reason": null, + "best_issue_reason": "46657 is the better cluster representative because it is a specific, actionable defect report with a clearer error class. 46692 is separate and should remain distinct.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46657", + "right": "issue:46692", + "accept": false, + "reason": "Different symptoms and likely different root causes: tensor size mismatch vs. memory exhaustion during RT-DETR v2 training." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 23, + "component_count": 24, + "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-46822-2", + "nodes": [ + "issue:46822", + "issue:46823" + ], + "soft_pairs": [ + "issue:46822|issue:46823" + ], + "prepared_review_unit_hash": "5055e8d779e9d8d8ef1390c548c74427c4a1548ce06d92b9a4fe6c7e766cffa8", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1672, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1672, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues both concern model-loading memory/device placement, but they describe different failure modes in different pathways: ZeRO-3 partitioning of a composite submodule vs 4-bit auto device mapping ignoring max_memory and pushing modules to CPU. They are not duplicates.", + "confidence": 0.96, + "canonical_issue_reason": "Issue #46822 is the more specific and actionable report: it names the exact failing path (ZeRO-3 zero.Init), the affected composite submodule, and the observed OOM during multi-GPU load.", + "canonical_pr_reason": null, + "best_issue_reason": "#46822 is the stronger canonical issue candidate because it describes a concrete partitioning bug with a clear model/component target and likely narrower root cause than #46823.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46822", + "right": "issue:46823", + "accept": false, + "reason": "Both are about loading-time memory placement, but #46822 is a ZeRO-3 partitioning/OOM bug while #46823 is a 4-bit device_map/max_memory dispatch bug. Different code paths and failure symptoms, so they should not be merged." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 24, + "component_count": 24, + "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-46906-3", + "nodes": [ + "issue:46906", + "issue:46940", + "issue:46948" + ], + "soft_pairs": [ + "issue:46906|issue:46940", + "issue:46906|issue:46948", + "issue:46940|issue:46948" + ], + "prepared_review_unit_hash": "1113d0b2d4a906c429c481a4e5c6990d1f9d7270b1849ff02504620fbd6ea0c9", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2807, + "estimated_input_tokens": 702, + "estimated_eval_tokens": 1660 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2807, + "estimated_input_tokens": 702, + "estimated_eval_tokens": 1660 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three items are nearly identical day-stamped integration triage issues for Serge, but the differing dates strongly suggest separate failure instances rather than duplicates of the same bug report.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue 46906 is the best representative/anchor because it is the earliest of the three and has the most inbound references, but it should not be treated as a duplicate-canonical merge target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46906", + "right": "issue:46940", + "accept": false, + "reason": "Same triage template and subsystem, but the different date in the title indicates a different daily failure instance, not the same underlying issue." + }, + { + "left": "issue:46906", + "right": "issue:46948", + "accept": false, + "reason": "High textual similarity, but this is a distinct date-stamped triage issue and likely corresponds to a separate integration run/failure." + }, + { + "left": "issue:46940", + "right": "issue:46948", + "accept": false, + "reason": "These are recurring daily triage reports with matching wording, yet the date-specific titles imply separate incidents rather than one shared bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 1, + "component_count": 38, + "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": "a7430a30ce33a0a618c32addac3de23617980e9efc9392f33234f2cd68a18e76", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7223, + "estimated_input_tokens": 1806, + "estimated_eval_tokens": 3868 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7223, + "estimated_input_tokens": 1806, + "estimated_eval_tokens": 3868 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two VibeVoice PRs overlap heavily in base model files, but they target different scopes: #40546 adds the initial VibeVoice implementation, while #42781 extends it to a separate realtime/acoustic-tokenizer feature set. They are related, but not duplicates.", + "confidence": 0.82, + "canonical_issue_reason": null, + "canonical_pr_reason": "#40546 is the better canonical PR for the cluster because it represents the original VibeVoice integration and has substantially more review activity; #42781 is a follow-up feature expansion rather than the same change.", + "best_issue_reason": null, + "best_pr_reason": "#40546 is the best representative of the underlying VibeVoice addition. It is the broader base implementation, whereas #42781 introduces a distinct realtime variant with extra tokenizer and generation changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Reject: despite many shared VibeVoice files, #42781 is not the same concrete change. It adds realtime support and an acoustic tokenizer, plus extra generation/auto wiring, while #40546 is the initial VibeVoice implementation. They are related follow-ups, 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": 2, + "component_count": 38, + "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": "d866530508a04d7047f6e760f1d80ebc0fb3e0439534a123648620f65a62f96c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two pull requests are unrelated: one adds covariate forecasting support to TimesFM, while the other adds DINOv3 image-classification support. They touch different models, different code paths, and different documentation/tests, so 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": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Different underlying changes: TimesFM forecasting with covariates vs DINOv3 image classification support. No shared bug or mergeable code-path overlap." + } + ] + }, + "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": 38, + "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": 4, + "component_count": 38, + "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-41855-13", + "nodes": [ + "pull_request:41855", + "pull_request:42668", + "pull_request:43251", + "pull_request:43254", + "pull_request:43424", + "pull_request:43823", + "pull_request:43995", + "pull_request:44827", + "pull_request:45170", + "pull_request:45213", + "pull_request:45655", + "pull_request:46048", + "pull_request:46267" + ], + "soft_pairs": [ + "pull_request:41855|pull_request:43823", + "pull_request:41855|pull_request:43424", + "pull_request:42668|pull_request:43823", + "pull_request:43251|pull_request:43823", + "pull_request:43251|pull_request:44827", + "pull_request:43424|pull_request:44827", + "pull_request:43424|pull_request:43823", + "pull_request:43823|pull_request:45655", + "pull_request:43823|pull_request:43995", + "pull_request:43995|pull_request:45213", + "pull_request:43995|pull_request:44827", + "pull_request:44827|pull_request:45170", + "pull_request:44827|pull_request:46267", + "pull_request:44827|pull_request:46048", + "pull_request:44827|pull_request:45655", + "pull_request:44827|pull_request:45213", + "pull_request:43823|pull_request:45213", + "pull_request:43251|pull_request:43254" + ], + "prepared_review_unit_hash": "0a0daf5d8432d703936e3c945841004ae98f3b047662915bee6dbec1b38233f8", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 18, + "serialized_chars": 34782, + "estimated_input_tokens": 8696, + "estimated_eval_tokens": 17648 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 18, + "serialized_chars": 34782, + "estimated_input_tokens": 8696, + "estimated_eval_tokens": 17648 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly false-positive similarity across unrelated PRs. The only true duplicate pair is the two `fixed_cross_entropy` PRs (`43251` and `43254`), which implement the same fix in the same function with nearly identical changes. All other pairs touch different subsystems and should stay separate.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43251 is the best canonical representative for the duplicate pair because it has the explicit issue link (#43240) and the more complete PR framing for the `fixed_cross_entropy` kwargs fix.", + "best_issue_reason": null, + "best_pr_reason": "PR #43251 is the strongest global representative among the items because it directly addresses a tracked bug with a clear code-path fix; however, the rest of the cluster is too heterogeneous to treat as one duplicate family.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Different changes: Mistral tokenizer helper methods vs adding MobileLLM model files. No shared bug or mergeable code-path overlap." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Tokenizer API additions are unrelated to Executorch dynamic-shape export tests." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor auto-loading robustness vs introducing a new model implementation are unrelated." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "`fixed_cross_entropy` kwargs support is unrelated to adding MobileLLM." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "Loss utility kwargs forwarding and Mistral4 test fixes are different code paths and bugs." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "Executorch dynamic-shape export tests do not match Mistral4 test fixes." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "Different areas: export test coverage vs new model addition." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "New model support and config-resolution order fix are unrelated changes." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "MobileLLM model addition and Falcon output-collection refactor do not fix the same concrete problem." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon model refactor is unrelated to the new-model creation skill PR." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Falcon interface refactor and Mistral4 test fixes are not the same underlying bug." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 test fixes and a LayerNorm typo rename across CLIP-like models are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46267", + "accept": false, + "reason": "Mistral4 tests do not overlap with chat response parsing updates." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46048", + "accept": false, + "reason": "Mistral4 tests and TDT loss kernel integration are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "Mistral4 test work is unrelated to config resolution ordering." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 tests and the model-creation skill are unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "A new model implementation and an AI skill for model creation are not the same change." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Falcon refactor and Mistral4 tests are different model-specific changes." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Both PRs fix the same concrete `fixed_cross_entropy` code path by adding support for passing supported kwargs (`weight`, `label_smoothing`) to `nn.functional.cross_entropy`; they are mergeable into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet and stays conservative: the non-43251/43254 pairs are plausibly unrelated, and the 43251/43254 pair is a legitimate same-function, same-bugfix duplicate that looks mergeable into one PR." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 5, + "component_count": 38, + "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": "676121bb0415cc6e29d2be0633dd53177569852d2e2dfb9976424e114a003b39", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5934, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5934, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs both update workflow action versions, but they are different upgrade efforts: one is a broad GitHub Actions/Node 24 compatibility sweep, the other is a docker/build-push-action version bump. They should not be merged as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: 42975 and 42976 are separate dependency upgrades affecting different action ecosystems and different code paths, so they are not the same change.", + "best_issue_reason": null, + "best_pr_reason": "PR 42975 is the better representative of the cluster because it is the broader GitHub Actions compatibility sweep across many workflows; 42976 is a narrower, separate build-action version bump.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "Reject: 42975 upgrades multiple GitHub Actions to newer versions for Node 24 compatibility, while 42976 only bumps docker/build-push-action in build-related workflows. Same broad area, but not the same concrete fix and not mergeable as one duplicate PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 6, + "component_count": 38, + "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-42978-6", + "nodes": [ + "pull_request:42978", + "pull_request:43448", + "pull_request:43451", + "pull_request:45490", + "pull_request:46562", + "pull_request:46658" + ], + "soft_pairs": [ + "pull_request:42978|pull_request:43448", + "pull_request:42978|pull_request:43451", + "pull_request:42978|pull_request:45490", + "pull_request:43448|pull_request:43451", + "pull_request:43448|pull_request:45490", + "pull_request:43451|pull_request:45490", + "pull_request:45490|pull_request:46562", + "pull_request:46562|pull_request:46658" + ], + "prepared_review_unit_hash": "840417c95fe3547e38f8df92c5cc6a89b58fc13d9eea9557f00360f983571ba1", + "component_budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 8, + "serialized_chars": 21197, + "estimated_input_tokens": 5300, + "estimated_eval_tokens": 10856 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 8, + "serialized_chars": 21197, + "estimated_input_tokens": 5300, + "estimated_eval_tokens": 10856 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are mostly unrelated model-addition efforts. The only near-match is Step3p7 Flash vs its mini-model/debugging companion, but that is still a different deliverable from the core Transformers implementation, so no duplicates are accepted.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46562 is the main implementation for Step 3.7 Flash in Transformers, with the broadest code surface across model, processors, auto mappings, docs, and tests.", + "best_issue_reason": null, + "best_pr_reason": "PR 46562 is the strongest standalone PR in this cluster: it is the substantive model integration, not just docs or support scripts, and has the clearest scope and review activity.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42978", + "right": "pull_request:43448", + "accept": false, + "reason": "Different models entirely: ViT NEPA vs Molmo. Shared docs/auto-registry edits are incidental." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "Different models entirely: ViT NEPA vs Molmo2. Similar repository scaffolding does not indicate the same change." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "Different models entirely: ViT NEPA vs CTSM. Only generic docs/auto files overlap." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are distinct model integrations, not one underlying bug or code-path fix." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo and CTSM are unrelated model additions; shared framework files are not enough to merge them." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM are separate model additions with different architectures and files." + }, + { + "left": "pull_request:45490", + "right": "pull_request:46562", + "accept": false, + "reason": "CTSM and Step3p7 Flash are unrelated model integrations; overlap is limited to generic registry/doc updates." + }, + { + "left": "pull_request:46562", + "right": "pull_request:46658", + "accept": false, + "reason": "Both are Step 3.7 Flash related, but 46562 adds the Transformers model while 46658 adds mini-model/conversion/debug scripts and artifacts; they are companion efforts, not the same concrete change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 7, + "component_count": 38, + "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-43291-7", + "nodes": [ + "pull_request:43291", + "pull_request:43665", + "pull_request:44815", + "pull_request:46380", + "pull_request:46539", + "pull_request:46886", + "pull_request:46922" + ], + "soft_pairs": [ + "pull_request:43291|pull_request:43665", + "pull_request:43291|pull_request:44815", + "pull_request:43291|pull_request:46380", + "pull_request:43291|pull_request:46922", + "pull_request:43665|pull_request:46539", + "pull_request:43665|pull_request:44815", + "pull_request:43665|pull_request:46380", + "pull_request:43665|pull_request:46922", + "pull_request:44815|pull_request:46539", + "pull_request:44815|pull_request:46380", + "pull_request:44815|pull_request:46922", + "pull_request:46380|pull_request:46539", + "pull_request:46380|pull_request:46922", + "pull_request:46539|pull_request:46922", + "pull_request:46886|pull_request:46922" + ], + "prepared_review_unit_hash": "d86633df9a370d81601f7405b4e87868c8372ac848a13fa1d5506167422e4a23", + "component_budget": { + "node_count": 7, + "item_count": 7, + "soft_pair_count": 15, + "serialized_chars": 19604, + "estimated_input_tokens": 4901, + "estimated_eval_tokens": 10058 + }, + "budget": { + "node_count": 7, + "item_count": 7, + "soft_pair_count": 15, + "serialized_chars": 19604, + "estimated_input_tokens": 4901, + "estimated_eval_tokens": 10058 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly unrelated PRs across different models and subsystems. The only clear duplicate pair is the two YOLOS fixes, which target the same device-mismatch bug in the same code path.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46922 is the strongest representative because it includes the concrete YOLOS runtime fix (`.to(hidden_states.device)`) and the same device-map stabilization change, making it the more complete version of the shared bug fix.", + "best_issue_reason": null, + "best_pr_reason": "PR 46922 is the best overall representative: it directly fixes the YOLOS device mismatch and also keeps the supporting `_no_split_modules` adjustment, so it subsumes the narrower 46886 change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Both are superficial test-fix PRs, but they touch different models (`whisper` vs `clip`/`vit`) and do not address the same underlying bug." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper tokenizer/test changes are unrelated to the FP8 dequantization/loading fix in 44815." + }, + { + "left": "pull_request:43291", + "right": "pull_request:46380", + "accept": false, + "reason": "Whisper test/code changes are unrelated to the generation-config watermarking fix in 46380." + }, + { + "left": "pull_request:43291", + "right": "pull_request:46922", + "accept": false, + "reason": "Whisper test changes and YOLOS device-mismatch fixes are different models and different bugs." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46539", + "accept": false, + "reason": "These are unrelated test changes for different areas: CLIP/ViT vs Phimoe CI/integration." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "The CLIP/ViT test-only PR is unrelated to the FP8 loading/dequant fix." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46380", + "accept": false, + "reason": "Test-only changes in CLIP/ViT do not match the generation watermarking bugfix." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46922", + "accept": false, + "reason": "Different models and different failure modes; no shared code-path bug." + }, + { + "left": "pull_request:44815", + "right": "pull_request:46539", + "accept": false, + "reason": "FP8 dequantization/loading changes are unrelated to Phimoe integration-test CI setup." + }, + { + "left": "pull_request:44815", + "right": "pull_request:46380", + "accept": false, + "reason": "These fixes target different subsystems: quantization/loading vs generation config validation." + }, + { + "left": "pull_request:44815", + "right": "pull_request:46922", + "accept": false, + "reason": "No common bug: FP8 kernel/padding work is unrelated to YOLOS device placement." + }, + { + "left": "pull_request:46380", + "right": "pull_request:46539", + "accept": false, + "reason": "Generation watermarking validation and Phimoe CI fixes are different issues in different paths." + }, + { + "left": "pull_request:46380", + "right": "pull_request:46922", + "accept": false, + "reason": "Generation config watermarking validation is unrelated to YOLOS device-mismatch behavior." + }, + { + "left": "pull_request:46539", + "right": "pull_request:46922", + "accept": false, + "reason": "Phimoe integration-test/CI adjustments do not share the same underlying bug as YOLOS device placement." + }, + { + "left": "pull_request:46886", + "right": "pull_request:46922", + "accept": true, + "reason": "Both PRs fix the same YOLOS device-mismatch problem on the same file and same issue target; 46922 is the more complete variant of the same change." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet. The only clearly mergeable duplicate pair is 46886/46922: both target the same YOLOS file, same issue target, and 46922 contains the same `_no_split_modules` change plus the additional device placement fix. The other pairwise rejections are conservative and consistent with the diffs, since they concern different models/subsystems and do not look like a single bugfix PR." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 8, + "component_count": 38, + "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-43488-3", + "nodes": [ + "pull_request:43488", + "pull_request:43532", + "pull_request:46747" + ], + "soft_pairs": [ + "pull_request:43488|pull_request:46747", + "pull_request:43488|pull_request:43532", + "pull_request:43532|pull_request:46747" + ], + "prepared_review_unit_hash": "3b526447904312f0aa6220e20920bf20e4628b27909b56d66496db7115e2b288", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 6348, + "estimated_input_tokens": 1587, + "estimated_eval_tokens": 3430 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 6348, + "estimated_input_tokens": 1587, + "estimated_eval_tokens": 3430 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The three PRs are all unrelated test/dummy changes and do not look like duplicates of one underlying bug or fix, even though two touch workflow-related files and one shares a Vitest test filename.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43532 is the most representative and broadest of the three, but it is still a distinct test-only workflow/diff-display change rather than the same fix as the others.", + "best_issue_reason": null, + "best_pr_reason": "PR #43532 has the largest and broadest diff surface, so it is the best single representative of the cluster if one must be chosen.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:46747", + "accept": false, + "reason": "Different concrete changes: setup/dependency/version tweaks versus a workflow security-gate change. They do not fix the same code-path problem." + }, + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "Only weak overlap via a shared Vitest test file; the actual changes are unrelated (dependency/version setup vs workflow/testing-utils edits)." + }, + { + "left": "pull_request:43532", + "right": "pull_request:46747", + "accept": false, + "reason": "Both touch workflows, but they alter different CI behaviors and files. They are not the same underlying bug/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": 38, + "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": "5f3cc201525aedeaa173bc1d8df92d97b0fee60dfa03bedbbc191467a2969bf7", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6082, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6082, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are related pipeline additions, but they implement different features: promptable concept segmentation vs promptable visual segmentation. They share scaffolding and docs structure, yet not the same concrete code path or change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: the two pull requests add different pipelines and should not be treated as duplicates.", + "best_issue_reason": null, + "best_pr_reason": "Neither PR is a duplicate candidate; if a representative is needed, 43613 is broader, but it is not a canonical match for 43612.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Different underlying features: concept segmentation vs visual segmentation. Shared docs/pipeline scaffolding is not enough to treat them as the same change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 10, + "component_count": 38, + "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": "d04bdee0700e80e1910b21d2616383b7bc682bb950d08d4278696793f10490fa", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4291, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4291, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two pull requests both edit `trainer.py`, but they address different trainer extension points: one changes gradient-accumulation loss scaling, the other factors out signature-column filtering for PEFT/custom trainers. They are not the same underlying bug or mergeable change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43651 is the more concrete behavior change: it fixes/extends training loss scaling logic in a single code path and has a clear user-facing effect. PR #43663 is also a refactor/extension point, but it targets a different mechanism.", + "best_issue_reason": null, + "best_pr_reason": "PR #43651 is the best representative because it changes a specific runtime behavior in `training_step` with a clear problem statement, whereas PR #43663 is about exposing a helper for signature-column filtering and is orthogonal.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Shared file only; the diffs touch different trainer subsystems and solve unrelated problems. They would not plausibly be merged as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 11, + "component_count": 38, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43656-2", + "nodes": [ + "pull_request:43656", + "pull_request:43842" + ], + "soft_pairs": [ + "pull_request:43656|pull_request:43842" + ], + "prepared_review_unit_hash": "e37f4f10366f93a04966c26ef0c6298f34331224b45cc6539e441e204010bf2e", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3662, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3662, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests touch `src/transformers/cli/serve.py` and are about a `TypeAdapter`-related NameError, but they fix different runtime paths: one is a broader annotation/type-hint cleanup, the other is specifically about pydantic not being installed. They are related, but not the same underlying change, so they should not be deduplicated as one PR.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43656 is the better canonical candidate because it is the earlier, more general fix for the CLI `TypeAdapter` NameError and has the explicit issue linkage plus more discussion activity.", + "best_issue_reason": null, + "best_pr_reason": "PR #43656 is the strongest standalone representative of the cluster: it targets the concrete CLI `TypeAdapter` NameError directly, has the explicit issue reference, and appears to address the broader code-path problem rather than only a narrower optional-dependency typing case.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": false, + "reason": "Do not merge as duplicates: #43656 appears to fix a general annotation/runtime NameError in `serve.py`, while #43842 specifically guards the `TypeAdapter` type for environments without pydantic. Same file and related symptom, but different concrete bug/patch." + } + ] + }, + "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": 38, + "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": 13, + "component_count": 38, + "review_unit_index": 1, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43996", + "pull_request:43997", + "pull_request:44007", + "pull_request:44010", + "pull_request:44013", + "pull_request:44017", + "pull_request:44018", + "pull_request:44019", + "pull_request:44024", + "pull_request:44044", + "pull_request:44056", + "pull_request:44059", + "pull_request:44066", + "pull_request:44068", + "pull_request:44071", + "pull_request:44072", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44085", + "pull_request:44086", + "pull_request:44098", + "pull_request:44101", + "pull_request:44116", + "pull_request:44129", + "pull_request:44154", + "pull_request:44161", + "pull_request:44722" + ], + "soft_pairs": [ + "pull_request:43996|pull_request:44085", + "pull_request:43996|pull_request:44044", + "pull_request:44066|pull_request:44085", + "pull_request:44007|pull_request:44072", + "pull_request:44072|pull_request:44722", + "pull_request:44066|pull_request:44072", + "pull_request:44013|pull_request:44044", + "pull_request:44066|pull_request:44086", + "pull_request:44018|pull_request:44068", + "pull_request:44066|pull_request:44071", + "pull_request:44066|pull_request:44068", + "pull_request:44018|pull_request:44066", + "pull_request:44019|pull_request:44722", + "pull_request:44086|pull_request:44722", + "pull_request:44019|pull_request:44085", + "pull_request:44018|pull_request:44019", + "pull_request:44019|pull_request:44071", + "pull_request:44068|pull_request:44722", + "pull_request:44018|pull_request:44116", + "pull_request:44068|pull_request:44116", + "pull_request:44071|pull_request:44722", + "pull_request:44018|pull_request:44085", + "pull_request:44018|pull_request:44086", + "pull_request:44068|pull_request:44086", + "pull_request:44071|pull_request:44085", + "pull_request:44085|pull_request:44086", + "pull_request:44059|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:44073|pull_request:44161", + "pull_request:44059|pull_request:44071", + "pull_request:44056|pull_request:44101", + "pull_request:44018|pull_request:44059", + "pull_request:44059|pull_request:44068", + "pull_request:44101|pull_request:44161", + "pull_request:43997|pull_request:44017", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:43997|pull_request:44018", + "pull_request:44056|pull_request:44161", + "pull_request:43997|pull_request:44161", + "pull_request:44074|pull_request:44161", + "pull_request:44018|pull_request:44161", + "pull_request:44010|pull_request:44071", + "pull_request:44018|pull_request:44129", + "pull_request:44010|pull_request:44073", + "pull_request:43997|pull_request:44074", + "pull_request:44010|pull_request:44018", + "pull_request:43997|pull_request:44129", + "pull_request:44071|pull_request:44129", + "pull_request:43996|pull_request:44161", + "pull_request:43996|pull_request:44017", + "pull_request:44010|pull_request:44044", + "pull_request:44044|pull_request:44074", + "pull_request:44066|pull_request:44722", + "pull_request:44085|pull_request:44722", + "pull_request:44085|pull_request:44129", + "pull_request:43996|pull_request:44007", + "pull_request:43996|pull_request:44129", + "pull_request:44044|pull_request:44129", + "pull_request:44044|pull_request:44154", + "pull_request:44024|pull_request:44076" + ], + "prepared_review_unit_hash": "779efc73c98441f309806e2125ff1409eb438e2d1c8aa81802eaa92bd9c8ba83", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84505, + "estimated_input_tokens": 21127, + "estimated_eval_tokens": 42510 + }, + "budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 72, + "serialized_chars": 76937, + "estimated_input_tokens": 19235, + "estimated_eval_tokens": 38726 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a batch of per-model output-tracing refactors, not a single duplicate. The only clear duplicate groups are the GPT-Neo pair and the GPT-J variants; most other pairs touch different model code paths and should remain separate.", + "confidence": 0.81, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44722 is the best representative for the cluster\u2019s actual duplicate subgroup: it is a concrete GPT-J/codegen output-tracing refactor, explicitly targets the tracking issue, and is the most complete version of that change path.", + "best_issue_reason": null, + "best_pr_reason": "PR 44722 is the strongest global representative because it cleanly captures the shared GPT-J output-tracing refactor and overlaps with the closest duplicates while being more complete than the narrower variants.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Different model families and code paths (CVT/FNet vs GPT-J); only share the broad output-tracing theme." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "CVT/FNet refactor vs DeBERTa refactor; same umbrella theme, but not the same concrete change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": true, + "reason": "Both are GPT-J output-tracing refactors against the same file and appear to be overlapping versions of the same change; 44066 simply also bundles CodeGen work." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "ResNet vs EfficientNet are separate model implementations; not the same underlying bug/change." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet output tracing and GPT-J/codegen output tracing are unrelated code paths." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen refactor vs EfficientNet refactor; too different to be duplicates." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "MobileNetV2 vs DeBERTa; same refactor pattern, but different implementations and fixes." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen refactor vs MGP-STR refactor; not the same concrete change." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44068", + "accept": true, + "reason": "Both are GPT-Neo output-tracing refactors on the same model file and appear to be the same change expressed in two PRs." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen vs MPT; different model code paths." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J/CodeGen refactor and GPT-Neo refactor are related only by the general decorator migration, not the same code path." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "GPT-Neo and GPT-J are different model implementations; not the same underlying change." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet output tracing is unrelated to GPT-J/codegen output tracing." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR refactor vs GPT-J/codegen refactor; same style, different change." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44085", + "accept": false, + "reason": "ResNet vs GPT-J; different models and code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet are separate implementations; not duplicates." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet vs MPT refactors are unrelated aside from the shared tracing theme." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo refactor vs GPT-J/codegen refactor; different model families." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo vs Flaubert/XLM; different code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT output tracing and GPT-J/codegen output tracing are distinct model changes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo vs GPT-J; same pattern, different implementation." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo vs MGP-STR; not the same underlying bug/change." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are unrelated model code paths." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT2 vs MGP-STR; both touch output handling, but not the same change." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT vs Flaubert/XLM; different model-specific refactors." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT2 and MPT are separate implementations." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "Segformer vs LongT5; unrelated model code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44722", + "accept": false, + "reason": "Segformer refactor is not the same concrete change as GPT-J/codegen." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44073", + "accept": false, + "reason": "Segformer vs VisualBert; different model families." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are different model implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44073", + "accept": false, + "reason": "RegNet vs VisualBert; same output-tracing theme, but different code paths." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT2 vs GPT-Neo are separate model code paths." + }, + { + "left": "pull_request:44101", + "right": "pull_request:44161", + "accept": false, + "reason": "Flaubert/XLM vs LongT5; different underlying changes." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44017", + "accept": false, + "reason": "RegNet vs Segformer are unrelated model-specific refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are different model implementations." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44076", + "accept": false, + "reason": "MPT vs ImageGPT; not the same code-path problem." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet vs Flaubert/XLM; different model-specific output changes." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet vs GPT-Neo; same broad migration, not the same duplicate." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet vs LongT5 are unrelated output-tracing refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet vs LongT5; different model code paths." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44161", + "accept": false, + "reason": "TextNet vs LongT5; not the same underlying change." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44161", + "accept": false, + "reason": "GPT-Neo vs LongT5; different model implementations." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBERT vs MPT; different code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo vs SpeechT5; unrelated model-specific changes." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBERT vs VisualBert; not the same concrete change." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet vs TextNet; separate model refactors." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44018", + "accept": false, + "reason": "SqueezeBERT vs GPT-Neo; same category, different model." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44129", + "accept": false, + "reason": "RegNet vs SpeechT5; different implementations." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT vs SpeechT5; unrelated output-tracing changes." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44161", + "accept": false, + "reason": "CVT/FNet vs LongT5; same pattern, different model code." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44017", + "accept": false, + "reason": "CVT/FNet vs Segformer; not the same change." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBERT vs DeBERTa; different model-specific refactors." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa vs TextNet; same output-tracing style, but distinct code paths." + }, + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": false, + "reason": "FocalNet and ImageGPT are different models; not a duplicate." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet and stays conservative about the soft edges. The only duplicate groupings it elevates are the GPT-Neo pair and the GPT-J/CodeGen variants, which are the clearest mergeable overlaps in the evidence. The many cross-model pairs are correctly left as non-duplicates." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 13, + "component_count": 38, + "review_unit_index": 2, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43997", + "pull_request:44010", + "pull_request:44056", + "pull_request:44073", + "pull_request:44074", + "pull_request:44129" + ], + "soft_pairs": [ + "pull_request:44073|pull_request:44074", + "pull_request:44056|pull_request:44073", + "pull_request:44056|pull_request:44074", + "pull_request:43997|pull_request:44010", + "pull_request:44010|pull_request:44129" + ], + "prepared_review_unit_hash": "82b80c6765e1a80e337691a8976f9bbd062d63f4bf9e6d67c91940e31f266f84", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84505, + "estimated_input_tokens": 21127, + "estimated_eval_tokens": 42510 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 5, + "serialized_chars": 13068, + "estimated_input_tokens": 3267, + "estimated_eval_tokens": 6790 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are all model-specific refactors to standardized output tracing/capture, but they target different architectures and code paths. They read as a family of similar cleanup PRs rather than duplicate fixes or mergeable single changes.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44073 is a good representative of the cluster because it is a straightforward, model-specific application of the standardized output tracing refactor and clearly reflects the shared theme across the set.", + "best_issue_reason": null, + "best_pr_reason": "PR 44073 is the best global representative because it is cleanly scoped, uses the new capture_outputs/can_return_tuple pattern, and most clearly captures the common refactor theme without extra model-specific complexity.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "Same refactor theme, but VisualBert and TextNet are different model implementations and the diffs touch different files/tests; not the same underlying change and not something to merge into one PR." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "Both are standardized output tracing refactors, but MPNet and VisualBert have distinct model-specific output plumbing; these are separate PRs for different code paths, not duplicates." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "Same broad migration pattern, but MPNet and TextNet are different architectures with different tracing hooks; the changes are not the same concrete fix." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are unrelated model code paths; despite similar decorator/import changes, they are separate refactors and not a single mergeable PR." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "Both modernize output handling, but SqueezeBert vs SpeechT5 are different models with different forward signatures and explicit targets; not the same bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 38, + "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": "6e5113c188ad6fbaae378d319ad0badca36ee7a3c2f7bc9346e1ccf3190fe773", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40167, + "estimated_input_tokens": 10042, + "estimated_eval_tokens": 20340 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40167, + "estimated_input_tokens": 10042, + "estimated_eval_tokens": 20340 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs all follow the same broad \"refactor output tracing\" theme, but they apply to different model families and different forward paths. They are not duplicates of one another, so none of the soft edges should be merged as the same change.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: the items are separate model-specific refactors, not one shared bug fix or implementation change.", + "best_issue_reason": null, + "best_pr_reason": "PR 44025 is the broadest single representative in the set (it touches two related Depth Anything files), but it is still just one of many independent refactors.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model families and forward paths; same refactor pattern, not the same concrete change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "timm_backbone vs vision_text_dual_encoder are unrelated code paths despite similar tracing cleanup." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "timm_backbone and univnet are separate models with different output behavior." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "Different architectures (timm_backbone vs mamba/falcon_mamba); not the same bug or fix." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Unrelated models; shared tracing refactor theme is too broad to merge as duplicates." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "mobilenet_v1 and univnet are unrelated forward implementations." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model families and distinct output-tracing code paths." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "MobileNetV1 vs Depth Anything are unrelated model implementations." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "Separate model wrappers; same refactor pattern, not one shared change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "mobilenet_v1 and speech_encoder_decoder are different code paths and model types." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "vision_text_dual_encoder and univnet are unrelated models." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model families; no shared concrete bug or fix." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "vision_text_dual_encoder vs mamba is only a shared refactor theme, not a duplicate change." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "Different forward paths in unrelated model families." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "univnet and upernet are separate models with separate output semantics." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Different architectures and code paths; not mergeable as one PR." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "UnivNet and Depth Anything are unrelated model implementations." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44029", + "accept": false, + "reason": "upernet and rwkv are unrelated; same refactor pattern is not enough." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "Different model families and different forward signatures." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44028", + "accept": false, + "reason": "mamba/falcon_mamba and superpoint are unrelated code paths." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "Different subsystems entirely: mamba vs codegen." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "timm_backbone and codegen are unrelated models." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "UnivNet and CodeGen fix different forward/output behaviors." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "upernet and codegen are separate changes in unrelated models." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "vision_text_dual_encoder vs codegen are unrelated code paths." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "Both are output-tracing refactors, but they target different wrapper classes and code paths." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "Depth Anything and Speech Encoder-Decoder are separate model families." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "Different tasks and forward APIs; not the same underlying change." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "Depth Anything and RWKV are unrelated implementations." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "Different model families with separate output handling." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "vision_encoder_decoder and rwkv are unrelated code paths." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "Both are encoder-decoder wrappers, but they are separate classes and not one duplicate fix." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "vision_encoder_decoder and superpoint are unrelated models." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "Different model families and distinct output APIs." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "speech_encoder_decoder and superpoint are unrelated." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "Separate model families with no shared concrete bug." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "speech_encoder_decoder and dpr are unrelated changes." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "superpoint and rwkv are unrelated implementations." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "superpoint and dpr fix different output paths." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "rwkv and dpr are different model families with different 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": 15, + "component_count": 38, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44662-2", + "nodes": [ + "pull_request:44662", + "pull_request:45401" + ], + "soft_pairs": [ + "pull_request:44662|pull_request:45401" + ], + "prepared_review_unit_hash": "e284d220010712ec4e4fe4135198590af701109725b59185177d86fd36fe81c9", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6552, + "estimated_input_tokens": 1638, + "estimated_eval_tokens": 3532 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6552, + "estimated_input_tokens": 1638, + "estimated_eval_tokens": 3532 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two unrelated model-integration pull requests: one adds PenguinVL, the other adds Voxtral TTS. They overlap only in shared Transformers registration/docs plumbing, not the same underlying feature or bug.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "Pick PR #45401 as the canonical PR: it is tied to an explicit issue (#45027), has a narrower and clearly scoped feature (Voxtral TTS), and is the more targetable integration request.", + "best_issue_reason": null, + "best_pr_reason": "PR #45401 is the strongest standalone representative because it has an explicit issue link and a focused model-specific implementation. PR #44662 is also a valid model add, but it is not the better canonical choice for the cluster.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Different model additions with different implementation stacks and targets; shared auto/docs registration is incidental and not enough to indicate 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": 16, + "component_count": 38, + "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-44965-2", + "nodes": [ + "pull_request:44965", + "pull_request:46845" + ], + "soft_pairs": [ + "pull_request:44965|pull_request:46845" + ], + "prepared_review_unit_hash": "73a7a5661400c66a132ba01e9542d93cb1eaa5f24e41a32c771c0e89fdb6b09c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4218, + "estimated_input_tokens": 1055, + "estimated_eval_tokens": 2366 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4218, + "estimated_input_tokens": 1055, + "estimated_eval_tokens": 2366 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs are not duplicates: one appears to rewrite a docs-related workflow into a draft-conversion action, while the other changes CI routing plus a model utility file in an unrelated smaller-runner experiment.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46845 is the better representative of this cluster because it has a clearer, coherent CI-oriented change set, whereas PR 44965 looks like an unrelated workflow rewrite with incomplete/garbled diff context.", + "best_issue_reason": null, + "best_pr_reason": "PR 46845 is the most suitable global representative only in the sense that it is the more substantive and internally consistent PR; however, it should not be treated as a duplicate of PR 44965.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44965", + "right": "pull_request:46845", + "accept": false, + "reason": "Different underlying changes: 44965 is about converting PRs to draft via a documentation workflow, while 46845 changes the PR CI caller and touches model code. They do not fix the same code-path problem and would not plausibly merge into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 17, + "component_count": 38, + "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": "406d73e7b80d504452faa6e73e26ec995bdc753b1aa70bf1ef2d76779d7ab8e7", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4408, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4408, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related fusion-feature PRs, but they address different layers: PR #44979 introduces a generic Module Fusion API, while PR #45363 implements n-to-1 kernel fusion through KernelConfig and related integration points. I would not treat them as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR stands out as a duplicate target: #44979 is the foundational module-fusion API, whereas #45363 is a separate KernelConfig-based fusion implementation with different files and scope.", + "best_issue_reason": null, + "best_pr_reason": "#45363 is the better representative for the cluster\u2019s concrete user-facing change because it wires fusion into model loading/integration code and includes an example; #44979 is more of an enabling API layer than the end-user feature.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "They share the broad theme of fusion, but the code paths and scope differ: #44979 adds a generic Module Fusion API, while #45363 implements n-to-1 kernel fusion via KernelConfig and integration/model-loading changes. They could plausibly coexist in one repo without being 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": 18, + "component_count": 38, + "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": "d7ce7bf95280189d61471821482be11022cbbf4d5e58dc0cc71556bca089cbc4", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4413, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4413, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs fix the same Mistral pre-tokenizer AttributeError by switching from `backend_tokenizer.pre_tokenizer` to `tokenizer.pre_tokenizer`; 45317 appears to be the more complete version because it adds tests.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "45317 is the stronger canonical PR because it contains the same code fix as 45086 plus test coverage, making it the more complete change set.", + "best_issue_reason": null, + "best_pr_reason": "45317 is the best PR to keep: it addresses the same concrete bug and includes validation in tests, while 45086 is the narrower earlier patch.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same underlying bug and same fix in the same file; 45317 is effectively a superset with tests, so these are duplicate PRs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 19, + "component_count": 38, + "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": "4104d51c2210dbe6e1cfd2fafa46f7bb5cd1314f48d67f367a1399b4165fa165", + "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": "Both PRs touch NemotronH internals, but they address different changes: #45115 is a narrow checkpoint-conversion/mapping refactor, while #45569 is a broader model split and documentation/auto-registry update. They do not look like the same underlying bug or mergeable single fix.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45569 is the better canonical PR for the cluster because it represents the more complete NemotronH architecture change, including new dense/sparse model packages, auto mappings, docs, and tests. PR #45115 is a narrower conversion-mapping follow-up and does not subsume the broader change.", + "best_issue_reason": null, + "best_pr_reason": "#45569 is the strongest representative of the cluster since it is the more substantial end-to-end NemotronH update; #45115 is too specialized and only overlaps on a few shared NemotronH files.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "Shared NemotronH files are incidental. #45115 only changes conversion mappings, while #45569 restructures the model into dense/sparse variants plus docs and auto-mapping updates. They fix different problems 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": 20, + "component_count": 38, + "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": "f9aedad3ef3bebe9d928406e27061da2fc1b7dd751e2de8273edf9af0f47c9d1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5953, + "estimated_input_tokens": 1489, + "estimated_eval_tokens": 3234 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5953, + "estimated_input_tokens": 1489, + "estimated_eval_tokens": 3234 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs are about the same heterogeneous-model feature family; #45333 is a narrower configuration-only subset of the broader end-to-end support in #45332.", + "confidence": 0.79, + "canonical_issue_reason": null, + "canonical_pr_reason": "#45332 is the broader, more complete PR: it includes the per-layer configuration work plus modeling, cache/masking, and integration changes needed for heterogeneous model support.", + "best_issue_reason": null, + "best_pr_reason": "#45332 is the best representative because it subsumes the config-only work in #45333 and covers the full concrete feature path rather than just one piece of it.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": true, + "reason": "Accept: #45333 looks like a narrower split-out of the same heterogeneous configuration feature covered by #45332; they share the core config files and the latter adds the modeling implementation on top." + } + ] + }, + "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": 38, + "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": "863f9a1c10407319a49d3f66da962fab2c570f37cb400b9d0473a3113576ad31", + "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": "Two separate typing-focused PRs with some overlapping files, but they address different changes: one is a broad repo-wide type-checking sweep, the other adds/adjusts shared typing protocols and modeling-utils-related annotations. Not a duplicate pair.", + "confidence": 0.87, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45425 is the better representative because it is more concrete and narrowly scoped around shared typing helpers and model-utils integration, rather than a broad type-checking sweep.", + "best_issue_reason": null, + "best_pr_reason": "PR 45425 is the strongest standalone candidate: it has a clearer, more cohesive technical change set and is less diffuse than the repo-wide typing PR.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "They overlap in typing-related files, but 45415 is a broad type-checking pass across many modules while 45425 adds specific typing protocols and targeted modeling_utils/peft annotations. Different concrete changes, not one mergeable duplicate PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 22, + "component_count": 38, + "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": "e5b7571779251ae3718d61ef10054500f4913bb6b00d13c43b3f17da7f423dac", + "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 PRs are related by expert-parallel loading, but they fix different bugs: #45548 is about preserving/loading EP config for DeepSpeed ZeRO-3 via accelerate, while #45662 fixes expert weights being overwritten during FSDP2/rank-0 broadcast.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "pull_request:45662 is the stronger representative of the cluster\u2019s EP-loading theme because it targets a concrete distributed-loading corruption bug with a specific code-path fix, whereas pull_request:45548 is a narrower compatibility/serialization patch.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Same broad EP loading area, but not the same underlying bug: #45548 handles config serialization/ZeRO-3 launch compatibility, while #45662 prevents rank-0 broadcast from overwriting experts under FSDP2. They would not plausibly merge into one single 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": 23, + "component_count": 38, + "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": "a2fbb038975fb9eaf1f9f8ad4098b3b5f184793e395ab7c3c4563fd38a3fff1c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4672, + "estimated_input_tokens": 1168, + "estimated_eval_tokens": 2592 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4672, + "estimated_input_tokens": 1168, + "estimated_eval_tokens": 2592 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs target the same xdist/captured_info problem and share the same testing utility files; one is a broader superset that also updates CI log collection for multiple output files.", + "confidence": 0.84, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45645 is the better canonical representative because it covers the same captured_info xdist collision fix and also updates the CI workflow to collect all split log files, making it the more complete change.", + "best_issue_reason": null, + "best_pr_reason": "PR 45645 is the strongest cluster member: it appears to subsume the narrower testing_utils-only fix in 45639 while preserving the same underlying debug-log issue.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45639", + "right": "pull_request:45645", + "accept": true, + "reason": "Both PRs fix the same captured_info/xdist artifact collision in the same testing utility path; 45645 is a plausible superset that adds the CI-side handling needed for 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": 24, + "component_count": 38, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45720-2", + "nodes": [ + "pull_request:45720", + "pull_request:46259" + ], + "soft_pairs": [ + "pull_request:45720|pull_request:46259" + ], + "prepared_review_unit_hash": "9f04de5e9eb363615620820b00ea32db9c7a858431670981cd4bcd149dfcc3f0", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7234, + "estimated_input_tokens": 1809, + "estimated_eval_tokens": 3874 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7234, + "estimated_input_tokens": 1809, + "estimated_eval_tokens": 3874 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are both linter-rule implementations in transformers modeling code, but they address different problems: TRF003 return-tuple/return_dict handling vs TRF018 model weight initialization. The overlap in filenames is incidental, not evidence of the same fix.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46259 is the better representative only because it is the larger, more central linter-rule change with much more review activity; it still does not duplicate #45720.", + "best_issue_reason": null, + "best_pr_reason": "PR #46259 is the stronger cluster representative due to broader diff scope and higher review activity, but it is not a duplicate of #45720.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45720", + "right": "pull_request:46259", + "accept": false, + "reason": "Different concrete changes: #45720 adds can_return_tuple/return_dict handling across model forwards, while #46259 adds TRF018 weight-init logic. Shared model files and linter context are too broad to indicate the same underlying bug or mergeable PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 25, + "component_count": 38, + "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": 26, + "component_count": 38, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45932-2", + "nodes": [ + "pull_request:45932", + "pull_request:46100" + ], + "soft_pairs": [ + "pull_request:45932|pull_request:46100" + ], + "prepared_review_unit_hash": "2e2cf026d6a51ca47251e160f40334a15501cb8acef36ee73c5d88446af5b8b1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5161, + "estimated_input_tokens": 1291, + "estimated_eval_tokens": 2838 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5161, + "estimated_input_tokens": 1291, + "estimated_eval_tokens": 2838 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two pull requests are not duplicates: one is a broad testing/telemetry and model-mapping change, while the other is a test-migration PR adding kernel and vision tests. They touch different code paths and would not plausibly be merged into one PR.", + "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:45932", + "right": "pull_request:46100", + "accept": false, + "reason": "Not the same underlying change. PR 45932 updates CI secrets plus several model files/configs; PR 46100 adds ddp config and test files for altclip/vit migration. Different goals, files, 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": 27, + "component_count": 38, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46126-2", + "nodes": [ + "pull_request:46126", + "pull_request:46269" + ], + "soft_pairs": [ + "pull_request:46126|pull_request:46269" + ], + "prepared_review_unit_hash": "3fba17f3e6637d34618d2b59c81b932a8ca392ee386fb3bc4cd4e767747db795", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5526, + "estimated_input_tokens": 1382, + "estimated_eval_tokens": 3020 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5526, + "estimated_input_tokens": 1382, + "estimated_eval_tokens": 3020 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related distributed-training changes, but they are not the same concrete fix/change: one is a focused EP + Trainer integration PR, while the other is a broad distributed branch base touching many unrelated areas. The overlap in filenames is too broad to merge as duplicates.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46126 is the more canonical, self-contained change: it targets a specific EP/Trainer integration path with a focused set of files and behavior. PR #46269 is a large base branch that spans many distributed subsystems and is not a clean representative of a single underlying change.", + "best_issue_reason": null, + "best_pr_reason": "PR #46126 is the best representative PR for this cluster because it has a narrower, well-defined purpose and concrete code-path changes. PR #46269 is too broad and includes many unrelated distributed updates, making it a poor canonical duplicate target.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46126", + "right": "pull_request:46269", + "accept": false, + "reason": "They share distributed filenames, but the PRs are not the same underlying change. #46126 is a specific EP + Trainer integration, while #46269 is a broad distributed branch base with many additional areas (FSDP, integrations, generation, CI, etc.). They would not plausibly be merged as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 28, + "component_count": 38, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46138-10", + "nodes": [ + "pull_request:46138", + "pull_request:46432", + "pull_request:46501", + "pull_request:46502", + "pull_request:46503", + "pull_request:46504", + "pull_request:46554", + "pull_request:46558", + "pull_request:46563", + "pull_request:46716" + ], + "soft_pairs": [ + "pull_request:46138|pull_request:46554", + "pull_request:46138|pull_request:46558", + "pull_request:46138|pull_request:46563", + "pull_request:46138|pull_request:46716", + "pull_request:46138|pull_request:46504", + "pull_request:46432|pull_request:46503", + "pull_request:46432|pull_request:46501", + "pull_request:46432|pull_request:46502", + "pull_request:46432|pull_request:46554", + "pull_request:46432|pull_request:46558", + "pull_request:46501|pull_request:46503", + "pull_request:46501|pull_request:46502", + "pull_request:46501|pull_request:46504", + "pull_request:46501|pull_request:46554", + "pull_request:46502|pull_request:46503", + "pull_request:46502|pull_request:46504", + "pull_request:46502|pull_request:46554", + "pull_request:46503|pull_request:46504", + "pull_request:46503|pull_request:46554", + "pull_request:46504|pull_request:46554", + "pull_request:46504|pull_request:46558", + "pull_request:46554|pull_request:46558", + "pull_request:46554|pull_request:46563", + "pull_request:46554|pull_request:46716", + "pull_request:46558|pull_request:46563", + "pull_request:46558|pull_request:46716", + "pull_request:46563|pull_request:46716", + "pull_request:46504|pull_request:46563", + "pull_request:46504|pull_request:46716" + ], + "prepared_review_unit_hash": "bf005833a957cda75d64e376ef483b4abb11f2c09e4a4a14ef1ad40eefe215e1", + "component_budget": { + "node_count": 10, + "item_count": 10, + "soft_pair_count": 29, + "serialized_chars": 25937, + "estimated_input_tokens": 6485, + "estimated_eval_tokens": 13226 + }, + "budget": { + "node_count": 10, + "item_count": 10, + "soft_pair_count": 29, + "serialized_chars": 25937, + "estimated_input_tokens": 6485, + "estimated_eval_tokens": 13226 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a set of unrelated workflow-only PRs. The only cohesive subcluster is several incremental updates to .github/workflows/pr-ci-security-gate.yml, but they address different hardening concerns rather than the same exact bug/change, so none of the soft edges should be merged.", + "confidence": 0.82, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46554 is the most substantive representative of the pr-ci-security-gate subcluster because it introduces the actual security-gate logic change (checkout plus changed-file checks), whereas the other PRs on that workflow are smaller permission/approval/comment tweaks.", + "best_issue_reason": null, + "best_pr_reason": "46554 is the strongest overall PR candidate: it makes the concrete security-gate behavior change and is more representative than the later incremental edits to the same workflow.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46138", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflow files and different behavior: stdin handling for self-comment-ci vs security-gate checks for fork PRs." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46558", + "accept": false, + "reason": "Unrelated workflow changes; one updates self-comment-ci argument passing, the other hardens pr-ci-security-gate permissions/approval flow." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46563", + "accept": false, + "reason": "Different CI workflows and different concerns; no shared underlying bug or change." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46716", + "accept": false, + "reason": "Different workflow files and different fixes; not the same code-path problem." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46504", + "accept": false, + "reason": "Different workflows and distinct shell-input handling changes in separate jobs." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46503", + "accept": false, + "reason": "SSH runner hardening vs benchmark workflow trigger change; unrelated changes." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46501", + "accept": false, + "reason": "Different workflows and different fixes: SSH runner secret handling vs anti-slop permissions structure." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46502", + "accept": false, + "reason": "Different workflow logic; no shared bug or concrete change." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46554", + "accept": false, + "reason": "SSH runner workflow hardening is unrelated to pr-ci-security-gate security checks." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46558", + "accept": false, + "reason": "Different workflow files and separate security concerns." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46503", + "accept": false, + "reason": "Anti-slop permissions vs benchmark dispatch trigger; unrelated workflow edits." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46502", + "accept": false, + "reason": "Different workflow files and different logic changes." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46504", + "accept": false, + "reason": "Both are workflow tweaks, but they fix different issues in different jobs and are not one mergeable change." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflows and different security/permission concerns." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46503", + "accept": false, + "reason": "Assign-reviewers checkout ref change vs benchmark trigger change are unrelated." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46504", + "accept": false, + "reason": "Different workflows with separate behavior changes; not the same underlying bug." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46554", + "accept": false, + "reason": "Assign-reviewers base-ref checkout fix is unrelated to pr-ci-security-gate changes." + }, + { + "left": "pull_request:46503", + "right": "pull_request:46504", + "accept": false, + "reason": "Benchmark dispatch support vs build-ci-docker-images commit-message handling; unrelated." + }, + { + "left": "pull_request:46503", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflow functionality and different code paths." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46554", + "accept": false, + "reason": "Same broad CI area, but one changes commit-message parsing for image tagging while the other changes fork PR security gating." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46558", + "accept": false, + "reason": "Different workflow logic and different concrete fixes." + }, + { + "left": "pull_request:46554", + "right": "pull_request:46558", + "accept": false, + "reason": "Same workflow file, but one adds file-change security checks while the other changes permissions and approval gating; these are separate fixes." + }, + { + "left": "pull_request:46554", + "right": "pull_request:46563", + "accept": false, + "reason": "The comment-only documentation change does not fix the same concrete behavior as the security checks added in 46554." + }, + { + "left": "pull_request:46554", + "right": "pull_request:46716", + "accept": false, + "reason": "Both touch pr-ci-security-gate.yml, but 46554 adds detection logic while 46716 only changes secret inheritance; different fixes." + }, + { + "left": "pull_request:46558", + "right": "pull_request:46563", + "accept": false, + "reason": "Permissions/approval gating vs explanatory comments in the same workflow; not the same bug or change." + }, + { + "left": "pull_request:46558", + "right": "pull_request:46716", + "accept": false, + "reason": "Different hardening changes to the same workflow, but not one concrete fix." + }, + { + "left": "pull_request:46563", + "right": "pull_request:46716", + "accept": false, + "reason": "Commenting on safety and changing secrets inheritance are different changes with different effects." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46563", + "accept": false, + "reason": "Unrelated workflow updates in different files and code paths." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46716", + "accept": false, + "reason": "Build-image tagging logic vs pr-ci-security-gate secret handling; unrelated changes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 29, + "component_count": 38, + "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-46174-2", + "nodes": [ + "pull_request:46174", + "pull_request:46780" + ], + "soft_pairs": [ + "pull_request:46174|pull_request:46780" + ], + "prepared_review_unit_hash": "a48ce4ca162f555682f79411d7229ac1ccc08d7822dfab654bb881639fccddbc", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6030, + "estimated_input_tokens": 1508, + "estimated_eval_tokens": 3272 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6030, + "estimated_input_tokens": 1508, + "estimated_eval_tokens": 3272 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two PRs target the same reported MPS crash in the sinusoidal position embedding helper, but they implement different fixes and are not safe to treat as duplicates.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46174 is the broader, more complete fix for the float64-on-MPS crash: it updates the helper across all affected model files and keeps the internal math on CPU float64 before casting back.", + "best_issue_reason": null, + "best_pr_reason": "PR 46174 is the best representative of the cluster because it is the original, more comprehensive fix and has more discussion/review context than the narrower follow-up.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46174", + "right": "pull_request:46780", + "accept": false, + "reason": "They share the same issue target and helper, but the patches are different solutions: 46174 avoids float64 on MPS by doing the arithmetic on CPU first, while 46780 switches the helper to use dtype for internal arithmetic. Those are alternative implementations, not one mergeable duplicate PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 30, + "component_count": 38, + "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-46196-18", + "nodes": [ + "pull_request:46196", + "pull_request:46707", + "pull_request:46836", + "pull_request:46891", + "pull_request:46907", + "pull_request:46908", + "pull_request:46912", + "pull_request:46926", + "pull_request:46927", + "pull_request:46928", + "pull_request:46941", + "pull_request:46949", + "pull_request:46950", + "pull_request:46951", + "pull_request:46953", + "pull_request:46954", + "pull_request:46956", + "pull_request:46957" + ], + "soft_pairs": [ + "pull_request:46196|pull_request:46957", + "pull_request:46196|pull_request:46836", + "pull_request:46196|pull_request:46956", + "pull_request:46196|pull_request:46891", + "pull_request:46196|pull_request:46953", + "pull_request:46707|pull_request:46953", + "pull_request:46707|pull_request:46891", + "pull_request:46707|pull_request:46836", + "pull_request:46707|pull_request:46957", + "pull_request:46707|pull_request:46912", + "pull_request:46836|pull_request:46957", + "pull_request:46836|pull_request:46891", + "pull_request:46836|pull_request:46956", + "pull_request:46836|pull_request:46953", + "pull_request:46891|pull_request:46957", + "pull_request:46891|pull_request:46956", + "pull_request:46891|pull_request:46912", + "pull_request:46907|pull_request:46926", + "pull_request:46907|pull_request:46941", + "pull_request:46907|pull_request:46908", + "pull_request:46907|pull_request:46927", + "pull_request:46907|pull_request:46928", + "pull_request:46908|pull_request:46926", + "pull_request:46908|pull_request:46928", + "pull_request:46908|pull_request:46951", + "pull_request:46908|pull_request:46941", + "pull_request:46912|pull_request:46957", + "pull_request:46836|pull_request:46912", + "pull_request:46912|pull_request:46956", + "pull_request:46926|pull_request:46941", + "pull_request:46926|pull_request:46928", + "pull_request:46926|pull_request:46927", + "pull_request:46927|pull_request:46949", + "pull_request:46927|pull_request:46951", + "pull_request:46927|pull_request:46950", + "pull_request:46927|pull_request:46941", + "pull_request:46928|pull_request:46941", + "pull_request:46928|pull_request:46951", + "pull_request:46949|pull_request:46951", + "pull_request:46949|pull_request:46950", + "pull_request:46891|pull_request:46949", + "pull_request:46836|pull_request:46949", + "pull_request:46950|pull_request:46951", + "pull_request:46908|pull_request:46950", + "pull_request:46926|pull_request:46950", + "pull_request:46891|pull_request:46951", + "pull_request:46953|pull_request:46956", + "pull_request:46953|pull_request:46957", + "pull_request:46891|pull_request:46953", + "pull_request:46891|pull_request:46954", + "pull_request:46954|pull_request:46957", + "pull_request:46954|pull_request:46956", + "pull_request:46953|pull_request:46954", + "pull_request:46836|pull_request:46954", + "pull_request:46956|pull_request:46957" + ], + "prepared_review_unit_hash": "bc87f73801481b6b039710b02dd83d3eac2d4cb935da3ec05c72fbb9af31bc35", + "component_budget": { + "node_count": 18, + "item_count": 18, + "soft_pair_count": 55, + "serialized_chars": 59205, + "estimated_input_tokens": 14802, + "estimated_eval_tokens": 29860 + }, + "budget": { + "node_count": 18, + "item_count": 18, + "soft_pair_count": 55, + "serialized_chars": 59205, + "estimated_input_tokens": 14802, + "estimated_eval_tokens": 29860 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This is not a single duplicate cluster; it mixes unrelated PRs across model bring-up, CI/test maintenance, distributed/FSDP infrastructure, generation/FP8 work, and many model-specific bug fixes.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: the items span unrelated changes (new models, CI, FSDP/FP8 infra, test refactors, and separate per-model fixes).", + "best_issue_reason": null, + "best_pr_reason": "No global best PR for duplicate collapse, because there is no shared underlying bug or change that would make one PR the canonical representative.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46196", + "right": "pull_request:46957", + "accept": false, + "reason": "wav2vec2 SDPA dispatch work vs finegrained FP8 decode optimization; unrelated subsystems." + }, + { + "left": "pull_request:46196", + "right": "pull_request:46836", + "accept": false, + "reason": "audio model SDPA/docs changes vs adding the Unlimited OCR model." + }, + { + "left": "pull_request:46196", + "right": "pull_request:46956", + "accept": false, + "reason": "model dispatch changes vs shared test-helper refactor." + }, + { + "left": "pull_request:46196", + "right": "pull_request:46891", + "accept": false, + "reason": "wav2vec2/audio model work vs PEFT CI/workflow changes." + }, + { + "left": "pull_request:46196", + "right": "pull_request:46953", + "accept": false, + "reason": "SDPA dispatch work vs FusedRMSNormGated dtype cleanup." + }, + { + "left": "pull_request:46707", + "right": "pull_request:46953", + "accept": false, + "reason": "FSDP2 migration vs RMSNorm module dtype removal; different code paths." + }, + { + "left": "pull_request:46707", + "right": "pull_request:46891", + "accept": false, + "reason": "FSDP runtime/migration work vs PEFT CI/test changes." + }, + { + "left": "pull_request:46707", + "right": "pull_request:46836", + "accept": false, + "reason": "distributed FSDP utilities vs new Unlimited OCR model bring-up." + }, + { + "left": "pull_request:46707", + "right": "pull_request:46957", + "accept": false, + "reason": "FSDP migration vs generation/FP8 optimization; not the same bug." + }, + { + "left": "pull_request:46707", + "right": "pull_request:46912", + "accept": false, + "reason": "FSDP infrastructure vs RoPE test coverage changes." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46957", + "accept": false, + "reason": "new OCR model addition vs generation decode optimization." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46891", + "accept": false, + "reason": "model bring-up vs PEFT CI workflow/test changes." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46956", + "accept": false, + "reason": "new model code vs shared floats_list test helper refactor." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46953", + "accept": false, + "reason": "Unlimited OCR addition vs RMSNorm dtype cleanup." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46957", + "accept": false, + "reason": "PEFT CI/workflow changes vs generation FP8 runtime changes." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46956", + "accept": false, + "reason": "PEFT CI work vs feature-extractor helper refactor." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46912", + "accept": false, + "reason": "PEFT CI changes vs RoPE test updates." + }, + { + "left": "pull_request:46907", + "right": "pull_request:46926", + "accept": false, + "reason": "EdgeTam kwarg cleanup vs MiniMax-M3-VL offload-folder fix." + }, + { + "left": "pull_request:46907", + "right": "pull_request:46941", + "accept": false, + "reason": "EdgeTam processor kwarg handling vs DAC distance-formula fix." + }, + { + "left": "pull_request:46907", + "right": "pull_request:46908", + "accept": false, + "reason": "EdgeTam import_or_config cleanup vs DeepSeek-V3.2 load/offload workaround." + }, + { + "left": "pull_request:46907", + "right": "pull_request:46927", + "accept": false, + "reason": "EdgeTam kwarg cleanup vs GLM46V input casting fix." + }, + { + "left": "pull_request:46907", + "right": "pull_request:46928", + "accept": false, + "reason": "EdgeTam model bug vs BLIP-2 tokenizer allowlist update." + }, + { + "left": "pull_request:46908", + "right": "pull_request:46926", + "accept": false, + "reason": "different model-specific load_error fixes (DeepSeek vs MiniMax)." + }, + { + "left": "pull_request:46908", + "right": "pull_request:46928", + "accept": false, + "reason": "load/offload workaround vs tokenizer allowlist change." + }, + { + "left": "pull_request:46908", + "right": "pull_request:46951", + "accept": false, + "reason": "DeepSeek loading workaround vs PEFT hotswap mismatch handling." + }, + { + "left": "pull_request:46908", + "right": "pull_request:46941", + "accept": false, + "reason": "load/offload issue vs DAC numerical fix." + }, + { + "left": "pull_request:46912", + "right": "pull_request:46957", + "accept": false, + "reason": "RoPE tests vs generation/FP8 optimization." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46912", + "accept": false, + "reason": "new OCR model bring-up vs RoPE test suite changes." + }, + { + "left": "pull_request:46912", + "right": "pull_request:46956", + "accept": false, + "reason": "RoPE test coverage vs shared float-list helper refactor." + }, + { + "left": "pull_request:46926", + "right": "pull_request:46941", + "accept": false, + "reason": "MiniMax load workaround vs DAC codebook-distance correction." + }, + { + "left": "pull_request:46926", + "right": "pull_request:46928", + "accept": false, + "reason": "MiniMax load issue vs BLIP-2 tokenizer allowlist." + }, + { + "left": "pull_request:46926", + "right": "pull_request:46927", + "accept": false, + "reason": "different model-specific fixes for unrelated failures." + }, + { + "left": "pull_request:46927", + "right": "pull_request:46949", + "accept": false, + "reason": "GLM46V input_id casting vs Whisper/AudioFlamingo cache-state fix." + }, + { + "left": "pull_request:46927", + "right": "pull_request:46951", + "accept": false, + "reason": "GLM46V fix vs PEFT hotswap loading behavior." + }, + { + "left": "pull_request:46927", + "right": "pull_request:46950", + "accept": false, + "reason": "GLM46V input/type fix vs Mamba2 OOM fix." + }, + { + "left": "pull_request:46927", + "right": "pull_request:46941", + "accept": false, + "reason": "GLM46V casting bug vs DAC numeric bug." + }, + { + "left": "pull_request:46928", + "right": "pull_request:46941", + "accept": false, + "reason": "tokenizer allowlist update vs DAC distance calculation fix." + }, + { + "left": "pull_request:46928", + "right": "pull_request:46951", + "accept": false, + "reason": "tokenizer allowlist vs PEFT hotswap mismatch handling." + }, + { + "left": "pull_request:46949", + "right": "pull_request:46951", + "accept": false, + "reason": "Whisper/AudioFlamingo cache initialization bug vs PEFT loading/mismatch logic." + }, + { + "left": "pull_request:46949", + "right": "pull_request:46950", + "accept": false, + "reason": "different model fixes: cache-state init vs Mamba2 OOM." + }, + { + "left": "pull_request:46950", + "right": "pull_request:46951", + "accept": false, + "reason": "Mamba2 kernel/OOM fix vs PEFT hotswap load behavior." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46949", + "accept": false, + "reason": "PEFT CI/workflow work vs Whisper/AudioFlamingo runtime fix." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46949", + "accept": false, + "reason": "new OCR model addition vs Whisper/AudioFlamingo bug fix." + }, + { + "left": "pull_request:46953", + "right": "pull_request:46956", + "accept": false, + "reason": "RMSNorm module cleanup vs shared test-helper refactor." + }, + { + "left": "pull_request:46953", + "right": "pull_request:46957", + "accept": false, + "reason": "RMSNorm dtype cleanup vs FP8 generation optimization." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46953", + "accept": false, + "reason": "PEFT CI changes vs RMSNorm cleanup." + }, + { + "left": "pull_request:46891", + "right": "pull_request:46954", + "accept": false, + "reason": "PEFT CI work vs GlmAsr Pytest v9 fix." + }, + { + "left": "pull_request:46954", + "right": "pull_request:46957", + "accept": false, + "reason": "GlmAsr pytest-related fix vs generation/FP8 optimization." + }, + { + "left": "pull_request:46954", + "right": "pull_request:46956", + "accept": false, + "reason": "GlmAsr model/test fix vs shared floats_list helper refactor." + }, + { + "left": "pull_request:46953", + "right": "pull_request:46954", + "accept": false, + "reason": "distinct fixes in different models; not the same underlying bug." + }, + { + "left": "pull_request:46836", + "right": "pull_request:46954", + "accept": false, + "reason": "Unlimited OCR model addition vs GlmAsr pytest fix." + }, + { + "left": "pull_request:46956", + "right": "pull_request:46957", + "accept": false, + "reason": "test helper refactor vs generation/FP8 runtime optimization." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 31, + "component_count": 38, + "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-46281-2", + "nodes": [ + "pull_request:46281", + "pull_request:46890" + ], + "soft_pairs": [ + "pull_request:46281|pull_request:46890" + ], + "prepared_review_unit_hash": "c246c6a23343ef9d1083f5d6d16f578c85cc6d038d98859cb6b64fa0ff1c62a9", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4271, + "estimated_input_tokens": 1068, + "estimated_eval_tokens": 2392 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4271, + "estimated_input_tokens": 1068, + "estimated_eval_tokens": 2392 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are related by theme (path traversal hardening) but they fix different untrusted inputs in different code paths, so they are not duplicates and should not be merged together.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "46281 is the cleaner representative of the cluster because it fixes a concrete traversal bug in sharded checkpoint loading with focused containment checks and tests; 46890 targets a separate `transformers_weights` config-field path.", + "best_issue_reason": null, + "best_pr_reason": "46281 is the strongest standalone PR to keep as the cluster representative, but only because it is a tighter, more explicit fix; it does not subsume the distinct bug fixed by 46890.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46281", + "right": "pull_request:46890", + "accept": false, + "reason": "Different vulnerable inputs and loaders: 46281 sanitizes shard filenames from the checkpoint index in `hub.py`, while 46890 validates `transformers_explicit_filename` from model config in `modeling_utils.py`. Same security class, but not the same underlying bug or 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": 32, + "component_count": 38, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46310-2", + "nodes": [ + "pull_request:46310", + "pull_request:46546" + ], + "soft_pairs": [ + "pull_request:46310|pull_request:46546" + ], + "prepared_review_unit_hash": "3458e8d1723e4efac7081d210033aee4c2d629590e5139b77c3dab65e52c792e", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4850, + "estimated_input_tokens": 1213, + "estimated_eval_tokens": 2682 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4850, + "estimated_input_tokens": 1213, + "estimated_eval_tokens": 2682 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs fix the same sliding-window right-boundary off-by-one in Moonshine streaming mask generation; PR 46546 is the broader, test-backed version and also applies the same fix to Gemma4.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46546 is the best canonical PR because it contains the Moonshine streaming boundary fix plus tests, and also applies the same correction to Gemma4; it is the more complete and representative change set.", + "best_issue_reason": null, + "best_pr_reason": "PR 46546 is the best global PR because it is the more complete fix for the same underlying boundary bug, with validation coverage and additional affected model handling.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46310", + "right": "pull_request:46546", + "accept": true, + "reason": "Both PRs change the same Moonshine streaming sliding-window mask from '< right_window_size' to '<= right_window_size', fixing the same off-by-one boundary bug; PR 46546 simply expands the same fix to another model and adds tests." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 33, + "component_count": 38, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46376-6", + "nodes": [ + "pull_request:46376", + "pull_request:46417", + "pull_request:46425", + "pull_request:46509", + "pull_request:46850", + "pull_request:46917" + ], + "soft_pairs": [ + "pull_request:46376|pull_request:46417", + "pull_request:46376|pull_request:46425", + "pull_request:46425|pull_request:46509", + "pull_request:46425|pull_request:46917", + "pull_request:46425|pull_request:46850", + "pull_request:46850|pull_request:46917" + ], + "prepared_review_unit_hash": "e7ffef7f5fa87a835c6ed97c3137c6f28567290d02b7e7ae70b646098030c586", + "component_budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 6, + "serialized_chars": 16834, + "estimated_input_tokens": 4209, + "estimated_eval_tokens": 8674 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 6, + "serialized_chars": 16834, + "estimated_input_tokens": 4209, + "estimated_eval_tokens": 8674 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is not a duplicate set: it mixes unrelated PRs for model additions, a continuous batching multimodal feature, a Gemma masking bugfix, and a deprecation cleanup. None of the soft-similarity pairs look like the same concrete change.", + "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:46376", + "right": "pull_request:46417", + "accept": false, + "reason": "Different features and codepaths: continuous batching multimodal support vs. adding HunYuan VL model files and docs." + }, + { + "left": "pull_request:46376", + "right": "pull_request:46425", + "accept": false, + "reason": "Unrelated model areas: continuous batching internals vs. RADIO model implementation and registration." + }, + { + "left": "pull_request:46425", + "right": "pull_request:46509", + "accept": false, + "reason": "Separate model additions for different architectures (RADIO vs. Nemotron Omni), not the same bug or change." + }, + { + "left": "pull_request:46425", + "right": "pull_request:46917", + "accept": false, + "reason": "RADIO model support and deprecated training-args cleanup are unrelated; shared repo touchpoints are too broad to merge." + }, + { + "left": "pull_request:46425", + "right": "pull_request:46850", + "accept": false, + "reason": "RADIO model addition vs. Gemma attention-masking fix are different subsystems and issues." + }, + { + "left": "pull_request:46850", + "right": "pull_request:46917", + "accept": false, + "reason": "Gemma sliding-window masking bugfix does not match the deprecation-removal PR; no shared concrete code-path problem." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 34, + "component_count": 38, + "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-46574-2", + "nodes": [ + "pull_request:46574", + "pull_request:46580" + ], + "soft_pairs": [ + "pull_request:46574|pull_request:46580" + ], + "prepared_review_unit_hash": "43570064594deb03d556be6e2877a1da0ed12cb333cf4336a6f25c93d4433ccb", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4410, + "estimated_input_tokens": 1103, + "estimated_eval_tokens": 2462 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4410, + "estimated_input_tokens": 1103, + "estimated_eval_tokens": 2462 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs are unrelated: one fixes CodeLlama decode-time leading-space stripping, the other fixes DeepSeek-Coder/DeepSeek-LLM being routed to the wrong tokenizer class. They touch different code paths and should not be merged as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: PR #46574 changes CodeLlama decoding behavior, while PR #46580 changes AutoTokenizer checkpoint-to-class routing for DeepSeek checkpoints.", + "best_issue_reason": null, + "best_pr_reason": "Neither PR is a good global best representative for a duplicate cluster because they solve different bugs in different subsystems.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46574", + "right": "pull_request:46580", + "accept": false, + "reason": "Different underlying problems: one is a decoder configuration bug in CodeLlama, the other is incorrect tokenizer-class selection for specific DeepSeek checkpoints." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 35, + "component_count": 38, + "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-46603-2", + "nodes": [ + "pull_request:46603", + "pull_request:46604" + ], + "soft_pairs": [ + "pull_request:46603|pull_request:46604" + ], + "prepared_review_unit_hash": "2fa60193f6c0aa6713574aca7daf9632d3517d8e90d051e4f511f93de8c0becb", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5496, + "estimated_input_tokens": 1374, + "estimated_eval_tokens": 3004 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5496, + "estimated_input_tokens": 1374, + "estimated_eval_tokens": 3004 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related Mistral tokenizer PRs, but they target different changes: one refactors/moves `MistralConverter` into the Mistral integration package, while the other adds tekken tokenizer detection, conversion, and save utilities. They share files, but not the same underlying fix/change.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46604 is the better representative because it introduces the broader functional change around tekken tokenizer support; PR #46603 is mainly a code organization/refactor move for an existing converter.", + "best_issue_reason": null, + "best_pr_reason": "PR #46604 is the strongest standalone PR in the cluster since it implements the main Mistral tokenizer feature work rather than just relocating code.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46603", + "right": "pull_request:46604", + "accept": false, + "reason": "Different underlying changes: #46603 moves `MistralConverter` into `integrations/mistral/`, while #46604 adds tekken tokenizer detection/conversion/save utilities. Shared files alone are not enough to treat them as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 36, + "component_count": 38, + "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-46741-3", + "nodes": [ + "pull_request:46741", + "pull_request:46773", + "pull_request:46781" + ], + "soft_pairs": [ + "pull_request:46741|pull_request:46773", + "pull_request:46741|pull_request:46781" + ], + "prepared_review_unit_hash": "beb55dd5fe8528ae2d9d6e41bbaeea06f6da2b88ad0273528a6adb93dbaf625f", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 8452, + "estimated_input_tokens": 2113, + "estimated_eval_tokens": 4482 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 8452, + "estimated_input_tokens": 2113, + "estimated_eval_tokens": 4482 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Three open PRs address different bugs in adjacent model/cache code; despite some shared filenames, they are not duplicates. The strongest representative of the cluster is the Mamba2 chunked-prefill/speculative-decoding fix.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46741 is the best canonical PR because it fixes a concrete runtime bug in the Mamba2-style cache path, with targeted code changes and tests across the affected model families.", + "best_issue_reason": null, + "best_pr_reason": "PR 46741 is the best overall PR to represent the cluster: it has the clearest single bug, the most directly related fix path, and the broadest but still coherent affected surface.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46741", + "right": "pull_request:46773", + "accept": false, + "reason": "Both touch linear-attention model files, but 46741 fixes chunked-prefill/speculative-decoding cache handling while 46773 fixes padding-mask application for batch-size-1; different bugs and different code paths." + }, + { + "left": "pull_request:46741", + "right": "pull_request:46781", + "accept": false, + "reason": "46741 is a model-specific cache/state bug fix, whereas 46781 changes QuantizedCache construction for hybrid SSM+attention models in cache_utils; these are unrelated failures." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 37, + "component_count": 38, + "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-46895-2", + "nodes": [ + "pull_request:46895", + "pull_request:46958" + ], + "soft_pairs": [ + "pull_request:46895|pull_request:46958" + ], + "prepared_review_unit_hash": "fdb7d9b457e40416e9df22f14c67b53b3b5bddf0d768aede8e15d3ca7ba84e59", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4790, + "estimated_input_tokens": 1198, + "estimated_eval_tokens": 2652 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4790, + "estimated_input_tokens": 1198, + "estimated_eval_tokens": 2652 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests target the same CohereASR training-loss double-shift bug under issue #46894, with differing implementation details. PR #46895 looks like the stronger canonical fix; #46958 is likely an alternate/less complete implementation of the same change.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46895 has the clearer, more complete fix path for the double-shift bug and more review activity, while still targeting the same issue as #46958.", + "best_issue_reason": null, + "best_pr_reason": "PR #46895 is the best representative of the underlying bug fix because it directly handles when labels should be shifted and appears to preserve the intended loss behavior.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46895", + "right": "pull_request:46958", + "accept": true, + "reason": "Same explicit issue target and same underlying CohereASR double-shift training-loss bug; the implementation differs, but both are fix attempts for the same code-path problem." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 38, + "component_count": 38, + "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-46933-2", + "nodes": [ + "pull_request:46933", + "pull_request:46942" + ], + "soft_pairs": [ + "pull_request:46933|pull_request:46942" + ], + "prepared_review_unit_hash": "15db6f7145d3893446235c4936314bff1b7d12c0ca0e670bb0fb95d6de5243cb", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4372, + "estimated_input_tokens": 1093, + "estimated_eval_tokens": 2442 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4372, + "estimated_input_tokens": 1093, + "estimated_eval_tokens": 2442 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same general anti-pattern\u2014casting activations to a quantized module\u2019s weight dtype\u2014but they fix different model-specific code paths in different architectures, so they are not duplicates.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR 46933 is the better representative of the cluster because it has an explicit issue link and fixes two Gemma4 embedder paths, but it is still not the same change as PR 46942.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46933", + "right": "pull_request:46942", + "accept": false, + "reason": "Similar fix pattern, but different concrete bugs and code paths: Gemma4 vision/audio embedders vs TimesFM 2.5 ResidualBlock. They 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 + } + ] +} diff --git a/snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/manifest.json b/snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..716dacd588b298ab26977d1f5a534e27bd18b9e8 --- /dev/null +++ b/snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260629t120018z", + "artifacts": { + "hybrid": "snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-06-29T12:11:21Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260629T120018Z", + "variant": "hybrid" +} diff --git a/snapshots/20260629T120018Z/manifest.json b/snapshots/20260629T120018Z/manifest.json index 8a4e5ad5302f04189f71e56291d5df8d41ac7e31..ecb03a0ac9ad69810fcffd8c6de8ca287effaedd 100644 --- a/snapshots/20260629T120018Z/manifest.json +++ b/snapshots/20260629T120018Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 1023 }, "extracted_at": "2026-06-29T12:00:18Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260629t120018z", + "runs": { + "hybrid-model-20260629t120018z": { + "analysis_id": "hybrid-model-20260629t120018z", + "artifacts": { + "hybrid": "snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260629T120018Z/analysis-runs/hybrid-model-20260629t120018z/manifest.json", + "model": null, + "published_at": "2026-06-29T12:11:21Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260629T120018Z", "watermark": {