diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index ad70ee3d28cb91c9203017afbefed5320c4ea978..1bef6deac9be49e21425ded88f627b895abba588 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": "20260620T060014Z", - "generated_at": "2026-06-20T06:04:12Z", + "snapshot_id": "20260620T120016Z", + "generated_at": "2026-06-20T12:05:21Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ @@ -1808,68 +1808,69 @@ ] }, { - "cluster_id": "cluster-46504-3", - "summary": "Cluster of 3 related pull requests linked by soft_similarity.", + "cluster_id": "cluster-46562-2", + "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": 46562, "issue_numbers": [], "pr_numbers": [ - 46504, - 46563, - 46716 + 46562, + 46658 ], "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, + "left_pr_number": 46562, + "right_pr_number": 46658, + "code_similarity": 0.153, + "size_similarity": 0.387, + "file_overlap": 0.024, "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, + "patch_similarity": 0.458, "shared_filenames": [ - ".github/workflows/pr-ci-security-gate.yml" + "src/transformers/models/auto/auto_mappings.py" ], "shared_file_areas": [ { - "filename": ".github/workflows/pr-ci-security-gate.yml", + "filename": "src/transformers/models/auto/auto_mappings.py", "left_ranges": [ [ - 13, - 19 + 441, + 447 + ], + [ + 591, + 598 + ], + [ + 823, + 829 + ], + [ + 883, + 889 + ], + [ + 1082, + 1088 + ], + [ + 1205, + 1211 ] ], "right_ranges": [ [ - 14, - 17 + 362, + 368 + ], + [ + 795, + 801 ] ] } @@ -1877,35 +1878,6 @@ } ] }, - { - "cluster_id": "cluster-43998-2", - "summary": "Cluster of 2 related pull requests linked by soft_similarity.", - "status": "open", - "confidence": 0.5, - "canonical_issue_number": null, - "canonical_pr_number": 43998, - "issue_numbers": [], - "pr_numbers": [ - 43998, - 44029 - ], - "evidence_types": [ - "soft_similarity" - ], - "pr_comparisons": [ - { - "left_pr_number": 43998, - "right_pr_number": 44029, - "code_similarity": 0.125, - "size_similarity": 0.43, - "file_overlap": 0.0, - "area_overlap": 0.0, - "patch_similarity": 0.259, - "shared_filenames": [], - "shared_file_areas": [] - } - ] - }, { "cluster_id": "cluster-44018-2", "summary": "Cluster of 2 related pull requests linked by soft_similarity.", @@ -2246,15 +2218,6 @@ "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-43998-2", - "canonical_pr_number": 43998, - "duplicate_pr_numbers": [ - 44029 - ], - "target_issue_number": null, - "reason": "PRs in cluster-43998-2 are treated as duplicates because they share soft_similarity evidence." - }, { "cluster_id": "cluster-44018-2", "canonical_pr_number": 44068, @@ -2319,14 +2282,13 @@ "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, + "cluster_id": "cluster-46562-2", + "canonical_pr_number": 46562, "duplicate_pr_numbers": [ - 46504, - 46563 + 46658 ], "target_issue_number": null, - "reason": "PRs in cluster-46504-3 are treated as duplicates because they share soft_similarity evidence." + "reason": "PRs in cluster-46562-2 are treated as duplicates because they share soft_similarity evidence." } ], "best_issue": { diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index 8a7828fa5cd6eaa5a8832b0dc1a7553c2df5b96d..f55cbaff88be3f6fab42fc9b89f44e039a09a961 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": "20260620T060014Z", - "generated_at": "2026-06-20T06:04:12Z", + "snapshot_id": "20260620T120016Z", + "generated_at": "2026-06-20T12:05:21Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -913,8 +913,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -1001,7 +1001,7 @@ "issue:43975|issue:44292", "issue:44292|issue:45507" ], - "prepared_review_unit_hash": "3ad5c355bce967cc8ec52d20614de340075ad009fbdd7859c574c91bf7b32f8a", + "prepared_review_unit_hash": "b69a0ee6dd1564d6cbe44e8e7d78d28d6715dbb368b830eca0282680bcaafaba", "component_budget": { "node_count": 484, "item_count": 484, @@ -1023,210 +1023,210 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous, with a few small duplicate-like subgroups: Qwen3.5 video-input regression, SigLIP2 bad-output regression, custom-module cache/path collision, and input-validation error messaging. Most other pairs are unrelated model-specific failures and should stay separate.", - "confidence": 0.72, - "canonical_issue_reason": "No single canonical issue cleanly covers the whole cluster; it spans several unrelated bug families rather than one duplicate thread.", + "summary": "This cluster is mostly heterogeneous transformer bug reports with only one strong near-duplicate pair around custom-module loading/caching. I would keep the cluster anchored on the open custom-code loading report and reject the rest as separate issues.", + "confidence": 0.78, + "canonical_issue_reason": "issue:45698 is the clearest and most general report in the only convincing duplicate pair; it describes the user-visible failure directly, while issue:45632 reads like a more specific cache-path collision manifestation of the same custom-module loading bug.", "canonical_pr_reason": null, - "best_issue_reason": "If forced to pick a broad umbrella, issue 45020 is the closest high-level regression report, but it is still too generic to serve as a true canonical duplicate target.", + "best_issue_reason": "issue:45698 is open, concrete, and broad enough to serve as the cluster anchor for the custom-code loading/caching problem.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44387", "right": "issue:45005", "accept": false, - "reason": "Different problems: int4 quantization OOM vs tied-weights handling on translation models." + "reason": "Different bugs: int4 quantization/OOM versus tied-weights behavior in translation models." }, { "left": "issue:44805", "right": "issue:44918", "accept": false, - "reason": "Different code paths: generic mask-shape IndexError vs Qwen3.5 unpacking with TRL SFT trainer." + "reason": "Both are shape/indexing failures, but they point to different model paths and failure modes." }, { "left": "issue:43541", "right": "issue:45161", "accept": false, - "reason": "Different models and failures: Mixtral grouped_mm tracing error vs GPT-OSS tensor-parallel issue." + "reason": "Grouped_mm tracing error in Mixtral is unrelated to tensor-parallel support in GPT-OSS MoE." }, { "left": "issue:43828", "right": "issue:45464", "accept": false, - "reason": "Different surfaces and root causes: autocast dtype mismatch on Phi MoE vs Qwen3.5 streaming chat/completions failure." + "reason": "Autocast dtype mismatch and streaming chat/completions failure are different code paths." }, { "left": "issue:43366", "right": "issue:45464", "accept": false, - "reason": "GGUF gpt-oss support request is unrelated to Qwen3.5 streaming inference." + "reason": "GGUF architecture support and Qwen3.5 streaming inference are not the same bug." }, { "left": "issue:44291", "right": "issue:45070", "accept": false, - "reason": "One is an init_empty_weights argument error; the other is a Pydantic/config-field regression." + "reason": "Unexpected init_empty_weights argument versus pydantic config-field breakage are separate regressions." }, { "left": "issue:43531", "right": "issue:45070", "accept": false, - "reason": "Qwen3-MoE sliding-window bug is unrelated to PretrainedConfig/Pydantic breakage." + "reason": "Qwen3-MoE sliding_window issue is unrelated to the pydantic model-field regression." }, { "left": "issue:43541", "right": "issue:44589", "accept": false, - "reason": "Different failures: tracing/grouped_mm vs missing Float8 storage object." + "reason": "Grouped_mm tracing failure and Float8 storage lookup error are different runtime problems." }, { "left": "issue:43716", "right": "issue:46294", "accept": false, - "reason": "Both are multimodal, but image-preprocessor dtype mismatch and video token/feature mismatch are different bugs." + "reason": "Both are multimodal/video-adjacent, but one is a dtype mismatch and the other is a video-feature/token-count mismatch." }, { "left": "issue:45020", "right": "issue:45800", "accept": false, - "reason": "Broad version-compatibility complaints, but different layers: remote_code breakage vs torch/transformers incompatibility." + "reason": "Remote-code breakage and torch/transformers version incompatibility are separate issues." }, { "left": "issue:41093", "right": "issue:43596", "accept": false, - "reason": "Different failures: mask/tensor shape mismatch vs DeepSpeed ZeRO-3 empty-tensor index error." + "reason": "Both are IndexErrors, but the mask-shape bug and the deepspeed-zero3 init bug are not the same failure." }, { "left": "issue:44479", "right": "issue:46294", - "accept": true, - "reason": "Same underlying Qwen3.5 video-input regression; 46294 looks like a specific manifestation of the broader v5.3.0 video/template bug." + "accept": false, + "reason": "Both mention video inputs, but one is a broad v5.3.0 regression across several models and the other is a specific chat-template mismatch." }, { "left": "issue:43054", "right": "issue:43994", - "accept": true, - "reason": "Same SigLIP2 regression family: both report broken/nonsensical outputs for the same base model variant." + "accept": false, + "reason": "Both concern siglip2, but one reports poor text embeddings while the other reports nonsensical pipeline/model output." }, { "left": "issue:45759", "right": "issue:45902", "accept": false, - "reason": "Different bugs: composite config unwrap issue vs uninitialized rotary-embedding memory after meta to_empty materialization." + "reason": "Unwrapping text_config in from_config and uninitialized inv_freq after to_empty are distinct initialization bugs." }, { "left": "issue:46640", "right": "issue:46661", "accept": false, - "reason": "Unrelated: FlashAttention 2 issue vs remote-code model_class/config_class AttributeError." + "reason": "Flash Attention 2 issue and remote-code model_class/config_class AttributeError are unrelated." }, { "left": "issue:43716", "right": "issue:44918", "accept": false, - "reason": "Different model families and failures: Mistral-3 multimodal dtype mismatch vs Qwen3.5 input-embedding unpacking in TRL." + "reason": "Different multimodal/training failures: image-preprocessor dtype mismatch versus TRL unpacking of Qwen3.5 input embeddings." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "FSDP 2-D weight requirement is unrelated to Qwen3.5/TRL embedding unpacking." + "reason": "FSDP 2-D weight error and Qwen3.5 input-embedding unpacking are unrelated." }, { "left": "issue:40990", "right": "issue:44960", "accept": false, - "reason": "Different models and symptoms: high perplexity on GPT-OSS vs GLM5 issue." + "reason": "High perplexity on gpt-oss and GLM5 are separate model-specific reports." }, { "left": "issue:43645", "right": "issue:45399", "accept": false, - "reason": "Notebook/custom-model initialization breakage is unrelated to FlashAttention fallback gating." + "reason": "Jupyter custom-model initialization and flash-attn2 fallback checks are unrelated." }, { "left": "issue:45632", "right": "issue:45698", "accept": true, - "reason": "Both point to custom-module loading/cache path collisions causing the wrong remote-code module to be loaded." + "reason": "Both point to the same custom-module loading/caching problem: a path/cache collision leading to the wrong module being loaded after save_pretrained/from_pretrained." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Specific lm_head tying bug in Mistral3 is not the same as the broader translation-model tied-weights issue." + "reason": "Tied-weight failure in Mistral3 is a different model-specific report from the broader translation-model tied-weights regression." }, { "left": "issue:42915", "right": "issue:45399", "accept": false, - "reason": "FP8 config failure on Qwen3-MoE is unrelated to FlashAttention fallback checks." + "reason": "FineGrainedFP8Config failure and flash-attn2 fallback blocking are unrelated." }, { "left": "issue:44530", "right": "issue:45005", "accept": false, - "reason": "PagedAttentionCache group-type crash is unrelated to tied-weights behavior." + "reason": "PagedAttentionCache linear_attention crash is not the same as the tied-weights regression." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantization offloading behavior is a different issue from tied weights." + "reason": "Quantization/offloading behavior and tied weights are different bugs." }, { "left": "issue:44360", "right": "issue:44512", "accept": false, - "reason": "DSA indexer/ReLU discussion is unrelated to the removed transformers run command docs issue." + "reason": "A DSA indexer/ReLU discussion and a docs typo are unrelated." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "NemotronH checkpoint loading failure is unrelated to tied-weights regressions." + "reason": "NemotronH checkpoint-loading failure is separate from the tied-weights regression." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "Typo in cache strategies is unrelated to stale v5 docs mentioning the removed run command." + "reason": "Cache-strategy typo and docs text about the removed run command are unrelated." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "Different models and paths: Qwen3.5 input embedding unpacking vs GraniteMoEHybrid invalid method call." + "reason": "Qwen3.5 embedding/trainer failure and GraniteMoE invalid-method calls are unrelated." }, { "left": "issue:45183", "right": "issue:46463", - "accept": true, - "reason": "Same validation-message bug family; 46463 appears to be the T5-specific instance of the broader XOR validation issue." + "accept": false, + "reason": "Both are validation-error complaints, but they describe different checks and different model paths." }, { "left": "issue:46489", "right": "issue:46491", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and likely different tokenization edge cases." + "reason": "Different tokenizer regressions affecting different model families and different round-trip symptoms." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "DeepSeek-Coder detokenization regression is unrelated to Qwen-3 NVFP4 execution failure." + "reason": "DeepSeek-Coder detokenization and Qwen-3 NVFP4 runtime error are unrelated." }, { "left": "issue:44292", "right": "issue:45507", "accept": false, - "reason": "Different model families and root causes: NVFP4 runtime failure vs GraniteMoEHybrid invalid method." + "reason": "Qwen NVFP4 runtime failure and GraniteMoE invalid-method bug are different issues." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet and stays conservative about the soft edges. The accepted pairs are plausibly the same bug family, and the rejected pairs are appropriately cautious rather than overreaching. No canonical target is forced beyond the evidence." + "feedback": "Grounded and conservative overall. The summary matches the packet\u2019s evidence, and the only accepted duplicate pair (issue:45632 \u2194 issue:45698) looks like the same custom-module loading/caching bug and is plausibly mergeable into one fix. The other rejections are cautious rather than overstated." }, "evaluator_used": true, "retried": false, @@ -2222,8 +2222,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -2307,7 +2307,7 @@ "issue:44387|issue:44756", "issue:45663|issue:45859" ], - "prepared_review_unit_hash": "0f45d23bbf0338efeca2d4a3c2522a5a4cd9873f380507a419698cedf35503ec", + "prepared_review_unit_hash": "425fc65696a75267e6ed61aa06d4b90d1b9ca165b314d7bb97f72677780a5e10", "component_budget": { "node_count": 484, "item_count": 484, @@ -2329,198 +2329,198 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a loose collection of unrelated Transformers bugs; only two soft pairs look like true duplicates or near-duplicates with the same underlying code-path bug.", + "summary": "Most candidate pairs are thematically related but not true duplicates. I\u2019d merge two pairs as the same underlying code-path bug: the `resize_token_embeddings` regressions and the missing `lm_head` weight-tying regressions. The rest are better treated as separate issues.", "confidence": 0.79, - "canonical_issue_reason": "Issue 43404 is the strongest representative of the cluster\u2019s clearest duplicate pair: a concrete missing lm_head weight-tying bug with an explicit affected API and a model-specific manifestation.", + "canonical_issue_reason": "Issue 45276 is the strongest canonical issue: it describes a concrete, user-visible regression in `resize_token_embeddings` with clear affected outputs, and its paired issue is the narrower variant of the same behavior on a related model family.", "canonical_pr_reason": null, - "best_issue_reason": "43404 is the best overall issue candidate because it is specific, actionable, and ties to a concrete generation correctness failure rather than a broad symptom-only report.", + "best_issue_reason": "Issue 45276 is the best global issue anchor because it is specific, actionable, and broad enough to represent the shared embedding-resize regression pattern across the duplicate pair.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45507", "right": "issue:46468", "accept": false, - "reason": "Different model families and different failures; both are broken examples, but not the same underlying bug." + "reason": "Different models and failure modes; both are generic broken-example reports, not the same underlying bug." }, { "left": "issue:45313", "right": "issue:45357", "accept": false, - "reason": "Both are Qwen3.5 regressions, but one is ZeRO-3 weight loading and the other is incorrect save_pretrained key handling; different code paths." + "reason": "Both involve Qwen3.5, but one is ZeRO-3 weight loading and the other is incorrect save_pretrained visual encoder keys; separate code paths." }, { "left": "issue:43644", "right": "issue:43645", "accept": false, - "reason": "Unrelated Transformers 5.0 regressions: junk non-persistent buffers vs Jupyter custom-model initialization." + "reason": "Both are Transformers 5.0 regressions, but one is non-persistent buffers and the other is Jupyter custom-model init; unrelated bugs." }, { "left": "issue:44488", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer/model loading failures on different models, but the titles point to distinct bugs rather than one shared root cause." + "reason": "Both are tokenizer/model loading failures, but they affect different models and likely different tokenizer code paths." }, { "left": "issue:43856", "right": "issue:43873", "accept": false, - "reason": "Memory/offloading symptoms are related broadly, but one is Qwen3 MoE training memory use and the other is quantization offloading behavior." + "reason": "One is Qwen3 MoE training memory usage, the other is quantization offloading; not the same bug." }, { "left": "issue:43856", "right": "issue:44387", "accept": false, - "reason": "Both mention memory/OOM, but the underlying causes differ: MoE training memory use vs int4 quantization reserved-memory growth." + "reason": "Both mention memory/OOM, but one is training memory inefficiency and the other is int4 quantization reserved-memory growth; distinct issues." }, { "left": "issue:39692", "right": "issue:43994", "accept": false, - "reason": "Same model family, but the first is a docs/example mismatch and quantization failure while the second is nonsensical outputs with AutoModel/pipeline." + "reason": "Shared SigLIP2 theme, but 39692 is a docs/example mismatch plus quantization failure while 43994 is bad outputs from AutoModel/pipeline; not the same concrete defect." }, { "left": "issue:42491", "right": "issue:45292", "accept": false, - "reason": "Different bugs: Qwen3 MoE LoRA compatibility on hf5.x vs resize_token_embeddings not updating output embeddings." + "reason": "Different models and different APIs; one is LoRA compatibility across branches, the other is resize_token_embeddings not updating output embeddings." }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Different model-loading issues: Qwen3 MoE weights not converted under accelerate/deepspeed vs GLM-4.7-Flash load failure." + "reason": "Different models and loading failures; no shared code-path evidence." }, { "left": "issue:43708", "right": "issue:44075", "accept": false, - "reason": "Trainer checkpoint-resume step calculation and SGD argument handling are unrelated." + "reason": "Trainer checkpoint resumption math and SGD argument handling are unrelated." }, { "left": "issue:43854", "right": "issue:43931", "accept": false, - "reason": "Different model-specific load failures with different symptoms and code paths." + "reason": "Different model-loading failures on different models; no sign of a shared root cause." }, { "left": "issue:42491", "right": "issue:43931", "accept": false, - "reason": "Unrelated: LoRA/qwen3_moe hf4\u2192hf5 compatibility versus Qwen3-VL weight-shape mismatch." + "reason": "LoRA branch-compatibility problem versus Qwen3-VL shape mismatch; separate issues." }, { "left": "issue:44960", "right": "issue:46050", "accept": false, - "reason": "Too vague to unify; GLM5 loading vs quantized model save/load corruption are not clearly the same bug." + "reason": "GLM5 model-loading issue versus sinq quantization save/load corruption; different bugs." }, { "left": "issue:44164", "right": "issue:44291", "accept": false, - "reason": "Both involve loading/saving, but extra_state handling and init_empty_weights argument propagation are distinct failures." + "reason": "Both touch save/load paths, but extra_state handling and unexpected `_is_hf_initialized` during init_empty_weights are different failures." }, { "left": "issue:45488", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions in v5, but they affect different tokenizers/models and do not clearly share one code-path bug." + "reason": "Both are tokenizer regressions, but they affect different tokenizer classes/models and the underlying changes are not clearly the same." }, { "left": "issue:45362", "right": "issue:45406", "accept": false, - "reason": "Different issues: Qwen3.5 chat crash vs Gemma4Processor missing _tokenizer in serve." + "reason": "Different frontends and different model families; no shared underlying bug is evident." }, { "left": "issue:45276", "right": "issue:45335", "accept": true, - "reason": "Both report resize_token_embeddings not propagating to model-specific embedding/output modules; same underlying weight-resize bug." + "reason": "Both report the same `resize_token_embeddings` regression: the resize does not propagate to downstream embeddings/output heads in related composite model classes." }, { "left": "issue:44910", "right": "issue:45588", "accept": false, - "reason": "Both touch flash_attention, but one is a Qwen3.5 packed-sequence position_ids crash and the other is an s_aux=None AttributeError." + "reason": "Both mention flash attention, but one is a position_ids packing crash and the other is an s_aux=None AttributeError; different code paths." }, { "left": "issue:44402", "right": "issue:44488", "accept": false, - "reason": "Different tokenizer/model-loading problems; no clear shared root cause." + "reason": "Both are tokenizer-loading complaints, but they concern different models and likely different special-token handling bugs." }, { "left": "issue:43425", "right": "issue:43957", "accept": false, - "reason": "Torch 2.10 incompatibility is a separate issue from meta-device loading breakage." + "reason": "General version incompatibility versus meta-device loading failures; too broad and not the same defect." }, { "left": "issue:43873", "right": "issue:44387", "accept": false, - "reason": "Both mention quantization/memory, but the concrete problems differ: offloading behavior vs CUDA reserved-memory OOM." + "reason": "Offloading/quantization behavior versus int4 reserved-memory OOM; separate memory problems." }, { "left": "issue:39692", "right": "issue:44960", "accept": false, - "reason": "Different models and different failure modes; not a duplicate." + "reason": "SigLIP2 docs/example issue and GLM5 loading issue are unrelated." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "Qwen3 auto device mapping cudaErrorAssert and AudioFlamingo3 track leakage are unrelated." + "reason": "Qwen3 auto device-map CUDA assert versus AudioFlamingo3 batched inference leakage; unrelated." }, { "left": "issue:42947", "right": "issue:45704", "accept": false, - "reason": "Gradient checkpointing/LoRA and apex RMSNorm memory leak are unrelated." + "reason": "Gradient checkpointing with PEFT LoRA is unrelated to the apex RMSNorm memory leak." }, { "left": "issue:45663", "right": "issue:45759", "accept": false, - "reason": "Different model/config bugs: Gemma4 FSDP2 shared_kv_states handling vs Qwen 3.5/3.6 text_config unwrapping." + "reason": "Gemma-4 FSDP2 KeyError and Qwen multimodal config unwrapping are different model-internal bugs." }, { "left": "issue:43404", "right": "issue:43454", "accept": true, - "reason": "Both describe missing lm_head weight tying in multimodal generation models leading to bad/garbage text output; same underlying bug class." + "reason": "Both describe missing `lm_head` weight tying in multimodal generation models, causing incorrect or garbage generation; same underlying fix class." }, { "left": "issue:44291", "right": "issue:44756", "accept": false, - "reason": "init_empty_weights argument handling and Strix Halo mmap/OOM are unrelated." + "reason": "Meta/init_empty_weights loading error versus disabling mmap to avoid OOM; unrelated." }, { "left": "issue:36010", "right": "issue:39404", "accept": false, - "reason": "ImportError for GenerationMixin and Whisper return_language pipeline regression are different issues." + "reason": "ImportError for GenerationMixin is unrelated to Whisper pipeline return_language behavior." }, { "left": "issue:44387", "right": "issue:44756", "accept": false, - "reason": "Both concern memory, but one is quantization reserved-memory growth and the other is mmap-related OOM on Strix Halo." + "reason": "Different OOM causes: quantization reserved memory versus mmap on Strix Halo." }, { "left": "issue:45663", "right": "issue:45859", "accept": false, - "reason": "Different model families and different failures: FSDP2 attention state rebuild vs CPU-offload rotary embedding incompatibility." + "reason": "Both involve Qwen/Gemma model internals, but one is FSDP2 shared_kv_states and the other is CPU-offload rotary embedding compatibility." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet, and the soft-edge calls are conservative overall. The two accepted pairs are well supported by the titles/body themes: 45276\u219445335 for resize_token_embeddings propagation, and 43404\u219443454 for lm_head tying / garbage generation. The rejected pairs mostly distinguish different model families or different code paths without overclaiming duplicate status. One mild caution: the canonical-issue wording should stay a bit softer about being a \u2018duplicate pair\u2019 unless the packet explicitly shows that, but the current phrasing is still reasonable." + "feedback": "The summary is grounded in the packet and stays mostly conservative. The two accepted pairs are well-supported by the issue titles: 45276/45335 both describe `resize_token_embeddings` failing to propagate to downstream embeddings, and 43404/43454 both describe missing `lm_head` weight tying causing bad generation. The remaining pairs are reasonably treated as separate issues based on the available evidence." }, "evaluator_used": true, "retried": false, @@ -3817,8 +3817,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -3902,7 +3902,7 @@ "issue:43701|issue:45325", "issue:42796|issue:45381" ], - "prepared_review_unit_hash": "82aa4b7cedd6fa54b33e07af764ea85fb06c77f8269a7d6ff3472b70720138ed", + "prepared_review_unit_hash": "0b9362b48509b4ebc9e8dcaffbc803795df87ed1150f9b97c4b57d2eb2fd9839", "component_budget": { "node_count": 484, "item_count": 484, @@ -3924,196 +3924,193 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous regression reports. The strongest duplicate groups are the Qwen2.5-VL position-id bugs (45325 \u2194 45381) and the SigLIP2 implementation mismatch (43493 \u2194 43994). The 40444 \u2194 45381 pair is related-sounding but not evidenced strongly enough to treat as a duplicate.", - "confidence": 0.78, - "canonical_issue_reason": "issue:45325 is the best canonical issue because it states a concrete Qwen2.5-VL `get_rope_index` / position-id regression, names the affected version, and is the clearest root-cause report in the strongest duplicate cluster.", + "summary": "No soft pair is a safe duplicate under strict triage. There are a few thematic overlaps (notably Qwen2.5-VL position-id bugs and several Transformers v5 loading regressions), but each pair appears to describe a different concrete failure mode or code path.", + "confidence": 0.9, + "canonical_issue_reason": "Issue 44291 is the most broadly representative loading regression in the set: it targets a concrete transformers>=5.0.0rc0 instantiation failure (`init_empty_weights` / unexpected `_is_hf_initialized`), which is more central and reusable than the narrower model-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45325 is the best representative issue for triage because it is the most specific, actionable, and appears to subsume the closely related Qwen2.5-VL video/position-id reports.", + "best_issue_reason": "Issue 44291 is the best single issue to anchor the cluster because it captures a clear upstream v5 loading breakage affecting many models, while the other candidates are either more specialized or only loosely related.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "Different failure modes: tokenizer-loading error messaging vs Apple Silicon allocator TypeError." + "reason": "Both are user-facing failures, but one is a tokenizer-loading error for Voxtral and the other is an AppleSilicon allocator TypeError; different subsystems and root causes." }, { "left": "issue:43531", "right": "issue:44291", "accept": false, - "reason": "Both are loading-related regressions, but they point to different concrete bugs: Qwen3-MoE sliding-window behavior vs `_is_hf_initialized` argument handling." + "reason": "Both mention Transformers v5-era regressions, but one is a Qwen3-MoE sliding-window bug and the other is an `init_empty_weights` loading-API break; not the same code path." }, { "left": "issue:43317", "right": "issue:43856", "accept": false, - "reason": "One is dequantized model offload loading failure; the other is Qwen3 MoE training memory usage." + "reason": "Both involve offload/quantization behavior, but one is a dequantized model load issue with `device_map=auto` and the other is inefficient Qwen3 MoE training memory usage; different bugs." }, { "left": "issue:42175", "right": "issue:45542", "accept": false, - "reason": "Packaging/backend overlap only; missing TensorFlow extra and tensorboard-without-TF runtime error are not the same bug." + "reason": "Both concern TensorFlow-related packaging/backend availability, but one is about `.[torch]` not including TensorFlow and the other is about a missing TF backend when only tensorboard is installed." }, { "left": "issue:43421", "right": "issue:44568", "accept": false, - "reason": "Both involve tokenizer special tokens, but one is runtime post-processor updates and the other is a BOS/EOS regression in a specific tokenizer." + "reason": "Both touch special-token handling, but one is runtime post-processor refresh in TokenizersBackend and the other is `add_special_tokens=True` failing for a specific tokenizer/model." }, { "left": "issue:44079", "right": "issue:44534", "accept": false, - "reason": "Unrelated regressions: ModelOutput key assignment versus non-persistent buffer initialization." + "reason": "These are unrelated v5 regressions: one is `ModelOutput` key assignment semantics, the other is non-persistent buffer initialization." }, { "left": "issue:41720", "right": "issue:45081", "accept": false, - "reason": "Different models and code paths: Qwen3 auto device mapping crash vs Mistral regex patch crash." + "reason": "One is a Qwen3 auto-device-map CUDA assert, the other is a Mistral tokenizer regex patch crash; different components and failure modes." }, { "left": "issue:44829", "right": "issue:45479", "accept": false, - "reason": "Both mention degenerate training, but the concrete causes differ: flash_attention_3 behavior vs `num_labels=1` loss handling." + "reason": "One is flash_attention_3 causing degenerate training, the other is a classification loss edge case with `num_labels=1`; not the same bug." }, { "left": "issue:44462", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but one ignores `tokenizer.json` and the other produces wrong token IDs for Granite models." + "reason": "Both are AutoTokenizer regressions, but one ignores `tokenizer.json` while the other returns wrong token IDs for Granite models; separate tokenizer paths." }, { "left": "issue:43611", "right": "issue:43646", "accept": false, - "reason": "Both are Transformers v5 init/load regressions, but the titles indicate distinct breakages rather than one underlying bug." + "reason": "Both are loading regressions in Transformers v5, but one is about `base_model_prefix` and the other about custom model initialization; too different to merge." }, { "left": "issue:43377", "right": "issue:45491", "accept": false, - "reason": "Both are batching/masking-related, but the affected model and root cause differ: MIMI padding masks vs Gemma3 sliding-window NaNs." + "reason": "One is a MIMI padding-mask batching discrepancy, the other is Gemma3 sliding-window NaNs; different models and mechanisms." }, { "left": "issue:43818", "right": "issue:46402", "accept": false, - "reason": "Different architectures and symptoms: Video-LLaVA tower wiring vs Sam3Model text encoder weights not loading." + "reason": "One is a Video-LLaVA vision-tower/temporal-attention problem, the other is Sam3Model text-encoder weights not loading." }, { "left": "issue:42175", "right": "issue:43296", "accept": false, - "reason": "Packaging/backend issue vs vLLM loading failure for PaddleOCR-VL; not the same bug." + "reason": "One is a packaging/backend install issue, the other is a PaddleOCR-VL load failure under a specific transformers/vLLM combination." }, { "left": "issue:43644", "right": "issue:44079", "accept": false, - "reason": "Non-persistent buffer junk fill and ModelOutput key assignment are unrelated regressions." + "reason": "One concerns non-persistent buffers being filled with junk; the other concerns `ModelOutput` key assignment after `None` values." }, { "left": "issue:43818", "right": "issue:45357", "accept": false, - "reason": "Different bugs in different models: Video-LLaVA tower setup vs Qwen3.5 visual-encoder save_pretrained keys." + "reason": "These are both multimodal model issues, but one is Video-LLaVA architecture/weights and the other is Qwen3.5 visual-encoder key serialization; different concrete defects." }, { "left": "issue:43749", "right": "issue:44079", "accept": false, - "reason": "FSDP CPU RAM-efficient loading is unrelated to ModelOutput key assignment." + "reason": "`FSDP_CPU_RAM_EFFICIENT_LOADING` and `ModelOutput` key assignment are unrelated transformers behaviors." }, { "left": "issue:44265", "right": "issue:45072", "accept": false, - "reason": "torch.export / `torch_compilable_check` failure is unrelated to bf16 dtype mismatches." + "reason": "One is `torch.export` failing on `torch_compilable_check`, the other is a dtype mismatch in bfloat16 inference; no shared underlying bug." }, { "left": "issue:43792", "right": "issue:43957", "accept": false, - "reason": "Different load paths and models: Whisper runtime failure vs meta-device loading regression." + "reason": "One is Whisper model execution failure, the other is `torch.device('meta')` loading breakage across some models; related to loading only at a very high level." }, { "left": "issue:45325", "right": "issue:45381", - "accept": true, - "reason": "Same underlying Qwen2.5-VL position-id regression: one reports still-image scaling and the other video position-id behavior, both pointing to the same broken multimodal position handling." + "accept": false, + "reason": "Both are Qwen2.5-VL position-id reports, but one is about still-image temporal scaling and the other about video-input vision position IDs; similar area, not enough evidence of one bug." }, { "left": "issue:43493", "right": "issue:43994", - "accept": true, - "reason": "Strong duplicate candidate: both describe the same SigLIP2 implementation mismatch, one from HF-vs-JAX comparison and one from the user-visible bad outputs." + "accept": false, + "reason": "Both involve SigLIP2 outputs, but one is an implementation discrepancy against JAX and the other is bad AutoModel/pipeline inference results; not clearly the same defect." }, { "left": "issue:43454", "right": "issue:44466", "accept": false, - "reason": "Both mention lm_head / weights, but one is tied-weight setup and the other is device-dependent serialization; not the same bug." + "reason": "One is a missing weight-tie bug in AyaVision, the other is inconsistent `lm_head.weight` serialization tied to device in v5." }, { "left": "issue:45325", "right": "issue:45834", "accept": false, - "reason": "Qwen2.5-VL position-id scaling and Kosmos2.5 long-OCR index errors are different failures." + "reason": "Both are multimodal/vision-related, but the bugs are different: Qwen2.5-VL rope-index scaling vs Kosmos2.5 OCR index overflow." }, { "left": "issue:40444", "right": "issue:45381", "accept": false, - "reason": "Only broad multimodal similarity is shown; the packet does not establish the same bug or root cause, so this is related but not a duplicate." + "reason": "Both mention Qwen2.5-VL and image/video inputs, but one is an IterableDataset multi-image training failure and the other is a video position-id regression." }, { "left": "issue:42915", "right": "issue:43482", "accept": false, - "reason": "Different model families and failure causes: Qwen3Moe FP8 config vs Qwen2.5-GGUF loading." + "reason": "One is a Qwen3MoE FineGrainedFP8Config failure, the other is a Qwen2.5-GGUF loading regression under transformers v5; distinct model/loading paths." }, { "left": "issue:44291", "right": "issue:44877", "accept": false, - "reason": "`init_empty_weights` argument mismatch is unrelated to strict config parsing for granite_speech." + "reason": "One is an unexpected-argument failure in model init under `init_empty_weights`, the other is strict config validation blocking `granite_speech` config loading." }, { "left": "issue:43632", "right": "issue:45072", "accept": false, - "reason": "`_is_hf_initialized` flag regression and bf16 dtype mismatch are separate issues." + "reason": "One is the `_is_hf_initialized` flag regression, the other is a dtype mismatch in bfloat16 inference; unrelated defects." }, { "left": "issue:43873", "right": "issue:45701", "accept": false, - "reason": "Quantization/offloading behavior and tokenization changes are unrelated." + "reason": "One is an offloading-with-quantization issue, the other is version-dependent tokenization changes; different subsystems entirely." }, { "left": "issue:43701", "right": "issue:45325", "accept": false, - "reason": "Checkpoint key mismatch is unrelated to Qwen2.5-VL position-id scaling." + "reason": "Checkpoint resume key mismatch is unrelated to Qwen2.5-VL rope/position-id handling." }, { "left": "issue:42796", "right": "issue:45381", "accept": false, - "reason": "Tokenizer BPE marker handling is unrelated to Qwen2.5-VL vision position IDs." + "reason": "A tokenizer BPE-marker bug and a Qwen2.5-VL video position-id bug are unrelated." } ] }, - "evaluator_result": { - "accept": false, - "feedback": "The cluster-level summary is mostly grounded, but one soft-edge acceptance is too strong for the evidence available. In particular, 40444 \u2194 45381 is only supported by broad Qwen2.5-VL multimodal similarity from the titles; the packet does not establish the same bug or root cause, so that should be treated more cautiously (related/uncertain rather than duplicate). The rest of the rejections look conservative, and 45325 \u2194 45381 / 43493 \u2194 43994 are the strongest duplicate candidates." - }, - "evaluator_used": true, - "retried": true, - "accepted_nontrivial_soft_edge": true, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -4125,8 +4122,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -4211,7 +4208,7 @@ "issue:41720|issue:43322", "issue:36331|issue:42617" ], - "prepared_review_unit_hash": "15f6718387fca34d2be9416c6816495c053ac1097751ba82cc9dc933c4daa6a4", + "prepared_review_unit_hash": "4373fe8c8aa347477fb9c5fe75a31297216038d825a72022496de98a4decc6bf", "component_budget": { "node_count": 484, "item_count": 484, @@ -4233,192 +4230,192 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a loose collection of unrelated bug reports across tokenizers, loading/offload, multimodal processors, training, and model-specific regressions. I would not merge any of the soft pairs; they share broad subsystem themes at most, not the same underlying bug.", - "confidence": 0.91, - "canonical_issue_reason": "No single true canonical issue stands out because the cluster is heterogeneous; if a representative is needed, issue 43873 is the broadest offloading/quantization report and has the most cross-cutting discussion, but it is not a duplicate anchor for the rest.", + "summary": "These items are largely a grab bag of distinct bugs/regressions across loading, offloading, tokenization, multimodal processors, and training/runtime paths. I don\u2019t see any soft pair that looks like the same underlying bug or a mergeable duplicate PR/change.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43873 is the broadest representative of the loading/offloading/quantization problems here and is still active with discussion, so it works best as the cluster anchor.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43873 is the best cluster representative because it is broad, currently open, and has the most inbound references, but it still describes a different bug than the other issues.", + "best_issue_reason": "issue:43873 is the most general and active issue in the set, with the clearest umbrella fit for nearby offloading/loading regressions.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43296", "right": "issue:45800", "accept": false, - "reason": "One is a PaddleOCR-VL/vLLM load failure; the other is a broad torch/transformers incompatibility report. Different failure modes and code paths." + "reason": "Different failures: model loading in vLLM vs a torch/transformers version incompatibility." }, { "left": "issue:44479", "right": "issue:45823", "accept": false, - "reason": "Both involve multimodal models, but one is a video-input regression and the other is a Gemma4 PLE device-mapping forward error. Not the same bug." + "reason": "Different subsystems and symptoms: video-input regression vs device-mapping/forward device mismatch." }, { "left": "issue:43606", "right": "issue:44265", "accept": false, - "reason": "CPU-offload device mismatch for Bark-small is unrelated to torch.export failures around torch_compilable_check." + "reason": "One is CPU offload device mismatch for Bark; the other is torch.export failing on torch_compilable_check." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "High perplexity on gpt-oss-20b and Whisper empty-transcription generation failure are unrelated symptoms in different models." + "reason": "Unrelated bugs: perplexity regression vs Whisper empty-transcription handling." }, { "left": "issue:43262", "right": "issue:43746", "accept": false, - "reason": "Audio chat-template sample-rate defaults and GraniteSpeech PEFT checkpoint loading are separate audio/loader issues." + "reason": "Audio template sample-rate default bug vs PEFT adapter checkpoint loading; not the same code path." }, { "left": "issue:35707", "right": "issue:41720", "accept": false, - "reason": "inputs_embeds/past_key_values progressive generation and Qwen3 auto device-map cuda asserts are different generation vs placement bugs." + "reason": "Both are generation-related, but one is inputs_embeds/past_key_values behavior and the other is auto device mapping crashing." }, { "left": "issue:43881", "right": "issue:44488", "accept": false, - "reason": "Both are load failures, but for different models and likely different causes; not enough to indicate one shared defect." + "reason": "Both are load failures, but for different models and likely different causes." }, { "left": "issue:34634", "right": "issue:35707", "accept": false, - "reason": "Bark voice_preset behavior is unrelated to progressive generation with inputs_embeds/past_key_values." + "reason": "BarkProcessor voice_preset issue is unrelated to progressive generation with inputs_embeds/past_key_values." }, { "left": "issue:43454", "right": "issue:43550", "accept": false, - "reason": "AyaVision weight tying and Bamba torch.compile SDPA failure are unrelated model internals." + "reason": "Weight-tying/generation bug vs torch.compile+SDPA failure; different concrete problems." }, { "left": "issue:43606", "right": "issue:43819", "accept": false, - "reason": "Bark CPU offload device mismatch is not the same as DAC.from_latents diverging from forward without STE." + "reason": "Device mismatch during CPU offload is unrelated to DAC.from_latents vs forward mismatch." }, { "left": "issue:39692", "right": "issue:40444", "accept": false, - "reason": "SigLIP2 documentation errors and Qwen2.5-VL iterable-dataset finetuning failures are different kinds of issues." + "reason": "Documentation/example errors for SigLIP2 vs finetuning failure on Qwen2.5-VL; not the same issue." }, { "left": "issue:41720", "right": "issue:45823", "accept": false, - "reason": "Both mention device mapping, but Qwen3 cudaErrorAssert and Gemma4 PLE forward mismatch are not the same concrete bug." + "reason": "Both involve device placement, but they hit different models and different failure modes." }, { "left": "issue:44186", "right": "issue:44265", "accept": false, - "reason": "LayoutLMv2 tokenizer crashes and torch.export/exportability failures are unrelated subsystems." + "reason": "Tokenizer crash on NER/padding vs export failure from torch_compilable_check; different bugs." }, { "left": "issue:43377", "right": "issue:44610", "accept": false, - "reason": "Missing padding-mask support in MIMI encoder and OmDet-Turbo processor/model size mismatch are different preprocessing bugs." + "reason": "MIMI batched-vs-single output mismatch is unrelated to OmDet-Turbo processor image-size mismatch." }, { "left": "issue:46305", "right": "issue:46315", "accept": false, - "reason": "Sliding-window off-by-one in MoonshineStreaming and slow convert_tokens_to_ids are unrelated." + "reason": "Sliding-window off-by-one vs tokenizer conversion slowdown; no shared underlying defect." }, { "left": "issue:43257", "right": "issue:45923", "accept": false, - "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed and a missing gradient-checkpointing flag on Nemotron are different model-loading/config issues." + "reason": "Weights not converted during accelerate+deepspeed loading vs missing supports_gradient_checkpointing flag; different fixes." }, { "left": "issue:44568", "right": "issue:45920", "accept": false, - "reason": "mdeberta-v3 special-token addition regression and broader AutoTokenizer wrong-token-ID regression are related only at a very high level, not the same bug." + "reason": "Specific add_special_tokens regression for one tokenizer vs broader wrong-token-id regression across multiple models." }, { "left": "issue:43454", "right": "issue:45072", "accept": false, - "reason": "Weight tying in AyaVision and dtype mismatches in SwitchTransformers/TimmWrapperModel are unrelated." + "reason": "lm_head tying bug is unrelated to dtype mismatch in bfloat16 inference." }, { "left": "issue:42796", "right": "issue:46209", "accept": false, - "reason": "Tokenizer BPE-marker handling and FP8 inference on B200 are entirely different problems." + "reason": "Tokenizer BPE marker handling vs FP8 inference on B200; unrelated." }, { "left": "issue:43317", "right": "issue:43873", "accept": false, - "reason": "Both touch offload/quantization, but one is dequantized-model loading on gpu+cpu offload while the other is a broader offloading regression; not the same concrete failure." + "reason": "Both mention offloading/quantization, but one is dequantized model loading on gpu+cpu offload and the other is a broader offloading regression; not clearly the same bug." }, { "left": "issue:43377", "right": "issue:44898", "accept": false, - "reason": "MIMI padding-mask output mismatch and Perceiver non-default-resolution failure are different model-specific issues." + "reason": "Different model components and symptoms: encoder padding-mask mismatch vs image-classification resolution handling." }, { "left": "issue:43454", "right": "issue:43720", "accept": false, - "reason": "AyaVision weight tying and BitNet packed-weight unpacking during accelerate loading are unrelated." + "reason": "Text generation weight-tying bug vs packed-weight unpacking during accelerate loading." }, { "left": "issue:40990", "right": "issue:42947", "accept": false, - "reason": "Perplexity regression on gpt-oss-20b and gradient checkpointing ineffectiveness with LoRA are unrelated." + "reason": "Perplexity regression is unrelated to gradient checkpointing ineffectiveness with PEFT LoRA." }, { "left": "issue:42617", "right": "issue:42915", "accept": false, - "reason": "3d_parallel.py execution failure and Qwen3Moe FineGrainedFP8Config failure are different model/runtime issues." + "reason": "3d_parallel.py runtime problem vs Qwen3 MoE FP8 config failure; different issues." }, { "left": "issue:43749", "right": "issue:44534", "accept": false, - "reason": "FSDP CPU RAM efficient loading and non-persistent buffer junk on v5 are both loading-related but not the same bug." + "reason": "FSDP CPU RAM efficient loading is unrelated to non-persistent buffers being filled with junk." }, { "left": "issue:43572", "right": "issue:45759", "accept": false, - "reason": "StableLm pad_token_idx regression and composite Qwen text_config unwrapping are separate config issues." + "reason": "Missing pad_token_idx after v5 update vs failing to unwrap text_config in composite multimodal configs." }, { "left": "issue:43526", "right": "issue:43883", "accept": false, - "reason": "BeitImageProcessorFast label reduction and Molmo missing all_tied_weights_keys are unrelated." + "reason": "BeitImageProcessorFast reduce_labels bug is unrelated to Molmo missing all_tied_weights_keys." }, { "left": "issue:43322", "right": "issue:46050", "accept": false, - "reason": "Llava Next segfault on load and sinq save/load quantization breakage are different failure modes." + "reason": "Llava Next loading segfault vs sinq save/load breakage; different code paths." }, { "left": "issue:41720", "right": "issue:43322", "accept": false, - "reason": "Both are model-loading problems, but Qwen3 auto device-map asserts and Llava Next segmentation faults are not the same code-path bug." + "reason": "Auto device mapping assert on Qwen3 vs segfault loading Llava Next; not the same defect." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "CustomTrainer.compute_loss signature mismatch and 3d_parallel.py runtime failure are unrelated." + "reason": "Trainer compute_loss signature breakage vs inability to run 3d_parallel.py; unrelated." } ] }, @@ -4437,8 +4434,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -4522,7 +4519,7 @@ "issue:28530|issue:33666", "issue:42915|issue:46291" ], - "prepared_review_unit_hash": "b164178f80a5b6b8e285d225411cefc8eda4742a1c98479e610935f643df7369", + "prepared_review_unit_hash": "8803e6e9097eb1f24f54ebe178e1691e4018da6d1505b92e460b4dcc912737f0", "component_budget": { "node_count": 484, "item_count": 484, @@ -4544,192 +4541,192 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The items are mostly unrelated regression reports; none of the soft pairs look like the same underlying bug/change closely enough to merge. The closest thematic overlap is around Transformers 5.x loading/init regressions, but the concrete failure modes differ.", - "confidence": 0.39, - "canonical_issue_reason": "issue:43646 is the broadest release-level regression in the set and the best umbrella representative for the cluster, even though the overall group is heterogeneous.", + "summary": "This cluster is highly heterogeneous: it mixes trainer/checkpointing bugs, tokenizer/processor regressions, quantization/offload issues, multimodal model failures, and serving crashes. None of the soft pairs look like the same underlying bug/change, so I would not merge any of them.", + "confidence": 0.94, + "canonical_issue_reason": "If a single issue must represent the cluster, issue 43708 is the clearest and most concrete bug report; however, the set is too diverse to have a strong true canonical issue.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43646 has the widest scope and reads like the most general user-facing breakage among the listed issues, so it is the least bad canonical issue choice.", + "best_issue_reason": "Issue 43708 is the best standalone issue because it is specific, reproducible, and tightly scoped to a real Trainer checkpointing bug, unlike the broader or more model-specific reports in the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43881", "right": "issue:44991", "accept": false, - "reason": "Both are load failures, but one is a model-loading failure and the other is a tokenizer-loading failure for a different model; no shared code-path bug is evident." + "reason": "Different problems: GLM-4V loading failure vs tokenizer loading failure for a different model family." }, { "left": "issue:44387", "right": "issue:45704", "accept": false, - "reason": "Quantization OOM from reserved memory and an apex RMSNorm memory leak are different failure modes and likely different fixes." + "reason": "Different failure modes: int4 quantization OOM vs T5 using apex FusedRMSNorm memory leak." }, { "left": "issue:41720", "right": "issue:44977", "accept": false, - "reason": "Different symptoms and code paths: cuda assert during auto device mapping vs flash-attention generation breakage." + "reason": "Different code paths: auto device-map CUDA assert on Qwen3 vs flash-attention generation issue on Qwen3.5." }, { "left": "issue:36331", "right": "issue:42915", "accept": false, - "reason": "Trainer API incompatibility and a Qwen3MoE FP8 failure are unrelated problems." + "reason": "Unrelated bugs: custom trainer loss signature change vs Qwen3Moe FineGrainedFP8Config failure." }, { "left": "issue:44877", "right": "issue:44991", "accept": false, - "reason": "Strict config loading and tokenizer loading are separate issues with different triggers." + "reason": "Different areas: strict config blocking granite_speech load vs tokenizer loading regression for est-roberta." }, { "left": "issue:44466", "right": "issue:45357", "accept": false, - "reason": "Both involve serialization, but one is tied-weight serialization and the other is incorrect visual encoder keys; different model paths." + "reason": "Both involve model saving, but one is lm_head serialization/tied weights and the other is incorrect visual encoder keys in Qwen3.5." }, { "left": "issue:43450", "right": "issue:45701", "accept": false, - "reason": "A batched video processor shape bug and a tokenization-version regression are unrelated." + "reason": "Different subsystems: batched video processor shape bug vs tokenization behavior changing across versions." }, { "left": "issue:40990", "right": "issue:45701", "accept": false, - "reason": "Perplexity regression on a model is not the same as tokenization changing across versions." + "reason": "Perplexity anomaly on gpt-oss is not the same issue as version-dependent tokenization changes." }, { "left": "issue:41720", "right": "issue:42617", "accept": false, - "reason": "Qwen3-VL device-mapping crash and a 3d_parallel.py execution failure do not point to the same underlying bug." + "reason": "Different failures: Qwen3 CUDA assert under auto device mapping vs 3d_parallel.py launch/runtime failure." }, { "left": "issue:43065", "right": "issue:46402", "accept": false, - "reason": "Both are SAM3-related, but one is a dummy Conv2d in the pixel decoder and the other is missing text encoder weights." + "reason": "Same model family, but different submodules and bugs: dummy Conv2d in pixel decoder vs text encoder weights not loading." }, { "left": "issue:43873", "right": "issue:44756", "accept": false, - "reason": "Quantization offload behavior and a Strix Halo mmap OOM are different system-level problems." + "reason": "Quantization offload behavior and Strix Halo mmap OOM are different root causes." }, { "left": "issue:44186", "right": "issue:45072", "accept": false, - "reason": "Tokenizer crashes on NER/padding and dtype mismatches in inference are not the same bug." + "reason": "Tokenizer crash on NER/padding is unrelated to dtype mismatches in inference." }, { "left": "issue:26413", "right": "issue:28530", "accept": false, - "reason": "Checkpoint resume failure and early-stopping metric resolution are separate Trainer issues." + "reason": "Both are Trainer-related, but checkpoint resume epoch-iterator failure is not the same as early-stopping metric lookup." }, { "left": "issue:43550", "right": "issue:43720", "accept": false, - "reason": "torch.compile+SDPA failure and packed-weight unpacking during accelerate loading are unrelated." + "reason": "Different code paths: torch.compile with SDPA on Bamba vs BitNet packed-weight unpacking during accelerate loading." }, { "left": "issue:43577", "right": "issue:43708", "accept": false, - "reason": "Blip2 dtype not propagating and resume max_steps miscalculation are different subsystems." + "reason": "Blip2 dtype propagation and resume_from_checkpoint max_steps miscalculation are unrelated." }, { "left": "issue:43550", "right": "issue:43606", "accept": false, - "reason": "Different models and failures: compile/SDPA vs CPU offload device mismatch." + "reason": "Different model-specific bugs: Bamba SDPA compile failure vs bark-small CPU offload device mismatch." }, { "left": "issue:43295", "right": "issue:45701", "accept": false, - "reason": "Processor/tokenizer regression and tokenization-version changes are not the same concrete bug." + "reason": "Processor.tokenizer regression is not the same as version-dependent tokenization output changes." }, { "left": "issue:43646", "right": "issue:46665", "accept": false, - "reason": "Both mention custom model loading, but one is a broad 5.0 initialization breakage and the other is a specific from_pretrained behavior issue; not enough evidence of one bug." + "reason": "Both concern custom code/init behavior, but one is Transformers 5.0 breaking model initialization and the other is from_pretrained behavior inside Python classes." }, { "left": "issue:44186", "right": "issue:44610", "accept": false, - "reason": "LayoutLMv2 tokenizer crashes and OmDet-Turbo image-size mismatch are unrelated." + "reason": "Different processor bugs: LayoutLMv2Tokenizer NER/padding crash vs OmDet-Turbo output-size mismatch." }, { "left": "issue:43404", "right": "issue:45357", "accept": false, - "reason": "lm_head tying/serialization and visual encoder key serialization are different model-specific bugs." + "reason": "Different serialization issues: untied lm_head weights in Mistral3 vs incorrect visual encoder keys in Qwen3.5 save_pretrained." }, { "left": "issue:42915", "right": "issue:43317", "accept": false, - "reason": "FP8 config failure and dequantized model offload failure are both loading issues but not the same path." + "reason": "Qwen3Moe FP8 config failure is unrelated to device_map auto loading failure for a dequantized model with GPU+CPU offload." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "Different vision-model preprocessing errors with different expected sizes and interpolation behavior." + "reason": "Both are image-size related, but they affect different models and different preprocessing/model expectations." }, { "left": "issue:44008", "right": "issue:45823", "accept": false, - "reason": "Gemma3n variable collision and Gemma4 device mismatch are separate model bugs." + "reason": "Gemma3n forward variable collision and Gemma4 PLE device mismatch are distinct bugs." }, { "left": "issue:36010", "right": "issue:42617", "accept": false, - "reason": "ImportError for GenerationMixin and a 3d_parallel execution issue are unrelated." + "reason": "Importing GenerationMixin and running 3d_parallel.py are unrelated failures." }, { "left": "issue:44423", "right": "issue:45406", "accept": false, - "reason": "Continuous-batching multimodal crash and missing _tokenizer on Gemma4Processor are different serve/processor bugs." + "reason": "Both hit serving/multimodal code, but the concrete attribute errors and processors differ." }, { "left": "issue:45632", "right": "issue:45684", "accept": false, - "reason": "Cache-path collision and read-only file permission propagation are unrelated save/load issues." + "reason": "Cache-path collisions for trust_remote_code and read-only permission propagation in save_pretrained are unrelated." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "Qwen3-VL auto device-map crash and GPT-OSS tensor parallel failure are different model/runtime problems." + "reason": "Auto device-map CUDA assert is not the same as tensor-parallel not working for GPT-OSS MoE." }, { "left": "issue:43531", "right": "issue:43572", "accept": false, - "reason": "Qwen3-MoE sliding-window behavior and a StableLmConfig field omission are unrelated." + "reason": "Sliding-window behavior for Qwen3-MoE and missing pad_token_idx in StableLmConfig are unrelated." }, { "left": "issue:28530", "right": "issue:33666", "accept": false, - "reason": "Early stopping metric handling and Qwen2-VL multi-GPU training are not the same issue." + "reason": "Early stopping metric resolution and Qwen2-VL multi-GPU training are different areas." }, { "left": "issue:42915", "right": "issue:46291", "accept": false, - "reason": "Model FP8 loading failure and pipelines import/version constraints are separate problems." + "reason": "FineGrainedFP8Config failure and kernels import/version requirement failure are unrelated." } ] }, @@ -4748,8 +4745,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -4831,7 +4828,7 @@ "issue:45759|issue:45859", "issue:44451|issue:44991" ], - "prepared_review_unit_hash": "982618eba3d98a366a7b6e6e557f20c5bed85260b5e83fd637a4844a9af7052c", + "prepared_review_unit_hash": "38463044b00f4772fe35a428d8dacd750c9faf9f79fe3b6c623752c90cf7ce57", "component_budget": { "node_count": 484, "item_count": 484, @@ -4853,180 +4850,180 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is not a true duplicate cluster: the issues span many unrelated models, trainers, tokenizers, and deployment paths. None of the soft pairs look like the same underlying bug/change, so all are rejected.", - "confidence": 0.9, - "canonical_issue_reason": "Issue 45701 is the broadest, most general item in an otherwise heterogeneous set, and it also has inbound references; it is the least bad placeholder canonical issue.", + "summary": "This cluster is highly heterogeneous: the soft pairs share only broad Transformers topical similarity, not the same underlying bug or change. None of the candidate pairs look safe to merge as duplicates, and there are no PRs to select.", + "confidence": 0.97, + "canonical_issue_reason": "If a representative issue is needed, issue 29159 is the broadest and most generally phrased user-facing regression report, but the set does not actually form one duplicate family.", "canonical_pr_reason": null, - "best_issue_reason": "45701 is the best representative only because its scope is generic enough to stand in for the cluster, but the cluster itself is too mixed for a strong canonical issue.", + "best_issue_reason": "Issue 29159 is the most generic/high-signal problem statement among the set, so it is the best stand-in representative even though it is not a true canonical duplicate.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44671", "right": "issue:44743", "accept": false, - "reason": "Both are model/regression bugs, but they affect different code paths and different models (CamemBERT MLM vs Qwen3 recurrent cache behavior)." + "reason": "Different models and failure modes: CamemBERT masked-LM prediction regression vs Qwen recurrent cache state reset." }, { "left": "issue:36331", "right": "issue:41720", "accept": false, - "reason": "A trainer API signature issue and a Qwen3 device-mapping CUDA assert are unrelated failures." + "reason": "Trainer API signature mismatch vs Qwen3 device-mapping CUDA assert; unrelated root causes." }, { "left": "issue:44186", "right": "issue:44361", "accept": false, - "reason": "Both are tokenizer-related, but they describe different tokenizers, errors, and execution paths; not the same bug." + "reason": "Both tokenizer-related, but one is LayoutLMv2 NER padding/truncation and the other is MLuke task AttributeError." }, { "left": "issue:43526", "right": "issue:43881", "accept": false, - "reason": "BeitImageProcessorFast label reduction and glm-4v-9b loading failure are unrelated." + "reason": "BeitImageProcessorFast label reduction bug vs glm-4v-9b load failure; no shared code-path." }, { "left": "issue:43316", "right": "issue:44913", "accept": false, - "reason": "Both are config/API issues, but they concern different fields and different model families; not the same underlying defect." + "reason": "Both are config/reload issues, but Gemma3TextConfig API discrepancy is unrelated to GPTNeoX rotary_pct persistence." }, { "left": "issue:28530", "right": "issue:30064", "accept": false, - "reason": "Early stopping metric resolution and void segmentation-map processing are different components with no shared failure mode." + "reason": "Early stopping metric resolution bug vs void segmentation-map processing bug; different subsystems." }, { "left": "issue:42915", "right": "issue:43322", "accept": false, - "reason": "Different models and different symptoms: FineGrainedFP8Config failure vs Llava Next segmentation fault." + "reason": "Qwen3Moe FP8 config failure vs Llava Next segmentation fault; different model families and symptoms." }, { "left": "issue:33357", "right": "issue:35707", "accept": false, - "reason": "A MacOS CLIP bus error and progressive generation with inputs_embeds/past_key_values are unrelated." + "reason": "MacOS CLIP bus error vs generation with inputs_embeds/past_key_values; unrelated failures." }, { "left": "issue:43540", "right": "issue:43782", "accept": false, - "reason": "Both involve Qwen3 variants, but one is video-input processing and the other is from_pretrained weight-only loading; not the same concrete bug." + "reason": "Different Qwen multimodal issues: video input processing error vs weight_only from_pretrained loading error." }, { "left": "issue:30064", "right": "issue:34634", "accept": false, - "reason": "Image segmentation-map handling and Bark voice_preset behavior are unrelated." + "reason": "Void segmentation maps vs BarkProcessor voice_preset handling; unrelated image vs audio processor bugs." }, { "left": "issue:43873", "right": "issue:45704", "accept": false, - "reason": "Quantization/offloading behavior and a T5 RMSNorm memory leak are different bugs in different subsystems." + "reason": "Quantization/offloading behavior vs T5 RMSNorm memory leak; not the same bug." }, { "left": "issue:43550", "right": "issue:43653", "accept": false, - "reason": "torch.compile with SDPA on Bamba and BigBirdTokenizer special-token registration are unrelated." + "reason": "torch.compile SDPA failure vs tokenizer special-token registration bug; completely different paths." }, { "left": "issue:44387", "right": "issue:45701", "accept": false, - "reason": "Both mention version regressions, but one is int4 memory/OOM and the other is tokenization behavior; too broad to be the same bug." + "reason": "Int4 quantization memory growth vs tokenization changes across versions; different concerns." }, { "left": "issue:29159", "right": "issue:36032", "accept": false, - "reason": "Slow-vs-fast tokenizer inconsistency and T5 tokenizer load failure are different tokenizer issues." + "reason": "Slow/fast tokenizer inconsistency vs T5 add_special_tokens name collision; both tokenizer issues but distinct root causes." }, { "left": "issue:43606", "right": "issue:43720", "accept": false, - "reason": "Both touch accelerate/offload loading, but one is a device mismatch and the other is packed-weight unpacking; not the same concrete path." + "reason": "CPU offload device mismatch vs accelerate loading of packed BitNet weights; related area but different bugs." }, { "left": "issue:44492", "right": "issue:44509", "accept": false, - "reason": "A documentation typo and stale docs about removed pipelines are not duplicates." + "reason": "Cache strategy typo vs docs referencing removed pipeline tasks; documentation-only and unrelated." }, { "left": "issue:44534", "right": "issue:46665", "accept": false, - "reason": "Non-persistent buffer initialization and from_pretrained behavior inside Python classes are different loading/runtime problems." + "reason": "Non-persistent buffer junk in v5 vs from_pretrained behavior inside Python classes; different code paths." }, { "left": "issue:43531", "right": "issue:45759", "accept": false, - "reason": "Both are Qwen3-related, but one is sliding_window behavior and the other is config unwrapping; distinct code paths." + "reason": "Qwen3-MoE sliding_window issue vs composite Qwen multimodal config unwrap bug; same family name only." }, { "left": "issue:41720", "right": "issue:43746", "accept": false, - "reason": "Qwen3 device-map CUDA assert and GraniteSpeech PEFT local-checkpoint loading are unrelated." + "reason": "Qwen3 auto device mapping CUDA assert vs GraniteSpeech PEFT local checkpoint loading; unrelated." }, { "left": "issue:42490", "right": "issue:43329", "accept": false, - "reason": "A trailing-slash import warning and an undefined-video-token-count bug are not the same underlying issue." + "reason": "Trailing-slash local model import warning vs multimodal token counting video-branch bug; different behaviors." }, { "left": "issue:43606", "right": "issue:43653", "accept": false, - "reason": "CPU offload device mismatch and BigBird special-token registration are unrelated." + "reason": "CPU offload mismatch vs BigBirdTokenizer mask-token registration; unrelated." }, { "left": "issue:43388", "right": "issue:45059", "accept": false, - "reason": "Batch metric gathering and SAM3 text/bounding-box behavior are different multimodal/data-handling problems." + "reason": "Metric gathering drops tuple labels vs SAM3 text/bounding-box weirdness; different training vs inference problems." }, { "left": "issue:45406", "right": "issue:45464", "accept": false, - "reason": "Different serving APIs, different models, and different failures; only superficially similar as runtime crashes." + "reason": "Gemma4Processor missing _tokenizer in serve vs Qwen3.5 streaming chat/completions failure; different models and layers." }, { "left": "issue:34567", "right": "issue:35707", "accept": false, - "reason": "Trainer input-token accounting and generation with inputs_embeds/past_key_values are unrelated." + "reason": "TrainerState token counter update bug vs generation with inputs_embeds/past_key_values; unrelated." }, { - "left": "issue:34634", - "right": "issue:41720", + "left": "issue:44991", + "right": "issue:44589", "accept": false, - "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping are completely separate issues." + "reason": "Tokenizer load failure for a specific model vs missing Float8 storage type; different failure classes." }, { "left": "issue:43531", "right": "issue:44589", "accept": false, - "reason": "Qwen3 sliding-window behavior and missing Float8 storage lookup are unrelated." + "reason": "Qwen3-MoE sliding-window behavior vs Float8 storage lookup error; no overlap beyond general Transformers scope." }, { "left": "issue:45759", "right": "issue:45859", "accept": false, - "reason": "Both are Qwen3.5-family config/runtime issues, but one is from_config text_config unwrapping and the other is CPU-offload incompatibility in rotary embedding; not mergeable as one bug." + "reason": "Composite Qwen config unwrap issue vs Qwen3_5MoeTextRotaryEmbedding CPU offload incompatibility; same ecosystem, different bug." }, { "left": "issue:44451", "right": "issue:44991", "accept": false, - "reason": "Two model-loading failures for different checkpoints and different tokenizers; no shared underlying defect." + "reason": "Model-specific loading regression for ScandiBERT vs tokenizer loading regression for est-roberta; unrelated root causes." } ] }, @@ -5045,8 +5042,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -5129,7 +5126,7 @@ "issue:43782|issue:45381", "issue:43421|issue:44186" ], - "prepared_review_unit_hash": "b37adac8864483d8038c3c646ae162dc33cf349440eb95d97e656b404b3daa9b", + "prepared_review_unit_hash": "3afebb1e27fe60d3aa947a116b68eab012472862e784904206d705e58fc477f5", "component_budget": { "node_count": 484, "item_count": 484, @@ -5151,192 +5148,192 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most pairs are only thematically similar (tokenizer/model-loading/config bugs) rather than true duplicates. The only clear duplicate in this set is the Qwen2_5_VLProcessor batched padding=False crash (44514/44545).", - "confidence": 0.89, - "canonical_issue_reason": "issue:44545 is the cleanest canonical for the one true duplicate pair: it reports the same `Qwen2_5_VLProcessor.apply_chat_template` batched `padding=False` crash as 44514, with slightly stronger supporting signals.", + "summary": "Most pairs are thematically related but not true duplicates; only the two Qwen2_5_VLProcessor issues appear to describe the same crash on batched `apply_chat_template` with `padding=False`.", + "confidence": 0.94, + "canonical_issue_reason": "Issue 44514 is the best canonical issue: it is the earlier, more detailed report of the exact batched-input `apply_chat_template` crash with `padding=False`.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43421 is the best overall representative issue here: it is a concrete tokenizer backend bug about special-token/post-processor state, and it sits closest to the cluster's recurring tokenizer-special-token theme.", + "best_issue_reason": "Issue 44514 is the strongest issue to keep as the reference because it has the clearer original report and fuller reproduction context for the shared bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45357", "right": "issue:46402", "accept": false, - "reason": "Different bugs: one is Qwen3.5 visual-encoder key saving, the other is Sam3 text-encoder weights not loading." + "reason": "Different models and different failures: incorrect visual encoder keys vs missing text encoder weights." }, { "left": "issue:44077", "right": "issue:44625", "accept": false, - "reason": "Unrelated failures: patchtsmixer post_init policy vs Qwen3.5 num_labels propagation." + "reason": "Unrelated problems: optional `post_init` policy vs `num_labels` propagation." }, { "left": "issue:44625", "right": "issue:45479", "accept": false, - "reason": "Different model/config issues: num_labels propagation vs zero-loss classification behavior." + "reason": "Both involve classification config, but one is missing propagation and the other is a zero-loss regression across models; not the same bug." }, { "left": "issue:43644", "right": "issue:46665", "accept": false, - "reason": "Different code paths: junk non-persistent buffers vs inconsistent from_pretrained-in-class behavior." + "reason": "Different subsystems and symptoms: buffer initialization vs `from_pretrained` behavior in Python classes." }, { "left": "issue:44898", "right": "issue:45072", "accept": false, - "reason": "Both are vision/model bugs, but the concrete failures differ: Perceiver interpolation vs bfloat16 dtype mismatch." + "reason": "Different model families and failure modes: image resolution handling vs dtype mismatch in inference." }, { "left": "issue:44062", "right": "issue:44451", "accept": false, - "reason": "Different causes: AddedToken special-kwarg conflict vs model loading failure for ScandiBERT." + "reason": "The first is a tokenizer constructor error; the second is a specific model load failure. Too broad to treat as the same bug." }, { "left": "issue:44987", "right": "issue:45042", "accept": false, - "reason": "Same broad area but different bug: one is loading a specific model, the other is PIL image processors incorrectly requiring torchvision." + "reason": "Unrelated: model loading failure for one package vs PIL processor dependency regression." }, { "left": "issue:43381", "right": "issue:43844", "accept": false, - "reason": "Both involve gradients/training, but one is checkpointing in eval mode and the other is a ZeRO-3 gradient regression." + "reason": "Different issues: eval-mode gradient checkpointing guard vs anomalous gradient growth under ZeRO-3." }, { "left": "issue:44368", "right": "issue:45357", "accept": false, - "reason": "Different issues: a config warning about tied embeddings vs incorrect visual-encoder save keys." + "reason": "A warning about `tie_word_embeddings` is unrelated to incorrect visual encoder keys during save/load." }, { "left": "issue:39692", "right": "issue:44610", "accept": false, - "reason": "Different bugs: SigLIP2 doc/example problems vs OmDet-Turbo processor/model input-size mismatch." + "reason": "Different docs/model mismatch issue vs processor image-size mismatch in OmDet-Turbo." }, { "left": "issue:43531", "right": "issue:43645", "accept": false, - "reason": "Different failures: Qwen3-MoE sliding_window handling vs custom model initialization in Jupyter." + "reason": "No shared underlying bug: sliding-window behavior in Qwen3-MoE vs notebook custom-model initialization." }, { "left": "issue:41720", "right": "issue:45704", "accept": false, - "reason": "No shared underlying bug: Qwen3 device-mapping cuda assert vs T5 apex RMSNorm memory leak." + "reason": "CUDA assert on Qwen3 device mapping is unrelated to T5 Apex RMSNorm memory leak." }, { "left": "issue:43746", "right": "issue:46050", "accept": false, - "reason": "Both concern load/save behavior, but one is GraniteSpeech PEFT local checkpoints and the other is sinq quantization persistence." + "reason": "Both involve checkpoint loading, but one is PEFT adapters and the other is sinq quantization; different code paths." }, { "left": "issue:45704", "right": "issue:45923", "accept": false, - "reason": "Different topics: T5/apex memory leak vs missing supports_gradient_checkpointing on a remote-code model." + "reason": "Completely different concerns: memory leak in T5 RMSNorm vs missing gradient-checkpointing capability flag." }, { "left": "issue:46655", "right": "issue:46693", "accept": false, - "reason": "Both are performance/processor-related, but one is video padding logic and the other is FA2 long-context generation regression." + "reason": "Video processor padding bug is unrelated to FA2 long-context performance regression." }, { "left": "issue:43381", "right": "issue:45704", "accept": false, - "reason": "Different bugs entirely: gradient checkpointing eval-mode restriction vs T5 apex memory leak." + "reason": "No substantive overlap; one is eval-mode checkpointing, the other is a T5 normalization leak." }, { "left": "issue:44918", "right": "issue:44936", "accept": false, - "reason": "Different trainer failures: unpacking Qwen3.5 embeddings in TRL SFT vs evaluate() after train()." + "reason": "Different trainer failures: embedding unpacking with TRL SFT vs post-train evaluate() failure." }, { "left": "issue:43421", "right": "issue:43927", "accept": false, - "reason": "Both involve tokenizer state, but one is runtime post-processor updates and the other is save/load losing custom token IDs." + "reason": "One is runtime special-token post-processing, the other is save/load loss of custom token IDs; related area but not same bug." }, { "left": "issue:42886", "right": "issue:44843", "accept": false, - "reason": "Both mention HF_HUB_OFFLINE, but the concrete reports are not clearly the same bug without stronger evidence." + "reason": "Both mention offline mode, but one is cache loading and the other is an unconditional hub metadata call in a specific regex patch." }, { "left": "issue:45759", "right": "issue:45923", "accept": false, - "reason": "Different composite-model issues: unwrapping text_config in from_config vs missing gradient-checkpointing support." + "reason": "Different bugs: unwrapping `text_config` in composite multimodal configs vs a trust_remote_code capability flag." }, { "left": "issue:44514", "right": "issue:44545", "accept": true, - "reason": "Exact duplicate: same `Qwen2_5_VLProcessor.apply_chat_template` batched `padding=False` crash with the same model/API path." + "reason": "Same crash, same API, same condition (`apply_chat_template` on batched input with `padding=False`); these are duplicates." }, { "left": "issue:46032", "right": "issue:46463", "accept": false, - "reason": "Different failures: Mamba2Mixer use_cache correctness vs a T5 decoder-input validation error message." + "reason": "Mamba cache correctness vs T5 decoder-input error messaging are unrelated." }, { "left": "issue:39692", "right": "issue:43873", "accept": false, - "reason": "Unrelated: SigLIP2 documentation/example errors vs quantization/offloading behavior." + "reason": "Documentation/model mismatch is not the same as quantization/offloading behavior." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "Different underlying problems: PEFT local checkpoint loading vs quantization offloading not behaving as expected." + "reason": "Checkpoint loading with PEFT adapters is unrelated to quantization offloading failures." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "Different model-specific regressions: SigLIP2 pipeline outputs vs CamemBERT masked LM predictions." + "reason": "Different models and symptoms: SigLIP2 output quality vs CamemBERT masked-LM predictions." }, { "left": "issue:43421", "right": "issue:43653", "accept": false, - "reason": "Related tokenizer area, but different bugs: runtime special-token updates vs BigBird mask token registration/decode output." + "reason": "Related to special tokens, but one is a framework feature gap and the other is a tokenizer-specific registration bug." }, { "left": "issue:30064", "right": "issue:36010", "accept": false, - "reason": "Completely different subsystems: image segmentation map processing vs an import error for GenerationMixin." + "reason": "Completely unrelated: image processor segmentation-map handling vs import error for `GenerationMixin`." }, { "left": "issue:43782", "right": "issue:45381", "accept": false, - "reason": "Different Qwen VL bugs: `weight_only=True` load error vs wrong vision_position_ids for video input." + "reason": "Both are multimodal Qwen issues, but one is `weight_only` loading and the other is video `vision_position_ids`; different failures." }, { "left": "issue:43421", "right": "issue:44186", "accept": false, - "reason": "Both tokenizer-related, but not the same defect: post-processor updates vs LayoutLMv2 NER/padding/truncation crashes." + "reason": "Different tokenizer issues with different triggers; not the same code-path bug." } ] }, "evaluator_result": { "accept": true, - "feedback": "The analyst summary is grounded in the packet and stays conservative on the soft edges. The only accepted duplicate pair, issue:44514/issue:44545, is clearly the same Qwen2_5_VLProcessor.apply_chat_template padding=False crash and is mergeable into one fix. The representative/canonical issue choices are also supported by the titles and linkage signals, and the rejected pairs do not overclaim equivalence." + "feedback": "The summary is grounded in the packet: 44514 is earlier and more detailed than 44545, and the duplicate call is conservative. The soft-edge rejections are generally cautious and align with the issue titles; I don\u2019t see overstatement that would require rejection." }, "evaluator_used": true, "retried": false, @@ -8106,8 +8103,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -8189,7 +8186,7 @@ "issue:41720|issue:46710", "issue:45406|issue:46121" ], - "prepared_review_unit_hash": "3e00e66f9392478f2ee32d45080d7f8bfadf40fd545be499951924aa4759fa2f", + "prepared_review_unit_hash": "48a32b1c77cd6a521bbe0c462ab21b5f6c09ce48cbfd243f2b7bc7c1de44b101", "component_budget": { "node_count": 484, "item_count": 484, @@ -8211,174 +8208,174 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is highly heterogeneous: the soft-similarity pairs mostly connect unrelated bugs across tokenization, serving, quantization, model-loading, and trainer code paths. I would not merge any of the proposed pairs as duplicates.", - "confidence": 0.96, - "canonical_issue_reason": "No single canonical issue stands out; the items span many unrelated failures, so a cluster-level canonical issue would be misleading.", - "canonical_pr_reason": "No pull requests were provided in this cluster.", - "best_issue_reason": "No global best issue: even the most central-looking reports are narrow, subsystem-specific bugs and do not represent the rest of the cluster.", - "best_pr_reason": "No pull requests were provided in this cluster.", + "summary": "This cluster is highly heterogeneous: tokenizer quirks, model-loading issues, serving crashes, quantization/offloading bugs, and training/runtime errors. None of the suggested soft pairs look like the same underlying bug or change.", + "confidence": 0.97, + "canonical_issue_reason": "No strong canonical issue; the listed issues span unrelated bugs, so any single representative would be arbitrary.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45229 is the most discussed and concrete report, with a clear reproduction and impact, so it is the best representative if one must be picked.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43572", "right": "issue:44291", "accept": false, - "reason": "Both are transformers v5 compatibility regressions, but one is a config field missing in StableLM and the other is an init_empty_weights argument error; different code paths and fixes." + "reason": "Config field regression vs an unexpected init_empty_weights argument; different code paths and symptoms." }, { "left": "issue:45245", "right": "issue:46710", "accept": false, - "reason": "An embedding/category cardinality error is unrelated to the DeepSeek accuracy regression after version upgrade." + "reason": "Category-cardinality runtime error vs an accuracy regression after upgrade; not the same bug." }, { "left": "issue:44493", "right": "issue:46285", "accept": false, - "reason": "Unexpected position-id keys vs CLIP text model structure drift are different bugs in different model families." + "reason": "Position-id key handling vs CLIP text model structure drift; unrelated model internals." }, { "left": "issue:43504", "right": "issue:43653", "accept": false, - "reason": "Legacy-field loading for BEiT semantic segmentation is unrelated to BigBirdTokenizer special-token registration." + "reason": "Legacy field loading failure in BEiT vs BigBird tokenizer special-token registration; different components." }, { "left": "issue:44464", "right": "issue:46424", "accept": false, - "reason": "Both mention generation/cache behavior, but one is compiled forward output inconsistency and the other is a cache-length config being ignored only for quantized cache; not the same bug." + "reason": "Compiled-forward generation mismatch vs cache-length config semantics; not one concrete defect." }, { "left": "issue:43937", "right": "issue:46701", "accept": false, - "reason": "A GenerationConfig validation failure is unrelated to the DiffusionGemma image-text-to-text pipeline failure." + "reason": "Invalid GenerationConfig vs a broken image-text pipeline; different failure surfaces." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency is not the same as Whisper empty-transcription failure after token alignment." + "reason": "Chunked generation with compiled forward is unrelated to Whisper empty-transcription decode failure." }, { "left": "issue:45200", "right": "issue:45229", "accept": false, - "reason": "Missing mm_token_type_ids defaults in text-only fine-tuning is unrelated to a multi-GPU OOM in Gemma4 inference." + "reason": "Missing mm_token_type_ids defaults vs multi-GPU CUDA OOM; separate Gemma4 issues." }, { "left": "issue:45229", "right": "issue:45823", "accept": false, - "reason": "Gemma4 OOM and Gemma4 PLE device mismatch are both in the same model family but describe different failure modes and code paths." + "reason": "OOM during inference vs device-mismatch during forward; not the same root cause." }, { "left": "issue:44869", "right": "issue:46752", "accept": false, - "reason": "Whisper timestamp decode crash on trailing replacement character is unrelated to apply_chat_template crashing on an empty message list." + "reason": "Whisper timestamp decode crash vs chat-template empty-list IndexError; different API paths." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "ZeRO-3 gradient growth during training is unrelated to a model load failure for Phi-4 multimodal." + "reason": "ZeRO-3 gradient anomaly vs model loading failure for Phi-4 multimodal; unrelated." }, { "left": "issue:43873", "right": "issue:46050", "accept": false, - "reason": "Quantization/offloading behavior issues are too different here: one is offloading not working as expected, the other is sinq save/load breakage." + "reason": "Offloading/quantization behavior vs sinq save/load breakage; same broad area but distinct bugs." }, { "left": "issue:29942", "right": "issue:30333", "accept": false, - "reason": "Failing Flash Attention 2 tests and MLFlow job-state reporting are completely unrelated subsystems." + "reason": "Flash Attention test failures vs MLflow reporting on job failure; no shared underlying defect." }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 nonsensical outputs are unrelated to the unexpected position-id key regression." + "reason": "SigLIP2 output nonsense vs unexpected position-id key handling; different problems." }, { "left": "issue:45588", "right": "issue:46050", "accept": false, - "reason": "A flash-attention integration crash on s_aux=None is not the same underlying bug as sinq quantized model save/load breakage." + "reason": "Flash-attention sink-less model crash vs quantized-model persistence bug; unrelated." }, { "left": "issue:29159", "right": "issue:32090", "accept": false, - "reason": "Tokenizer slow/fast inconsistency is unrelated to the Trainer broadcast TypeError." + "reason": "Tokenizer inconsistency vs Trainer broadcast NoneType error; different layers entirely." }, { "left": "issue:42490", "right": "issue:43696", "accept": false, - "reason": "Import warnings from trailing-slash paths have nothing to do with GPT-oss CUDA OOM." + "reason": "Path trailing-slash import warning vs CUDA OOM; no duplicate signal." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "A compile-time template-node error and a Python 3.13 flash_attn import KeyError are different failures." + "reason": "Template-node compilation error vs flash_attn import KeyError on Python 3.13; unrelated." }, { "left": "issue:44977", "right": "issue:46701", "accept": false, - "reason": "Qwen3.5 flash-attention generation issues are not the same as the DiffusionGemma pipeline failure." + "reason": "Qwen3.5 flash-attention generation issue vs image-text pipeline failure; different models and code paths." }, { "left": "issue:28218", "right": "issue:34567", "accept": false, - "reason": "Extra tokenizer spacing after added tokens is unrelated to TrainerState token-count tracking." + "reason": "Tokenizer spacing bug vs TrainerState token counter not updating; not the same bug." }, { "left": "issue:43065", "right": "issue:46369", "accept": false, - "reason": "A dummy Conv2d in Sam3PixelDecoder is a code-structure issue, not the same as a memory overflow when calling sam3-litetext." + "reason": "Dummy Conv2d in Sam3PixelDecoder vs memory overflow on sam3-litetext; different SAM3 issues." }, { "left": "issue:44554", "right": "issue:45446", "accept": false, - "reason": "MPS attention correctness with mismatched head dims is unrelated to an incorrect PyTorch version check in flex_attention import logic." + "reason": "MPS attention correctness issue vs incorrect PyTorch version check for flex_attention import; unrelated." }, { "left": "issue:15354", "right": "issue:38175", "accept": false, - "reason": "TorchScript GeneratorExp export failure is unrelated to SigLIP2 zero-probability output behavior." + "reason": "TorchScript GeneratorExp export failure vs SigLIP2 zero-probability output; different functionality." }, { "left": "issue:44423", "right": "issue:44734", "accept": false, - "reason": "Both are serve crashes, but one is multimodal continuous-batching string handling and the other is KV-cache continuation tensor indexing; not the same code-path problem." + "reason": "Serve crash from multimodal input handling vs KV-cache continuation indexing bug; both in serve, but distinct bugs." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "SigLIP2 implementation discrepancy and resume_from_checkpoint key mismatch are unrelated." + "reason": "SigLIP2 HF/JAX discrepancy vs resume_from_checkpoint key mismatch; unrelated training/model issues." }, { "left": "issue:41720", "right": "issue:46710", "accept": false, - "reason": "A device-mapping cudaErrorAssert on A800 is not the same bug as an accuracy regression after upgrading Transformers." + "reason": "Device-mapping CUDA assert on A800 vs an accuracy regression after upgrade; not the same defect." }, { "left": "issue:45406", "right": "issue:46121", "accept": false, - "reason": "Gemma4Processor missing _tokenizer in serve is unrelated to rope-parameter JSON list type handling." + "reason": "Gemma4 processor missing _tokenizer vs rope-parameter conversion type error; unrelated." } ] }, @@ -8397,8 +8394,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -8483,7 +8480,7 @@ "issue:28218|issue:35141", "issue:42915|issue:43824" ], - "prepared_review_unit_hash": "2bfe277a3719762074468a58710baf755c856ffc522339f89137d1288dd667bc", + "prepared_review_unit_hash": "26a22a119cc76027a63c2acfbb1d765214d432029b55502c470c878ac87c4b2f", "component_budget": { "node_count": 484, "item_count": 484, @@ -8505,192 +8502,192 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is not a single duplicate cluster; it\u2019s a broad set of unrelated Transformers issues. The soft links mostly share a subsystem or buzzwords (tokenizers, generation, flash attention, multimodal, quantization) but describe different bugs or regressions, so they should be rejected.", - "confidence": 0.98, - "canonical_issue_reason": null, + "summary": "The items are a loose mix of unrelated Transformers issues. None of the soft pairs look like the same concrete bug/change, so I would not merge any of them. If forced to pick a single issue as the most representative umbrella report, the multimodal processor/tokenizer regression in #43295 is the closest fit.", + "confidence": 0.39, + "canonical_issue_reason": "#43295 is the most central of the weakly related reports: it describes a broad multimodal processor/tokenizer regression and has the most soft links to other processor-facing issues.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "#43295 is the best umbrella issue among this set because it is a reproducible user-facing regression with the widest apparent overlap to other multimodal processor/tokenizer complaints.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43404", "right": "issue:46710", "accept": false, - "reason": "Different problems: tied lm_head weights in Mistral3 multimodal generation vs accuracy regression in DeepSeek-R1 after upgrade." + "reason": "Different bugs: one is tied-weight handling in a specific multimodal generation model, the other is an accuracy regression in DeepSeek after an upgrade." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "Different models and failure modes: dtype mismatches in inference vs Wav2Vec2 save/tokenization failure." + "reason": "Bfloat16 dtype mismatches during inference are unrelated to save_pretrained/tokenization failures." }, { "left": "issue:37428", "right": "issue:43957", "accept": false, - "reason": "Both are load/import related, but one is a missing flash-attention helper import and the other is a meta-device loading breakage across models." + "reason": "An import error for a flash-attention helper is not the same as a meta-device model-loading regression." }, { "left": "issue:45588", "right": "issue:45759", "accept": false, - "reason": "Different code paths: flash_attention integration crashes on sink-less models vs AutoModelForCausalLM config unwrapping for multimodal Qwen configs." + "reason": "Flash-attention sink handling and AutoModelForCausalLM text_config unwrapping are separate code paths and separate bugs." }, { "left": "issue:43295", "right": "issue:43450", "accept": false, - "reason": "Different regressions: processor.tokenizer/image handling vs video processor batch shape output." + "reason": "Both are multimodal, but one is a processor/tokenizer regression and the other is a batched video-shape bug." }, { "left": "issue:43540", "right": "issue:45200", "accept": false, - "reason": "Different failures: Qwen3OmniMoe video processing ValueError vs Gemma4 mm_token_type_ids defaults for text-only fine-tuning." + "reason": "Qwen3OmniMoe video processing failure and Gemma4 mm_token_type_ids defaults are different model-specific issues." }, { "left": "issue:44186", "right": "issue:46192", "accept": false, - "reason": "LayoutLMv2Tokenizer NER/batching crash is unrelated to return_tensors being ignored when text_kwargs is passed." + "reason": "LayoutLMv2 tokenizer crashes on NER/padding are unrelated to return_tensors being ignored when text_kwargs is passed." }, { "left": "issue:45200", "right": "issue:45823", "accept": false, - "reason": "Different multimodal bugs: missing mm_token_type_ids default vs PLE device mismatch during forward." + "reason": "One is about missing/default mm_token_type_ids, the other is a device placement mismatch in Gemma4 forward." }, { "left": "issue:43295", "right": "issue:43873", "accept": false, - "reason": "Different issues: processor.tokenizer regression vs quantization/offloading not working as expected." + "reason": "Processor/tokenizer regression and quantization offloading behavior are not the same underlying defect." }, { "left": "issue:45823", "right": "issue:46192", "accept": false, - "reason": "Device placement bug in Gemma4 PLE is unrelated to return_tensors/text_kwargs behavior." + "reason": "A Gemma4 device_map bug does not match a text_kwargs/return_tensors precedence issue." }, { "left": "issue:43295", "right": "issue:44610", "accept": false, - "reason": "processor.tokenizer/image input regression is unrelated to OmDet-Turbo processor/model resolution mismatch." + "reason": "Custom model code breaking on processor.tokenizer/images is different from OmDet-Turbo's 640px vs 224px preprocessing mismatch." }, { "left": "issue:29870", "right": "issue:29942", "accept": false, - "reason": "Both are generation/test failures, but one is a BART decoder_attention_mask bug and the other is Flash Attention 2 test failures." + "reason": "Bart generate decoder_attention_mask initialization and Flash Attention 2 test failures are distinct problems." }, { "left": "issue:45520", "right": "issue:46178", "accept": false, - "reason": "Import backend errors in different modules: flash_attn KeyError on Python 3.13 vs tensorflow_text backend mapping error." + "reason": "Python 3.13 flash_attn import metadata and tensorflow_text backend mapping are unrelated import-system errors." }, { "left": "issue:43572", "right": "issue:44987", "accept": false, - "reason": "Config default missing pad_token_idx in StableLm is unrelated to loading physical-intelligence/fast on transformers>=5.1.0." + "reason": "Missing pad_token_idx in StableLmConfig is not the same as failing to load a specific model checkpoint." }, { "left": "issue:43450", "right": "issue:43575", "accept": false, - "reason": "Video processor batch-shape bug is unrelated to tensor-parallel OOM when loading Qwen2-57B-A14B-Instruct." + "reason": "A batched video processor shape bug is unrelated to a tensor-parallel OOM when loading Qwen2-57B." }, { "left": "issue:43550", "right": "issue:43844", "accept": false, - "reason": "torch.compile SDPA failure in Bamba is unrelated to gradient explosion with HfDeepSpeedConfig + ZeRO-3." + "reason": "torch.compile SDPA failure and ZeRO-3 gradient growth are different runtime/training issues." }, { "left": "issue:43232", "right": "issue:44849", "accept": false, - "reason": "Different generation/model bugs: sync_gpus model_kwargs update vs Qwen3.5 output_hidden_states behavior." + "reason": "Generation kwargs handling after sync_gpus is unrelated to an output_hidden_states bug in Qwen3.5." }, { "left": "issue:43572", "right": "issue:43957", "accept": false, - "reason": "StableLm config default omission is unrelated to meta-device loading regressions across some models." + "reason": "StableLmConfig missing pad_token_idx and meta-device loading regressions are not the same bug." }, { "left": "issue:37428", "right": "issue:43824", "accept": false, - "reason": "Different import issues: missing flash_attention helper vs missing Qwen2.5-VL class export." + "reason": "These are both import errors, but they involve different missing symbols and different models." }, { "left": "issue:44554", "right": "issue:45356", "accept": false, - "reason": "MPS attention correctness when head dims differ is unrelated to Kimi-K2.5 tokenizer codec/regex regression." + "reason": "An MPS attention correctness issue is unrelated to Kimi-K2.5 tokenizer codec/warning regressions." }, { "left": "issue:42490", "right": "issue:43493", "accept": false, - "reason": "Local-path trailing slash import warning is unrelated to SigLIP2 HF vs JAX implementation discrepancy." + "reason": "A local-path import warning and a SigLIP2 HF-vs-JAX discrepancy are different issues." }, { "left": "issue:38175", "right": "issue:43493", "accept": false, - "reason": "Unexpected zero probabilities in SigLIP2 and SigLIP2 implementation mismatch are related in theme, but not the same concrete bug." + "reason": "Unexpected zero probabilities in one SigLIP2 model are not the same as a broader implementation discrepancy report." }, { "left": "issue:39692", "right": "issue:46192", "accept": false, - "reason": "SigLIP2 docs example errors are documentation/example problems, not the return_tensors/text_kwargs API behavior." + "reason": "Documentation example errors for SigLIP2 are unrelated to return_tensors being ignored with text_kwargs." }, { "left": "issue:43653", "right": "issue:44112", "accept": false, - "reason": "Both are CI/test failures, but one is BigBirdTokenizer special-token registration and the other is a stale device override test in GraniteSpeech." + "reason": "BigBirdTokenizer special-token registration and GraniteSpeech CI device-override failure are unrelated." }, { "left": "issue:44466", "right": "issue:45127", "accept": false, - "reason": "Both involve tied weights serialization, but one is inconsistent lm_head.weight serialization by device and the other is LoRA merge collapse with extended vocabulary." + "reason": "Both touch tied weights, but one is serialization inconsistency and the other is LoRA merge collapse with extended vocab." }, { "left": "issue:43278", "right": "issue:43856", "accept": false, - "reason": "Different training/eval issues: BF16-to-FP32 embedding dtype change vs Qwen3 MoE memory inefficiency." + "reason": "Embedding dtype drift during eval and Qwen3 MoE memory inefficiency are different failures." }, { "left": "issue:45381", "right": "issue:46209", "accept": false, - "reason": "Qwen2.5-VL video vision_position_ids bug is unrelated to FP8 inference failure on B200." + "reason": "Qwen2.5-VL video position ids and FP8 inference failure on B200 are unrelated." }, { "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "Both are apply_chat_template issues, but one is all-zero assistant masks for multimodal inputs and the other is a crash on tool-call assistant messages with no content." + "reason": "All-zero assistant masks for multimodal inputs and tool-call tokenize crashes are different apply_chat_template bugs." }, { "left": "issue:28218", "right": "issue:35141", "accept": false, - "reason": "Tokenizer added-space bug and output-embedding reinitialization on resize are different behaviors in different code paths." + "reason": "Tokenizer extra spacing after added tokens is unrelated to embedding reinitialization after resizing." }, { "left": "issue:42915", "right": "issue:43824", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure is unrelated to missing Qwen2.5-VL import/export." + "reason": "A FineGrainedFP8Config failure in Qwen3Moe is unrelated to the Qwen2_5_VL import error." } ] }, @@ -9613,8 +9610,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -9698,7 +9695,7 @@ "issue:29942|issue:30064", "issue:43756|issue:45440" ], - "prepared_review_unit_hash": "c89bd0843746345959638a139ddb82970cfbbe3742ca8b67325f9e445f5cc1ee", + "prepared_review_unit_hash": "7bbff46ca77ff682346e4c3064f526a8bfb893993a43b44c31b714bfb20a3f12", "component_budget": { "node_count": 484, "item_count": 484, @@ -9720,186 +9717,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft edge looks like a true duplicate: the candidates mostly share broad themes (tokenizers, generation, quantization, Gemma/Qwen, etc.) but describe different concrete failure modes and code paths. I\u2019d anchor the cluster on issue 44561 as the most broadly impactful, well-scoped v5 regression report.", - "confidence": 0.86, - "canonical_issue_reason": "Issue 44561 is the best representative anchor: it\u2019s a concrete, cross-cutting v5 compatibility break with downstream impact and clear symptom/reporting, whereas the other issues are narrower or more fragmented.", + "summary": "The cluster is a loose semantic grouping of unrelated issues, mostly spanning model loading, tokenizer/processor bugs, generation regressions, and memory/offloading problems. None of the soft pairs look like true duplicates or the same concrete code-path change.", + "confidence": 0.91, + "canonical_issue_reason": "Issue #43873 is the broadest and most central representative of the loose memory/device/quantization subset, with active discussion and inbound references; the rest are narrower, model-specific regressions or unrelated bugs.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44561 is the strongest overall issue candidate because it captures a clear release regression affecting trust_remote_code models and is easier to use as a cluster representative than the more model-specific reports.", + "best_issue_reason": "Issue #43873 best represents the cluster because it is a prominent, ongoing infrastructure-level bug around offloading/quantization, whereas most other items are isolated model- or feature-specific failures.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44561", "right": "issue:44779", "accept": false, - "reason": "Both are v5 regressions, but one is an API removal breaking trust_remote_code models and the other is a tokenizer correctness issue for DeepSeek; different bug classes and fix paths." + "reason": "Different failures: a removed utility breaking trust_remote_code models vs a DeepSeek tokenizer regression." }, { "left": "issue:43873", "right": "issue:45229", "accept": false, - "reason": "Quantization/offloading behavior is unrelated to a Gemma4 multi-GPU CUDA OOM; different model, symptom, and likely root cause." + "reason": "Both are memory-related, but one is quantized offloading behavior and the other is a Gemma4 multi-GPU OOM; not the same bug." }, { "left": "issue:25251", "right": "issue:32090", "accept": false, - "reason": "Pipeline top_k nesting behavior and Trainer GPU broadcast NoneType error are unrelated subsystems and failures." + "reason": "Pipeline top_k nesting and Trainer _gpu_broadcast_one TypeError are unrelated code paths." }, { "left": "issue:44448", "right": "issue:44466", "accept": false, - "reason": "One concerns generation output differences for Pegasus, the other serialization of lm_head.weight; both are v5-related but not the same concrete bug." + "reason": "One is a generation-output regression, the other is device-dependent serialization of lm_head.weight." }, { "left": "issue:42490", "right": "issue:43873", "accept": false, - "reason": "Local-path import warnings have nothing to do with quantization offloading behavior." + "reason": "Local-path import warning is unrelated to quantization/offloading behavior." }, { "left": "issue:43976", "right": "issue:45440", "accept": false, - "reason": "Python version compatibility and DeepSeekV3MoE implementation divergence are unrelated issues." + "reason": "Python-version compatibility break and DeepSeekV3MoE divergence are unrelated." }, { "left": "issue:43278", "right": "issue:43575", "accept": false, - "reason": "Training/eval dtype drift is a different problem from tensor-parallel OOM while loading a large Qwen2 model." + "reason": "Embedding dtype drift in eval is not the same as tp-induced OOM on a large Qwen model." }, { "left": "issue:45588", "right": "issue:45823", "accept": false, - "reason": "Flash-attention crashing on s_aux=None is a different code path from Gemma4 PLE device placement mismatch in forward." + "reason": "Flash-attention s_aux=None crash and Gemma4 PLE device mismatch affect different subsystems and failure modes." }, { "left": "issue:41720", "right": "issue:45229", "accept": false, - "reason": "Different models and failure modes: Qwen3 auto device mapping asserts on A800, while Gemma4 hits OOM in multi-GPU inference." + "reason": "Different models and errors: Qwen3 auto device mapping assert vs Gemma4 OOM." }, { "left": "issue:44568", "right": "issue:45491", "accept": false, - "reason": "Tokenizer special-token insertion failure is unrelated to Gemma3 NaN embeddings from mixed-length sliding-window attention." + "reason": "Tokenizer special-token handling is unrelated to Gemma3 NaN embeddings from sliding-window attention." }, { "left": "issue:43381", "right": "issue:43701", "accept": false, - "reason": "Gradient checkpointing in eval mode and resume_from_checkpoint key mismatch are distinct trainer-state issues." + "reason": "Eval-mode gradient-checkpointing restriction is not the same as a resume_from_checkpoint key mismatch." }, { "left": "issue:33357", "right": "issue:39692", "accept": false, - "reason": "MacOS bus error on a community CLIP model and a SigLIP2 doc example mismatch/quantization failure are not the same bug." + "reason": "A MacOS bus error on a community CLIP model and a broken SigLIP2 docs example are different issues." }, { "left": "issue:44977", "right": "issue:45563", "accept": false, - "reason": "Flash-attention generation regression and a stale warning in paged generate are different symptoms and code paths." + "reason": "Qwen3.5 flash-attention generation regression vs a stale warning in paged generate are different." }, { "left": "issue:29127", "right": "issue:34567", "accept": false, - "reason": "LayoutLMv3 error-message clarity and TrainerState token-count updating are unrelated." + "reason": "LayoutLMv3 error-message clarity and TrainerState token counter updates are unrelated." }, { "left": "issue:44871", "right": "issue:44977", "accept": false, - "reason": "Gemma eos_token_id configuration inconsistency is unrelated to Qwen3.5 flash-attention generation failure." + "reason": "Gemma eos_token_id config mismatch is not the same as Qwen3.5 flash-attention generation failure." }, { "left": "issue:45198", "right": "issue:45803", "accept": false, - "reason": "Wav2Vec2 save/tokenization failure and Fuyu's bare except swallowing errors are different defects in different models." + "reason": "Wav2Vec2 save/tokenization failures are unrelated to a bare-except bug in FuyuBatchFeature.convert_to_tensors()." }, { "left": "issue:43653", "right": "issue:44743", "accept": false, - "reason": "BigBirdTokenizer special-token registration and Qwen3.5 recurrent-state reset are unrelated." + "reason": "BigBirdTokenizer special-token registration and Qwen3.5 recurrent-state reset are different bugs." }, { "left": "issue:44361", "right": "issue:45198", "accept": false, - "reason": "MLukeTokenizer task crashes and Wav2Vec2 save/tokenization failures are separate tokenizer bugs." + "reason": "Different tokenizers and failure modes: MLuke AttributeError vs Wav2Vec2 save/tokenization issues." }, { "left": "issue:45372", "right": "issue:45588", "accept": false, - "reason": "Gemma4 processor import/dependency failure is not the same as a flash-attention AttributeError on sink-less models." + "reason": "Gemma4 processor import/dependency failure and flash_attention s_aux=None crash are separate code paths." }, { "left": "issue:44186", "right": "issue:45198", "accept": false, - "reason": "LayoutLMv2 NER padding/truncation crashes and Wav2Vec2 save/tokenization failures do not share the same root cause." + "reason": "LayoutLMv2 NER/padding crashes and Wav2Vec2 save/tokenization failures are not the same underlying bug." }, { "left": "issue:43611", "right": "issue:46121", "accept": false, - "reason": "Base-model-prefix loading regression and rope-parameter validation TypeError are unrelated changes." + "reason": "Model loading with base_model_prefix and rope-validation JSON list handling are unrelated." }, { "left": "issue:42915", "right": "issue:42994", "accept": false, - "reason": "FP8 configuration failure in Qwen3Moe and quantized model saving failure are both quantization-adjacent but distinct." + "reason": "Qwen3Moe FP8 config failure and quantized model saving failure are different quantization problems." }, { "left": "issue:29870", "right": "issue:34689", "accept": false, - "reason": "BART generate() unbound local error and Llama 3.2 Vision model loading breakage are different model-loading/generation problems." + "reason": "BART generate-time UnboundLocalError and Llama 3.2 model-loading regression are unrelated." }, { "left": "issue:43412", "right": "issue:45588", "accept": false, - "reason": "Executorch export failures and flash-attention s_aux=None crashes are unrelated." + "reason": "Executorch export failures and flash-attention runtime crashes do not share the same code path." }, { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Feature-extractor unsupported-arg crash and video-input regression for Qwen VL models are different modality regressions." + "reason": "Feature-extractor center-arg crash and Qwen video-input regression are separate regressions." }, { "left": "issue:45372", "right": "issue:45406", "accept": false, - "reason": "Both involve Gemma4, but one is an import-time dependency issue and the other is a serve-time missing _tokenizer attribute; not the same bug." + "reason": "Both involve Gemma4, but one is an import-time dependency problem and the other is a serve-time missing _tokenizer attribute." }, { "left": "issue:43653", "right": "issue:44442", "accept": false, - "reason": "BigBird decode emptiness and AutoTokenizer failing to load FastSpeech2ConformerTokenizer are different tokenizer problems." + "reason": "BigBird special-token registration and AutoTokenizer loading FastSpeech2ConformerTokenizer are unrelated tokenizer bugs." }, { "left": "issue:29942", "right": "issue:30064", "accept": false, - "reason": "Flash Attention 2 test failures and void segmentation-map processing failure are unrelated." + "reason": "Flash Attention 2 CI failures and void segmentation-map processing are different subsystems." }, { "left": "issue:43756", "right": "issue:45440", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch and DeepSeekV3MoE divergence are different model-architecture issues." + "reason": "Smollm3 RoPE-layer mismatch and DeepSeekV3MoE divergence are different model-specific implementation issues." } ] }, @@ -10251,8 +10248,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -10333,7 +10330,7 @@ "issue:43761|issue:44188", "issue:43931|issue:45216" ], - "prepared_review_unit_hash": "515617d23246ad9183bf43fbccd0ff391cf7e810803cb4f419b9baa2659ce20c", + "prepared_review_unit_hash": "78a18e03751e77b5a62e416781fb6387251a0787f7e37e06f6c1d9e4ab7fa87f", "component_budget": { "node_count": 484, "item_count": 484, @@ -10355,174 +10352,174 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are distinct bug reports across different models, processors, configs, and loading paths. None of the soft pairs look like true duplicates or mergeable into one change.", - "confidence": 0.83, - "canonical_issue_reason": "No clear canonical issue: the cluster spans unrelated regressions and feature bugs, so any single issue would be only a loose representative.", + "summary": "This cluster is not a duplicate set; it contains many distinct bugs/regressions across unrelated areas (generation, audio, configs, loading, quantization, docs, and model-specific logic). The soft pairs are mostly only superficially similar by subsystem or model family.", + "confidence": 0.97, + "canonical_issue_reason": "No single issue is a clear canonical representative; the items span unrelated bug reports rather than one underlying defect.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the broadest active bug report in the set, but it is still just a representative issue, not a real duplicate anchor for the cluster.", + "best_issue_reason": "If one issue must anchor the cluster, issue 43749 has the strongest activity/signal (discussion and inbound references), but it is still just one distinct report among many unrelated ones.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43381", "right": "issue:43873", "accept": false, - "reason": "Both are regression reports, but they describe different failures in different code paths (gradient checkpointing eval-mode vs quantization/offloading)." + "reason": "Both involve training/runtime behavior, but one is gradient-checkpointing in eval mode and the other is quantization offloading; different code paths and fixes." }, { "left": "issue:44514", "right": "issue:45381", "accept": false, - "reason": "Same Qwen VL family, but one is a batched chat-template crash and the other is wrong video position ids; not the same bug." + "reason": "Same model family, but one is batched chat-template padding behavior and the other is video RoPE/position-id computation." }, { "left": "issue:43646", "right": "issue:44077", "accept": false, - "reason": "Unrelated areas: AutoConfig trust_remote_code field loss vs patchtsmixer post_init policy. No shared underlying defect." + "reason": "Missing fields from trust_remote_code config loading is unrelated to patchtsmixer post-init validation." }, { "left": "issue:43262", "right": "issue:44206", "accept": false, - "reason": "Both involve audio processors, but one is a sampling-rate default mismatch and the other is an unsupported center argument crash." + "reason": "Both are audio-related, but one is sampling-rate defaulting and the other is an unsupported feature-extractor argument." }, { "left": "issue:44977", "right": "issue:45245", "accept": false, - "reason": "Completely different failures: Qwen3.5 flash-attention generation vs category cardinality limit error." + "reason": "Completely unrelated: flash-attention generation bug versus a categorical-cardinality runtime error." }, { "left": "issue:44442", "right": "issue:44898", "accept": false, - "reason": "Tokenizer loading for FastSpeech2Conformer is unrelated to Perceiver image classification padding/interpolation behavior." + "reason": "Tokenizer loading and image-classification positional interpolation are different subsystems with different failure modes." }, { "left": "issue:44743", "right": "issue:44811", "accept": false, - "reason": "Different subsystems and bugs: recurrent state reset in modular Qwen vs Whisper batch_decode skip_special_tokens handling." + "reason": "Recurrent cache reset in modular Qwen code is unrelated to Whisper batch_decode skip_special_tokens handling." }, { "left": "issue:42490", "right": "issue:44743", "accept": false, - "reason": "Local-path import warning is unrelated to Qwen recurrent-state cache reset." + "reason": "Path-trailing-slash import warning is unrelated to Qwen recurrent-state cache behavior." }, { "left": "issue:43720", "right": "issue:44442", "accept": false, - "reason": "BitNet packed-weight loading and tokenizer loading failure are separate issues with no shared code path." + "reason": "BitNet packed-weight loading and FastSpeech2Conformer tokenizer loading are different code paths." }, { "left": "issue:45003", "right": "issue:45440", "accept": false, - "reason": "Generic sys.modules access bug vs DeepSeekV3 MoE divergence are not the same defect." + "reason": "Generic sys.modules access safety issue is not the same as DeepSeekV3MoE implementation divergence." }, { "left": "issue:42175", "right": "issue:46234", "accept": false, - "reason": "Different dependency/backend problems: missing tensorflow in torch install vs missing import_utils helper." + "reason": "Both touch backend/import utilities, but they are about different missing backends and unrelated symbols." }, { "left": "issue:27957", "right": "issue:29870", "accept": false, - "reason": "XLMRoberta Flash Attention 2 support and BART generate decoder_attention_mask error are unrelated." + "reason": "Flash Attention 2 support for XLM-Roberta is unrelated to BART generate() decoder_attention_mask assignment." }, { "left": "issue:43316", "right": "issue:43525", "accept": false, - "reason": "Both are config/API issues, but they concern different models and different missing/mismatched fields." + "reason": "Both are config/API discrepancies, but they concern different models and different missing fields." }, { "left": "issue:43493", "right": "issue:43827", "accept": false, - "reason": "SigLIP2 implementation mismatch and stale docs referencing pipeline() are unrelated." + "reason": "SigLIP2 implementation mismatch is unrelated to documentation still referencing removed pipeline() APIs." }, { "left": "issue:43723", "right": "issue:45479", "accept": false, - "reason": "Tokenizer loading in v5 and zero-loss sequence-classification behavior are different bugs." + "reason": "Tokenizer-loading regression and degenerate classification loss are unrelated." }, { "left": "issue:44077", "right": "issue:44479", "accept": false, - "reason": "PatchTSMixer post_init policy and Qwen video input regression do not share an underlying code-path problem." + "reason": "patchtsmixer post_init validation is unrelated to Qwen video-input regression." }, { "left": "issue:45325", "right": "issue:46697", "accept": false, - "reason": "Both are image/video preprocessing issues, but one is Qwen rope indexing and the other is Swin2SR over-padding." + "reason": "Both are preprocessing issues, but one is video temporal-position scaling and the other is image padding behavior." }, { "left": "issue:43873", "right": "issue:44871", "accept": false, - "reason": "Quantization/offloading bug and Gemma-3 eos_token_id config mismatch are unrelated." + "reason": "Quantization/offloading behavior and Gemma-3 eos_token_id configuration are unrelated." }, { "left": "issue:44062", "right": "issue:45907", "accept": false, - "reason": "AddedToken keyword duplication and list_repo_files revision forwarding are different API bugs." + "reason": "AddedToken constructor argument duplication is unrelated to composite Qwen text_config unwrapping." }, { "left": "issue:43761", "right": "issue:44079", "accept": false, - "reason": "Both touch outputs, but one is CLIPVisionModel hidden_states regression and the other is generic ModelOutput key assignment behavior." + "reason": "A model failing to return hidden_states is not the same as ModelOutput key assignment when a prior key was None." }, { "left": "issue:43749", "right": "issue:43931", "accept": false, - "reason": "FSDP CPU RAM efficient loading and Qwen3-VL shape mismatch are distinct model-loading problems." + "reason": "FSDP CPU RAM-efficient loading and Qwen3-VL shape mismatch are different loading problems." }, { "left": "issue:43643", "right": "issue:44913", "accept": false, - "reason": "Missing fields with trust_remote_code and rotary_pct not persisting on reload are separate config serialization issues." + "reason": "trust_remote_code missing fields is unrelated to GPTNeoX rotary_pct not persisting on reload." }, { "left": "issue:43316", "right": "issue:45759", "accept": false, - "reason": "Gemma3TextConfig API discrepancy and AutoModelForCausalLM.from_config not unwrapping text_config are not the same bug." + "reason": "Both involve config handling for multimodal/text configs, but they are distinct API regressions with different fixes." }, { "left": "issue:44987", "right": "issue:46178", "accept": false, - "reason": "Different backend/loading failures: physical-intelligence/fast loading vs tensorflow_text backend mapping error." + "reason": "One is a model-loading regression, the other is a backend-mapping/import error for tensorflow_text." }, { "left": "issue:44855", "right": "issue:45137", "accept": false, - "reason": "Python 3.13 TorchScript parsing error and DeepSpeed ZeRO3 deque underflow are unrelated runtime issues." + "reason": "Python 3.13 torch.jit parsing issue is unrelated to DeepSpeed ZeRO3 deque underflow." }, { "left": "issue:43761", "right": "issue:44188", "accept": false, - "reason": "Hidden_states regression in CLIPVisionModel and attention-kernel divergence under torch.compile are separate problems." + "reason": "Hidden-states output regression is unrelated to attention-kernel divergence under torch.compile." }, { "left": "issue:43931", "right": "issue:45216", "accept": false, - "reason": "Weight-shape mismatch on model loading and incorrect save_pretrained checkpointing are different directions of failure." + "reason": "A loading shape mismatch for Qwen3-VL is not the same as a save_pretrained regression for Qwen3.5." } ] }, @@ -13264,8 +13261,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -13349,7 +13346,7 @@ "issue:44945|issue:45753", "issue:41628|issue:44351" ], - "prepared_review_unit_hash": "7d0d68181c43b6f376f1982e64464b0ca89a466ae3a21f0725c69cbfaaef41ee", + "prepared_review_unit_hash": "4513848cce2fccb68e7b7bc0ba0cdf0d731d70e376f566a01b0759fc8128abfd", "component_budget": { "node_count": 484, "item_count": 484, @@ -13371,195 +13368,195 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of distinct Transformers bugs/regressions. The only clear duplicate-looking pair is the AutoProcessor/vision-backend crash without torchvision; the rest are related only at a broad subsystem or model-family level and should stay separate.", - "confidence": 0.86, - "canonical_issue_reason": "issue:43986 is the strongest representative of the one true duplicate pair: it is the earlier report, has more inbound references, and describes the same AutoProcessor crash when a required vision backend is missing.", + "summary": "Most pairs are superficially similar but refer to different underlying bugs. The only strong duplicate-like match is the AutoProcessor/vision-backend missing-dependency crash between issue:43986 and issue:46726.", + "confidence": 0.81, + "canonical_issue_reason": "issue:43986 is the better canonical issue because it states the broader AutoProcessor crash pattern more generally and was reported earlier, making it a better umbrella for the duplicate pair.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43986 is the best issue anchor because it is the earliest and most referenced instance of the duplicate AutoProcessor/vision-backend failure, making it the most suitable canonical issue in the cluster.", + "best_issue_reason": "issue:43986 is the strongest issue anchor overall because it captures the shared AutoProcessor failure in the broadest, most reusable wording for the duplicate pair.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43646", "right": "issue:44162", "accept": false, - "reason": "Both are regression-style complaints, but one is about custom model initialization and the other is an ESM2-specific breakage; they do not describe the same code-path failure." + "reason": "Both are Transformers v5 regressions, but one is a generic custom-model-init breakage and the other is an ESM2-specific failure; different code paths and fixes." }, { "left": "issue:44749", "right": "issue:45656", "accept": false, - "reason": "One reports preprocessing slowdown after upgrade; the other is about DeepSpeed causing optimizer.step to run twice. Different subsystems and symptoms." + "reason": "A filtering slowdown is unrelated to DeepSpeed causing optimizer.step to run twice." }, { "left": "issue:43701", "right": "issue:44749", "accept": false, - "reason": "Checkpoint resume key mismatch and token-filtering slowdown are unrelated failures." + "reason": "Checkpoint resume key mismatch is not the same bug as a data filtering slowdown." }, { "left": "issue:44871", "right": "issue:46132", "accept": false, - "reason": "EOS-token config mismatch in Gemma is unrelated to AttentionInterface.register changing callback behavior." + "reason": "EOS token config inconsistency is unrelated to AttentionInterface.register changing registered-function behavior." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Both mention quantization/gguf, but one is model saving and the other is model loading; not the same bug." + "reason": "Quantized model saving failure and GGUF loading failure are different operations with different fixes." }, { "left": "issue:44623", "right": "issue:45656", "accept": false, - "reason": "Missing files from processor.save_pretrained is unrelated to the DeepSpeed double optimizer-step report." + "reason": "Missing files in processor.save_pretrained is unrelated to a DeepSpeed optimizer-step double-call." }, { "left": "issue:43441", "right": "issue:46752", "accept": false, - "reason": "FlashAttention failure for Ministral and apply_chat_template IndexError are different code paths with no clear overlap." + "reason": "FlashAttention failure for Ministral-3 is a different path from apply_chat_template crashing on an empty list." }, { "left": "issue:43232", "right": "issue:43673", "accept": false, - "reason": "Both are generation-related, but one is a sync_gpus kwargs update issue and the other is missing cache during chunked_prefill; too different to merge." + "reason": "Both touch generation internals, but sync_gpus keyword handling and chunked_prefill cache absence are distinct failures." }, { "left": "issue:43475", "right": "issue:43643", "accept": false, - "reason": "A missing field on Sam3VisionEncoderOutput is not the same as trust_remote_code dropping fields from AutoConfig." + "reason": "SAM 3 missing an output attribute is not the same as trust_remote_code dropping fields from AutoConfig results." }, { "left": "issue:45375", "right": "issue:45478", "accept": false, - "reason": "Same model family, but one is a missing config field silently dropped by strict parsing and the other is a generic from_pretrained error; insufficient evidence of one bug." + "reason": "Same model family, but one is a strict/config-field omission and the other is a from_pretrained error; too little evidence of one underlying bug." }, { "left": "issue:28282", "right": "issue:30990", "accept": false, - "reason": "PyTorch import failure for AutoModel and a Sentence Transformers loading stall are unrelated." + "reason": "ImportError for missing PyTorch and Sentence Transformers loading hang are unrelated." }, { "left": "issue:44964", "right": "issue:45137", "accept": false, - "reason": "Model loading failure and DeepSpeed ZeRO3 deque underflow are different problems." + "reason": "Phi-4 multimodal loading and DeepSpeed ZeRO3 deque underflow are unrelated." }, { "left": "issue:43986", "right": "issue:46726", "accept": true, - "reason": "Both describe the same AutoProcessor failure when a required vision backend is missing, producing a confusing unrecognized-image-processor crash." + "reason": "Both describe the same underlying AutoProcessor failure when a required vision backend is missing; one is the specific video-model/torchvision case and the other is the generic error-message variant." }, { "left": "issue:41628", "right": "issue:44933", "accept": false, - "reason": "These are both import/export complaints, but they reference different symbols and likely different API surfaces." + "reason": "Different missing-import symbols and different import locations; not clearly the same bug." }, { "left": "issue:45310", "right": "issue:45375", "accept": false, - "reason": "The former is a general Qwen3.5 Moe from_pretrained error; the latter is a specific missing vision-config field. They may be related, but not enough to treat as the same bug." + "reason": "Both involve Qwen3.5 Moe, but one is a from_pretrained error and the other is a missing config field silently dropped; likely separate issues." }, { "left": "issue:44869", "right": "issue:45356", "accept": false, - "reason": "Whisper timestamp decode crash and Kimi tokenizer regression are distinct tokenizer/decoding issues." + "reason": "Whisper timestamp decoding crash and Kimi tokenizer codec regression are different tokenizer problems." }, { "left": "issue:42994", "right": "issue:43525", "accept": false, - "reason": "Quantized model saving and missing pad_token_id on Llama4Config are unrelated." + "reason": "Quantized model saving and Llama4Config missing pad_token_id are unrelated." }, { "left": "issue:44038", "right": "issue:44623", "accept": false, - "reason": "Qwen3-VL-Moe v5 incompatibility and processor.save_pretrained missing files do not share a concrete failure path." + "reason": "Qwen3-VL-Moe loading and processor.save_pretrained file omission are different code paths." }, { "left": "issue:45230", "right": "issue:45397", "accept": false, - "reason": "The generic bug report provides no specific overlap with the Gemma-4 ZeRO3 from_pretrained problem." + "reason": "A placeholder bug report does not establish the same problem as Gemma-4 zero3 from_pretrained." }, { "left": "issue:39692", "right": "issue:46726", "accept": false, - "reason": "SigLIP2 doc example errors and the missing-vision-backend AutoProcessor crash are different issues." + "reason": "Documentation/example errors for SigLIP2 are not the same as missing-vision-backend processor loading behavior." }, { "left": "issue:43723", "right": "issue:46326", "accept": false, - "reason": "Tokenizer loading in v5 and continue_final_message on mistral_common tokenizers are separate behaviors." + "reason": "AutoTokenizer loading in v5 and continue_final_message for mistral_common tokenizers are separate tokenizer bugs." }, { "left": "issue:43475", "right": "issue:45381", "accept": false, - "reason": "Both involve vision/video models, but the missing output attribute and wrong vision_position_ids are different concrete bugs." + "reason": "Both are video/vision-related, but one is a missing attribute in SAM 3 and the other is wrong vision_position_ids in Qwen2.5-VL." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF32 settings guidance and InternVL2 meta-initialization incompatibility are unrelated." + "reason": "TF32 settings guidance and remote-code checkpoint meta-initialization incompatibility are unrelated." }, { "left": "issue:43784", "right": "issue:46399", "accept": false, - "reason": "Import-time NameError in sentence-transformers and a Qwen3.5 kernel-registration ValueError are unrelated." + "reason": "Import-time NameError in sentence-transformers is unrelated to Qwen3.5 kernel registration failure." }, { "left": "issue:43232", "right": "issue:44188", "accept": false, - "reason": "Generation kwargs update after sync_gpus and divergent attention kernels under torch.compile are different generation internals issues." + "reason": "Generation kwargs update after sync_gpus and attention-kernel divergence under torch.compile are different regressions." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Offloading/quantization behavior and wrong LLM output with pipeline parallelism are distinct problems." + "reason": "Quantization offloading behavior and pipeline-parallelism output correctness are distinct problems." }, { "left": "issue:45406", "right": "issue:46202", "accept": false, - "reason": "Gemma4Processor missing _tokenizer and Gemma pipeline roundtrip failure on \u2581 are unrelated." + "reason": "Gemma4Processor missing _tokenizer and Gemma pipeline roundtrip failure are not the same bug." }, { "left": "issue:44945", "right": "issue:45753", "accept": false, - "reason": "Pipeline-parallel output corruption and Qwen3_5 infinite-loop on a specific image are not the same bug." + "reason": "Pipeline parallelism incorrect output and Qwen3_5 infinite loop on a specific image are unrelated symptoms." }, { "left": "issue:41628", "right": "issue:44351", "accept": false, - "reason": "Different missing imports: AutoImageProcessor versus HybridCache; likely separate export regressions." + "reason": "Both are import errors from transformers, but they concern different symbols and likely different missing exports." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet, and the only accepted soft-edge pair is a clearly close AutoProcessor/vision-backend duplicate candidate. The rest of the verdicts stay conservative and appropriately avoid merging unrelated subsystem regressions. I\u2019d only soften the wording from certainty to \u2018appears to be the same bug\u2019 if you want maximal caution." + "accept": false, + "feedback": "The pair-level judgment for issue:43986 vs issue:46726 is conservative and grounded, but the canonical-issue rationale overstates the evidence: issue:43986 does not clearly have \"more discussion/inbound context\" than issue:46726 (46726 has more inbound references, while 43986 has more discussion activity). If you want 43986 as canonical, justify it by its broader wording / earlier report instead of a comparative context claim. The rest of the summary looks fine." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -16065,8 +16062,8 @@ "review_unit_count": 58, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-484", "nodes": [ @@ -16155,7 +16152,7 @@ "issue:42175|issue:46365", "issue:43064|issue:44380" ], - "prepared_review_unit_hash": "8e2020ffc14ba78765f628ff8a5068561837b77ae35c1a5d80c1d2f202473bc0", + "prepared_review_unit_hash": "391e600863c3bde8a91e56732f6acb4405f57c6e7588196531c7ec7746eb2ed8", "component_budget": { "node_count": 484, "item_count": 484, @@ -16177,216 +16174,216 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is a loose grab-bag of Transformers issues with only superficial keyword similarity; I do not see any soft pair that looks like the same underlying bug or change.", - "confidence": 0.29, - "canonical_issue_reason": "Issue 46365 is the most connected item in the set and a clear user-facing runtime/import failure, so it is the best anchor if one issue must represent this loose cluster.", + "summary": "These items are a loose set of unrelated Transformers issues rather than a duplicate cluster. I found no soft-edge pair that clearly describes the same underlying bug/change, so all candidate links should be rejected.", + "confidence": 0.41, + "canonical_issue_reason": "issue:43873 has the strongest standalone signal in this set: active discussion, inbound references, and a concrete user-facing runtime bug with a clear reproduction around quantization/offloading.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 46365 has the strongest soft-similarity connectivity and is a concrete, reproducible failure report, making it the best overall triage representative here.", + "best_issue_reason": "issue:43873 is the best representative issue here because it is specific, actionable, and broadly impactful compared with the mostly unrelated import/test/config complaints in the rest of the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46298", "right": "issue:46752", "accept": false, - "reason": "A noisy logging message and an empty-list chat-template crash are unrelated symptoms and code paths." + "reason": "A noisy logging message is unrelated to an IndexError in apply_chat_template on an empty list." }, { "left": "issue:42890", "right": "issue:46298", "accept": false, - "reason": "Missing set_seed coverage in a SAM-HQ test suite is unrelated to logging noise." + "reason": "Test flakiness from missing set_seed is unrelated to logging output." }, { "left": "issue:46172", "right": "issue:46710", "accept": false, - "reason": "A specific dtype/transpose bug in DeepSeek-v4 is not the same as an accuracy regression after upgrading Transformers." + "reason": "These involve different failure modes: a model-specific tensor operation bug versus an accuracy regression after upgrade." }, { "left": "issue:38617", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but they refer to different missing symbols and likely different fixes." + "reason": "Both are import errors, but for different missing symbols and different code paths." }, { "left": "issue:43097", "right": "issue:43873", "accept": false, - "reason": "Removal of a config attribute and quantization offloading behavior are different issues." + "reason": "Removed API warning behavior and quantization offloading failure are different issues." }, { "left": "issue:44062", "right": "issue:44117", "accept": false, - "reason": "An AddedToken argument conflict is not the same as TOKENIZER_MAPPING_NAMES returning None." + "reason": "Tokenizer AddedToken construction error is unrelated to TOKENIZER_MAPPING_NAMES returning None." }, { "left": "issue:44117", "right": "issue:46701", "accept": false, - "reason": "Tokenizer mapping fallback behavior and an image-text pipeline failure are unrelated." + "reason": "A tokenizer mapping assumption bug is unrelated to a pipeline failure for DiffusionGemma output." }, { "left": "issue:44936", "right": "issue:46365", "accept": false, - "reason": "trainer.evaluate() failing after train() is a different code path from importing Trainer being affected by datasets." + "reason": "trainer.evaluate() after train() and failing to import Trainer due to datasets are different problems." }, { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "local_files_only still making API requests is unrelated to a missing embed_tokens.weight on UMT5Encoder load." + "reason": "local_files_only network leakage is unrelated to a missing embed_tokens.weight during UMT5Encoder loading." }, { "left": "issue:44704", "right": "issue:45103", "accept": false, - "reason": "Passing kwargs to cached_file and auto_docstring crashes with future annotations are unrelated failures." + "reason": "Passing kwargs to cached_file and auto_docstring crashes from future annotations are unrelated code paths." }, { "left": "issue:44380", "right": "issue:46421", "accept": false, - "reason": "Ignoring GPT-2 attention scaling under SDPA/FlashAttention is unrelated to cache lazy initialization causing recompilation." + "reason": "Attention scaling being ignored and cache lazy initialization causing recompilation are different bugs." }, { "left": "issue:43986", "right": "issue:44117", "accept": false, - "reason": "AutoProcessor/video-model loading without torchvision is not the same bug as tokenizer mapping returning None." + "reason": "AutoProcessor/torchvision crash is unrelated to tokenizer mapping returning None." }, { "left": "issue:44938", "right": "issue:46178", "accept": false, - "reason": "Python 3.14 load failure and a backend mapping error for tensorflow_text are different compatibility problems." + "reason": "Python 3.14 import failure and missing BACKENDS_MAPPING entry for tensorflow_text are separate compatibility issues." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Failing to save dequantized models and failing to save Pipeline objects are different save_pretrained paths." + "reason": "Saving dequantized models and saving Pipeline objects both involve serialization, but they are not the same defect." }, { "left": "issue:42907", "right": "issue:45753", "accept": false, - "reason": "Saving dequantized models is unrelated to an infinite-loop generation bug on a specific image." + "reason": "Model save failure and an infinite-loop generation bug are unrelated." }, { "left": "issue:43825", "right": "issue:44655", "accept": false, - "reason": "A misleading pipeline error message is not the same as pipeline save_pretrained failure." + "reason": "A v5 pipeline error message and save_pretrained on Pipeline objects are different behaviors." }, { "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "xLSTM training bugs and a Bamba torch.compile SDPA failure are unrelated model-specific issues." + "reason": "xLSTM training bugs and Bamba torch.compile SDPA failure are model-specific and distinct." }, { "left": "issue:43012", "right": "issue:43550", "accept": false, - "reason": "A bfloat16 precision warning during compile is not the same as a torch.compile SDPA failure." + "reason": "A precision warning during bf16 compilation is not the same as a compile failure with SDPA." }, { "left": "issue:45712", "right": "issue:46298", "accept": false, - "reason": "Leftover dummy classes in dummy_pt_objects.py are unrelated to a noisy logging message." + "reason": "Dummy class leakage in no-torch environments is unrelated to a logging message." }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "A compilation precision warning and a tie_word_embeddings config warning are different warnings with different causes." + "reason": "These are different warnings with different causes and remediation." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "Missing weights on from_pretrained and the 50GB shard-size question are unrelated save/load concerns." + "reason": "Missing weights on load is unrelated to the default 50GB shard size in save_pretrained." }, { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "Wrong tokenizer decoder type in v5 and an isolated reproduction of another issue do not match as the same bug." + "reason": "Wrong tokenizer decoder type and an isolated reproduction of another issue are not the same bug." }, { "left": "issue:43352", "right": "issue:43976", "accept": false, - "reason": "Flash Attention 2 support for a model and a Python version compatibility issue are unrelated." + "reason": "Flash Attention support for a model and Python version compatibility are separate problems." }, { "left": "issue:43352", "right": "issue:44380", "accept": false, - "reason": "Flash Attention support gating for NemotronH and GPT-2 attention scaling under SDPA are different backend/model bugs." + "reason": "They both mention attention backends, but one is missing FA2 support for a model while the other ignores attention scaling config." }, { "left": "issue:43097", "right": "issue:46298", "accept": false, - "reason": "A removed config attribute and a noisy logging message are not the same underlying problem." + "reason": "Removed tie_embeddings_and_encoder_decoder and noisy logging are unrelated." }, { "left": "issue:44162", "right": "issue:46172", "accept": false, - "reason": "ESM2 breakage and a DeepSeek-v4 tensor operation bug are unrelated model failures." + "reason": "ESM2 breakage and a DeepSeek tensor operation issue are unrelated model-specific bugs." }, { "left": "issue:44380", "right": "issue:46132", "accept": false, - "reason": "Ignored GPT-2 attention scaling and AttentionInterface.register behavior changes are different attention-system changes." + "reason": "Ignoring GPT2 attention scaling and AttentionInterface.register changing function behavior are different code-path issues." }, { "left": "issue:44077", "right": "issue:46365", "accept": false, - "reason": "patchtsmixer post_init validation and importing Trainer being affected by datasets are unrelated." + "reason": "A patchtsmixer post_init policy issue is unrelated to importing Trainer failing because of datasets." }, { "left": "issue:45584", "right": "issue:46365", "accept": false, - "reason": "Whisper empty-transcription generation and a Trainer import conflict are different failures." + "reason": "Whisper empty-transcription generation failure is unrelated to a Trainer import problem." }, { "left": "issue:44222", "right": "issue:45478", "accept": false, - "reason": "FP8 save_pretrained for moe and a Qwen3.5 MoE from_pretrained error are related by model family only, not the same concrete bug." + "reason": "FP8 save_pretrained and Qwen3.5 MoE from_pretrained errors are different save/load bugs." }, { "left": "issue:36683", "right": "issue:43582", "accept": false, - "reason": "Gemma3Config missing vocab_size and an Apple Silicon allocator TypeError are unrelated." + "reason": "A missing config attribute and an Apple Silicon TypeError are unrelated." }, { "left": "issue:44380", "right": "issue:45479", "accept": false, - "reason": "GPT-2 attention scaling under backends and a zero-loss classification bug are unrelated model behaviors." + "reason": "Attention backend config being ignored is unrelated to a degenerate zero-loss classification setup." }, { "left": "issue:42175", "right": "issue:46365", "accept": false, - "reason": "TensorFlow backend packaging and Trainer import failing because of datasets are different dependency issues." + "reason": "Backend installation packaging and Trainer import interference from datasets are different problems." }, { "left": "issue:43064", "right": "issue:44380", "accept": false, - "reason": "FSDP2/PEFT optimizer-state corruption and GPT-2 attention scaling under SDPA are unrelated training vs inference bugs." + "reason": "Distributed optimizer state corruption and attention scaling backend behavior are unrelated." } ] }, @@ -20292,8 +20289,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-46657-2", "nodes": [ @@ -20303,7 +20300,7 @@ "soft_pairs": [ "issue:46657|issue:46692" ], - "prepared_review_unit_hash": "428b7cbcd76b48dff0757f56d0ba45d19eb0ad3af2c7349bfc5c017ff65fa9a8", + "prepared_review_unit_hash": "8bbb237adefcde7eb3a297ee3c8ae2d434a85e42df093b5291675d8632ca7e66", "component_budget": { "node_count": 2, "item_count": 2, @@ -20325,18 +20322,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues do not look like duplicates: one reports a tensor shape mismatch during object-detection training, while the other reports out-of-memory after several epochs in RT-DETR v2 training. They point to different failure modes and likely different root causes.", - "confidence": 0.96, - "canonical_issue_reason": "Issue #46657 is the more specific, concrete bug report and reads like a distinct code-path failure; it is a better representative issue than the separate OOM report.", + "summary": "These two issues are not duplicates: one reports a tensor shape mismatch during object-detection training, while the other reports out-of-memory after several epochs in RT-DETR v2. They may be broadly related to training instability, but they describe different failure modes and likely different fixes.", + "confidence": 0.95, + "canonical_issue_reason": "Issue #46657 is the more concrete, narrowly scoped bug report (tensor size mismatch during object detection training) and is the better representative if a single issue must be chosen.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #46657 is the best issue candidate because it describes a precise, actionable error (tensor size mismatch) rather than a broad resource symptom.", + "best_issue_reason": "Issue #46657 is the best issue anchor because it describes a specific reproducible error rather than a broader resource problem.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46657", "right": "issue:46692", "accept": false, - "reason": "Reject: one is a tensor-shape mismatch error and the other is an out-of-memory failure after epochs; these are different underlying problems, not the same bug." + "reason": "Different underlying problems: #46657 is a tensor dimension mismatch during training, while #46692 is an out-of-memory failure after some epochs in RT-DETR v2. Shared training context is not enough to treat them as the same bug." } ] }, @@ -20366,12 +20363,12 @@ "soft_pairs": [ "pull_request:40546|pull_request:42781" ], - "prepared_review_unit_hash": "6e6ec09523764e0aacf87e83a28f56e57412512bdd23e965f4e7b948a920fe16", + "prepared_review_unit_hash": "ede7dd2766c805e2ddc64a1845b151f97a7b2ed5a18c48486a14a9cb49c89387", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7223, + "serialized_chars": 7224, "estimated_input_tokens": 1806, "estimated_eval_tokens": 3868 }, @@ -20379,7 +20376,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7223, + "serialized_chars": 7224, "estimated_input_tokens": 1806, "estimated_eval_tokens": 3868 }, @@ -20388,18 +20385,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related VibeVoice PRs, but they target different changes: PR 40546 is the core VibeVoice implementation, while PR 42781 is a later draft for VibeVoice Realtime and the acoustic tokenizer. They share files because the second builds on the same subsystem, but they are not duplicates.", - "confidence": 0.89, + "summary": "The two PRs are related to VibeVoice but are not the same change: #40546 is the base VibeVoice implementation, while #42781 adds VibeVoice Realtime plus acoustic-tokenizer/feature-extraction and generation-path changes. They overlap on shared model integration files, but they target different concrete functionality and should not be deduplicated.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 40546 is the better canonical representative of the cluster\u2019s main feature work: it is the non-draft, core VibeVoice implementation with broader integration and more review activity. PR 42781 is a separate realtime/tokenizer extension rather than the same change.", + "canonical_pr_reason": "#40546 is the more canonical base PR: it introduces the core VibeVoice model/docs integration and is the broader foundational implementation, whereas #42781 is a separate realtime extension built on top of that work.", "best_issue_reason": null, - "best_pr_reason": "PR 40546 is the best single representative of the underlying VibeVoice feature. It covers the primary code path and package integration, whereas PR 42781 adds a distinct realtime-oriented extension and should not be merged as the same PR.", + "best_pr_reason": "#40546 is the best representative of the cluster because it is the foundational VibeVoice implementation and has the clearest standalone scope; #42781 is a later, narrower realtime follow-up with additional tokenizer/feature-extraction work.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Shared VibeVoice files do not make these the same change. 40546 implements the base model, while 42781 adds realtime/acoustic-tokenizer work and additional auto/generation plumbing, so they are related but not duplicates." + "reason": "Reject: these are related VibeVoice PRs, but not the same underlying change. #40546 implements VibeVoice itself; #42781 adds a separate realtime variant and extra acoustic-tokenizer/feature-extraction/generation changes, so they would not plausibly be merged into one PR." } ] }, @@ -20429,12 +20426,12 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "e4d722668d2a6aa8b9cf92ec66abc43e9e11cc1e95dd6b5f9ba1410f9134c0eb", + "prepared_review_unit_hash": "8438154b2b81ce848874db2f56ef18b21b67707af0a789569ca2b5b2466a353a", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4674, + "serialized_chars": 4676, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -20442,7 +20439,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4674, + "serialized_chars": 4676, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -20451,18 +20448,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are unrelated feature additions in different model families, so they should not be clustered as duplicates.", - "confidence": 0.98, + "summary": "The two pull requests are unrelated feature additions in different model families: TimesFM covariate forecasting versus DINOv3 image classification support. They do not fix the same bug or implement the same change path, so they should not be merged as duplicates.", + "confidence": 0.99, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: #40755 adds TimesFM covariate forecasting support, while #41224 adds DINOv3 image classification support; they are separate model-specific changes.", + "canonical_pr_reason": "41224 is the more focused and self-contained PR, adding a concrete DINOv3 image-classification path; 40755 is a broader TimesFM forecasting feature in a different subsystem.", "best_issue_reason": null, - "best_pr_reason": "Neither PR is a duplicate candidate for the other, and there is no single best representative of a shared underlying change.", + "best_pr_reason": "41224 is the better representative PR because it is narrower, more concrete, and easier to review/merge as a single feature change than the broader TimesFM covariate work.", "soft_edge_verdicts": [ { "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different models and code paths: TimesFM forecasting-with-covariates vs DINOv3 image classification. They do not fix the same underlying bug or implement the same change." + "reason": "Different model families and different code paths: TimesFM covariate forecasting vs DINOv3 image classification support. They are not the same underlying change." } ] }, @@ -20585,154 +20582,166 @@ "pull_request:44827|pull_request:45655", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "9a521bd931588280fee17bcff22923802f7a1b2e389f5924619c1fa8879d1821", + "prepared_review_unit_hash": "f78b61df4dd42171716fd28ae526d107ddc36b33b007730e62600aa3ba73d1b9", "component_budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 19, - "serialized_chars": 36837, - "estimated_input_tokens": 9210, - "estimated_eval_tokens": 18676 + "serialized_chars": 36844, + "estimated_input_tokens": 9211, + "estimated_eval_tokens": 18678 }, "budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 19, - "serialized_chars": 36837, - "estimated_input_tokens": 9210, - "estimated_eval_tokens": 18676 + "serialized_chars": 36844, + "estimated_input_tokens": 9211, + "estimated_eval_tokens": 18678 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This set is mostly unrelated PRs grouped by loose soft-similarity; the only clear duplicate pair is the two `fixed_cross_entropy` kwargs PRs. One PR in that pair should be treated as the canonical representative, while the rest should stay separate.", - "confidence": 0.93, + "summary": "Only one soft edge looks like a true duplicate: the two cross-entropy loss PRs both change the same helper in the same way. The rest are merely in the same broad Transformers codebase or thematically similar, but they fix unrelated models, docs, tests, or subsystems.", + "confidence": 0.78, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43251 is the better canonical representative for the duplicate loss-fix pair: it has the same concrete code change as #43254, an explicit issue target, and slightly more review activity/clearer framing.", + "canonical_pr_reason": "pull_request:43251 is the better canonical PR for the only true duplicate pair: it has the explicit issue target, a concrete loss-utils fix, and a slightly fuller implementation of the same cross-entropy kwargs change.", "best_issue_reason": null, - "best_pr_reason": "PR #43251 is the strongest standalone PR in this set because it fixes a concrete bug in one code path (`fixed_cross_entropy` passthrough to `nn.functional.cross_entropy`), has an explicit tracking issue, and is narrowly scoped. Most other PRs are unrelated feature/docs/model additions or test-only changes.", + "best_pr_reason": "pull_request:43251 is the strongest standalone PR in this set because it addresses a specific bug in a single code path (`fixed_cross_entropy`), has an explicit issue link, and is tightly scoped compared with the more speculative similarity-only matches.", "soft_edge_verdicts": [ { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Different model areas and changes: Mistral tokenizer API additions vs. a new MobileLLM model implementation. No shared bug or mergeable combined fix." + "reason": "Different areas and goals: Mistral tokenizer API additions vs adding a new MobileLLM model. No shared bug or code path." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Tokenizer-method additions are unrelated to executorch dynamic-shape export tests." + "reason": "Tokenizer method completion vs Executorch dynamic-shape export testing; unrelated changes." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Both touch processor/model-loading infrastructure and model code, but they are about different multimodal model families and different failure modes." + "reason": "Processor-loading robustness across multiple models vs adding a brand-new model. Broad framework overlap only." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Loss-function kwarg passthrough is unrelated to adding MobileLLM." + "reason": "Loss-function kwargs fix vs new model addition; entirely different functionality." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "`fixed_cross_entropy` kwarg support is unrelated to Mistral4 test fixes and model wiring." + "reason": "Both touch training/modeling code, but one fixes `cross_entropy` kwargs and the other is a Mistral4 model/test update. Not the same change." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "Executorch export tests and Mistral4 fixes are not the same underlying change." + "reason": "Executorch export testing and Mistral4 test fixes are unrelated test coverage in different subsystems." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "Dynamic-shape export tests do not overlap with MobileLLM model addition." + "reason": "Dynamic export test vs new model implementation; no shared underlying bug." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "New MobileLLM model code is unrelated to `cls.config` resolution ordering." + "reason": "Adding MobileLLM vs fixing config annotation resolution in `modeling_utils.py`; unrelated." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "MobileLLM addition and Falcon output-capture refactor are separate model-level changes." + "reason": "New model addition vs Falcon output-interface refactor; different models and different code paths." }, { "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "Falcon refactor and the model-creation skill are unrelated." + "reason": "Falcon model refactor vs model-creation skill/docs; no shared bug or mergeable overlap." }, { "left": "pull_request:43995", "right": "pull_request:44827", "accept": false, - "reason": "Falcon output-interface refactor is unrelated to Mistral4 test/model fixes." + "reason": "Falcon output-capture refactor and Mistral4 test fixes are distinct model-specific changes." }, { "left": "pull_request:44827", "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 test fixes and `layrnorm` renaming across CLIP-like models are different subsystems and fixes." + "reason": "Mistral4 test fixes vs a CLIP-like `layrnorm` rename/conversion mapping fix; unrelated models and bugs." }, { "left": "pull_request:44827", "right": "pull_request:46267", "accept": false, - "reason": "Mistral4-specific fixes do not match serve chat parsing updates." + "reason": "Mistral4 tests vs chat-response parsing overhaul; completely different functionality." }, { "left": "pull_request:44827", "right": "pull_request:46744", "accept": false, - "reason": "Model test fixes and Apple Silicon docs are unrelated." + "reason": "Mistral4 tests vs Apple Silicon/safetensors docs update; not the same underlying change." }, { "left": "pull_request:44827", "right": "pull_request:46048", "accept": false, - "reason": "Mistral4 test fixes and TDT loss-kernel integration are different changes." + "reason": "Mistral4 tests vs TDT loss kernel integration; unrelated code paths." }, { "left": "pull_request:44827", "right": "pull_request:45213", "accept": false, - "reason": "Mistral4 changes and the model-creation skill are not the same change." + "reason": "Mistral4 tests/modeling vs model-creation skill/dataset/doc tooling. The shared auto-model file is incidental, not a duplicate bug fix." }, { "left": "pull_request:43823", "right": "pull_request:45213", "accept": false, - "reason": "MobileLLM addition and the model-creation skill are unrelated." + "reason": "New MobileLLM model implementation vs a repository skill/docs PR. Similar 'model creation' theme only." }, { "left": "pull_request:44827", "right": "pull_request:45655", "accept": false, - "reason": "Mistral4 test/model work is unrelated to `cls.config` resolution ordering." + "reason": "Mistral4 test/model changes vs config-resolution logic in the base modeling utility; different layers and problems." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46048", + "accept": false, + "reason": "Mistral4 test fixes vs kernel-loading for TDT loss; unrelated subsystems." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Model test/modeling changes vs model-creation tooling; no shared underlying code change." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "Both PRs implement the same concrete fix in `src/transformers/loss/loss_utils.py`: pass supported kwargs through to `nn.functional.cross_entropy`, including `weight` and `label_smoothing`. They are direct alternatives for the same underlying bug." + "reason": "Same file, same helper (`fixed_cross_entropy`), same added kwargs (`weight`, `label_smoothing`), and same issue target 43240. These are the same underlying fix." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet, and the only true duplicate pair identified is the two `fixed_cross_entropy` PRs. The other pairwise rejections are conservative and do not overclaim mergeability into the same bugfix PR." + "feedback": "The summary stays within the packet evidence, and the non-duplicate verdicts are conservative: they treat broad thematic overlap as non-duplication. The only accepted soft edge, pull_request:43251 vs pull_request:43254, is well-supported by the shared file, shared issue target, and same kwargs-to-cross_entropy fix." }, "evaluator_used": true, "retried": false, @@ -20759,7 +20768,7 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "c462d362048c9aa9ae1c374e33df0148cc8ae41131c74a8de8c72cfd2fc2eacc", + "prepared_review_unit_hash": "575e438ebec92de7d9851f707d94123fb53d190790fa7e699f0f615dddb85214", "component_budget": { "node_count": 2, "item_count": 2, @@ -20781,18 +20790,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs both do broad GitHub Actions version bumps, but they target different action families and different upgrade goals, so they should remain separate.", - "confidence": 0.89, + "summary": "Both pull requests are broad GitHub Actions dependency bumps, but they target different underlying changes: PR 42975 is a Node 24 compatibility sweep across many workflow actions, while PR 42976 specifically upgrades docker/build-push-action and related workflow versions. They overlap in files and theme, but not in the same concrete fix.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #42975 is the cleaner canonical PR because it has a narrower, explicit goal: upgrading workflow actions for Node 24 compatibility across many workflows.", + "canonical_pr_reason": "PR 42975 is the better canonical representative because it has the clearest unifying purpose (Node 24 compatibility) and covers the wider workflow-action upgrade set.", "best_issue_reason": null, - "best_pr_reason": "PR #42975 is the best representative of the cluster because it is the most clearly scoped maintenance upgrade and affects the widest set of workflow actions.", + "best_pr_reason": "PR 42975 is the best single PR to represent the cluster since it expresses a concrete compatibility goal rather than a narrower action-version refresh.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "Both are GitHub Actions version bumps, but they change different concrete actions: #42975 upgrades checkout/cache/upload-artifact for Node 24 compatibility, while #42976 upgrades docker/build-push-action and a few other actions. They are related maintenance, not the same underlying fix, and would not naturally collapse into one PR." + "reason": "They are related GitHub Actions maintenance PRs, but they fix different things: one is a Node 24 compatibility upgrade across workflow actions, the other is a docker/build-push-action version bump. Shared workflow files are not enough to treat them as the same change." } ] }, @@ -20838,106 +20847,106 @@ "pull_request:45490|pull_request:46705", "pull_request:46562|pull_request:46658" ], - "prepared_review_unit_hash": "816c56cbae63e865315de22bc8e1ea65e4c28195cea0ae79fb26f7f5f22fd97b", + "prepared_review_unit_hash": "8701d24d4d5f226878fde60ba7db3806947945b9e6c1bb61da50997d3b49683f", "component_budget": { "node_count": 7, "item_count": 7, "soft_pair_count": 12, - "serialized_chars": 24620, - "estimated_input_tokens": 6155, - "estimated_eval_tokens": 12566 + "serialized_chars": 24621, + "estimated_input_tokens": 6156, + "estimated_eval_tokens": 12568 }, "budget": { "node_count": 7, "item_count": 7, "soft_pair_count": 12, - "serialized_chars": 24620, - "estimated_input_tokens": 6155, - "estimated_eval_tokens": 12566 + "serialized_chars": 24621, + "estimated_input_tokens": 6156, + "estimated_eval_tokens": 12568 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates; they are separate model-addition or infrastructure PRs. The only superficially close pair is the Step 3.7 Flash model PR and its companion mini-model/debug script PR, but they do not implement the same mergeable code change.", - "confidence": 0.96, + "summary": "Most PRs are separate model-addition or infrastructure changes that only overlap on common registration/docs files. The only likely duplicate-like pair is the Step 3.7 Flash feature PR and its WIP companion script/model package; everything else should stay separate.", + "confidence": 0.78, "canonical_issue_reason": null, - "canonical_pr_reason": "If one PR must represent the cluster, 43451 is the strongest standalone feature PR: it has the broadest model implementation surface, the most review activity, and a complete docs/code/tests package. That said, the cluster does not appear to contain duplicates.", + "canonical_pr_reason": "PR #46562 is the best canonical PR because it is the main library-side Step 3.7 Flash model integration with docs/tests; #46658 looks like supporting WIP conversion/build tooling for the same feature, not the primary implementation.", "best_issue_reason": null, - "best_pr_reason": "43451 is the best representative PR because it is the most complete and actively reviewed model addition in the set; the other PRs are unrelated additions or support tooling.", + "best_pr_reason": "PR #46562 is the strongest representative of the only accepted duplicate-like pair and is the most self-contained concrete model change in the set.", "soft_edge_verdicts": [ { "left": "pull_request:42978", "right": "pull_request:43448", "accept": false, - "reason": "Different model additions (ViT NEPA vs Molmo). Shared docs/auto-registry edits are generic scaffolding, not the same underlying change." + "reason": "Both are unrelated new model additions; shared docs/auto-registration files are boilerplate and do not indicate the same underlying change." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "Different model additions (ViT NEPA vs Molmo2). Overlap is only in common Transformers registry/docs files." + "reason": "ViT NEPA and Molmo2 are different models with different code paths; overlap is only in shared framework registration/docs." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "Different features/models (ViT NEPA vs CTSM). Similar auto-mapping/docs edits are incidental." + "reason": "Different model families entirely (ViT NEPA vs CTSM); shared auto/docs edits are generic integration work." }, { "left": "pull_request:42978", "right": "pull_request:46705", "accept": false, - "reason": "Distributed-config infrastructure PR is unrelated to adding ViT NEPA; no shared bug or concrete code path." + "reason": "ViT NEPA model addition is unrelated to the distributed config PR; no shared bug or concrete change." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are distinct model families and separate implementations, not one fix/change." + "reason": "Molmo and Molmo2 are distinct model additions with different architectures/files; overlap is just model-registration boilerplate." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Different model additions (Molmo vs CTSM); only broad registry/docs files overlap." + "reason": "Molmo2 and CTSM are different models; common files like docs and auto mappings are not enough to treat them as the same change." }, { "left": "pull_request:43448", "right": "pull_request:46705", "accept": false, - "reason": "Molmo model-add PR and distributed-config PR are unrelated." + "reason": "Molmo2 model work is unrelated to distributed configuration support." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 and CTSM are separate model additions with no shared concrete bug or implementation." + "reason": "Molmo2 and CTSM are separate model integrations with different implementations and tests." }, { "left": "pull_request:43451", "right": "pull_request:46705", "accept": false, - "reason": "Molmo2 model-add PR and distributed-config PR are unrelated." + "reason": "Molmo2 model addition and distributed config are distinct changes with no shared code-path problem." }, { "left": "pull_request:45490", "right": "pull_request:46562", "accept": false, - "reason": "CTSM and Step 3.7 Flash are different model additions. Shared auto-registry/docs changes are not enough to merge them as duplicates." + "reason": "CTSM and Step 3.7 Flash are unrelated model additions; shared auto/docs files are normal repo integration overlap." }, { "left": "pull_request:45490", "right": "pull_request:46705", "accept": false, - "reason": "CTSM model addition is unrelated to distributed config changes." + "reason": "CTSM model addition does not match the distributed config PR; no same underlying bug/change." }, { "left": "pull_request:46562", "right": "pull_request:46658", - "accept": false, - "reason": "Both mention Step 3.7 Flash, but 46562 adds the Transformers model while 46658 is a separate mini-model/debug/conversion script PR; they do not fix the same concrete code path and would not cleanly merge into one PR." + "accept": true, + "reason": "Both are Step 3.7 Flash work and overlap on the same model family and registration plumbing; #46562 is the core model implementation while #46658 is WIP conversion/build scaffolding for the same feature, so they could plausibly be consolidated." } ] }, @@ -20986,118 +20995,118 @@ "pull_request:46380|pull_request:46539", "pull_request:46380|pull_request:46579" ], - "prepared_review_unit_hash": "6da3bc6a0cabf56cf73f452288a06df45a792591556f766d17d70d6cbe815f90", + "prepared_review_unit_hash": "c75cd616b33a8c70760607a4c297e0085816ead8dbc2d77b3131568dc5476db8", "component_budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 14, - "serialized_chars": 23162, - "estimated_input_tokens": 5791, - "estimated_eval_tokens": 11838 + "serialized_chars": 23167, + "estimated_input_tokens": 5792, + "estimated_eval_tokens": 11840 }, "budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 14, - "serialized_chars": 23162, - "estimated_input_tokens": 5791, - "estimated_eval_tokens": 11838 + "serialized_chars": 23167, + "estimated_input_tokens": 5792, + "estimated_eval_tokens": 11840 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The cluster is heterogeneous: it mixes unrelated PRs for Whisper, test cleanup, dequantization, doctest docs, generation validation, CI infrastructure, and typo fixes. The soft similarities are superficial; none look like the same underlying change or bug fix.", - "confidence": 0.91, + "summary": "This cluster is a mix of unrelated open PRs: Whisper test fixes, tie_weights backward-compatibility, vision test cleanup, FP8 dequant changes, doctest updates, generation watermarking validation, CI infra changes, and typo/documentation cleanup. None of the soft edges look like true duplicates of the same underlying change.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #44815 is the clearest concrete bug fix in the set: it has a coherent dequantization-related code path change across loading/FP8 integration and test coverage, unlike the mostly broad cleanup/test/docs/typo PRs.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR #44815 is the strongest standalone PR in this group because it targets a specific runtime code-path problem and makes coordinated code changes, making it the best representative candidate overall.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "Both touch tests, but one is Whisper tokenization/decode behavior and the other is CLIP/ViT test assertions; they are not the same bug or change." + "reason": "Both are test-related, but they target different models and bugs: Whisper tokenizer decode behavior vs CLIP/ViT test edits. No shared code path." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper decode/test cleanup is unrelated to FP8 dequantization and model-loading changes." + "reason": "Whisper decoding tests are unrelated to FP8 dequant/loading logic. Different subsystem and fix scope." }, { "left": "pull_request:43291", "right": "pull_request:46380", "accept": false, - "reason": "Whisper tokenizer/test work is unrelated to generation watermarking validation and broader generation fixes." + "reason": "Whisper test fix vs generation watermarking/validation and broad model test changes. Only superficial similarity as test fixes." }, { "left": "pull_request:43498", "right": "pull_request:46579", "accept": false, - "reason": "A backward-compatibility shim for tie_weights is unrelated to a sweeping typo-fix PR across many files." + "reason": "Backward-compatibility for tie_weights is unrelated to typo cleanup across many files. No common bug or mergeable change." }, { "left": "pull_request:43665", "right": "pull_request:46539", "accept": false, - "reason": "Test assertion edits in model tests are unrelated to CI runner memory/device-map handling for Phimoe integration tests." + "reason": "Vision model test tweaks are unrelated to Phimoe CI/integration test infrastructure changes." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "Model test-only assertions do not match the dequantization/runtime fix in FP8/model loading." + "reason": "These touch different areas: test assertions for CLIP/ViT versus dequant/fp8 model loading and quantization code." }, { "left": "pull_request:43665", "right": "pull_request:46380", "accept": false, - "reason": "These are different test/model areas and different problems; one is test placeholder edits, the other is generation config validation plus Whisper-related code." + "reason": "Both mention tests, but the underlying changes are different: CLIP/ViT test files versus generation config/utils and Whisper behavior." }, { "left": "pull_request:43665", "right": "pull_request:46579", "accept": false, - "reason": "Local test assertion changes are not the same as repository-wide typo cleanup." + "reason": "Vision test file edits and code-typo cleanup are not the same bug or change." }, { "left": "pull_request:44815", - "right": "pull_request:46380", + "right": "pull_request:46539", "accept": false, - "reason": "Both involve model code, but the concrete problems differ: FP8 dequantization versus generation watermarking/config validation." + "reason": "FP8/dequant logic changes are unrelated to Phimoe integration-test CI adjustments." }, { "left": "pull_request:44815", - "right": "pull_request:46539", + "right": "pull_request:46380", "accept": false, - "reason": "Dequantization/model-loading logic is unrelated to CI memory limits and Phimoe integration test setup." + "reason": "Different code paths: finegrained FP8 loading/quantization versus generation configuration and Whisper-related test fixes." }, { "left": "pull_request:44815", "right": "pull_request:46579", "accept": false, - "reason": "A targeted runtime fix is not the same underlying change as a broad typo sweep across many modules." + "reason": "FP8/model-loading changes do not match a broad typo-fix cleanup PR." }, { "left": "pull_request:45114", "right": "pull_request:46579", "accept": false, - "reason": "Doctest documentation edits are unrelated to code typo fixes across the library." + "reason": "Doctest documentation updates are not the same underlying change as code typo fixes across multiple modules." }, { "left": "pull_request:46380", "right": "pull_request:46539", "accept": false, - "reason": "Generation validation/Whisper fixes do not match CI infrastructure changes for Phimoe integration tests." + "reason": "Generation/watermarking and Whisper test fixes are unrelated to Phimoe CI environment changes." }, { "left": "pull_request:46380", "right": "pull_request:46579", "accept": false, - "reason": "Generation config/Whisper code changes are unrelated to broad typo-only cleanup." + "reason": "Broad generation/Whisper bugfixes are not the same as a multi-file typo cleanup." } ] }, @@ -21130,52 +21139,52 @@ "pull_request:43488|pull_request:43532", "pull_request:43532|pull_request:46747" ], - "prepared_review_unit_hash": "cb903ad1a24b24d4cda141e6a820a445dc6589f7870289c1cc9eaf8cb40485b3", + "prepared_review_unit_hash": "466decdcca13ae52aaf9bd3c927219e036c85c2fdaee60e8729322f44fd2daf2", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 6349, - "estimated_input_tokens": 1588, - "estimated_eval_tokens": 3432 + "serialized_chars": 6348, + "estimated_input_tokens": 1587, + "estimated_eval_tokens": 3430 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 6349, - "estimated_input_tokens": 1588, - "estimated_eval_tokens": 3432 + "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": "Three separate '[don't merge]' PRs for repo-bot/CI testing: one tweaks packaging/versioning and a Vit test, one changes self-comment CI and related helper files, and one switches a security-gate workflow trigger. The overlap is superficial; they are not the same underlying change.", + "summary": "Three unrelated \"don't merge\"/demo PRs that only overlap at a very broad level (repo-bot, workflow, and test-related changes). They do not look like the same underlying bug or change.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: each PR targets a different code path or workflow and appears to be an isolated test change rather than a duplicate of the others.", + "canonical_pr_reason": "PR 43532 is the most representative of the cluster as a generic diff/demo PR: it touches several workflow and test files and best matches the overall placeholder 'show diff' theme.", "best_issue_reason": null, - "best_pr_reason": "43532 is the broadest CI/bot-related PR, but it is still not a good duplicate anchor because the other PRs modify different workflows and packaging behavior.", + "best_pr_reason": "PR 43532 is the best global representative because it has the broadest, most illustrative change set and the clearest 'show diff' intent among the three.", "soft_edge_verdicts": [ { "left": "pull_request:43488", "right": "pull_request:46747", "accept": false, - "reason": "One is a setup.py/versioning and dependency test; the other changes a PR security-gate workflow trigger. Different subsystems and different intended behavior." + "reason": "Different concrete changes: 43488 tweaks setup/dependency/test files, while 46747 changes a security-gate workflow trigger. Same broad repo-maintenance theme, but not the same bug or change." }, { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "The shared Vit test file is incidental; 43488 is about packaging/version format, while 43532 changes self-comment CI/workflow plumbing and helper code." + "reason": "They both touch testing-related files, but 43488 is about setup/version/dependency and a Vit test, whereas 43532 mainly rewires comment/notification workflows and test utilities. Not one fix." }, { "left": "pull_request:43532", "right": "pull_request:46747", "accept": false, - "reason": "Both touch workflows, but they address different concrete paths: self-comment CI/diff display versus the PR security-gate trigger. Not the same change." + "reason": "43532 is a multi-file workflow/test utility demo, while 46747 is a single workflow security-gate trigger change. Overlap is only superficial at the workflow level." } ] }, @@ -21205,12 +21214,12 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "8052e85306e905de3f9e0a390ff002bf4042f1d28fee533dd116a51bd6e231d8", + "prepared_review_unit_hash": "e94c39db0b6341f74b5bfafb4f46a06c9eebabfaf86c6cb7c339bd1903f335a6", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6083, + "serialized_chars": 6082, "estimated_input_tokens": 1521, "estimated_eval_tokens": 3298 }, @@ -21218,7 +21227,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6083, + "serialized_chars": 6082, "estimated_input_tokens": 1521, "estimated_eval_tokens": 3298 }, @@ -21227,18 +21236,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related but distinct feature PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They share documentation and registry scaffolding, but they target different model/pipeline code paths and should not be deduplicated.", - "confidence": 0.97, + "summary": "Both PRs add new, similarly named segmentation pipelines, but they target different features and codepaths: promptable concept segmentation vs promptable visual segmentation. They share docs/registry scaffolding, yet the underlying implementations, tests, and model integrations are distinct, so they should not be merged as duplicates.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: the two PRs implement different segmentation pipelines with different filenames, model integrations, and tests.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR #43613 is the broader representative of this cluster because it covers the larger visual-segmentation pipeline integration, but it is not a duplicate of #43612.", + "best_pr_reason": "PR #43613 is the stronger representative of this cluster because it is broader and more integrated, covering multiple model test suites and a dedicated pipeline implementation; #43612 is a narrower, separate feature.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Different concrete features and code paths: concept segmentation vs visual segmentation. Shared scaffolding/docs does not make them the same change, and they would not plausibly be merged into one PR." + "reason": "Similar naming and shared docs/registry changes, but they implement different pipelines with different filenames, model integrations, and test targets. They are related feature additions, not the same concrete change or bug fix." } ] }, @@ -21268,7 +21277,7 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "39a3de1db386ee6b95c72fe9f8c2e0f04c9a293c514b7931f033564bcbe3e937", + "prepared_review_unit_hash": "f55c1cb5e5eb64505a908e61a242c4a1369b423c2a62d8e6ec92f140b48b3f83", "component_budget": { "node_count": 2, "item_count": 2, @@ -21290,18 +21299,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs both touch `trainer.py`, but they fix different Trainer behaviors: gradient-accumulation loss scaling vs signature-column filtering for custom/PEFT models. They are not duplicates and should not be merged together.", + "summary": "These two PRs both modify `trainer.py` and both mention custom trainer extensibility, but they address different problems: one changes gradient-accumulation loss scaling, the other refactors signature-column filtering. They are not the same underlying fix and should not be merged as duplicates.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: the two changes are orthogonal and address different Trainer code paths.", + "canonical_pr_reason": "PR 43651 is the more concrete bug fix: it changes `training_step` loss normalization for gradient accumulation and introduces `_loss_is_scaled_for_ga`, which is a specific behavior change rather than a broad refactor.", "best_issue_reason": null, - "best_pr_reason": "No single PR is a good global representative here; 43651 is about GA loss scaling, while 43663 is about column filtering, so neither subsumes the other.", + "best_pr_reason": "PR 43651 is the better representative of the cluster because it is a narrowly scoped correctness fix in the training path, whereas PR 43663 is a separate extensibility refactor for input-column filtering.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Different underlying changes in the same file: 43651 adjusts loss scaling during gradient accumulation, while 43663 refactors signature-column detection for input filtering. Shared filename alone is not enough." + "reason": "Different code paths and different behaviors: 43651 fixes loss scaling during gradient accumulation, while 43663 adds an overridable helper for signature-column filtering. Shared file/subsystem alone is not enough to treat them as the same change." } ] }, @@ -21331,7 +21340,7 @@ "soft_pairs": [ "pull_request:43656|pull_request:43842" ], - "prepared_review_unit_hash": "f965de3d9650d5d99f3f485d2c92569cddbf05c29ef2b95a243e01a412f2fd21", + "prepared_review_unit_hash": "3192996f63d910653cf95ac283b226f8ca5f1a01dfc34d1b1beedb79ab21c9ee", "component_budget": { "node_count": 2, "item_count": 2, @@ -21353,18 +21362,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs touch `src/transformers/cli/serve.py` and aim to fix `NameError`s in the CLI, but they target different failure modes: one is about forward-reference/typing issues around `TypeAdapter` and related annotations, while the other is specifically about `TypeAdapter` when `pydantic` is not installed. They are related, but not the same underlying change.", - "confidence": 0.92, + "summary": "Both PRs touch `src/transformers/cli/serve.py` and are about a TypeAdapter-related NameError in the CLI, but they use different fixes: PR 43656 is a broader postponed-annotations/type-hint cleanup, while PR 43842 specifically guards the optional `pydantic.TypeAdapter` import with `TYPE_CHECKING` and quotes the annotation. They are related, but not clearly the same concrete change.", + "confidence": 0.78, "canonical_issue_reason": null, - "canonical_pr_reason": "No true canonical PR stands out because these are distinct fixes; if one must be chosen as the cluster representative, `pull_request:43842` is slightly more specific and clearly scoped to the `TypeAdapter` import/annotation failure case.", + "canonical_pr_reason": "PR 43842 is the tighter, more explicit fix for the concrete failure mode: it directly addresses `TypeAdapter` being unavailable when `pydantic` is not installed. That makes it a better representative of the underlying bug than the broader annotation cleanup in PR 43656.", "best_issue_reason": null, - "best_pr_reason": "`pull_request:43842` is the better representative for the cluster theme because its title and diff clearly describe a concrete `TypeAdapter`-related `NameError` fix in the CLI, with a narrowly scoped change.", + "best_pr_reason": "PR 43842 is the best single PR in the cluster because it targets the specific import-time NameError path and is narrowly scoped to the optional dependency problem.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", "accept": false, - "reason": "Do not merge: despite the same file and similar symptom class, the diffs fix different bugs. `43656` adjusts future annotations and type hints for CLI serving objects, while `43842` adds a `TypeAdapter` import under `TYPE_CHECKING` and quotes the annotation to avoid a pydantic-related `NameError`. They are related but not the same code-path problem." + "reason": "They overlap in file and theme, but the fixes are not the same concrete change: 43656 applies a broader `from __future__ import annotations`/typing cleanup, while 43842 specifically makes `TypeAdapter` safe when `pydantic` is absent. They could coexist, but this is not strong enough to treat as a duplicate pair." } ] }, @@ -21511,8 +21520,8 @@ "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:44059|pull_request:44722", "pull_request:44017|pull_request:44161", "pull_request:44017|pull_request:44722", "pull_request:44017|pull_request:44073", @@ -21556,94 +21565,94 @@ "pull_request:44044|pull_request:44129", "pull_request:44044|pull_request:44154" ], - "prepared_review_unit_hash": "89cbe04309a55a3e7ec700734611a081c88477a82af8be3f85154de6b9de0e4b", + "prepared_review_unit_hash": "30001924bb5826d219991b7c9f09996b9981a2196cee95192a0fa263a7ebc8d8", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84899, - "estimated_input_tokens": 21225, - "estimated_eval_tokens": 42706 + "serialized_chars": 84884, + "estimated_input_tokens": 21221, + "estimated_eval_tokens": 42698 }, "budget": { "node_count": 27, "item_count": 27, "soft_pair_count": 72, - "serialized_chars": 75094, - "estimated_input_tokens": 18774, - "estimated_eval_tokens": 37804 + "serialized_chars": 75081, + "estimated_input_tokens": 18771, + "estimated_eval_tokens": 37798 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a family of similar output-tracing refactors across different model implementations, not a single duplicate bug. The only clear duplicate-like groups are the GPT-Neo pair and the GPT-J/CodeGen pair; the rest are same-theme but different code paths/models.", - "confidence": 0.86, + "summary": "This cluster is mostly a set of model-specific refactors to standardized output tracing. The only clear duplicate subclusters are GPT-J/CodeGen and GPT-Neo variants; the rest are similar but separate model PRs, not the same code change.", + "confidence": 0.78, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44722 is the strongest representative of the GPT-J/CodeGen output-tracing refactor: it touches the shared GPT-J/CodeGen code paths, has the broadest scope in that subcluster, and looks like the most complete version of the same change.", + "canonical_pr_reason": "No issue artifact to anchor on; PR 44722 is the broadest/latest representative of the GPT-J/CodeGen output-tracing refactor subcluster.", "best_issue_reason": null, - "best_pr_reason": "PR 44722 is the best single PR to anchor the duplicate triage because it most clearly captures the concrete GPT-J/CodeGen decorator migration, while the other PRs are separate model-specific siblings rather than true duplicates of that change.", + "best_pr_reason": "PR 44722 best represents the duplicate subcluster because it covers the same GPT-J code path as earlier GPT-J-only PRs and also includes the related CodeGen refactor.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Same broad output-tracing theme, but different model families and different code paths." + "reason": "Different model families (CVT/FNet vs GPT-J); same issue target is not enough to call these duplicates." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "Different models/files; only the standardized tracing pattern is shared." + "reason": "Unrelated model refactors (CVT/FNet vs DeBERTa); similar output-tracing theme only." }, { "left": "pull_request:44066", "right": "pull_request:44085", "accept": true, - "reason": "Both are GPT-J output-tracing refactors in the same file and could plausibly be merged into one change." + "reason": "Same GPT-J output-tracing change on the same file; 44066 is the broader version and 44085 is GPT-J-only." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "Different vision backbones; same refactor style is not enough to make them duplicates." + "reason": "Different model code paths (RegNet/ResNet vs EfficientNet); both are output-tracing refactors but not the same change." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "Different models and code paths; only the output-tracing pattern overlaps." + "reason": "EfficientNet is a separate model from GPT-J/CodeGen; shared refactor style only." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen versus EfficientNet; these are separate refactors." + "reason": "GPT-J/CodeGen vs EfficientNet are distinct model-specific changes." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 and DeBERTa-v2 are unrelated implementation paths." + "reason": "MobileNetV2 and DeBERTa are unrelated code paths." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "Different models and different output-capture implementations." + "reason": "GPT-J/CodeGen vs MGP-STR are unrelated model implementations." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "GPT-J/CodeGen refactor versus MPT refactor; not the same underlying change." + "reason": "GPT-J/CodeGen and MPT are different model refactors." }, { "left": "pull_request:44018", "right": "pull_request:44068", "accept": true, - "reason": "Same GPT-Neo file/model and same output-tracing migration, with overlapping test/file edits." + "reason": "Both are GPT-Neo output-tracing refactors on the same model/file and appear to target the same concrete change." }, { "left": "pull_request:44066", @@ -21653,183 +21662,207 @@ }, { "left": "pull_request:44018", - "right": "pull_request:44019", + "right": "pull_request:44066", "accept": false, - "reason": "Different architectures and unrelated forward implementations." + "reason": "GPT-Neo vs GPT-J/CodeGen are distinct model-specific PRs." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet and GPT-J/CodeGen are unrelated changes." + "reason": "ResNet is unrelated to GPT-J/CodeGen." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "MGP-STR and GPT-J/CodeGen are different models and different refactors." + "reason": "MGP-STR and GPT-J/CodeGen are unrelated model changes." }, { "left": "pull_request:44019", "right": "pull_request:44085", "accept": false, - "reason": "ResNet and GPT-J are different code paths." + "reason": "ResNet and GPT-J are different model refactors." }, { "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet do not share the same concrete code path." + }, + { + "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo versus MPT; same tracing pattern only." + "reason": "ResNet and MPT are separate model implementations." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo and GPT-J/CodeGen are separate model changes." + "reason": "GPT-Neo and GPT-J/CodeGen are different models; similar decorator migration only." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo and Flaubert are unrelated implementations." + "reason": "GPT-Neo and Flaubert are unrelated model code paths." }, { "left": "pull_request:44018", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo versus Flaubert; not the same bug/change." + "reason": "GPT-Neo and Flaubert are unrelated model code paths." }, { "left": "pull_request:44071", "right": "pull_request:44722", "accept": false, - "reason": "MPT and GPT-J/CodeGen are different code paths." + "reason": "MPT and GPT-J/CodeGen are distinct model changes." }, { "left": "pull_request:44018", "right": "pull_request:44085", "accept": false, - "reason": "GPT-Neo and GPT-J are different model refactors." + "reason": "GPT-Neo vs GPT-J are different model-specific refactors." }, { "left": "pull_request:44018", "right": "pull_request:44086", "accept": false, - "reason": "Different models and different output-tracing implementations." + "reason": "GPT-Neo and MGP-STR are unrelated." }, { "left": "pull_request:44068", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are unrelated code paths." + "reason": "GPT-Neo and MGP-STR are unrelated model code paths." }, { "left": "pull_request:44071", "right": "pull_request:44085", "accept": false, - "reason": "MPT and GPT-J are separate model implementations." + "reason": "MPT and GPT-J are unrelated model implementations." }, { "left": "pull_request:44085", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J and MGP-STR are different models." + "reason": "GPT-J and MGP-STR are different models; same issue target does not make them duplicates." }, { "left": "pull_request:44059", "right": "pull_request:44086", "accept": false, - "reason": "GPT-2 and MGP-STR are unrelated refactors." + "reason": "GPT2 and MGP-STR are unrelated model-specific changes." }, { "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "ViLT and Flaubert/XLM are distinct model code paths." + "reason": "ViLT and XLM/Flaubert are separate code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are different models." }, { "left": "pull_request:44059", "right": "pull_request:44722", "accept": false, - "reason": "GPT-2 and GPT-J/CodeGen are separate model changes." + "reason": "GPT2 and GPT-J/CodeGen are separate model refactors." }, { "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "SegFormer and LongT5 are unrelated model changes." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44101", + "accept": false, + "reason": "VisualBERT and XLM/Flaubert are different model code paths." + }, + { + "left": "pull_request:44059", "right": "pull_request:44071", "accept": false, - "reason": "Segformer and MPT are unrelated implementations." + "reason": "GPT2 and MPT are unrelated model refactors." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44101", + "accept": false, + "reason": "MPNet and XLM/Flaubert are different model implementations." }, { "left": "pull_request:43997", - "right": "pull_request:44073", + "right": "pull_request:44018", "accept": false, - "reason": "RegNet and VisualBERT are different architectures." + "reason": "RegNet/ResNet and GPT-Neo are unrelated." }, { "left": "pull_request:43997", - "right": "pull_request:44056", + "right": "pull_request:44161", "accept": false, - "reason": "RegNet and MPNet are separate refactors." + "reason": "RegNet/ResNet and LongT5 are unrelated model changes." }, { "left": "pull_request:44071", "right": "pull_request:44076", "accept": false, - "reason": "MPT and ImageGPT are unrelated model code paths." + "reason": "MPT and ImageGPT are different model code paths." }, { "left": "pull_request:44074", "right": "pull_request:44101", "accept": false, - "reason": "TextNet and Flaubert/XLM are different models." + "reason": "TextNet and XLM/Flaubert are unrelated model changes." }, { "left": "pull_request:44056", "right": "pull_request:44161", "accept": false, - "reason": "MPNet and LongT5 are unrelated implementations." + "reason": "MPNet and LongT5 are separate model refactors." }, { "left": "pull_request:43997", - "right": "pull_request:44018", + "right": "pull_request:44017", "accept": false, - "reason": "RegNet and GPT-Neo are different model families." + "reason": "RegNet/ResNet and SegFormer are unrelated." }, { "left": "pull_request:43997", - "right": "pull_request:44161", + "right": "pull_request:44056", "accept": false, - "reason": "RegNet and LongT5 are unrelated changes." + "reason": "RegNet/ResNet and MPNet are distinct code paths." }, { - "left": "pull_request:44074", - "right": "pull_request:44161", + "left": "pull_request:43996", + "right": "pull_request:44017", "accept": false, - "reason": "TextNet and LongT5 are different code paths." + "reason": "CVT/FNet and SegFormer are different models." }, { - "left": "pull_request:44018", + "left": "pull_request:43996", "right": "pull_request:44161", "accept": false, - "reason": "GPT-Neo and LongT5 are unrelated model refactors." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44071", - "accept": false, - "reason": "RegNet and MPT are separate backbones." + "reason": "CVT/FNet and LongT5 are unrelated changes." }, { "left": "pull_request:44010", "right": "pull_request:44071", "accept": false, - "reason": "SqueezeBert and MPT are unrelated implementations." + "reason": "SqueezeBERT and MPT are unrelated model implementations." }, { "left": "pull_request:44010", "right": "pull_request:44073", "accept": false, - "reason": "SqueezeBert and VisualBERT are different code paths." + "reason": "SqueezeBERT and VisualBERT are different code paths." }, { "left": "pull_request:44018", @@ -21841,85 +21874,61 @@ "left": "pull_request:43997", "right": "pull_request:44074", "accept": false, - "reason": "RegNet and TextNet are different architectures." + "reason": "RegNet/ResNet and TextNet are separate model changes." }, { "left": "pull_request:44010", "right": "pull_request:44018", "accept": false, - "reason": "SqueezeBert and GPT-Neo are unrelated implementations." + "reason": "SqueezeBERT and GPT-Neo are unrelated model code paths." }, { "left": "pull_request:44071", "right": "pull_request:44129", "accept": false, - "reason": "MPT and SpeechT5 are different model code paths." - }, - { - "left": "pull_request:43996", - "right": "pull_request:44129", - "accept": false, - "reason": "FNet/CVT and SpeechT5 are unrelated refactors." + "reason": "MPT and SpeechT5 are different model implementations." }, { - "left": "pull_request:43996", - "right": "pull_request:44017", + "left": "pull_request:43997", + "right": "pull_request:44071", "accept": false, - "reason": "FNet/CVT and Segformer are different models." + "reason": "RegNet/ResNet and MPT are unrelated." }, { "left": "pull_request:44010", "right": "pull_request:44044", "accept": false, - "reason": "SqueezeBert and DeBERTa-v2 are separate implementations." + "reason": "SqueezeBERT and DeBERTa are different model refactors." }, { "left": "pull_request:44044", "right": "pull_request:44074", "accept": false, - "reason": "DeBERTa-v2 and TextNet are unrelated changes." - }, - { - "left": "pull_request:44066", - "right": "pull_request:44722", - "accept": true, - "reason": "Same GPT-J/CodeGen output-tracing migration with overlapping files; one is a broader version of the other." - }, - { - "left": "pull_request:44085", - "right": "pull_request:44722", - "accept": true, - "reason": "Same GPT-J output-tracing refactor in the same file; they could plausibly be merged into one PR." - }, - { - "left": "pull_request:44085", - "right": "pull_request:44129", - "accept": false, - "reason": "GPT-J and SpeechT5 are unrelated model implementations." + "reason": "DeBERTa and TextNet are unrelated code paths." }, { - "left": "pull_request:43996", - "right": "pull_request:44007", + "left": "pull_request:44017", + "right": "pull_request:44071", "accept": false, - "reason": "FNet/CVT and RegNet/ResNet are separate model refactors." + "reason": "SegFormer and MPT are unrelated model changes." }, { - "left": "pull_request:43996", - "right": "pull_request:44129", + "left": "pull_request:44073", + "right": "pull_request:44161", "accept": false, - "reason": "Different model families and different code paths." + "reason": "VisualBERT and LongT5 are different models." }, { - "left": "pull_request:44044", - "right": "pull_request:44129", + "left": "pull_request:44059", + "right": "pull_request:44068", "accept": false, - "reason": "DeBERTa-v2 and SpeechT5 are unrelated changes." + "reason": "GPT2 and GPT-Neo are different model refactors." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: it correctly treats this as a broader output-tracing refactor cluster, and it only elevates the GPT-Neo and GPT-J/CodeGen groups as duplicate-like. The accepted soft-edge pairs are conservative and look mergeable into single bugfix PRs within the same model code paths. No evidence is overstated in a way that would require rejection." + "feedback": "The summary stays within the packet: it identifies the GPT-J/CodeGen and GPT-Neo clusters as the only clearly duplicate-looking groups and treats the rest conservatively as separate model-specific refactors. The two accepted soft edges are plausibly mergeable subset/broader pairs on the same code path, and the rejected pairs are appropriately cautious." }, "evaluator_used": true, "retried": false, @@ -21957,70 +21966,70 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "f9ba1b210536093a3962133e5d219873fc3b9a54030a027a3d171e55a8a88783", + "prepared_review_unit_hash": "fdb40461182ac596bcdadd2ed17d2c61520a08c418b7beea7845c0a270658962", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84899, - "estimated_input_tokens": 21225, - "estimated_eval_tokens": 42706 + "serialized_chars": 84884, + "estimated_input_tokens": 21221, + "estimated_eval_tokens": 42698 }, "budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 6, - "serialized_chars": 17126, - "estimated_input_tokens": 4282, - "estimated_eval_tokens": 8820 + "serialized_chars": 17124, + "estimated_input_tokens": 4281, + "estimated_eval_tokens": 8818 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a set of separate, model-specific PRs that all migrate different transformers models to standardized output tracing/capture helpers. They share a common refactor pattern, but they are not duplicates of the same concrete code-path change.", - "confidence": 0.94, + "summary": "These PRs are all model-specific refactors around standardized output tracing/capture_outputs, but they touch different architectures and code paths. They are related by theme, not duplicates of the same change.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "pull_request:44073 is the clearest representative of the cluster\u2019s shared theme: it explicitly refactors one model to use can_return_tuple/capture_outputs and adds output-recording metadata, making it the most canonical example of the standardized output-tracing migration.", + "canonical_pr_reason": "PR 44073 is a good representative canonical PR: it clearly shows the standardized output-tracing migration pattern, includes the modern decorators/imports, and is scoped to a single model implementation.", "best_issue_reason": null, - "best_pr_reason": "pull_request:44073 is the best single representative PR because its title and diff are explicit about the standardized output-tracing refactor, and its implementation pattern matches the rest of the cluster closely.", + "best_pr_reason": "PR 44073 is the best exemplar in the set because its title and diff directly express the core refactor pattern (capture_outputs + can_return_tuple) without conflating it with unrelated model changes.", "soft_edge_verdicts": [ { "left": "pull_request:44024", "right": "pull_request:44076", "accept": false, - "reason": "Both are output-tracing migrations, but for different models (FocalNet vs ImageGPT) and different code paths. Similar refactor pattern is not enough to treat them as the same change." + "reason": "Both are output-tracing refactors, but for different models (FocalNet vs ImageGPT) and different internals. They are not the same bug/change and would not be merged as one PR." }, { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "VisualBert and TextNet are separate model implementations with distinct forward paths; these are parallel refactors, not the same underlying bug or change." + "reason": "VisualBert and TextNet are separate model implementations. The changes are similar refactors, but not the same underlying code-path problem." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "Both update standardized output tracing, but in different models (MPNet vs VisualBert). They cannot plausibly be merged as one concrete fix." + "reason": "Both modernize output recording, but MPNet and VisualBert are unrelated model code paths. Similar pattern, not duplicate change." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "MPNet and TextNet touch different model code paths; the similarity is only the shared migration pattern, so this is not a duplicate pair." + "reason": "MPNet vs TextNet are distinct model refactors with no shared concrete bug or patch to combine." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are unrelated model-specific refactors. They share helper imports/output-capture migration, but not the same code-path problem." + "reason": "RegNet and SqueezeBert are different architectures; these are separate standardized-output migrations, not one duplicate fix." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "SqueezeBert and SpeechT5 are different model implementations with separate output-handling changes; this is not the same concrete PR-worthy fix." + "reason": "SqueezeBert and SpeechT5 touch different model implementations and the SpeechT5 PR is tied to a specific issue. Same broad theme, but not the same underlying change." } ] }, @@ -22100,12 +22109,12 @@ "pull_request:44028|pull_request:44030", "pull_request:44029|pull_request:44030" ], - "prepared_review_unit_hash": "a925d17a4907cb3e10516a095e6278f028a6f2bf9c856563bae134e66817aa07", + "prepared_review_unit_hash": "af1df07749f8b683c6f32f8baf6741f585fffede22b4c73a4c18ff5906f734c6", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40173, + "serialized_chars": 40174, "estimated_input_tokens": 10044, "estimated_eval_tokens": 20344 }, @@ -22113,7 +22122,7 @@ "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40173, + "serialized_chars": 40174, "estimated_input_tokens": 10044, "estimated_eval_tokens": 20344 }, @@ -22122,252 +22131,252 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is a batch of similarly titled but model-specific output-tracing refactors. They share a pattern (adding can_return_tuple/capture_outputs or adjusting forward signatures), but each PR targets a different model family and code path, so they should not be deduplicated together.", - "confidence": 0.88, + "summary": "This cluster is a set of similar-looking but mostly unrelated refactor PRs across different model families, all centered on output tracing / `can_return_tuple` / `capture_outputs`. They share theme, not the same underlying change, so they should stay separate.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "44025 is the best representative of the cluster\u2019s shared refactor style: it applies the output-tracing change across two closely related depth-anything models and clearly shows the intended decorator-based pattern.", + "canonical_pr_reason": "PR #44003 is the strongest representative of the cluster: it is the broadest single change here, spanning closely related Mamba/FalconMamba code and adding the output-capture machinery used by the refactor.", "best_issue_reason": null, - "best_pr_reason": "44025 is the strongest single representative PR because it captures the common refactor pattern cleanly without being as narrow as the one-file changes or as cross-family as the broader model refactors.", + "best_pr_reason": "PR #44003 best fits the cluster\u2019s overall theme and has the most substantial, cross-file implementation among the candidates.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Different model families and forward paths (timm_backbone vs upernet); same tracing-refactor theme only." + "reason": "Different model families (`timm_backbone` vs `upernet`); same refactor theme, but not the same bug or mergeable change." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "Unrelated architectures (timm_backbone vs vision_text_dual_encoder); not the same underlying code change." + "reason": "Different code paths and models (`timm_backbone` vs `vision_text_dual_encoder`); only a broad output-tracing refactor in common." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "timm_backbone and univnet are separate code paths; these are analogous refactors, not duplicates." + "reason": "Unrelated model implementations (`timm_backbone` vs `univnet`); no shared concrete fix." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "Different subsystems (timm_backbone vs mamba/falcon_mamba); cannot plausibly merge into one fix." + "reason": "Different architectures and files; both are output-refactor work, but not the same underlying change." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "timm_backbone and codegen are unrelated models; only the refactor style matches." + "reason": "`timm_backbone` and `codegen` are unrelated code paths; similar refactoring pattern only." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families (`timm_backbone` vs `rwkv`); no evidence of one shared fix." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "MobileNetV1 and UnivNet have distinct forward logic and outputs; not the same bug/change." + "reason": "`mobilenet_v1` and `univnet` are unrelated; they only share the generic output-tracing refactor motif." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "MobileNetV1 vs UperNet are unrelated model implementations; same output-tracing pattern only." + "reason": "Different subsystems and files (`mobilenet_v1` vs `upernet`); not the same concrete change." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "MobileNetV1 and depth_anything target different architectures and files; not a duplicate PR." + "reason": "`mobilenet_v1` and `depth_anything` are separate model families; same style of refactor, not duplicate work." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "MobileNetV1 and vision_encoder_decoder are separate code paths; no shared concrete fix." + "reason": "`mobilenet_v1` vs `vision_encoder_decoder`; no shared code path or bug." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "MobileNetV1 and speech_encoder_decoder are distinct model stacks; only broadly similar refactoring." + "reason": "`mobilenet_v1` and `speech_encoder_decoder` are different models; only superficial similarity." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "vision_text_dual_encoder and univnet are different model families; not the same underlying issue." + "reason": "`vision_text_dual_encoder` and `univnet` are unrelated implementations; same refactor pattern is not enough." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "vision_text_dual_encoder and upernet touch different forward implementations; same theme, not same change." + "reason": "Different model families (`vision_text_dual_encoder` vs `upernet`); no shared concrete fix." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "vision_text_dual_encoder vs mamba/falcon_mamba are unrelated code paths; cannot be one PR." + "reason": "Different code paths and architectures; these are separate refactors, not one mergeable PR." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "Different model families (vision_text_dual_encoder vs depth_anything); only the output-tracing refactor matches." + "reason": "`vision_text_dual_encoder` and `depth_anything` do not share the same underlying bug/change." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "UnivNet and UperNet are unrelated models with separate forward semantics." + "reason": "`univnet` vs `upernet` are unrelated model code paths; only the generic output-tracing theme matches." }, { "left": "pull_request:44001", "right": "pull_request:44003", "accept": false, - "reason": "UnivNet does not share the mamba/falcon_mamba code path; these are not duplicates." + "reason": "Different model families (`univnet` vs `mamba`); not the same fix." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "UnivNet and depth_anything are separate model implementations; same refactor motif only." + "reason": "`univnet` and `depth_anything` are separate changes in different model stacks." }, { "left": "pull_request:44002", "right": "pull_request:44029", "accept": false, - "reason": "UperNet and RWKV are different architectures and output plumbing; not mergeable as one fix." + "reason": "`upernet` and `rwkv` are unrelated code paths; no evidence of the same underlying issue." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "UperNet and mamba/falcon_mamba affect different layers of the stack; no shared concrete bug." + "reason": "Different architectures and files; similar refactor intent only." }, { "left": "pull_request:44003", "right": "pull_request:44028", "accept": false, - "reason": "Falcon/Mamba internals and SuperPoint are unrelated code paths; same generic tracing cleanup only." + "reason": "`mamba`/`falcon_mamba` vs `superpoint` are different model families; not one shared bug or change." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "mamba/falcon_mamba and codegen are unrelated model implementations; not the same change." + "reason": "`mamba` and `codegen` touch different model implementations; same refactor category, not a duplicate." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "timm_backbone and codegen are separate model families; no common code-path fix." + "reason": "Different code paths (`timm_backbone` vs `codegen`); the overlap is only in output-tracing cleanup." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "UnivNet and CodeGen are unrelated; these PRs share only a refactor theme." + "reason": "`univnet` and `codegen` are unrelated models; not mergeable as one fix." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "UperNet and CodeGen have distinct forward logic and outputs; not duplicates." - }, - { - "left": "pull_request:44000", - "right": "pull_request:44004", - "accept": false, - "reason": "vision_text_dual_encoder and codegen are different models; same general tracing refactor only." + "reason": "`upernet` and `codegen` do not address the same underlying issue." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "depth_anything and vision_encoder_decoder are unrelated model families; not the same underlying change." + "reason": "`depth_anything` and `vision_encoder_decoder` are different model stacks; broad theme match only." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "depth_anything and speech_encoder_decoder target different code paths; not duplicates." + "reason": "`depth_anything` and `speech_encoder_decoder` are separate refactors, not the same bug fix." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "depth_anything and superpoint are separate model implementations; only the refactor style overlaps." + "reason": "`depth_anything` vs `superpoint`; different files and model logic." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "depth_anything and RWKV are unrelated; no shared concrete bug or fix." + "reason": "`depth_anything` and `rwkv` are unrelated implementations." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "depth_anything and DPR have different forward signatures and output handling; not the same change." + "reason": "`depth_anything` and `dpr` do not share a concrete code-path problem." }, { "left": "pull_request:44026", - "right": "pull_request:44029", + "right": "pull_request:44027", "accept": false, - "reason": "vision_encoder_decoder and RWKV are separate model stacks; not a duplicate." + "reason": "Encoder-decoder wrappers for different modalities; same style of cleanup, not the same change." }, { "left": "pull_request:44026", - "right": "pull_request:44027", + "right": "pull_request:44028", "accept": false, - "reason": "vision_encoder_decoder and speech_encoder_decoder are different encoder-decoder wrappers; not the same PR." + "reason": "`vision_encoder_decoder` and `superpoint` are different subsystems; no duplicate evidence." }, { "left": "pull_request:44026", - "right": "pull_request:44028", + "right": "pull_request:44029", "accept": false, - "reason": "vision_encoder_decoder and superpoint are unrelated model code paths; only broadly similar cleanup." + "reason": "`vision_encoder_decoder` vs `rwkv` are unrelated model implementations." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "vision_encoder_decoder and DPR are distinct implementations; no shared concrete fix." + "reason": "`vision_encoder_decoder` and `dpr` are separate refactors in different code paths." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "speech_encoder_decoder and superpoint are unrelated models; this is just the same output-tracing pattern." + "reason": "`speech_encoder_decoder` and `superpoint` are unrelated model families; not one underlying fix." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "speech_encoder_decoder and RWKV are different architectures and forward paths." + "reason": "`speech_encoder_decoder` vs `rwkv`; same output-tracing motif only." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "speech_encoder_decoder and DPR do not share the same code path; not mergeable as one PR." + "reason": "`speech_encoder_decoder` and `dpr` do not appear to fix the same code-path issue." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "SuperPoint and RWKV are unrelated model families; same refactor theme only." + "reason": "`superpoint` and `rwkv` are separate model implementations; no shared concrete bug." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "SuperPoint and DPR are separate implementations; not duplicates." + "reason": "`superpoint` and `dpr` touch different subsystems; not duplicates." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "RWKV and DPR are different model stacks with different output semantics; not the same underlying fix." + "reason": "`rwkv` and `dpr` are unrelated model families; only broad refactor similarity." } ] }, @@ -22397,12 +22406,12 @@ "soft_pairs": [ "pull_request:44662|pull_request:45401" ], - "prepared_review_unit_hash": "ee127a0fcab052569a9d749e9a04bb6371e93750063fafb9348a758f3473f4a5", + "prepared_review_unit_hash": "d05a02297d867ef2099f202c2917cf70d822d346ef43a6afe813753edbd712aa", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6551, + "serialized_chars": 6552, "estimated_input_tokens": 1638, "estimated_eval_tokens": 3532 }, @@ -22410,7 +22419,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6551, + "serialized_chars": 6552, "estimated_input_tokens": 1638, "estimated_eval_tokens": 3532 }, @@ -22419,18 +22428,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated model-addition PRs. Both touch shared Transformers registration/docs scaffolding, but PR #44662 implements PenguinVL while PR #45401 adds Voxtral TTS support, so they are not the same change and should not be merged as duplicates.", - "confidence": 0.98, + "summary": "Two unrelated model-addition PRs: PenguinVL and Voxtral TTS. They overlap mainly in transformer registry/docs scaffolding, but they implement different models and code paths.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: #44662 and #45401 implement different models and different code paths, despite overlapping boilerplate files.", + "canonical_pr_reason": "PR 44662 is the better representative PR here: it is non-draft, has extensive review activity, and contains the more complete model integration.", "best_issue_reason": null, - "best_pr_reason": "No single best PR for the cluster because the PRs are not duplicates; if forced, #44662 is the more mature implementation PR, while #45401 is draft and issue-linked but still unrelated.", + "best_pr_reason": "PR 44662 is the strongest standalone PR in the cluster because it is a fully fleshed-out implementation with substantial review history, whereas 45401 is still a draft tied to a specific issue.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different model implementations (PenguinVL vs Voxtral TTS) with only shared auto-registration/docs boilerplate; they do not fix the same concrete bug or add the same feature." + "reason": "Reject: these add different models (PenguinVL vs Voxtral TTS) with different architectures and tasks. Shared registry/docs changes are incidental and not the same underlying bug or change." } ] }, @@ -22460,12 +22469,12 @@ "soft_pairs": [ "pull_request:44979|pull_request:45363" ], - "prepared_review_unit_hash": "11b3df9352ca8d23ad61e782d2cd7a9a38bfb333c6bb10b0ac37e9fb04844105", + "prepared_review_unit_hash": "c857ad2b9689ac484e634d46ec50b70bb32fe6e80694aeb8c2ff2808bf738625", "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 }, @@ -22473,7 +22482,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4408, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -22482,18 +22491,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two PRs are related by a shared fusion theme, but they implement different layers of functionality: one introduces a generic module-fusion API, while the other wires up KernelConfig-based n-to-1 kernel fusion and integrations. They are not duplicates.", - "confidence": 0.89, + "summary": "These are related feature PRs around module/kernel fusion, but they target different layers and code paths: one introduces a generic Module Fusion API, the other implements n-to-1 kernel fusion via `KernelConfig` and integration plumbing. I do not treat them as duplicates.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44979 is the better representative of the cluster theme because it introduces the reusable module_fusion infrastructure and tests; PR 45363 is a separate feature built around KernelConfig and additional integration paths.", + "canonical_pr_reason": "No single canonical PR stands out as a duplicate target; the PRs are adjacent but distinct feature work rather than the same change.", "best_issue_reason": null, - "best_pr_reason": "PR 44979 is the best single representative of the shared fusion work because it defines the foundational API surface, but it should not be treated as covering PR 45363's distinct kernel-fusion implementation.", + "best_pr_reason": "PR 45363 is the more concrete end-user feature implementation, but it is still separate from the generic API work in PR 44979.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "Related only at a high level. 44979 adds a generic module-fusion API in src/transformers/module_fusion.py, while 45363 adds KernelConfig-driven n-to-1 kernel fusion across multiple integration files and an example. Different concrete code paths and merge scopes." + "reason": "Related fusion efforts, but not the same underlying change: 44979 adds a generic module fusion API, while 45363 wires in n-to-1 kernel fusion through `KernelConfig` and integration/modeling changes. They touch different abstractions and would not plausibly be merged into one PR." } ] }, @@ -22523,7 +22532,7 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "51cbb04d2481383a54d8b2756158de36907fd3b673850336f903ca26e146d5a5", + "prepared_review_unit_hash": "d48f8ba93b0b37e26d1f1a88357dd2c8808eb2673912f80351586fb25ab62150", "component_budget": { "node_count": 2, "item_count": 2, @@ -22545,18 +22554,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests patch the same Mistral regex AttributeError in the same file against the same tracking issue, with the later PR adding tests and a slightly broader implementation cleanup. They look like duplicate attempts at the same fix.", - "confidence": 0.97, + "summary": "Both pull requests target the same Mistral regex AttributeError in `tokenization_utils_tokenizers.py` and make the same code-path fix; `45317` is the more complete version because it adds regression tests.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "pull_request:45317 is the better canonical PR because it is the more complete variant: same bug fix, same target issue, plus test coverage and a clearer final title.", + "canonical_pr_reason": "`45317` is the stronger representative: it fixes the same `_patch_mistral_regex` AttributeError as `45086` but also includes a regression test, making it the more complete patch for the same underlying bug.", "best_issue_reason": null, - "best_pr_reason": "pull_request:45317 is the best overall representative since it covers the same code-path fix as 45086 while also including tests, making it the more merge-ready duplicate candidate.", + "best_pr_reason": "`45317` is the best overall candidate because it addresses the same concrete bug, keeps the same fix, and adds tests that validate the behavior change.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Accept: both PRs fix the same AttributeError in _patch_mistral_regex in src/transformers/tokenization_utils_tokenizers.py, target the same issue, and could plausibly be merged as one change; 45317 is the more complete version." + "reason": "Same underlying bug and same code-path fix in `_patch_mistral_regex` (`backend_tokenizer.pre_tokenizer` -> `tokenizer.pre_tokenizer`); the second PR is just a more complete version with tests, so they are plausibly mergeable into one change." } ] }, @@ -22588,7 +22597,7 @@ "pull_request:45115|pull_request:45569", "pull_request:45569|pull_request:46565" ], - "prepared_review_unit_hash": "d64056e3a1ac03b4a2f3cb41866b57cc44e62776abe0d06efedc446e71d509ed", + "prepared_review_unit_hash": "bc2adbb4b9608ca4346a28ae079cf797a3436322dd2778cb0d5aeabf7169a3ed", "component_budget": { "node_count": 3, "item_count": 3, @@ -22610,24 +22619,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two PRs are NemotronH-family refactors, but they target different concrete changes; the Nemotron3.5 ASR PR is unrelated. I would keep PR 45569 as the representative NemotronH architecture-split change.", - "confidence": 0.84, + "summary": "All three items are pull requests in the Nemotron family, but they target different changes: one is a conversion-mapping/refactor PR, one is a larger NemotronH split/architecture PR, and one is an unrelated Nemotron3.5 ASR addition. No soft pair looks like the same concrete change.", + "confidence": 0.86, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45569 is the broadest and most central NemotronH change here: it introduces the dedicated dense/sparse model split and backward-compatible dispatcher, which is the clearest cluster anchor.", + "canonical_pr_reason": "PR 45569 is the most central and substantial NemotronH implementation/refactor item in the set, with the broadest code and docs/test coverage.", "best_issue_reason": null, - "best_pr_reason": "PR 45569 is the best standalone PR in the set because it covers the main architecture refactor for NemotronH rather than only narrower conversion-mapping work.", + "best_pr_reason": "PR 45569 is the best representative of the cluster because it covers the main NemotronH rework, while 45115 is narrower conversion-mapping work and 46565 is a different model family entirely.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "Related by model family and some shared files, but they are not the same concrete change: 45115 focuses on conversion-mapping and inheritance adjustments, while 45569 performs the larger NemotronH dense/sparse split and dispatcher refactor. They could not cleanly be merged as one PR without losing scope clarity." + "reason": "Related NemotronH work, but not the same underlying change: 45115 is a conversion-mapping and inheritance refactor, while 45569 is a broader model split/addition with docs and tests." }, { "left": "pull_request:45569", "right": "pull_request:46565", "accept": false, - "reason": "Different model family and different feature area: NemotronH refactor versus Nemotron3.5 ASR streaming. No shared underlying bug or change." + "reason": "Different model families and code paths: NemotronH architecture work versus Nemotron3.5 ASR streaming. They are not mergeable as one duplicate PR." } ] }, @@ -22657,7 +22666,7 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "fbf0d1ca51385a2604ecc18eafc372fdc950b01bd0423d06b9fa86c05c631186", + "prepared_review_unit_hash": "638bf2b526d636f0827c07b1c1fab086b014bdc8a478fcf1f3f499d7e184b1a8", "component_budget": { "node_count": 2, "item_count": 2, @@ -22679,18 +22688,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs in the same new heterogeneous-modeling feature area, but they appear to be different layers of the implementation: one adds per-layer config support, the other adds broader modeling/cache/masking integration on top of that foundation.", - "confidence": 0.89, + "summary": "These are related but not duplicates: one PR introduces per-layer heterogeneous config support, while the other adds broader heterogeneous model/runtime changes that build on that foundation.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45332 is the broader implementation and touches the core runtime paths (cache, masking, modeling, heterogeneity helpers), so it is the better representative for the cluster.", + "canonical_pr_reason": "PR #45333 is the cleaner canonical representative because it focuses on the foundational per-layer configuration layer; #45332 is a broader follow-on that includes modeling/cache integration.", "best_issue_reason": null, - "best_pr_reason": "PR #45332 is the best single PR to keep because it covers the main end-to-end heterogeneous model support rather than only the config layer.", + "best_pr_reason": "PR #45333 is the best single representative for the cluster if one must be chosen, since it is the more narrowly scoped core change and is likely prerequisite infrastructure for the larger modeling PR.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "They share the heterogeneous feature theme and some config helpers, but #45333 is specifically per-layer configuration while #45332 is the larger modeling/runtime integration. They are related dependencies, not the same concrete change that should be merged as duplicates." + "reason": "They are adjacent pieces of the heterogeneous-model effort, but not the same concrete change: #45333 adds per-layer config plumbing, while #45332 adds modeling/cache support on top of that. They could be separate, sequential PRs rather than one duplicate PR." } ] }, @@ -22720,40 +22729,40 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "6dfe5d5bbd430531e9eff7ac3d18509ca321a328a0194f013dd8fe5f8ddf0df2", + "prepared_review_unit_hash": "f6cf2316e6a74ed0ecf8ada272a4f03b07835d93e4b8e48f4bea45f81c01d508", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5512, - "estimated_input_tokens": 1378, - "estimated_eval_tokens": 3012 + "serialized_chars": 5514, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5512, - "estimated_input_tokens": 1378, - "estimated_eval_tokens": 3012 + "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 open PRs in the same general typing-cleanup area, but they appear to be different changes: one is a broad repo-wide type-checking sweep, the other is a narrower typing support update around modeling_utils and related protocols.", - "confidence": 0.88, + "summary": "These are both typing-related PRs, but they address different work: one is a broad type-checking sweep across many transformer modules, while the other adds specific typing protocols/helpers and touches a smaller set of modeling/PEFT-related code paths. They should not be treated as duplicates.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45415 is the broader umbrella change: it updates type-checking config and applies type-checking fixes across many transformers modules, so it is the most representative PR in this cluster.", + "canonical_pr_reason": "No single canonical PR: 45415 is a broad type-checking cleanup across many files, while 45425 is a separate typing-helper change focused on protocol definitions and related model utilities.", "best_issue_reason": null, - "best_pr_reason": "PR 45415 is the best global representative because it covers the main repo-wide typing effort, while PR 45425 is a narrower follow-up focused on specific typing helpers and modeling_utils integration.", + "best_pr_reason": "PR 45415 is the better cluster representative because it is the broader umbrella typing sweep and subsumes more of the generic type-checking cleanup work.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Both are typing-related, but they address different scopes: 45415 is a broad type-checking pass with config changes across many files, while 45425 adds specific typing protocols/aliases and touches a smaller set of modeling-related files. They are related follow-ups, not the same concrete change." + "reason": "Both are about typing, but they fix different concrete changes: 45415 is a wide annotation/type-checking pass, while 45425 adds typing protocol aliases and helper interfaces for specific modeling/PEFT code paths. They would not plausibly merge into one PR as the same underlying change." } ] }, @@ -22783,7 +22792,7 @@ "soft_pairs": [ "pull_request:45513|pull_request:46084" ], - "prepared_review_unit_hash": "08c481ee85c720168404b70aee08de39ac62b3e6e244d24948d8764ab82abbf8", + "prepared_review_unit_hash": "e4614d0b70699a9da9507774ba28318ba97497761e4d43e082b5c8ce9536c273", "component_budget": { "node_count": 2, "item_count": 2, @@ -22805,18 +22814,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: one PR fixes multi-token cached forward behavior in Qwen3.5/Qwen3 Next/Olmo Hybrid linear attention, while the other fixes Mamba2 use_cache when seq_len > 1. They target different model families and different code paths, so they should stay separate.", - "confidence": 0.97, + "summary": "These are two unrelated pull requests about cached/multi-token forward behavior in different model families: one for Qwen3.5/OLMo hybrid/Qwen3-next linear attention, and one for Mamba2 use_cache with seq_len > 1. They do not look like the same underlying bug or a PR that could plausibly be merged with the other.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "PR #46084 is the cleaner canonical representative: it has an explicit issue target (#46032), a narrowly scoped Mamba2 fix, and a focused code-path change around cached forward with seq_len > 1.", "best_issue_reason": null, - "best_pr_reason": "PR #46084 is the clearest standalone fix: it has an explicit issue target, a focused title, and a concrete seq_len > 1 cache bug in one code path.", + "best_pr_reason": "PR #46084 is the strongest standalone representative because it is tightly scoped, explicitly tied to a specific bug report, and its diff clearly addresses one concrete cached-forward edge case.", "soft_edge_verdicts": [ { "left": "pull_request:45513", "right": "pull_request:46084", "accept": false, - "reason": "Different model families and different bugs: Qwen/Olmo cached multi-token forward vs Mamba2 use_cache with seq_len > 1. Not the same underlying change." + "reason": "Different model families and different bugs: #45513 fixes multi-token cached forward for GDN linear attention in Qwen3.5/OLMo/Qwen3-next, while #46084 fixes Mamba2 use_cache behavior when seq_len > 1. They are not the same underlying change." } ] }, @@ -22846,7 +22855,7 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "74c93d7845a9c6110376730492d0b1f6f0e41ed0b448ddda59ed821283ef277f", + "prepared_review_unit_hash": "b6a70245f6ca0e42e45dc632971ec2874053148b579ebfc89f39f31f1eba1291", "component_budget": { "node_count": 2, "item_count": 2, @@ -22868,18 +22877,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs concern expert-parallel loading, but they fix different bugs on different code paths: #45548 targets EP + DeepSpeed ZeRO-3 / accelerate launch config serialization and a `has_ep` check, while #45662 fixes EP + FSDP2 expert parameters being overwritten during distributed loading/broadcast. They are related, but not duplicates.", - "confidence": 0.93, + "summary": "Both PRs touch expert-parallel loading, but they fix different bugs: one removes `distributed_config` from serialization and adds `has_ep` for Accelerate/ZeRO-3 loading, while the other prevents expert parameters from being overwritten during FSDP2/rank-0 broadcast by post-wrapping sharded tensors. Related area, not the same change.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "#45662 is the better canonical PR for this cluster because it addresses the more concrete loading-path corruption bug and includes the broader set of fixes around distributed parameter materialization.", + "canonical_pr_reason": "PR #45662 is the better representative because it targets a concrete distributed-loading overwrite bug with broader code-path coverage, whereas #45548 is a narrower serialization/detection fix.", "best_issue_reason": null, - "best_pr_reason": "#45662 is the strongest representative PR: it fixes a specific, reproducible expert-parallel parameter overwrite in the core loading path and is broader than the config-only change in #45548.", + "best_pr_reason": "PR #45662 is the strongest standalone PR in this cluster: it addresses a specific expert-parameter corruption path during distributed loading and includes the main fix in the loading stack.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "Different underlying failures: #45548 is about EP config serialization / ZeRO-3 loading via accelerate, while #45662 prevents expert weights from being overwritten during FSDP2-style distributed loading. Related subsystem, but not the same bug or a single mergeable PR." + "reason": "Do not merge: they address different underlying failures. #45548 fixes EP/ZeRO-3 loading via Accelerate by excluding `distributed_config` from serialization and exposing `has_ep`; #45662 fixes expert parameters being overwritten after sharding/broadcast in FSDP2. Same subsystem, different bug and code path." } ] }, @@ -22909,12 +22918,12 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "2327802cf187c2aff5a067845d45bd0a56caec04f535c0d226054944dec90257", + "prepared_review_unit_hash": "92c690e818ecabaffa04db1f810812356454f46b0948ee84d348d87b90d07bb6", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4590, + "serialized_chars": 4589, "estimated_input_tokens": 1148, "estimated_eval_tokens": 2552 }, @@ -22922,7 +22931,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4590, + "serialized_chars": 4589, "estimated_input_tokens": 1148, "estimated_eval_tokens": 2552 }, @@ -22931,18 +22940,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs address the same xdist-related captured_info/debug-log collision problem in testing_utils; PR 45645 is the broader, more complete fix because it also updates CI workflow handling of multiple artifact files.", - "confidence": 0.86, + "summary": "Two open PRs address the same xdist collision around captured_info test logs; one is a narrow fix in testing_utils, the other extends it with CI handling for multiple log files.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45645 is the best canonical PR because it covers the core xdist-safe logging fix and the downstream CI/artifact aggregation changes needed to consume multiple captured_info files.", + "canonical_pr_reason": "PR #45645 is the more complete representative: it fixes the xdist-safe output filename in testing_utils and updates CI/log collection to handle the resulting per-worker files.", "best_issue_reason": null, - "best_pr_reason": "PR 45645 is the strongest representative of the cluster since it subsumes the narrower testing_utils change from 45639 and adds the related workflow handling.", + "best_pr_reason": "PR #45645 best matches the cluster because it covers the core bug fix plus the downstream artifact collection changes needed for the new behavior.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", "accept": true, - "reason": "Same underlying xdist collision for captured_info/debug logging, with overlapping testing_utils/test changes; 45645 is a superset that could plausibly absorb 45639." + "reason": "Both PRs fix the same concrete problem: patched testing logs colliding under pytest-xdist. #45645 appears to subsume #45639 with the same core code-path change plus related CI adjustments, so they could plausibly be merged into one PR." } ] }, @@ -22972,7 +22981,7 @@ "soft_pairs": [ "pull_request:45720|pull_request:46259" ], - "prepared_review_unit_hash": "c9277c08db6e205233bbc48813ca0976daf6b72bd0e52b1673a058674be0bbc6", + "prepared_review_unit_hash": "ee95f657f424021aa7e8a4a034eaba3b34dc6afb88ee7b5b6d4b557bfaccb570", "component_budget": { "node_count": 2, "item_count": 2, @@ -22994,18 +23003,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two PRs are both linter-driven model refactors, but they address different rules and different code paths: TRF003 changes forward return behavior, while TRF018 changes weight initialization. They are not duplicates.", + "summary": "The two PRs are both broad linter-driven changes across many Transformer model files, but they target different underlying problems: one updates forward signatures/tuple returns, the other refactors weight initialization rules. They are not the same fix and should not be merged together.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46259 is the broader representative of the cluster because it has the larger, more general modeling-rule patch across many classes, but it is not a duplicate of PR 45720.", + "canonical_pr_reason": "PR #45720 is the better canonical PR because it is a more concrete, narrowly scoped change: implementing part of TRF003 by adding `can_return_tuple` and adjusting forward signatures across affected models.", "best_issue_reason": null, - "best_pr_reason": "PR 46259 is the stronger anchor only in scope/coverage, but it targets a different underlying change than PR 45720.", + "best_pr_reason": "PR #45720 is the best representative of the cluster since it has the clearest single code-path change, whereas PR #46259 is a separate initialization-rule refactor with different behavior.", "soft_edge_verdicts": [ { "left": "pull_request:45720", "right": "pull_request:46259", "accept": false, - "reason": "Different underlying fixes: 45720 implements TRF003 return-tuple behavior, while 46259 implements TRF018 initialization changes. Shared files are incidental broad-model coverage, not the same change." + "reason": "Reject: they touch overlapping model files, but the underlying changes are unrelated. #45720 changes forward/return-tuple behavior; #46259 changes `_init_weights`/initialization logic. Shared filenames and linter theme are not enough to treat them as the same PR." } ] }, @@ -23101,7 +23110,7 @@ "soft_pairs": [ "pull_request:45932|pull_request:46100" ], - "prepared_review_unit_hash": "a551c8d184f3ffb958ae12fec3a85d1be1c991ea714877e88025d6da12a01ffb", + "prepared_review_unit_hash": "600bb0c74e7cb2ab5ce235f159e9a9a854e5af41e16240c68886d3ae651e1dce", "component_budget": { "node_count": 2, "item_count": 2, @@ -23123,18 +23132,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two pull requests are not the same change: PR 45932 is a mixed Grafana/CI/model-code tweak, while PR 46100 is a broad test-migration PR adding kernel and model tests. The soft similarity signal is too weak to treat them as duplicates or mergeable into one PR.", - "confidence": 0.94, + "summary": "Two open PRs, but they are not duplicates: one is a CI/graphana-style testing and model-code PR, the other is a test-migration PR for kernel/DDP and vision model tests. They touch different code paths and aim at different changes.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46100 is the cleaner representative of the cluster because it has a single coherent purpose (test migration) and a consistent test-focused scope.", + "canonical_pr_reason": "PR #45932 is the better representative of the cluster only because it has broader scope and more discussion activity; it is not a duplicate of #46100.", "best_issue_reason": null, - "best_pr_reason": "PR 46100 is the best standalone PR in this set: it is self-contained and clearly about test migration, whereas PR 45932 mixes unrelated CI secret plumbing with model code edits and does not match the same concrete change.", + "best_pr_reason": "PR #45932 is the strongest PR to keep as the cluster representative since it has more activity and a more central code-change footprint, while #46100 is a narrower test migration.", "soft_edge_verdicts": [ { "left": "pull_request:45932", "right": "pull_request:46100", "accept": false, - "reason": "Different underlying changes and code paths. 45932 touches CI secrets plus model implementation/mapping files under a Grafana/testing setup, while 46100 adds new kernel and model tests for migration. They are not plausibly one merged PR." + "reason": "Different underlying work: #45932 updates CI/secrets and several model implementations, while #46100 adds DDP kernel test fixtures and migrates AltCLIP/ViT tests. They do not fix the same concrete bug or change the same code path, so they should not be merged as duplicates." } ] }, @@ -23164,40 +23173,40 @@ "soft_pairs": [ "pull_request:46126|pull_request:46269" ], - "prepared_review_unit_hash": "6c2250321687814006ad56971dc53f8fe01515ebd13b9710715013cd37926f77", + "prepared_review_unit_hash": "94b2c7f3d2b6d97a8fda48c022f431f52c25f39cadeba5bc797688367a75a167", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5525, - "estimated_input_tokens": 1382, - "estimated_eval_tokens": 3020 + "serialized_chars": 5524, + "estimated_input_tokens": 1381, + "estimated_eval_tokens": 3018 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5525, - "estimated_input_tokens": 1382, - "estimated_eval_tokens": 3020 + "serialized_chars": 5524, + "estimated_input_tokens": 1381, + "estimated_eval_tokens": 3018 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related distributed-work PRs, but not duplicates: one is a focused EP/Trainer feature patch, the other is a broad distributed branch base touching many subsystems and CI. They share some filenames, but not the same concrete change.", - "confidence": 0.92, + "summary": "These two PRs are not duplicates. PR #46126 is a focused EP + Trainer integration change, while PR #46269 is a broad distributed branch base that touches many unrelated distributed components and infrastructure. They share some filenames because #46269 likely subsumes the same area, but they do not appear to be the same concrete bugfix/change and should not be merged as duplicates.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46269 is the broader umbrella change and the better representative of the cluster; PR 46126 is a narrower EP + Trainer integration patch built on top of distributed plumbing.", + "canonical_pr_reason": "PR #46126 is the better canonical PR for this cluster because it is the narrower, clearly scoped change centered on EP/Trainer integration. PR #46269 is too broad and looks like a branch-base/meta integration PR rather than a single discrete fix.", "best_issue_reason": null, - "best_pr_reason": "PR 46269 is the best cluster anchor because it covers the wider distributed branch base and subsumes the shared infrastructure touched by the smaller PR.", + "best_pr_reason": "PR #46126 is the best representative artifact here: it has a specific title, small targeted diff, and a coherent change set in distributed tensor parallel / trainer integration. PR #46269 spans many files and multiple distributed subsystems, so it is not a good single canonical change.", "soft_edge_verdicts": [ { "left": "pull_request:46126", "right": "pull_request:46269", "accept": false, - "reason": "Reject: shared distributed files are not enough. 46126 adds a specific EP router/Trainer integration path, while 46269 is a much larger distributed branch base with CI and cross-cutting changes. They do not appear to fix the same concrete bug or implement the same change, and they would not plausibly collapse into one PR." + "reason": "Shared distributed filenames are not enough to treat these as the same PR. #46126 is a focused EP + Trainer integration change; #46269 is a large branch-base PR touching many unrelated areas, so they do not look like the same underlying code-path fix." } ] }, @@ -23263,12 +23272,12 @@ "pull_request:46504|pull_request:46563", "pull_request:46504|pull_request:46716" ], - "prepared_review_unit_hash": "f88552103d28c89d5075138a999e2e5bca291b33034557d26b3c44a4134404dd", + "prepared_review_unit_hash": "40439c2bb126b9af18f6a54b74f25f305f195eda8a56b8cca16837fae5d37f2a", "component_budget": { "node_count": 10, "item_count": 10, "soft_pair_count": 29, - "serialized_chars": 25927, + "serialized_chars": 25926, "estimated_input_tokens": 6482, "estimated_eval_tokens": 13220 }, @@ -23276,7 +23285,7 @@ "node_count": 10, "item_count": 10, "soft_pair_count": 29, - "serialized_chars": 25927, + "serialized_chars": 25926, "estimated_input_tokens": 6482, "estimated_eval_tokens": 13220 }, @@ -23285,174 +23294,186 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are mostly unrelated workflow-chore pull requests. The only tight grouping is the pr-ci-security-gate workflow, but the diffs there are follow-up edits to the same file rather than the same exact change, so I would not collapse them as duplicates.", - "confidence": 0.93, + "summary": "This cluster is mostly a grab bag of unrelated workflow-maintenance PRs. The only tight subgroup is the pr-ci-security-gate.yml changes, but even those are separate hardening edits rather than the same duplicate fix.", + "confidence": 0.83, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46554 is the strongest representative anchor: it makes the substantive security-gate workflow change and is the base file later follow-ups also modify. The other PRs are either different workflows or small, separate adjustments.", + "canonical_pr_reason": "PR 46554 is the most substantive and central workflow-security change in the set, and it anchors the only close mini-cluster around pr-ci-security-gate.yml.", "best_issue_reason": null, - "best_pr_reason": "46554 is the best global representative because it contains the main workflow logic change, while the other candidates are narrower edits or unrelated workflow updates.", + "best_pr_reason": "PR 46554 best represents the strongest concrete change in this cluster: a real hardening update to the PR CI security gate rather than a minor workflow tweak.", "soft_edge_verdicts": [ { "left": "pull_request:46138", "right": "pull_request:46554", "accept": false, - "reason": "Different workflows and different concerns: self-comment CI message piping vs pr-ci-security-gate hardening." + "reason": "Different workflows and different fixes: stdin handling for self-comment CI vs PR security-gate hardening." }, { "left": "pull_request:46138", "right": "pull_request:46558", "accept": false, - "reason": "Different files and different behavior; not the same underlying fix." + "reason": "Only a broad workflow-safety theme in common; the concrete changes are unrelated." }, { "left": "pull_request:46138", "right": "pull_request:46563", "accept": false, - "reason": "Different workflows; one updates self-comment CI, the other adjusts PR security gating." + "reason": "One updates comment parsing, the other adds a safety note in the PR security gate wrapper." }, { "left": "pull_request:46138", "right": "pull_request:46716", "accept": false, - "reason": "Unrelated workflow changes; no shared concrete bug or code path." + "reason": "No shared bug/change beyond generic workflow hardening." }, { "left": "pull_request:46138", "right": "pull_request:46504", "accept": false, - "reason": "Different workflow files and separate shell-safety tweaks." + "reason": "Different files and different shell-safety fixes in separate workflows." }, { "left": "pull_request:46432", "right": "pull_request:46503", "accept": false, - "reason": "SSH runner hardening and benchmark trigger changes are unrelated." + "reason": "Unrelated workflows: SSH runner security tweaks versus benchmark trigger logic." }, { "left": "pull_request:46432", "right": "pull_request:46501", "accept": false, - "reason": "Different workflows; no same bug/change." + "reason": "Different workflow files and different permissions/schedule changes." }, { "left": "pull_request:46432", "right": "pull_request:46502", "accept": false, - "reason": "Different workflow jobs and different fixes." + "reason": "No same underlying bug; SSH runner cleanup versus reviewer assignment behavior." }, { "left": "pull_request:46432", "right": "pull_request:46554", "accept": false, - "reason": "SSH-runner changes are not the same as PR security-gate logic changes." + "reason": "Both are workflow hardening, but they fix different workflows and code paths." }, { "left": "pull_request:46432", "right": "pull_request:46558", "accept": false, - "reason": "Different workflows and different security concerns." + "reason": "Distinct workflow changes with no concrete duplication." }, { "left": "pull_request:46501", "right": "pull_request:46503", "accept": false, - "reason": "Anti-slop permissions change and benchmark trigger change are separate edits." + "reason": "Separate workflows with different operational behavior changes." }, { "left": "pull_request:46501", "right": "pull_request:46502", "accept": false, - "reason": "Different workflow files; no concrete overlap beyond being chores." + "reason": "Both are chore edits, but they modify different workflow logic entirely." }, { "left": "pull_request:46501", "right": "pull_request:46504", "accept": false, - "reason": "Different workflow behavior in unrelated jobs." + "reason": "No duplicate bug/change; one alters benchmark triggers, the other shell quoting in build image CI." }, { "left": "pull_request:46501", "right": "pull_request:46554", "accept": false, - "reason": "Anti-slop workflow permissions and PR CI security-gate changes are not the same fix." + "reason": "Different workflow areas and no shared concrete fix." }, { "left": "pull_request:46502", "right": "pull_request:46503", "accept": false, - "reason": "Different workflows and different trigger/checkout behavior." + "reason": "Reviewer assignment checkout behavior is unrelated to benchmark dispatch changes." }, { "left": "pull_request:46502", "right": "pull_request:46504", "accept": false, - "reason": "Unrelated workflow edits; not a duplicate change." + "reason": "Different workflow files and different intent." }, { "left": "pull_request:46502", "right": "pull_request:46554", "accept": false, - "reason": "Assign-reviewers checkout target and PR security-gate logic are different code paths." + "reason": "No same underlying code-path problem." }, { "left": "pull_request:46503", "right": "pull_request:46504", "accept": false, - "reason": "Benchmark dispatch support and build-ci command handling are separate changes." + "reason": "Benchmark dispatch and build-image message parsing are unrelated changes." }, { "left": "pull_request:46503", "right": "pull_request:46554", "accept": false, - "reason": "Different workflows and different operational intent." + "reason": "Different workflows and different fixes." }, { "left": "pull_request:46504", "right": "pull_request:46554", "accept": false, - "reason": "Build image trigger shell quoting vs PR security gate rewrite are unrelated." + "reason": "Similar security posture, but different workflows and different concrete changes." }, { "left": "pull_request:46504", "right": "pull_request:46558", "accept": false, - "reason": "Different files and distinct changes; not one underlying bug." + "reason": "No duplicate fix; build-image shell quoting versus PR security-gate permissions." }, { "left": "pull_request:46554", "right": "pull_request:46558", "accept": false, - "reason": "Same file, but different concrete edits: one adds checkout/diff checks, the other changes permissions and approval gating." + "reason": "Same file, but these are complementary hardening changes rather than the same concrete fix." }, { "left": "pull_request:46554", "right": "pull_request:46563", "accept": false, - "reason": "Same workflow file but different change sets; not the same fix." + "reason": "One changes security checks, the other only documents trust assumptions in the wrapper." }, { "left": "pull_request:46554", "right": "pull_request:46716", "accept": false, - "reason": "Distinct follow-up edit to the same workflow, not the same underlying change." + "reason": "Actual security-gate logic change vs secrets-passing change; related but not the same fix." }, { "left": "pull_request:46558", "right": "pull_request:46563", "accept": false, - "reason": "Both touch the same workflow, but they modify different aspects of it." + "reason": "Different edits to the same workflow with different purposes." }, { "left": "pull_request:46558", "right": "pull_request:46716", "accept": false, - "reason": "Permissions/approval tweak vs secrets inheritance removal are different fixes." + "reason": "Permissions/approval gating versus secrets inheritance are distinct changes." }, { "left": "pull_request:46563", "right": "pull_request:46716", "accept": false, - "reason": "Comment clarification and secrets isolation are not the same change." + "reason": "Comment-only safety note is not the same as changing secrets inheritance." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46563", + "accept": false, + "reason": "Different files and unrelated workflow fixes." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46716", + "accept": false, + "reason": "Shell quoting in build CI is unrelated to PR CI secrets handling." } ] }, @@ -23482,7 +23503,7 @@ "soft_pairs": [ "pull_request:46152|pull_request:46207" ], - "prepared_review_unit_hash": "2ff8e7fca99885f4671ee1beaec9130c214ce60543797135c3fd0e5dcdd4e841", + "prepared_review_unit_hash": "561b3c91a7ee420e22ea692f1cdc35ba46ce0cfcbd9447e218469ab759ea657f", "component_budget": { "node_count": 2, "item_count": 2, @@ -23504,18 +23525,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate PRs that both replace direct file-opening/source-reading patterns, but they address different code paths and bugs: one in import utilities for locating package resources, the other in modeling utilities for checking attention-implementation support. They are not the same underlying change.", - "confidence": 0.94, + "summary": "Two open PRs share a broad theme of replacing file-based source inspection with safer source lookup, but they target different code paths and should not be deduplicated.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46207 is the cleaner representative of the cluster because it has a concrete behavior change in a specific model-path heuristic and includes tests; PR #46152 is a separate import-resource helper refactor in a different subsystem.", + "canonical_pr_reason": "PR 46152 is the better canonical anchor: it introduces a focused helper in import_utils to avoid reading Transformers\u2019 own source via open(), which is a self-contained utility-level fix.", "best_issue_reason": null, - "best_pr_reason": "PR #46207 is the best standalone PR in the cluster because it targets a well-scoped bug fix in _can_set_attn_implementation and appears more directly mergeable and verifiable than the resource-resolution helper change in PR #46152.", + "best_pr_reason": "PR 46152 is the clearest representative of the cluster\u2019s theme because it is a direct, reusable source-access change rather than a separate model-internal heuristic update.", "soft_edge_verdicts": [ { - "left": "pull_request:46152", - "right": "pull_request:46207", + "left": "Avoid reading self source code via file open()", + "right": "Use inspect.getsource instead of open() for source-reading in _can_set_*_implementation", "accept": false, - "reason": "Both avoid open() for source-reading, but they fix different mechanisms in different modules: import_utils resource traversal vs modeling_utils attention-implementation detection. Not the same underlying bug or change, and not mergeable as one PR." + "reason": "Related pattern, but different concrete code paths: import_utils package resource resolution vs modeling_utils attention-implementation detection. They are not the same underlying bug/change and would not plausibly merge into one PR." } ] }, @@ -23545,7 +23566,7 @@ "soft_pairs": [ "pull_request:46154|pull_request:46196" ], - "prepared_review_unit_hash": "cb039d8347523f179368636a16c4847df2ebc0d20347fc149bc9c06605ae98d1", + "prepared_review_unit_hash": "402cbd316d8ea27d683ad83cf680a822482ee50d2d56d69a9fc832a2c44f0598", "component_budget": { "node_count": 2, "item_count": 2, @@ -23567,18 +23588,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs add SDPA attention dispatch support, but they target different model families and code paths: T5 versus the wav2vec2/audio stack. They are related in theme, not duplicates of the same change.", - "confidence": 0.93, + "summary": "Both pull requests are about SDPA attention dispatch, but they target different model families and code paths (T5 vs wav2vec2/audio models). They are related by theme, not duplicates, and should not be merged as one PR.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "pull_request:46196 is the better representative of the cluster because it is the broader, more central SDPA dispatch implementation across the audio models, while pull_request:46154 is a narrower T5-specific adaptation.", + "canonical_pr_reason": "No canonical PR for duplication purposes: these are separate model-specific changes rather than two copies of the same fix. If a representative must be chosen, neither is clearly canonical over the other.", "best_issue_reason": null, - "best_pr_reason": "pull_request:46196 is the strongest standalone example of the underlying work because it spans multiple audio architectures and reflects the main dispatch integration effort more than the single-model T5 PR.", + "best_pr_reason": "PR #46196 is the broader and more substantial SDPA-dispatch change set, touching many audio models and associated docs/tests, so it is the better representative of the cluster theme.", "soft_edge_verdicts": [ { "left": "pull_request:46154", "right": "pull_request:46196", "accept": false, - "reason": "Reject: they implement the same general feature flag (attn_implementation=sdpa) but for different model families and different concrete code paths. This is thematic similarity, not a single duplicate change that should be merged into one PR." + "reason": "Same high-level theme (SDPA dispatch) but different concrete implementations and model families; not the same underlying bug/change and not plausibly one merged PR." } ] }, @@ -23674,12 +23695,12 @@ "soft_pairs": [ "pull_request:46310|pull_request:46546" ], - "prepared_review_unit_hash": "da5ed938a73e993f164661236f7ff7161a8d0a3613613751ab476f3a9f690084", + "prepared_review_unit_hash": "9faa1aeadd68146d36ec14b61b72665e01c038a1534c8858fa67708060c9bbbe", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4851, + "serialized_chars": 4850, "estimated_input_tokens": 1213, "estimated_eval_tokens": 2682 }, @@ -23687,7 +23708,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4851, + "serialized_chars": 4850, "estimated_input_tokens": 1213, "estimated_eval_tokens": 2682 }, @@ -23696,18 +23717,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests fix the same Moonshine streaming sliding-window boundary bug (the right window was excluding the last allowed token). PR 46546 is the broader, more complete version because it also applies the same fix to Gemma4 and adds tests.", - "confidence": 0.95, + "summary": "Both PRs fix the same sliding-window boundary bug: the right-side mask used a strict `< right_window_size` check instead of including the edge. They also target the same issue and share the same Moonshine streaming code path, with 46546 extending the fix to Gemma4 and tests.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46546 is the better canonical PR because it contains the Moonshine streaming boundary fix plus the same correction in Gemma4 and test coverage, making it the more complete superset of the shared bug fix.", + "canonical_pr_reason": "PR 46546 is the better canonical representative because it addresses the same boundary bug in Moonshine streaming, also applies the same fix to Gemma4, and includes test coverage.", "best_issue_reason": null, - "best_pr_reason": "PR 46546 is the best overall PR candidate because it addresses the same Moonshine bug as 46310 and also broadens the fix to another model with tests, making it the most complete implementation.", + "best_pr_reason": "PR 46546 is the strongest overall candidate: it covers the shared concrete bug, touches the relevant implementations, and adds tests; 46310 is the narrower subset patch.", "soft_edge_verdicts": [ { "left": "pull_request:46310", "right": "pull_request:46546", "accept": true, - "reason": "Both PRs fix the same concrete Moonshine streaming right-window boundary condition (`-dist <= right_window_size` vs `<`) in the same code path and target the same issue. PR 46546 is a superset, but this is still the same underlying fix." + "reason": "Same underlying boundary-condition fix in the same sliding-window mask logic, and the smaller PR is a subset of the broader one; they could plausibly be merged into one change." } ] }, @@ -23737,12 +23758,12 @@ "soft_pairs": [ "pull_request:46376|pull_request:46417" ], - "prepared_review_unit_hash": "a41d25d0f2e5074342e4879d996a5b23902ee665ca70ad4a90d647d3cb804433", + "prepared_review_unit_hash": "ca6f763e662935a8caf7f447b8503d39170eefb03c6a1b8b47e28e283c07e8ad", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5938, + "serialized_chars": 5939, "estimated_input_tokens": 1485, "estimated_eval_tokens": 3226 }, @@ -23750,7 +23771,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5938, + "serialized_chars": 5939, "estimated_input_tokens": 1485, "estimated_eval_tokens": 3226 }, @@ -23759,18 +23780,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two pull requests are unrelated: one adds multimodal support to continuous batching internals, while the other adds a new HunYuan VL model with docs, modeling, processors, and tests.", - "confidence": 0.99, + "summary": "These are two unrelated pull requests: one adds multimodal support to continuous batching, the other adds the HunYuan VL model and its docs/tests. There is no meaningful evidence they fix the same underlying bug or implement the same change.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "PR 46417 is the better canonical PR candidate because it is a complete, well-scoped model addition with substantial review activity; PR 46376 is a draft WIP for a different subsystem and not a duplicate of 46417.", "best_issue_reason": null, - "best_pr_reason": "PR 46417 is the more self-contained, end-to-end feature addition (model, docs, tests, auto-mappings) and is a better representative PR than the draft continuous-batching refactor.", + "best_pr_reason": "PR 46417 is the strongest representative for the cluster only because it is the more mature and self-contained change; however, the two PRs should not be merged as duplicates since they target different code paths and feature areas.", "soft_edge_verdicts": [ { "left": "pull_request:46376", "right": "pull_request:46417", "accept": false, - "reason": "Different areas and different goals: continuous batching multimodal cache/scheduler changes vs. a new HunYuan VL model implementation. They do not fix the same code-path problem and would not be merged into one PR." + "reason": "Different goals and file sets: continuous batching/multimodal infrastructure vs. a new HunYuan VL model integration. They do not appear to fix the same concrete problem or be mergeable into one PR." } ] }, @@ -23800,7 +23821,7 @@ "soft_pairs": [ "pull_request:46574|pull_request:46580" ], - "prepared_review_unit_hash": "2d2a69a9d775f83945eefbb384ff028f560c935ddf5519bae34b24d4991d054f", + "prepared_review_unit_hash": "b5b6886031257f63c89ec2448c269c459910a8b5f8e1970d72f9fb0e8ef2594e", "component_budget": { "node_count": 2, "item_count": 2, @@ -23822,18 +23843,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs address different tokenizer problems in different code paths: CodeLlama decode behavior vs. DeepSeek checkpoint-to-tokenizer-class resolution. They are not duplicates and should not be merged into one cluster.", - "confidence": 0.98, + "summary": "The two PRs are unrelated: one fixes CodeLlama decode behavior that strips real leading spaces, while the other fixes DeepSeek-Coder/DeepSeek-LLM being routed to the wrong tokenizer class. They should not be clustered as duplicates.", + "confidence": 0.99, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: 46574 fixes CodeLlama leading-space stripping on decode, while 46580 fixes DeepSeek-Coder/DeepSeek-LLM loading as the wrong tokenizer class.", + "canonical_pr_reason": "No single canonical PR: these address different code paths and different bugs, so neither supersedes the other.", "best_issue_reason": null, - "best_pr_reason": "Neither PR is a strong global representative of the other; the cluster is heterogeneous. If forced, 46580 is broader in scope, but it is still unrelated to 46574's decoder fix.", + "best_pr_reason": "No global best PR for deduping this cluster; PR #46574 is a focused decode fix and PR #46580 is a separate tokenizer-class routing fix.", "soft_edge_verdicts": [ { "left": "pull_request:46574", "right": "pull_request:46580", "accept": false, - "reason": "Different bugs and code paths: one changes CodeLlama tokenizer decode stripping, the other changes AutoTokenizer checkpoint-prefix handling for DeepSeek models." + "reason": "Different underlying problems: CodeLlama decoder assignment/leading-space stripping vs DeepSeek checkpoint-to-tokenizer-class override in auto loading. They are not mergeable into one fix." } ] }, @@ -23863,12 +23884,12 @@ "soft_pairs": [ "pull_request:46603|pull_request:46604" ], - "prepared_review_unit_hash": "3ba62653df0da727a689be3fbb3c1f72f07477c6578db59cb56102b8b68fee1e", + "prepared_review_unit_hash": "efb8323e3a772f5d212cf91f06e9d94d7a2910638fa431894a2ea71d15f9e882", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5494, + "serialized_chars": 5495, "estimated_input_tokens": 1374, "estimated_eval_tokens": 3004 }, @@ -23876,7 +23897,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5494, + "serialized_chars": 5495, "estimated_input_tokens": 1374, "estimated_eval_tokens": 3004 }, @@ -23885,18 +23906,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs are related to the Mistral tokenizer area but they address different underlying changes: one is a refactor/move of the existing MistralConverter into a package, while the other adds tekken tokenizer detection, conversion, and save utilities. They should not be treated as duplicates.", + "summary": "The two PRs overlap on Mistral tokenizer files, but they are not the same change: one is a refactor that moves `MistralConverter` into the Mistral integration package, while the other adds tekken tokenizer detection/conversion/save utilities. Not a duplicate pair.", "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR duplicate here; the closest cluster anchor is PR #46604 because it introduces the substantive tokenizer functionality, whereas PR #46603 is mainly an internal code move/refactor.", + "canonical_pr_reason": "No exact PR duplicate here; if one PR has to represent the cluster, `#46604` is the better canonical PR because it is the broader, feature-adding change, while `#46603` is a supporting refactor/move.", "best_issue_reason": null, - "best_pr_reason": "PR #46604 is the best representative for the cluster because it covers the main functional change to Mistral tokenizer handling. PR #46603 is ancillary packaging/refactoring work and not the same change.", + "best_pr_reason": "`#46604` is the most substantive and central change in the set, with new tokenizer detection/conversion/save behavior. `#46603` is narrower and mainly reorganizes existing Mistral conversion code.", "soft_edge_verdicts": [ { "left": "pull_request:46603", "right": "pull_request:46604", "accept": false, - "reason": "Different code changes and different goals: #46603 moves MistralConverter into a package, while #46604 adds tekken tokenizer detection/conversion/save utilities. They touch some of the same files but not the same underlying bug or mergeable change." + "reason": "Shared Mistral files alone are not enough: `#46603` relocates `MistralConverter`, while `#46604` implements tekken tokenizer support and save utilities. Different underlying code-path changes, so not mergeable as one PR." } ] }, @@ -23918,63 +23939,78 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-46741-3", + "cluster_id": "cluster-46741-4", "nodes": [ "pull_request:46741", "pull_request:46748", - "pull_request:46773" + "pull_request:46773", + "pull_request:46781" ], "soft_pairs": [ + "pull_request:46741|pull_request:46781", "pull_request:46741|pull_request:46773", "pull_request:46741|pull_request:46748", + "pull_request:46748|pull_request:46781", "pull_request:46748|pull_request:46773" ], - "prepared_review_unit_hash": "2719ee14d6ee7263ee0ed4980ca22961cba939f62a5667e26410b46b53b4e274", + "prepared_review_unit_hash": "fc138322b12c5095c135a8c72037ba61b5679b6d22485da2f59e86efae1bc528", "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": 4, + "item_count": 4, + "soft_pair_count": 5, + "serialized_chars": 10650, + "estimated_input_tokens": 2663, + "estimated_eval_tokens": 5582 }, "budget": { - "node_count": 3, - "item_count": 3, - "soft_pair_count": 3, - "serialized_chars": 7882, - "estimated_input_tokens": 1971, - "estimated_eval_tokens": 4198 + "node_count": 4, + "item_count": 4, + "soft_pair_count": 5, + "serialized_chars": 10650, + "estimated_input_tokens": 2663, + "estimated_eval_tokens": 5582 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three pull requests touch neighboring Mamba/linear-attention caching paths, but they fix different bugs: chunked-prefill/speculative decoding, offloaded-cache device placement, and padding-mask handling for batch size 1. They should not be deduplicated.", - "confidence": 0.97, + "summary": "Four distinct PRs about hybrid-model/runtime edge cases: Mamba2 chunked-prefill, cache prefetch on offloaded hybrid layers, padding mask handling for batch size 1, and QuantizedCache layer construction. They share a loose theme but do not fix the same bug or code path.", + "confidence": 0.86, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: the changes target different failure modes and different code paths, despite overlapping model/cache infrastructure.", + "canonical_pr_reason": "pull_request:46748 is the best representative anchor: it is a focused core cache_utils fix for hybrid models, has a clear runtime bug description, and sits in the shared infrastructure area, while the other PRs are separate model- or cache-specific bugs.", "best_issue_reason": null, - "best_pr_reason": "No PR is a clear cluster representative because none subsumes the others; 46741 is the most narrowly concrete fix, but it does not represent the other two bugs.", + "best_pr_reason": "pull_request:46748 is the strongest standalone PR for the cluster\u2019s general theme because it touches shared cache infrastructure and addresses a concrete hybrid-model failure mode with tests.", "soft_edge_verdicts": [ + { + "left": "pull_request:46741", + "right": "pull_request:46781", + "accept": false, + "reason": "Both relate to hybrid/SSM models, but one fixes Mamba2 chunked-prefill/speculative decode in model forward paths while the other fixes QuantizedCache construction in cache_utils; different bug and code path." + }, { "left": "pull_request:46741", "right": "pull_request:46773", "accept": false, - "reason": "Different bugs: 46741 fixes Mamba2 chunked-prefill/speculative decoding behavior, while 46773 fixes padding-mask application for batch size 1 across linear-attention models." + "reason": "Different failure modes: chunked-prefill/speculative decoding vs padding-mask application for batch-size-1 linear-attention models. Broadly related only by model family." }, { "left": "pull_request:46741", "right": "pull_request:46748", "accept": false, - "reason": "Different issues: 46741 is about cached convolution handling during chunked-prefill/decoding; 46748 is about offloaded cache prefetch device mismatch in hybrid models." + "reason": "One is a Mamba2/Nemotron-H forward-path fix; the other is cache prefetch device handling. Same ecosystem, not the same underlying bug." + }, + { + "left": "pull_request:46748", + "right": "pull_request:46781", + "accept": false, + "reason": "Both modify cache_utils for hybrid models, but one fixes offloaded cache prefetch device mismatch and the other fixes QuantizedCache layer selection/crash. Different concrete code paths." }, { "left": "pull_request:46748", "right": "pull_request:46773", "accept": false, - "reason": "Both concern cache/attention infrastructure, but 46748 fixes offloaded-cache device placement, whereas 46773 fixes padding-mask skipping for batch size 1." + "reason": "Shared hybrid-model context is too broad; cache prefetch mismatch and padding-mask skipping are unrelated bugs." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index 50d73ca63d52376e929a5554198954609f16ce6b..71712f1336c5821c37bc99e00366de5c01029eff 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260620t060014z", + "analysis_id": "hybrid-model-20260620t120016z", "archived_artifacts": { - "hybrid": "snapshots/20260620T060014Z/analysis-runs/hybrid-model-20260620t060014z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260620T060014Z/analysis-runs/hybrid-model-20260620t060014z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-06-20T06:04:14Z", + "published_at": "2026-06-20T12:05:23Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260620T060014Z", + "snapshot_id": "20260620T120016Z", "variant": "hybrid" } diff --git a/snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.json b/snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..1bef6deac9be49e21425ded88f627b895abba588 --- /dev/null +++ b/snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.json @@ -0,0 +1,2306 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260620T120016Z", + "generated_at": "2026-06-20T12:05:21Z", + "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-45290-4", + "summary": "Cluster of 2 issues and 2 PRs centered on issue #46752.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 46752, + "canonical_pr_number": 46753, + "issue_numbers": [ + 45290, + 46752 + ], + "pr_numbers": [ + 46753, + 46778 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46753, + "right_pr_number": 46778, + "code_similarity": 0.825, + "size_similarity": 0.882, + "file_overlap": 1.0, + "area_overlap": 0.846, + "patch_similarity": 0.346, + "shared_filenames": [ + "src/transformers/tokenization_utils_base.py", + "tests/test_tokenization_common.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/tokenization_utils_base.py", + "left_ranges": [ + [ + 3086, + 3094 + ] + ], + "right_ranges": [ + [ + 3086, + 3096 + ] + ] + }, + { + "filename": "tests/test_tokenization_common.py", + "left_ranges": [ + [ + 1010, + 1022 + ] + ], + "right_ranges": [ + [ + 1010, + 1024 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-41115-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41115.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 41115, + "canonical_pr_number": 41116, + "issue_numbers": [ + 41115 + ], + "pr_numbers": [ + 41116, + 45613 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 41116, + "right_pr_number": 45613, + "code_similarity": 0.811, + "size_similarity": 0.9, + "file_overlap": 0.75, + "area_overlap": 0.882, + "patch_similarity": 0.648, + "shared_filenames": [ + "docs/source/en/_toctree.yml", + "docs/source/en/model_doc/minicpm3.md", + "src/transformers/models/auto/modeling_auto.py", + "src/transformers/models/minicpm3/__init__.py", + "src/transformers/models/minicpm3/configuration_minicpm3.py", + "src/transformers/models/minicpm3/modeling_minicpm3.py", + "src/transformers/models/minicpm3/modular_minicpm3.py", + "tests/models/minicpm3/__init__.py", + "tests/models/minicpm3/test_modeling_minicpm3.py" + ], + "shared_file_areas": [ + { + "filename": "docs/source/en/_toctree.yml", + "left_ranges": [ + [ + 737, + 744 + ] + ], + "right_ranges": [ + [ + 730, + 737 + ] + ] + }, + { + "filename": "docs/source/en/model_doc/minicpm3.md", + "left_ranges": [ + [ + 1, + 67 + ] + ], + "right_ranges": [ + [ + 1, + 45 + ] + ] + }, + { + "filename": "src/transformers/models/auto/modeling_auto.py", + "left_ranges": [ + [ + 301, + 307 + ], + [ + 742, + 748 + ], + [ + 1360, + 1366 + ] + ], + "right_ranges": [ + [ + 720, + 726 + ], + [ + 1331, + 1337 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/__init__.py", + "left_ranges": [ + [ + 1, + 29 + ] + ], + "right_ranges": [ + [ + 1, + 29 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/configuration_minicpm3.py", + "left_ranges": [ + [ + 1, + 150 + ] + ], + "right_ranges": [ + [ + 1, + 126 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 575 + ] + ], + "right_ranges": [ + [ + 1, + 522 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modular_minicpm3.py", + "left_ranges": [ + [ + 1, + 354 + ] + ], + "right_ranges": [ + [ + 1, + 342 + ] + ] + }, + { + "filename": "tests/models/minicpm3/test_modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 135 + ] + ], + "right_ranges": [ + [ + 1, + 136 + ] + ] + } + ] + } + ] + }, + { + "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-45987-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45987.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 45987, + "canonical_pr_number": 46173, + "issue_numbers": [ + 45987 + ], + "pr_numbers": [ + 46173, + 46474 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 46173, + "right_pr_number": 46474, + "code_similarity": 0.65, + "size_similarity": 1.0, + "file_overlap": 1.0, + "area_overlap": 0.0, + "patch_similarity": 1.0, + "shared_filenames": [ + "src/transformers/cache_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/cache_utils.py", + "left_ranges": [ + [ + 69, + 75 + ], + [ + 375, + 381 + ] + ], + "right_ranges": [ + [ + 79, + 85 + ], + [ + 461, + 467 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45561-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45561.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45561, + "canonical_pr_number": 45645, + "issue_numbers": [ + 45561 + ], + "pr_numbers": [ + 45639, + 45645 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45639, + "right_pr_number": 45645, + "code_similarity": 0.494, + "size_similarity": 0.619, + "file_overlap": 0.5, + "area_overlap": 0.434, + "patch_similarity": 0.451, + "shared_filenames": [ + "src/transformers/testing_utils.py", + "tests/utils/test_testing_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/testing_utils.py", + "left_ranges": [ + [ + 3595, + 3602 + ], + [ + 3819, + 3853 + ] + ], + "right_ranges": [ + [ + 3525, + 3558 + ], + [ + 3782, + 3788 + ] + ] + }, + { + "filename": "tests/utils/test_testing_utils.py", + "left_ranges": [ + [ + 1, + 86 + ] + ], + "right_ranges": [ + [ + 1, + 114 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46305-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46305.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 46305, + "canonical_pr_number": 46546, + "issue_numbers": [ + 46305 + ], + "pr_numbers": [ + 46310, + 46546 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46310, + "right_pr_number": 46546, + "code_similarity": 0.566, + "size_similarity": 0.444, + "file_overlap": 0.5, + "area_overlap": 0.867, + "patch_similarity": 0.16, + "shared_filenames": [ + "src/transformers/models/moonshine_streaming/modeling_moonshine_streaming.py", + "src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/moonshine_streaming/modeling_moonshine_streaming.py", + "left_ranges": [ + [ + 356, + 362 + ] + ], + "right_ranges": [ + [ + 355, + 361 + ] + ] + }, + { + "filename": "src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py", + "left_ranges": [ + [ + 269, + 275 + ] + ], + "right_ranges": [ + [ + 269, + 275 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45081-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45081.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45081, + "canonical_pr_number": 45317, + "issue_numbers": [ + 45081 + ], + "pr_numbers": [ + 45086, + 45317 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45086, + "right_pr_number": 45317, + "code_similarity": 0.589, + "size_similarity": 0.34, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.136, + "shared_filenames": [ + "src/transformers/tokenization_utils_tokenizers.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/tokenization_utils_tokenizers.py", + "left_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ], + "right_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43698-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43698.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43698, + "canonical_pr_number": 43779, + "issue_numbers": [ + 43698 + ], + "pr_numbers": [ + 43779, + 43816 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43779, + "right_pr_number": 43816, + "code_similarity": 0.418, + "size_similarity": 0.538, + "file_overlap": 1.0, + "area_overlap": 0.02, + "patch_similarity": 0.02, + "shared_filenames": [ + "src/transformers/integrations/integration_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/integration_utils.py", + "left_ranges": [ + [ + 2227, + 2241 + ], + [ + 2303, + 2309 + ] + ], + "right_ranges": [ + [ + 2278, + 2291 + ], + [ + 2309, + 2322 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46562-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 46562, + "issue_numbers": [], + "pr_numbers": [ + 46562, + 46658 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46562, + "right_pr_number": 46658, + "code_similarity": 0.153, + "size_similarity": 0.387, + "file_overlap": 0.024, + "area_overlap": 0.0, + "patch_similarity": 0.458, + "shared_filenames": [ + "src/transformers/models/auto/auto_mappings.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/auto/auto_mappings.py", + "left_ranges": [ + [ + 441, + 447 + ], + [ + 591, + 598 + ], + [ + 823, + 829 + ], + [ + 883, + 889 + ], + [ + 1082, + 1088 + ], + [ + 1205, + 1211 + ] + ], + "right_ranges": [ + [ + 362, + 368 + ], + [ + 795, + 801 + ] + ] + } + ] + } + ] + }, + { + "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-45290-4", + "canonical_issue_number": 46752, + "duplicate_issue_numbers": [ + 45290 + ], + "reason": "Issues in cluster-45290-4 are treated as duplicates because they share closing_reference, shared_issue_target, soft_similarity evidence." + } + ], + "duplicate_prs": [ + { + "cluster_id": "cluster-41115-3", + "canonical_pr_number": 41116, + "duplicate_pr_numbers": [ + 45613 + ], + "target_issue_number": 41115, + "reason": "PRs in cluster-41115-3 are treated as duplicates because they converge on issue #41115 with closing_reference evidence." + }, + { + "cluster_id": "cluster-41211-3", + "canonical_pr_number": 44339, + "duplicate_pr_numbers": [ + 41356 + ], + "target_issue_number": 41211, + "reason": "PRs in cluster-41211-3 are treated as duplicates because they converge on issue #41211 with closing_reference evidence." + }, + { + "cluster_id": "cluster-43240-3", + "canonical_pr_number": 43251, + "duplicate_pr_numbers": [ + 43254 + ], + "target_issue_number": 43240, + "reason": "PRs in cluster-43240-3 are treated as duplicates because they converge on issue #43240 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43698-3", + "canonical_pr_number": 43779, + "duplicate_pr_numbers": [ + 43816 + ], + "target_issue_number": 43698, + "reason": "PRs in cluster-43698-3 are treated as duplicates because they converge on issue #43698 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43979-11", + "canonical_pr_number": 44007, + "duplicate_pr_numbers": [ + 43996, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "target_issue_number": 43979, + "reason": "PRs in cluster-43979-11 are treated as duplicates because they converge on issue #43979 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-44018-2", + "canonical_pr_number": 44068, + "duplicate_pr_numbers": [ + 44018 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-45081-3", + "canonical_pr_number": 45317, + "duplicate_pr_numbers": [ + 45086 + ], + "target_issue_number": 45081, + "reason": "PRs in cluster-45081-3 are treated as duplicates because they converge on issue #45081 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45290-4", + "canonical_pr_number": 46753, + "duplicate_pr_numbers": [ + 46778 + ], + "target_issue_number": 46752, + "reason": "PRs in cluster-45290-4 are treated as duplicates because they converge on issue #46752 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45561-3", + "canonical_pr_number": 45645, + "duplicate_pr_numbers": [ + 45639 + ], + "target_issue_number": 45561, + "reason": "PRs in cluster-45561-3 are treated as duplicates because they converge on issue #45561 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45647-3", + "canonical_pr_number": 45738, + "duplicate_pr_numbers": [ + 45737 + ], + "target_issue_number": 45647, + "reason": "PRs in cluster-45647-3 are treated as duplicates because they converge on issue #45647 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-45987-3", + "canonical_pr_number": 46173, + "duplicate_pr_numbers": [ + 46474 + ], + "target_issue_number": 45987, + "reason": "PRs in cluster-45987-3 are treated as duplicates because they converge on issue #45987 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-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-46562-2", + "canonical_pr_number": 46562, + "duplicate_pr_numbers": [ + 46658 + ], + "target_issue_number": null, + "reason": "PRs in cluster-46562-2 are treated as duplicates because they share soft_similarity evidence." + } + ], + "best_issue": { + "cluster_id": "cluster-45290-4", + "issue_number": 46752, + "reason": "Issue #46752 is the strongest global issue candidate because it is open, belongs to a cluster with 4 artifacts, and carries 0 discussion comments plus 4 inbound references.", + "score": 144.0 + }, + "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/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..f55cbaff88be3f6fab42fc9b89f44e039a09a961 --- /dev/null +++ b/snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,24091 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260620T120016Z", + "generated_at": "2026-06-20T12:05:21Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 1, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "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:46304", + "issue:46468", + "issue:46515", + "issue:46661" + ], + "soft_pairs": [ + "issue:44322|issue:45464", + "issue:45237|issue:45685", + "issue:43716|issue:45237", + "issue:43828|issue:45237", + "issue:45237|issue:46468", + "issue:45237|issue:45507", + "issue:30277|issue:35545", + "issue:35545|issue:42915", + "issue:35545|issue:41720", + "issue:45901|issue:46015", + "issue:44509|issue:44512", + "issue:35545|issue:36010", + "issue:44918|issue:46165", + "issue:44805|issue:45823", + "issue:43638|issue:44805", + "issue:43638|issue:46165", + "issue:43596|issue:44805", + "issue:44805|issue:46294", + "issue:44805|issue:45161", + "issue:46165|issue:46468", + "issue:44560|issue:46165", + "issue:43596|issue:46165", + "issue:34634|issue:35545", + "issue:30277|issue:43638", + "issue:43596|issue:43638", + "issue:43039|issue:45161", + "issue:43975|issue:44457", + "issue:44292|issue:44457", + "issue:44841|issue:46661", + "issue:24643|issue:30277", + "issue:34928|issue:44805", + "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: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:43716|issue:45685", + "issue:44292|issue:44918", + "issue:45507|issue:45685", + "issue:43716|issue:44292", + "issue:44918|issue:45161", + "issue:43039|issue:44918", + "issue:41762|issue:43638", + "issue:43827|issue:44512", + "issue:43828|issue:45685", + "issue:44918|issue:45685", + "issue:43039|issue:43828", + "issue:30277|issue:36331", + "issue:34928|issue:36010", + "issue:45237|issue:46015", + "issue:43039|issue:44805", + "issue:45800|issue:46304" + ], + "prepared_review_unit_hash": "5d7288ba8d76086e36b1d4d7e0a1c1611bf8776194d0c7b459a4c53db7ac39da", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 63, + "serialized_chars": 46594, + "estimated_input_tokens": 11649, + "estimated_eval_tokens": 23554 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous, with only two clear duplicate families: the DeepSpeed ZeRO-3/BertModel index-out-of-bounds reports, and the torch 2.4.1 incompatibility reports.", + "confidence": 0.72, + "canonical_issue_reason": "Issue 43596 is the best anchor for the strongest duplicate subgroup: it is a concrete ZeRO-3/BertModel index-out-of-bounds report and is closely matched by 43638.", + "canonical_pr_reason": null, + "best_issue_reason": "43596 is the most representative issue in the only clearly duplicate family, with the broadest wording and the cleanest nearby duplicate match.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Different failures: missing Qwen3.5 config attribute vs streaming chat/completions inference error." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Different models and code paths: AMD GPU failure for GPT-OSS vs MPS histogram op not implemented for Int." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Different subsystems: multimodal dtype mismatch vs GPT-OSS AMD GPU runtime failure." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Both involve runtime errors, but the models and underlying bug are unrelated." + }, + { + "left": "issue:45237", + "right": "issue:46468", + "accept": false, + "reason": "Different issue types: AMD GPU incompatibility vs broken official Nemotron example." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Unrelated failures in different model families; no shared concrete code path." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "DeepSpeed collective mismatch is unrelated to ModernBERT ONNX export." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "ONNX export error and FineGrainedFP8 failure are different bugs." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "ModernBERT export and Qwen3 auto device mapping crash are unrelated." + }, + { + "left": "issue:45901", + "right": "issue:46015", + "accept": false, + "reason": "Table QA crash and outdated RAG docs are different changes." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs cleanup for v5 removals, but they target different removed items and are not the same concrete change." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "ONNX export failure is unrelated to the GenerationMixin import error." + }, + { + "left": "issue:44918", + "right": "issue:46165", + "accept": false, + "reason": "Different training issues: TRL SFT embedding unpacking vs resume-training with remote code." + }, + { + "left": "issue:44805", + "right": "issue:45823", + "accept": false, + "reason": "Mask-shape mismatch and Gemma4 device mismatch are different forward-pass bugs." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Different symptoms: Bert ZeRO-3 empty-index failure vs mask/tensor shape mismatch." + }, + { + "left": "issue:43638", + "right": "issue:46165", + "accept": false, + "reason": "ZeRO-3 Bert index error is unrelated to resume-training with local remote-code models." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "Both are index/shape runtime errors, but they affect different models and failure mechanisms." + }, + { + "left": "issue:44805", + "right": "issue:46294", + "accept": false, + "reason": "Mask shape mismatch is not the same as Qwen3.5 video token/feature mismatch." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Different bug classes: tensor mask mismatch vs GPT-OSS tensor-parallel issue." + }, + { + "left": "issue:46165", + "right": "issue:46468", + "accept": false, + "reason": "Resume-training problem and broken Nemotron example are unrelated." + }, + { + "left": "issue:44560", + "right": "issue:46165", + "accept": false, + "reason": "Video embedding StopIteration and resume-training with remote code are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:46165", + "accept": false, + "reason": "ZeRO-3 Bert index error is unrelated to the remote-code resume-training issue." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "BarkProcessor voice_preset and ModernBERT ONNX export are unrelated." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "DeepSpeed collective mismatch is a different failure from the Bert ZeRO-3 empty-index bug." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "These describe the same underlying ZeRO-3/Bert empty-index failure with very similar reproduction context." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger Kernel cross-entropy invocation and GPT-OSS tensor-parallel failure are different bugs." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "Incorrect detokenization and LoRA merge/save/reload mismatch are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Qwen-3 NVFP4 runtime error and LoRA merge consistency issue are different problems." + }, + { + "left": "issue:44841", + "right": "issue:46661", + "accept": false, + "reason": "Voxtral processor failure and remote-code config_class AttributeError are unrelated." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "Both are DeepSpeed-related, but the weight-dimension error and collective mismatch are different bugs." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "Activation-checkpointing/FSDP recomputation mismatch is unrelated to mask-shape indexing failure." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": false, + "reason": "Both concern Qwen3.5 MoE rotary embedding, but the reported failures are different enough that they do not look like the same concrete bug." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "Bert ZeRO-3 empty-index failure and GPT-OSS TP failure are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "FSDP checkpointing tensor-size mismatch is unrelated to CustomTrainer.compute_loss signature breakage." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "Different models and failure modes; no shared code path." + }, + { + "left": "issue:43716", + "right": "issue:45507", + "accept": false, + "reason": "Mistral-3 dtype mismatch and GraniteMoEHybrid invalid method are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Qwen-3 NVFP4 runtime error and GPT-OSS AMD GPU failure are different issues." + }, + { + "left": "issue:45874", + "right": "issue:46294", + "accept": false, + "reason": "Gemma4 reverse-embedding lookup and Qwen3.5 video-token mismatch are different multimodal bugs." + }, + { + "left": "issue:39290", + "right": "issue:45081", + "accept": false, + "reason": "Gemma3Config missing attribute and Mistral tokenizer regex patch crash are unrelated." + }, + { + "left": "issue:45823", + "right": "issue:45874", + "accept": false, + "reason": "Gemma4 device-map mismatch and expensive reverse embedding lookup are different bugs." + }, + { + "left": "issue:45800", + "right": "issue:46515", + "accept": true, + "reason": "Both report the same torch 2.4.1 incompatibility, just phrased differently." + }, + { + "left": "issue:43425", + "right": "issue:44292", + "accept": false, + "reason": "Torch 2.10 incompatibility and Qwen-3 NVFP4 runtime failure are not the same underlying problem." + }, + { + "left": "issue:34928", + "right": "issue:41720", + "accept": false, + "reason": "FSDP activation checkpointing mismatch is unrelated to Qwen3 auto device mapping cuda assert." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Different dtype/runtime failures in different models and devices." + }, + { + "left": "issue:43828", + "right": "issue:44292", + "accept": false, + "reason": "Phi-tiny-MoE autocast dtype mismatch and Qwen-3 NVFP4 runtime failure are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:43716", + "accept": false, + "reason": "Mixtral grouped_mm tracing failure and Mistral-3 image dtype mismatch are different code paths." + }, + { + "left": "issue:43716", + "right": "issue:46468", + "accept": false, + "reason": "Mistral-3 multimodal dtype issue and Nemotron example breakage are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:45685", + "accept": false, + "reason": "Different models, devices, and failure modes." + }, + { + "left": "issue:44292", + "right": "issue:44918", + "accept": false, + "reason": "Qwen-3 NVFP4 runtime error is unrelated to TRL SFT embedding unpacking failure." + }, + { + "left": "issue:44918", + "right": "issue:45161", + "accept": false, + "reason": "Different subsystems: TRL embedding unpacking vs GPT-OSS tensor parallel." + }, + { + "left": "issue:41762", + "right": "issue:43638", + "accept": false, + "reason": "Both are ZeRO-3/Bert-related, but the titles point to different concrete scenarios and I would not merge them as the same bug without stronger evidence." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Both are documentation fixes for v5 removals, but they cover different removed features and different pages." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Repeated pair, but still unrelated bugs: autocast dtype mismatch vs MPS histogram op limitation." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "TRL SFT embedding issue and MPS histogram error are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Liger Kernel cross_entropy and Phi-tiny-MoE autocast mismatch are different problems." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch is unrelated to the trainer signature change." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "Activation checkpointing tensor-size mismatch and GenerationMixin import error are unrelated." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "GPT-OSS AMD GPU failure and outdated RAG docs are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Liger Kernel cross-entropy path and mask/tensor indexing mismatch are unrelated." + }, + { + "left": "issue:45800", + "right": "issue:46304", + "accept": false, + "reason": "Torch 2.4.1 incompatibility and v5.9.0 USE_DISTRIBUTED=0 breakage are different compatibility issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall: the cluster is presented as mostly heterogeneous, and the two accepted duplicate families are supported by the packet (43596/43638 and 45800/46515). The soft-edge rejections stay conservative and do not overclaim matches." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 2, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:36010", + "issue:39290", + "issue:41093", + "issue:41720", + "issue:41762", + "issue:43039", + "issue:43296", + "issue:43366", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43638", + "issue:43792", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43872", + "issue:43901", + "issue:43975", + "issue:44292", + "issue:44512", + "issue:44560", + "issue:44589", + "issue:44841", + "issue:44863", + "issue:45020", + "issue:45070", + "issue:45084", + "issue:45237", + "issue:45362", + "issue:45399", + "issue:45464", + "issue:45685", + "issue:45800", + "issue:45812", + "issue:45823", + "issue:45901", + "issue:45920", + "issue:46015", + "issue:46294", + "issue:46304", + "issue:46402", + "issue:46661", + "issue:46772" + ], + "soft_pairs": [ + "issue:43866|issue:46402", + "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:43039|issue:44560", + "issue:45823|issue:46294", + "issue:43975|issue:46015", + "issue:44841|issue:45823", + "issue:43866|issue:46772", + "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:39290|issue:43531", + "issue:41093|issue:43638", + "issue:43792|issue:45901", + "issue:43854|issue:43866", + "issue:43425|issue:45070", + "issue:43975|issue:45901", + "issue:43596|issue:45800", + "issue:43792|issue:45070", + "issue:45362|issue:45464", + "issue:44292|issue:45901", + "issue:39290|issue:45823", + "issue:45464|issue:45685", + "issue:43541|issue:45685", + "issue:36010|issue:39290", + "issue:43541|issue:43828", + "issue:44841|issue:45084", + "issue:43404|issue:43866", + "issue:41093|issue:41762", + "issue:45399|issue:46661", + "issue:43572|issue:45070", + "issue:45800|issue:45823", + "issue:43296|issue:43366" + ], + "prepared_review_unit_hash": "ac4e3bf453bff129aa5b79b871dc3f990894d9c8696d2f3804a930d4bca4a3c8", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 40, + "serialized_chars": 38902, + "estimated_input_tokens": 9726, + "estimated_eval_tokens": 19708 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly heterogeneous issue reports with no clear PR cluster. The only strong duplicate-like pair is the v5 AutoTokenizer wrong-token-ID regression affecting Granite and other models.", + "confidence": 0.77, + "canonical_issue_reason": "issue:45920 is the best representative issue: it states the broad v5 AutoTokenizer regression across multiple model families, which cleanly subsumes the narrower Granite-only report in issue:45812.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45920 is the most suitable global issue to keep because it is broad, concrete, and clearly captures the shared tokenizer regression seen in the only accepted soft match.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43866", + "right": "issue:46402", + "accept": false, + "reason": "Different model-loading failures: a corrupted Ovis2 checkpoint versus missing Sam3 text encoder weights." + }, + { + "left": "issue:43296", + "right": "issue:46304", + "accept": false, + "reason": "Unrelated failures: PaddleOCR-VL loading in vLLM versus a USE_DISTRIBUTED=0 build/runtime break." + }, + { + "left": "issue:41762", + "right": "issue:43596", + "accept": false, + "reason": "Both mention ZeRO-3, but one is Gemma3-specific loading and the other is a BertModel init failure; not the same bug." + }, + { + "left": "issue:39290", + "right": "issue:41720", + "accept": false, + "reason": "Different subsystems and symptoms: Gemma3TextConfig sliding_window attribute error versus Qwen3 auto device mapping cuda assert." + }, + { + "left": "issue:43901", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs issues, but they concern different docs and different changes." + }, + { + "left": "issue:44589", + "right": "issue:46304", + "accept": false, + "reason": "Different runtime incompatibilities: Float8 storage lookup versus distributed build support." + }, + { + "left": "issue:45237", + "right": "issue:45901", + "accept": false, + "reason": "Completely different domains: AMD GPU execution versus table-question-answering crash." + }, + { + "left": "issue:43039", + "right": "issue:44560", + "accept": false, + "reason": "Liger Kernel cross_entropy routing and Qwen3-vl video StopIteration are unrelated." + }, + { + "left": "issue:45823", + "right": "issue:46294", + "accept": false, + "reason": "Both are model-specific forward/input issues, but the model families and failure modes differ." + }, + { + "left": "issue:43975", + "right": "issue:46015", + "accept": false, + "reason": "DeepSeek detokenization regression and RAG docs problems are unrelated." + }, + { + "left": "issue:44841", + "right": "issue:45823", + "accept": false, + "reason": "Different model paths: Voxtral processor failure versus Gemma4 device placement mismatch." + }, + { + "left": "issue:43866", + "right": "issue:46772", + "accept": false, + "reason": "Corrupted Ovis2 checkpoint is unrelated to DiffusionGemma NVFP4 compatibility." + }, + { + "left": "issue:44863", + "right": "issue:46015", + "accept": false, + "reason": "NemotronH checkpoint loading and RAG documentation are unrelated." + }, + { + "left": "issue:45020", + "right": "issue:46304", + "accept": false, + "reason": "Remote-code model breakage and USE_DISTRIBUTED=0 support are separate regressions." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same underlying v5 AutoTokenizer wrong-token-ID regression; 45920 is the broader report and 45812 is a narrower Granite-specific instance." + }, + { + "left": "issue:43872", + "right": "issue:46304", + "accept": false, + "reason": "bitsandbytes Int8Params incompatibility is unrelated to distributed build support." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Corrupted checkpoint versus checkpoint loading implementation failure are different causes." + }, + { + "left": "issue:44292", + "right": "issue:46015", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime failure is unrelated to RAG example docs." + }, + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Both mention sliding_window, but they involve different models and different error contexts." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Both are index errors under ZeRO-3, but one is a mask/tensor shape mismatch and the other is an init-time zero-size tensor issue." + }, + { + "left": "issue:43792", + "right": "issue:45901", + "accept": false, + "reason": "Whisper loading failure and table QA task crash are unrelated." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "GLM-4.7-Flash test loading failure and Ovis2 checkpoint corruption are different issues." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Torch version incompatibility and a pydantic PretrainedConfig regression are different failures." + }, + { + "left": "issue:43975", + "right": "issue:45901", + "accept": false, + "reason": "Tokenizer detokenization regression and table QA crashes are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "ZeRO-3 Bert init error and torch/transformers version incompatibility are not the same bug." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "Whisper loading failure and pydantic config breakage are unrelated." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Both involve Qwen3.5 chat use, but the reported crashes and code paths are different enough to keep separate." + }, + { + "left": "issue:44589", + "right": "issue:45685", + "accept": false, + "reason": "Float8 storage lookup and MPS histogram-on-Int failure are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:45685", + "accept": false, + "reason": "Torch dynamo grouped_mm tracing failure and MPS histogram support are different code paths." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "ImportError for GenerationMixin and Gemma3 sliding_window config error are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Different model/runtime problems: grouped_mm tracing versus autocast dtype mismatch." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Voxtral processor failure and template-node compilation error are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "Mistral3 weight-tying bug and Ovis2 checkpoint corruption are unrelated." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both are errors, but one is a mask shape mismatch and the other is a Gemma3 ZeRO-3 loading issue." + }, + { + "left": "issue:45399", + "right": "issue:46661", + "accept": false, + "reason": "Flash-attn fallback logic and remote-code config_class handling are different problems." + }, + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "StableLm pad_token_idx regression and pydantic PretrainedConfig breakage are unrelated." + }, + { + "left": "issue:45800", + "right": "issue:45823", + "accept": false, + "reason": "Torch version incompatibility and Gemma4 device mismatch are separate issues." + }, + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "PaddleOCR-VL loading in vLLM and GGUF gpt-oss support are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well-grounded in the packet, and the only accepted soft pair (issue:45812 / issue:45920) is supported by the titles as the same broad v5 AutoTokenizer regression. The other verdicts are generally conservative, with no clear overstatement of evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 3, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:39186", + "issue:40990", + "issue:41093", + "issue:42915", + "issue:43054", + "issue:43366", + "issue:43404", + "issue:43531", + "issue:43541", + "issue:43596", + "issue:43645", + "issue:43716", + "issue:43828", + "issue:43873", + "issue:43975", + "issue:43994", + "issue:44291", + "issue:44292", + "issue:44360", + "issue:44387", + "issue:44479", + "issue:44492", + "issue:44512", + "issue:44530", + "issue:44589", + "issue:44805", + "issue:44863", + "issue:44918", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45161", + "issue:45183", + "issue:45399", + "issue:45464", + "issue:45507", + "issue:45632", + "issue:45698", + "issue:45759", + "issue:45800", + "issue:45902", + "issue:46294", + "issue:46463", + "issue:46489", + "issue:46491", + "issue:46640", + "issue:46661" + ], + "soft_pairs": [ + "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:43541|issue:44589", + "issue:43716|issue:46294", + "issue:45020|issue:45800", + "issue:41093|issue:43596", + "issue:44479|issue:46294", + "issue:43054|issue:43994", + "issue:45759|issue:45902", + "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:44530|issue:45005", + "issue:43873|issue:45005", + "issue:44360|issue:44512", + "issue:44863|issue:45005", + "issue:44492|issue:44512", + "issue:44918|issue:45507", + "issue:45183|issue:46463", + "issue:46489|issue:46491", + "issue:43975|issue:44292", + "issue:44292|issue:45507" + ], + "prepared_review_unit_hash": "b69a0ee6dd1564d6cbe44e8e7d78d28d6715dbb368b830eca0282680bcaafaba", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37142, + "estimated_input_tokens": 9286, + "estimated_eval_tokens": 18828 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous transformer bug reports with only one strong near-duplicate pair around custom-module loading/caching. I would keep the cluster anchored on the open custom-code loading report and reject the rest as separate issues.", + "confidence": 0.78, + "canonical_issue_reason": "issue:45698 is the clearest and most general report in the only convincing duplicate pair; it describes the user-visible failure directly, while issue:45632 reads like a more specific cache-path collision manifestation of the same custom-module loading bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45698 is open, concrete, and broad enough to serve as the cluster anchor for the custom-code loading/caching problem.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Different bugs: int4 quantization/OOM versus tied-weights behavior in translation models." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Both are shape/indexing failures, but they point to different model paths and failure modes." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Grouped_mm tracing error in Mixtral is unrelated to tensor-parallel support in GPT-OSS MoE." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "Autocast dtype mismatch and streaming chat/completions failure are different code paths." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "GGUF architecture support and Qwen3.5 streaming inference are not the same bug." + }, + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "Unexpected init_empty_weights argument versus pydantic config-field breakage are separate regressions." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Qwen3-MoE sliding_window issue is unrelated to the pydantic model-field regression." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Grouped_mm tracing failure and Float8 storage lookup error are different runtime problems." + }, + { + "left": "issue:43716", + "right": "issue:46294", + "accept": false, + "reason": "Both are multimodal/video-adjacent, but one is a dtype mismatch and the other is a video-feature/token-count mismatch." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "Remote-code breakage and torch/transformers version incompatibility are separate issues." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Both are IndexErrors, but the mask-shape bug and the deepspeed-zero3 init bug are not the same failure." + }, + { + "left": "issue:44479", + "right": "issue:46294", + "accept": false, + "reason": "Both mention video inputs, but one is a broad v5.3.0 regression across several models and the other is a specific chat-template mismatch." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": false, + "reason": "Both concern siglip2, but one reports poor text embeddings while the other reports nonsensical pipeline/model output." + }, + { + "left": "issue:45759", + "right": "issue:45902", + "accept": false, + "reason": "Unwrapping text_config in from_config and uninitialized inv_freq after to_empty are distinct initialization bugs." + }, + { + "left": "issue:46640", + "right": "issue:46661", + "accept": false, + "reason": "Flash Attention 2 issue and remote-code model_class/config_class AttributeError are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Different multimodal/training failures: image-preprocessor dtype mismatch versus TRL unpacking of Qwen3.5 input embeddings." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "FSDP 2-D weight error and Qwen3.5 input-embedding unpacking are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "High perplexity on gpt-oss and GLM5 are separate model-specific reports." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Jupyter custom-model initialization and flash-attn2 fallback checks are unrelated." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": true, + "reason": "Both point to the same custom-module loading/caching problem: a path/cache collision leading to the wrong module being loaded after save_pretrained/from_pretrained." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Tied-weight failure in Mistral3 is a different model-specific report from the broader translation-model tied-weights regression." + }, + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "FineGrainedFP8Config failure and flash-attn2 fallback blocking are unrelated." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash is not the same as the tied-weights regression." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization/offloading behavior and tied weights are different bugs." + }, + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "A DSA indexer/ReLU discussion and a docs typo are unrelated." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint-loading failure is separate from the tied-weights regression." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "Cache-strategy typo and docs text about the removed run command are unrelated." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Qwen3.5 embedding/trainer failure and GraniteMoE invalid-method calls are unrelated." + }, + { + "left": "issue:45183", + "right": "issue:46463", + "accept": false, + "reason": "Both are validation-error complaints, but they describe different checks and different model paths." + }, + { + "left": "issue:46489", + "right": "issue:46491", + "accept": false, + "reason": "Different tokenizer regressions affecting different model families and different round-trip symptoms." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "DeepSeek-Coder detokenization and Qwen-3 NVFP4 runtime error are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:45507", + "accept": false, + "reason": "Qwen NVFP4 runtime failure and GraniteMoE invalid-method bug are different issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded and conservative overall. The summary matches the packet\u2019s evidence, and the only accepted duplicate pair (issue:45632 \u2194 issue:45698) looks like the same custom-module loading/caching bug and is plausibly mergeable into one fix. The other rejections are cautious rather than overstated." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 4, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:24643", + "issue:28530", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:41553", + "issue:41762", + "issue:42491", + "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:44368", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44589", + "issue:44661", + "issue:44910", + "issue:44928", + "issue:44960", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45356", + "issue:45399", + "issue:45520", + "issue:45663", + "issue:45685", + "issue:45701", + "issue:45823", + "issue:45859", + "issue:45902", + "issue:45920", + "issue:46489" + ], + "soft_pairs": [ + "issue:43950|issue:44534", + "issue:39692|issue:43054", + "issue:43644|issue:43950", + "issue:36331|issue:41762", + "issue:43872|issue:44589", + "issue:43828|issue:44928", + "issue:45920|issue:46489", + "issue:43054|issue:45701", + "issue:38175|issue:43054", + "issue:43950|issue:45902", + "issue:41553|issue:43883", + "issue:44910|issue:45859", + "issue:43054|issue:43493", + "issue:42947|issue:44387", + "issue:45663|issue:45902", + "issue:41553|issue:43881", + "issue:44403|issue:44451", + "issue:43975|issue:45520", + "issue:44403|issue:44488", + "issue:24643|issue:28530", + "issue:42947|issue:43856", + "issue:43643|issue:45020", + "issue:41762|issue:42915", + "issue:44991|issue:45399", + "issue:43824|issue:43872", + "issue:45685|issue:45823", + "issue:43782|issue:44960", + "issue:43881|issue:43883", + "issue:45020|issue:45823", + "issue:45081|issue:45399", + "issue:44661|issue:45020", + "issue:43828|issue:45823", + "issue:42491|issue:44368", + "issue:45020|issue:45356" + ], + "prepared_review_unit_hash": "2dcd6e6a049060b2b21eb03adb4c8a11a80699f21097aef020fbd3db8c7b74bb", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 38206, + "estimated_input_tokens": 9552, + "estimated_eval_tokens": 19360 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is dominated by one clear duplicate family around v5 non-persistent-buffer corruption, centered on issue:43950. Most other pairs are only broad-theme related; in particular issue:43643 vs issue:45020 should be treated as related-not-duplicate, not merged as the same bug.", + "confidence": 0.96, + "canonical_issue_reason": "issue:43950 is the strongest duplicate hub: it states the core non-persistent-buffer corruption regression directly and has the clearest re-report trail from nearby variants.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43950 is the best representative issue because it captures the underlying regression cleanly rather than a downstream symptom, and it anchors the tight duplicate cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "Same v5 non-persistent-buffer corruption/junk-buffer regression." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "Both are SigLIP2-related, but one is a docs/example failure and the other is a model quality complaint." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": true, + "reason": "Both describe the same non-persistent-buffer corruption during loading in Transformers 5.x." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "CustomTrainer signature breakage and Gemma3 ZeRO-3 loading failure are different code paths." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "bitsandbytes init-kwarg incompatibility and Float8 storage lookup failure are distinct issues." + }, + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Both are Qwen3.5/dtype-adjacent, but one is autocast mismatch and the other is a flash-attn crash from 3D position_ids." + }, + { + "left": "issue:45920", + "right": "issue:46489", + "accept": true, + "reason": "46489 looks like a specific DeepSeek-Coder tokenizer regression within the broader wrong-token-IDs family reported in 45920." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "Embedding quality drift and tokenizer-version behavior change are not the same bug." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "Unexpected zero probabilities and worse text embeddings are different symptoms with no clear shared root cause." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": true, + "reason": "Same non-persistent-buffer materialization/corruption problem, just reported with different models." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Bad AutoTokenizer error for Voxtral and missing all_tied_weights_keys on Molmo are unrelated." + }, + { + "left": "issue:44910", + "right": "issue:45859", + "accept": false, + "reason": "Both involve Qwen3.5 rotary/attention code, but the concrete failure modes differ." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": false, + "reason": "SigLIP2 output-quality complaint versus HF-vs-JAX implementation discrepancy are related only at a high level." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "Gradient-checkpointing ineffectiveness and int4 reserved-memory growth are different memory problems." + }, + { + "left": "issue:45663", + "right": "issue:45902", + "accept": false, + "reason": "Gemma4 FSDP2 shared_kv_states rebuild bug is separate from the Qwen3_5Moe uninitialized buffer issue." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Different model-loading failures with different root causes." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic load-noise complaint and inability to load a specific model are not the same bug." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "DeepSeek-Coder detokenization regression and Python 3.13 flash_attn import failure are unrelated." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "These are separate model-loading complaints without a shared concrete code path." + }, + { + "left": "issue:24643", + "right": "issue:28530", + "accept": false, + "reason": "DeepSpeed weight-shape error and early-stopping metric lookup are unrelated training issues." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "LoRA checkpointing inefficiency and Qwen3 MoE memory usage are both memory-related but not the same bug." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "They share a broad remote_code theme, but the packet only supports related loading regressions, not the same concrete defect." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Gemma3 ZeRO-3 loading failure and Qwen3Moe FineGrainedFP8 failure are different model/config problems." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer loading failure and flash-attn2 fallback gating are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "Qwen2.5-VL import failure and bitsandbytes Int8Params incompatibility are separate issues." + }, + { + "left": "issue:45685", + "right": "issue:45823", + "accept": false, + "reason": "MPS histogram op-not-implemented and Gemma4 device mismatch are different runtime/model failures." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Qwen3VL weight_only load error and GLM5 are unrelated." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "Both are loading failures, but on different models and different causes." + }, + { + "left": "issue:45020", + "right": "issue:45823", + "accept": false, + "reason": "Remote_code loading regressions and Gemma4 PLE device mismatch are not the same bug." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Mistral tokenizer regex crash and flash-attn2 fallback gating are separate tokenizer vs attention-selection problems." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "Tokenizer-mapping add-new-model-like failure and remote_code regression are different code paths." + }, + { + "left": "issue:43828", + "right": "issue:45823", + "accept": false, + "reason": "Autocast dtype mismatch in Phi-tiny-MoE and Gemma4 device placement mismatch are unrelated." + }, + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "LoRA incompatibility across versions and a tie_word_embeddings warning are not the same underlying issue." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "Broad remote_code breakage and Kimi-K2.5 tokenizer regression are different bugs despite both touching v5 loading behavior." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Mostly grounded, but one soft-edge acceptance is too strong for the packet: issue:43643 vs issue:45020 share a broad `remote_code` theme, yet the packet only supports them as related loading regressions, not clearly the same concrete bug. I\u2019d keep that pair as related-not-duplicate. The non-persistent-buffer cluster summaries look well supported." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 5, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36331", + "issue:41093", + "issue:41553", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43366", + "issue:43404", + "issue:43531", + "issue:43596", + "issue:43638", + "issue:43792", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43872", + "issue:43940", + "issue:43950", + "issue:44220", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44530", + "issue:44560", + "issue:44805", + "issue:44863", + "issue:44928", + "issue:44991", + "issue:45020", + "issue:45305", + "issue:45313", + "issue:45325", + "issue:45488", + "issue:45588", + "issue:45663", + "issue:45715", + "issue:45759", + "issue:45823", + "issue:45902", + "issue:45923", + "issue:46315", + "issue:46402", + "issue:46489", + "issue:46515", + "issue:46661" + ], + "soft_pairs": [ + "issue:45715|issue:46315", + "issue:43827|issue:44509", + "issue:45488|issue:46489", + "issue:45759|issue:46661", + "issue:43872|issue:44291", + "issue:43404|issue:43950", + "issue:42915|issue:43872", + "issue:45020|issue:46661", + "issue:43792|issue:44220", + "issue:43257|issue:45305", + "issue:41553|issue:44991", + "issue:42947|issue:45923", + "issue:36331|issue:41093", + "issue:43940|issue:45313", + "issue:43854|issue:44863", + "issue:44479|issue:44560", + "issue:44928|issue:45588", + "issue:41553|issue:44488", + "issue:41093|issue:44805", + "issue:24643|issue:30064", + "issue:44530|issue:44863", + "issue:44863|issue:46402", + "issue:43366|issue:43828", + "issue:43596|issue:46515", + "issue:43366|issue:43531", + "issue:45325|issue:45902", + "issue:43638|issue:46515", + "issue:45663|issue:45823" + ], + "prepared_review_unit_hash": "ca5a4ced789ca38f1cbb68176da25ab3740e33a55301f9a73099c8372bb12df5", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36634, + "estimated_input_tokens": 9159, + "estimated_eval_tokens": 18574 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: many unrelated Transformers bugs around loading, tokenizers, DeepSpeed, and model-specific regressions. The only clear duplicates are the docs pipeline-removal pair, the mask-shape IndexError pair, and the bitsandbytes/init_empty_weights _is_hf_initialized pair.", + "confidence": 0.9, + "canonical_issue_reason": "Issue 45020 is the broadest representative of the cluster\u2019s dominant theme: recent v5 regressions that break model loading, especially for remote-code models. It\u2019s more central than the highly specific model- or error-message-only reports.", + "canonical_pr_reason": null, + "best_issue_reason": "45020 is the best single issue to anchor this cluster because it captures the widest shared failure mode across several nearby loading/regression reports, even though most items are not true duplicates of it.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45715", + "right": "issue:46315", + "accept": false, + "reason": "Both are tokenizer performance regressions, but they hit different APIs and different slow-paths (skip_special_tokens iteration vs many added tokens)." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": true, + "reason": "Same docs regression: v5 removed pipeline tasks but the documentation still references summarization/translation pipeline usage." + }, + { + "left": "issue:45488", + "right": "issue:46489", + "accept": false, + "reason": "Both concern DeepSeek tokenization, but they describe different model families and different failure modes; not the same concrete bug." + }, + { + "left": "issue:45759", + "right": "issue:46661", + "accept": false, + "reason": "Both are model-loading/config issues, but one is composite Qwen multimodal config handling and the other is remote-code model_class/config_class handling." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "Same root cause: unexpected _is_hf_initialized handling in model init/bitsandbytes paths under transformers 5.x / init_empty_weights." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "One is an untied lm_head weight in a specific model class; the other is silent corruption of non-persistent buffers in from_pretrained." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "Different failure classes: FineGrainedFP8Config on Qwen3Moe vs bitsandbytes/_is_hf_initialized constructor incompatibility." + }, + { + "left": "issue:45020", + "right": "issue:46661", + "accept": false, + "reason": "Related broadly to remote-code loading, but not the same bug or code path." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": false, + "reason": "Both involve Whisper/audio, but the titles point to different layers: a model failing to run vs a specific feature-extraction helper issue." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Both involve DeepSpeed and Qwen3-family training/loading, but one is weight conversion/loading and the other is gradient averaging with GAS." + }, + { + "left": "issue:41553", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading problems, but they concern different models and different symptoms." + }, + { + "left": "issue:42947", + "right": "issue:45923", + "accept": false, + "reason": "Both mention gradient checkpointing, but one is a PEFT/LoRA behavior bug and the other is a missing model capability flag on a remote-code variant." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "Completely different errors: Trainer compute_loss signature mismatch vs tensor mask shape mismatch." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Both are ZeRO-3 weight-loading issues, but the affected model families and missing parameters differ enough that they are not clearly the same bug." + }, + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "Both are model loading failures, but for unrelated architectures and no shared concrete code path." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": false, + "reason": "Both are Qwen video regressions in v5.3.0, but one is a broad input regression and the other is a StopIteration in a specific embedding variant." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "Different attention-related failures: gradient explosion from position_ids/SDPA fallback vs flash_attention crash on s_aux=None." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Different tokenizer-loading reports for different models; only the theme overlaps." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": true, + "reason": "These are the same mask/tensor shape mismatch IndexError, just with different concrete shapes reported." + }, + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "Unrelated: DeepSpeed training weight shape error vs segmentation-map preprocessing bug." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "Different bugs in different models: PagedAttentionCache/group type failure vs checkpoint loading failure." + }, + { + "left": "issue:44863", + "right": "issue:46402", + "accept": false, + "reason": "Both are loading problems, but one is NemotronH checkpoint loading and the other is Sam3 text encoder weight loading." + }, + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "Unrelated: GGUF gpt-oss support vs autocast dtype mismatch in Phi-tiny-MoE." + }, + { + "left": "issue:43596", + "right": "issue:46515", + "accept": false, + "reason": "No common bug: ZeRO-3/BertModel index error vs torch 2.4.1 incompatibility." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "Different model-support issues: gpt-oss GGUF vs Qwen3-MoE sliding_window behavior." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "Both are position/rotary-embedding issues, but they affect different model families and distinct code paths." + }, + { + "left": "issue:43638", + "right": "issue:46515", + "accept": false, + "reason": "Different root causes and symptoms: ZeRO-3/Bert init index error vs torch version incompatibility." + }, + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Both are Gemma4 bugs, but one is an FSDP2 shared_kv_states KeyError and the other is a device placement mismatch." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet and stays conservative about heterogeneity. The three accepted soft edges are well supported by the issue titles and descriptions: the docs pipeline-removal pair, the mask-shape IndexError pair, and the bitsandbytes/init_empty_weights _is_hf_initialized pair. The rejected links are generally framed cautiously as related themes rather than duplicates, which fits the evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 6, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:36010", + "issue:38175", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42617", + "issue:42915", + "issue:43317", + "issue:43366", + "issue:43550", + "issue:43596", + "issue:43632", + "issue:43638", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43720", + "issue:43824", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44155", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44683", + "issue:44756", + "issue:44805", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:44960", + "issue:44977", + "issue:45030", + "issue:45276", + "issue:45584", + "issue:45685", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:46459", + "issue:46489", + "issue:46512", + "issue:46515" + ], + "soft_pairs": [ + "issue:43366|issue:44589", + "issue:43716|issue:44560", + "issue:44155|issue:45584", + "issue:43824|issue:43957", + "issue:43950|issue:45834", + "issue:42915|issue:43957", + "issue:44560|issue:45685", + "issue:43645|issue:43646", + "issue:44560|issue:44918", + "issue:43550|issue:44910", + "issue:44683|issue:46515", + "issue:42617|issue:43366", + "issue:36010|issue:42915", + "issue:45812|issue:46489", + "issue:45823|issue:46515", + "issue:44560|issue:44805", + "issue:44661|issue:45823", + "issue:38175|issue:43994", + "issue:43632|issue:45834", + "issue:44877|issue:45030", + "issue:43638|issue:44661", + "issue:43596|issue:44661", + "issue:43317|issue:44756", + "issue:44661|issue:44805", + "issue:42491|issue:43720", + "issue:43644|issue:44534", + "issue:38175|issue:39692", + "issue:41720|issue:42915", + "issue:42915|issue:44756", + "issue:42491|issue:45276", + "issue:44451|issue:44488", + "issue:44960|issue:44977", + "issue:46459|issue:46512" + ], + "prepared_review_unit_hash": "bcb268c67da88bf3e3e8f6d2778ed1775fa2ea56f07fafde88e0c9f0bcf44519", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37470, + "estimated_input_tokens": 9368, + "estimated_eval_tokens": 18992 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated issue reports with a few strong duplicate clusters around the non-persistent-buffer v5 regression, SigLIP2 output issues, custom-model initialization, and config-validation load failures. I downgraded the BERT pair 44451/44488 to a likely-related symptom rather than a confirmed duplicate, since the titles alone don\u2019t establish the same underlying regression.", + "confidence": 0.81, + "canonical_issue_reason": "issue:43644 is the clearest and earliest statement of the non-persistent-buffer corruption regression, and issue:44534 is the same bug described later with slightly different wording.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43644 best anchors the cluster because it states the core regression directly and also has the broadest wording among the duplicate candidates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "Different bug families: GGUF gpt-oss support vs a Float8 storage lookup failure." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Different model/path: image preprocessor dtype mismatch vs Qwen3-VL video StopIteration." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "Different audio/embedding leak vs Whisper empty-transcription handling." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "Import/export failure for a missing class vs meta-device model loading issues." + }, + { + "left": "issue:43950", + "right": "issue:45834", + "accept": false, + "reason": "Different regressions: non-persistent buffer corruption vs long-OCR index error." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "Qwen3Moe FP8 failure and meta-device loading are distinct code paths." + }, + { + "left": "issue:44560", + "right": "issue:45685", + "accept": false, + "reason": "Different subsystems: video inference StopIteration vs MPS histogram on Int." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "Same custom-model initialization regression; the notebook report is a narrower manifestation of the broader issue." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Unrelated failures: video StopIteration vs mask/tensor shape mismatch." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different attention/compile bugs affecting different models and failure modes." + }, + { + "left": "issue:44683", + "right": "issue:46515", + "accept": false, + "reason": "Both mention torch versions, but they are different compatibility problems on different releases." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "Different features: 3d_parallel script failure vs gpt-oss GGUF support." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "Missing GenerationMixin import is unrelated to Qwen3Moe FP8 failures." + }, + { + "left": "issue:45812", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer regressions, but the affected model families and likely fixes differ." + }, + { + "left": "issue:45823", + "right": "issue:46515", + "accept": false, + "reason": "Gemma4 device mismatch is unrelated to a general torch 2.4.1 incompatibility report." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Different concrete bugs: video StopIteration vs tensor/mask index mismatch." + }, + { + "left": "issue:44661", + "right": "issue:45823", + "accept": false, + "reason": "add-new-model-like filtering and Gemma4 device placement are unrelated." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": true, + "reason": "Same SigLIP2 regression on google/siglip2-base-patch16-224; both describe incorrect outputs from the same model." + }, + { + "left": "issue:43632", + "right": "issue:45834", + "accept": false, + "reason": "_is_hf_initialized flag regression is unrelated to Kosmos2.5 long-input indexing." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": true, + "reason": "Both report stricter config validation blocking otherwise valid model configs, so they fit one fixable load-failure class." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "Zero3/Bert initialization failure is unrelated to TOKENIZER_MAPPING_NAMES handling." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Same as above: a zero-sized tensor error under deepspeed is not the tokenizer-mapping bug." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "Device-map/offload loading failure and mmap OOM on Strix Halo are different problems." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer-mapping logic and mask-shape mismatch are separate bugs." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "Qwen3 MoE LoRA compatibility and BitNet packed-weight unpacking are unrelated." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Exact same non-persistent-buffer junk regression in Transformers v5, just reworded." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "Same model family, but one is an actual inference regression while the other is a documentation/example problem." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Different Qwen3 issues: device-mapping cuda assert vs FineGrainedFP8Config failure." + }, + { + "left": "issue:42915", + "right": "issue:44756", + "accept": false, + "reason": "Qwen3Moe FP8 failure is unrelated to Strix Halo mmap/OOM behavior." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "Different model families and different failures: LoRA reuse vs resize_token_embeddings not propagating." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": false, + "reason": "These are similar BERT load failures, but the titles alone don\u2019t show the same underlying loading regression, so this is only a likely-related symptom pair." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "GLM5 and Qwen3.5 flash-attention failures are different model/code-path issues." + }, + { + "left": "issue:46459", + "right": "issue:46512", + "accept": false, + "reason": "Silent dtype handling for composite checkpoints is distinct from stale config.dtype after from_config()." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Most of the summary is grounded in the packet, but it overstates at least one soft-edge accept: issue:44451 vs issue:44488 are clearly similar BERT load failures, but the titles alone do not justify claiming the same underlying loading regression. I\u2019d downgrade that to a cautious \u2018likely related / similar symptom\u2019 rather than a true duplicate. The rest of the accepted pairs look more defensible." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 7, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:36010", + "issue:39404", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42947", + "issue:43257", + "issue:43404", + "issue:43425", + "issue:43454", + "issue:43644", + "issue:43645", + "issue:43708", + "issue:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44075", + "issue:44155", + "issue:44164", + "issue:44291", + "issue:44387", + "issue:44402", + "issue:44488", + "issue:44756", + "issue:44910", + "issue:44960", + "issue:44991", + "issue:45276", + "issue:45292", + "issue:45313", + "issue:45335", + "issue:45357", + "issue:45362", + "issue:45406", + "issue:45488", + "issue:45507", + "issue:45588", + "issue:45663", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45859", + "issue:46050", + "issue:46468" + ], + "soft_pairs": [ + "issue:45507|issue:46468", + "issue:45313|issue:45357", + "issue:43644|issue:43645", + "issue:44488|issue:44991", + "issue:43856|issue:43873", + "issue:43856|issue:44387", + "issue:39692|issue:43994", + "issue:42491|issue:45292", + "issue:43257|issue:43854", + "issue:43708|issue:44075", + "issue:43854|issue:43931", + "issue:42491|issue:43931", + "issue:44960|issue:46050", + "issue:44164|issue:44291", + "issue:45488|issue:45812", + "issue:45362|issue:45406", + "issue:45276|issue:45335", + "issue:44910|issue:45588", + "issue:44402|issue:44488", + "issue:43425|issue:43957", + "issue:43873|issue:44387", + "issue:39692|issue:44960", + "issue:41720|issue:44155", + "issue:42947|issue:45704", + "issue:45663|issue:45759", + "issue:43404|issue:43454", + "issue:44291|issue:44756", + "issue:36010|issue:39404", + "issue:44387|issue:44756", + "issue:45663|issue:45859" + ], + "prepared_review_unit_hash": "425fc65696a75267e6ed61aa06d4b90d1b9ca165b314d7bb97f72677780a5e10", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36207, + "estimated_input_tokens": 9052, + "estimated_eval_tokens": 18360 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most candidate pairs are thematically related but not true duplicates. I\u2019d merge two pairs as the same underlying code-path bug: the `resize_token_embeddings` regressions and the missing `lm_head` weight-tying regressions. The rest are better treated as separate issues.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 45276 is the strongest canonical issue: it describes a concrete, user-visible regression in `resize_token_embeddings` with clear affected outputs, and its paired issue is the narrower variant of the same behavior on a related model family.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45276 is the best global issue anchor because it is specific, actionable, and broad enough to represent the shared embedding-resize regression pattern across the duplicate pair.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45507", + "right": "issue:46468", + "accept": false, + "reason": "Different models and failure modes; both are generic broken-example reports, not the same underlying bug." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "Both involve Qwen3.5, but one is ZeRO-3 weight loading and the other is incorrect save_pretrained visual encoder keys; separate code paths." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "Both are Transformers 5.0 regressions, but one is non-persistent buffers and the other is Jupyter custom-model init; unrelated bugs." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer/model loading failures, but they affect different models and likely different tokenizer code paths." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "One is Qwen3 MoE training memory usage, the other is quantization offloading; not the same bug." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both mention memory/OOM, but one is training memory inefficiency and the other is int4 quantization reserved-memory growth; distinct issues." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Shared SigLIP2 theme, but 39692 is a docs/example mismatch plus quantization failure while 43994 is bad outputs from AutoModel/pipeline; not the same concrete defect." + }, + { + "left": "issue:42491", + "right": "issue:45292", + "accept": false, + "reason": "Different models and different APIs; one is LoRA compatibility across branches, the other is resize_token_embeddings not updating output embeddings." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Different models and loading failures; no shared code-path evidence." + }, + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Trainer checkpoint resumption math and SGD argument handling are unrelated." + }, + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Different model-loading failures on different models; no sign of a shared root cause." + }, + { + "left": "issue:42491", + "right": "issue:43931", + "accept": false, + "reason": "LoRA branch-compatibility problem versus Qwen3-VL shape mismatch; separate issues." + }, + { + "left": "issue:44960", + "right": "issue:46050", + "accept": false, + "reason": "GLM5 model-loading issue versus sinq quantization save/load corruption; different bugs." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "Both touch save/load paths, but extra_state handling and unexpected `_is_hf_initialized` during init_empty_weights are different failures." + }, + { + "left": "issue:45488", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different tokenizer classes/models and the underlying changes are not clearly the same." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Different frontends and different model families; no shared underlying bug is evident." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": true, + "reason": "Both report the same `resize_token_embeddings` regression: the resize does not propagate to downstream embeddings/output heads in related composite model classes." + }, + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Both mention flash attention, but one is a position_ids packing crash and the other is an s_aux=None AttributeError; different code paths." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "Both are tokenizer-loading complaints, but they concern different models and likely different special-token handling bugs." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "General version incompatibility versus meta-device loading failures; too broad and not the same defect." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Offloading/quantization behavior versus int4 reserved-memory OOM; separate memory problems." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "SigLIP2 docs/example issue and GLM5 loading issue are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "Qwen3 auto device-map CUDA assert versus AudioFlamingo3 batched inference leakage; unrelated." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing with PEFT LoRA is unrelated to the apex RMSNorm memory leak." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "Gemma-4 FSDP2 KeyError and Qwen multimodal config unwrapping are different model-internal bugs." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": true, + "reason": "Both describe missing `lm_head` weight tying in multimodal generation models, causing incorrect or garbage generation; same underlying fix class." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "Meta/init_empty_weights loading error versus disabling mmap to avoid OOM; unrelated." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "ImportError for GenerationMixin is unrelated to Whisper pipeline return_language behavior." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "Different OOM causes: quantization reserved memory versus mmap on Strix Halo." + }, + { + "left": "issue:45663", + "right": "issue:45859", + "accept": false, + "reason": "Both involve Qwen/Gemma model internals, but one is FSDP2 shared_kv_states and the other is CPU-offload rotary embedding compatibility." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays mostly conservative. The two accepted pairs are well-supported by the issue titles: 45276/45335 both describe `resize_token_embeddings` failing to propagate to downstream embeddings, and 43404/43454 both describe missing `lm_head` weight tying causing bad generation. The remaining pairs are reasonably treated as separate issues based on the available evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 8, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:39404", + "issue:42796", + "issue:42947", + "issue:43335", + "issue:43381", + "issue:43425", + "issue:43611", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43716", + "issue:43828", + "issue:43854", + "issue:43906", + "issue:43957", + "issue:44155", + "issue:44164", + "issue:44292", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44530", + "issue:44589", + "issue:44683", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44877", + "issue:44912", + "issue:44928", + "issue:45084", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45588", + "issue:45663", + "issue:45684", + "issue:45685", + "issue:45698", + "issue:45701", + "issue:45812", + "issue:45823", + "issue:46315", + "issue:46326", + "issue:46468", + "issue:46491", + "issue:46665" + ], + "soft_pairs": [ + "issue:44292|issue:44912", + "issue:43957|issue:44756", + "issue:44155|issue:45823", + "issue:42947|issue:43381", + "issue:43335|issue:45663", + "issue:44863|issue:46468", + "issue:45276|issue:45292", + "issue:45685|issue:46468", + "issue:44479|issue:44877", + "issue:44164|issue:44756", + "issue:43381|issue:44928", + "issue:45588|issue:45663", + "issue:45684|issue:45698", + "issue:43854|issue:45216", + "issue:42796|issue:46326", + "issue:43828|issue:43957", + "issue:45292|issue:45335", + "issue:43646|issue:43708", + "issue:44402|issue:44451", + "issue:43645|issue:44164", + "issue:44589|issue:44756", + "issue:43425|issue:43828", + "issue:45812|issue:46315", + "issue:46315|issue:46491", + "issue:39404|issue:45701", + "issue:43611|issue:43645", + "issue:39404|issue:46665", + "issue:44530|issue:45084", + "issue:44683|issue:45823", + "issue:43425|issue:45823", + "issue:43425|issue:43716", + "issue:44589|issue:45084", + "issue:44479|issue:44805", + "issue:43906|issue:44530" + ], + "prepared_review_unit_hash": "2b12845dfcf2cc21fc9bb260407d4b879534089da837b66d82ec2a174cf65f10", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37883, + "estimated_input_tokens": 9471, + "estimated_eval_tokens": 19198 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: tokenizer regressions, checkpointing bugs, model-loading breakages, and hardware/dtype issues do not look like one duplicate set. The only clear near-duplicate subcluster is the resize_token_embeddings/output_embeddings reports around 45292.", + "confidence": 0.66, + "canonical_issue_reason": "Issue 45292 is the broadest and cleanest representative of the resize_token_embeddings/output_embeddings bug family; the Gemma4 and t5gemma reports look like narrower variants of the same underlying propagation problem.", + "canonical_pr_reason": null, + "best_issue_reason": "45292 is the strongest standalone issue to keep as the representative bug report because it captures the shared API-level failure without overfitting to one model family.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Different models and different quantization/loading symptoms; no shared code-path evidence." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "Meta-device loading breakage and mmap/OOM mitigation are unrelated problems." + }, + { + "left": "issue:44155", + "right": "issue:45823", + "accept": false, + "reason": "Audio track token leakage and Gemma4 device mismatch are distinct bugs in different subsystems." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Both mention gradient checkpointing, but one is LoRA inefficacy and the other is eval-mode incompatibility." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "Sparse-layer construction and Gemma4 shared_kv_states reuse are different model-specific defects." + }, + { + "left": "issue:44863", + "right": "issue:46468", + "accept": false, + "reason": "NemotronH checkpoint loading and a broken Nemotron example are related only by model family, not by bug." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same resize_token_embeddings propagation bug; 45276 is just a Gemma4-specific manifestation of the broader output_embeddings failure." + }, + { + "left": "issue:45685", + "right": "issue:46468", + "accept": false, + "reason": "MPS histogram support and the Nemotron example issue are unrelated." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Video input regression for Qwen VL models and strict granite_speech config loading are separate issues." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "save/from_pretrained extra_state handling and Strix Halo mmap OOM are different code paths." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Eval-mode checkpointing and RLHF NaN explosions from 3D position_ids are not the same bug." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "flash_attention.py s_aux=None crash and Gemma4 FSDP2 shared_kv_states rebuilding are unrelated." + }, + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "File permission propagation and wrong custom-module loading after save_pretrained are distinct failures." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "GLM-4.7-Flash test-loading failure and Qwen3.5 save_pretrained regression do not share a concrete bug." + }, + { + "left": "issue:42796", + "right": "issue:46326", + "accept": false, + "reason": "BPE marker handling and continue_final_message behavior are different tokenizer problems." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Autocast dtype mismatch and torch.device(\"meta\") loading issues are different failures." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Both report the same resize_token_embeddings propagation bug affecting downstream embedding/output layers; model name differs, not the underlying defect." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model initialization in notebooks and Trainer resume max_steps miscalculation are unrelated." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Tokenizer vocab-size mismatch and ScandiBERT loading failure are not the same underlying bug." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Notebook custom-model initialization and extra_state save/load handling are separate regressions." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Missing Float8 storage support and mmap OOM avoidance are unrelated." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Torch version incompatibility and autocast dtype mismatch are different issues." + }, + { + "left": "issue:45812", + "right": "issue:46315", + "accept": false, + "reason": "Granite token ID regression is correctness; convert_tokens_to_ids slowdown is performance only." + }, + { + "left": "issue:46315", + "right": "issue:46491", + "accept": false, + "reason": "Slow added-token lookup and CodeLlama leading-space round-trip are distinct tokenizer behaviors." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper return_language pipeline breakage and tokenization-version differences are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "Both involve custom models, but one is base_model_prefix loading and the other is notebook initialization semantics." + }, + { + "left": "issue:39404", + "right": "issue:46665", + "accept": false, + "reason": "Whisper pipeline regression and from_pretrained inside Python classes are unrelated." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache linear_attention failure and compile non-template-node error are different code paths." + }, + { + "left": "issue:44683", + "right": "issue:45823", + "accept": false, + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 device mismatch are unrelated." + }, + { + "left": "issue:43425", + "right": "issue:45823", + "accept": false, + "reason": "Torch 2.10 compatibility and Gemma4 forward device placement are different problems." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "Torch compatibility and Mistral-3 image-preprocessor dtype mismatch are not the same bug." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage lookup and template-node compilation errors are unrelated." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "Qwen video-input regression and mask-shape IndexError are different defects." + }, + { + "left": "issue:43906", + "right": "issue:44530", + "accept": false, + "reason": "A reproduction of another issue and a PagedAttentionCache crash do not match the same underlying bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet: most pairs are clearly different bug families, and the resize_token_embeddings/output_embeddings trio around 45292 is the only convincingly related subcluster. The accepted soft-edge pairs are conservative and supported by the issue titles, and none of the reject verdicts overstate similarity." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 9, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:32090", + "issue:36032", + "issue:39404", + "issue:42796", + "issue:42915", + "issue:43299", + "issue:43404", + "issue:43425", + "issue:43452", + "issue:43541", + "issue:43596", + "issue:43611", + "issue:43643", + "issue:43644", + "issue:43716", + "issue:43856", + "issue:43901", + "issue:43906", + "issue:43931", + "issue:43994", + "issue:44079", + "issue:44164", + "issue:44265", + "issue:44462", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44530", + "issue:44661", + "issue:44671", + "issue:44683", + "issue:44737", + "issue:44756", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45759", + "issue:45859", + "issue:46050", + "issue:46285", + "issue:46326", + "issue:46402", + "issue:46489" + ], + "soft_pairs": [ + "issue:43901|issue:44509", + "issue:43425|issue:44918", + "issue:43994|issue:44079", + "issue:44265|issue:44898", + "issue:44671|issue:44991", + "issue:36032|issue:43452", + "issue:39404|issue:44912", + "issue:44164|issue:45759", + "issue:46050|issue:46326", + "issue:43299|issue:43931", + "issue:43906|issue:46326", + "issue:44991|issue:45081", + "issue:43856|issue:44756", + "issue:43596|issue:44683", + "issue:44164|issue:44479", + "issue:44488|issue:45081", + "issue:44488|issue:44987", + "issue:43611|issue:43644", + "issue:44462|issue:46489", + "issue:42796|issue:43404", + "issue:43716|issue:45084", + "issue:36032|issue:42915", + "issue:44737|issue:45859", + "issue:32090|issue:39404", + "issue:46285|issue:46402", + "issue:44661|issue:44683", + "issue:43643|issue:44530", + "issue:43541|issue:45084" + ], + "prepared_review_unit_hash": "fb61fa34205899f846421398544040c07b191a8c0f6e1551922283273956fc90", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35663, + "estimated_input_tokens": 8916, + "estimated_eval_tokens": 18088 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a loose collection of Transformers v5-era regressions, mostly around loading/tokenizers/model plumbing, but the soft pairs are not true duplicates of one another. They share broad subsystem similarity, not the same underlying bug.", + "confidence": 0.79, + "canonical_issue_reason": "issue:43611 is the broadest and most representative issue here: a general Transformers 5.0 model-loading regression centered on core loading behavior. It best matches the cluster\u2019s overall theme without overfitting to one model family.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43611 is the best issue to anchor this cluster because it is a high-level loading regression that reflects the common pattern across the set, while still being concrete enough to label the group.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "Both are docs-only mismatches, but they concern different documentation problems: return_all_scores vs removed pipeline task names." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "One is a Torch 2.10 compatibility issue; the other is a Qwen3.5 SFT trainer embedding-unpacking failure. Different root causes and code paths." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "SigLIP2 bad outputs is a model/pipeline behavior bug; ModelOutput key assignment with None is a generic container bug. Not the same bug." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "torch.export failure from torch_compilable_check is unrelated to Perceiver interpolation / non-default resolution handling." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "CamemBERT MLM prediction regressions and est-roberta tokenizer loading are different model-specific failures, not one shared defect." + }, + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "T5Tokenizer method-name conflict and gguf_file loading breakage are different tokenizer/loading paths with different failure modes." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper return_language pipeline regression and MXFP4 quantization fallback are unrelated model-specific issues." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "save/from_pretrained extra_state handling and text_config unwrapping in from_config are distinct serialization/config bugs." + }, + { + "left": "issue:46050", + "right": "issue:46326", + "accept": false, + "reason": "Quantized save/load corruption and continue_final_message behavior for mistral_common tokenizers are different concrete problems." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3VL MoE loading breakage from a dev version and Qwen3-VL-30B-A3B-Instruct shape mismatch are not the same loading defect." + }, + { + "left": "issue:43906", + "right": "issue:46326", + "accept": false, + "reason": "Isolated reproduction of an earlier issue and mistral_common continue_final_message are unrelated." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer loading for EMBEDDIA/est-roberta and _patch_mistral_regex backend_tokenizer crashes are different tokenizer failures." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "Both mention memory, but one is Qwen3 MoE training efficiency and the other is mmap/OOM on Strix Halo. Different causes and mitigations." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "Deepspeed ZeRO3/BertModel index error and compiled flex_attention torch>=2.9 failure are unrelated runtime/compiler issues." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "extra_state save/load handling and video-input regression in v5.3 are different bugs." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Loading cjvt/sleng-bert and mistral regex patch crashes are separate tokenizer-loading failures." + }, + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "cjvt/sleng-bert loading and physical-intelligence/fast loading failures are distinct model/tokenizer issues." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "Both are v5 regressions, but one is base_model_prefix loading while the other is junk-filled non-persistent buffers. Not the same code-path bug." + }, + { + "left": "issue:44462", + "right": "issue:46489", + "accept": false, + "reason": "Ignoring tokenizer.json in AutoTokenizer and DeepSeek-Coder tokenizer corruption are different tokenizer bugs." + }, + { + "left": "issue:42796", + "right": "issue:43404", + "accept": false, + "reason": "Ministral tokenizer BPE marker handling and Mistral3 lm_head weight tying are unrelated issues in the same general family." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Image preprocessor dtype mismatch and template-node compilation failure are different multimodal/compilation problems." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "T5 tokenizer special-token conflict and Qwen3Moe FineGrainedFP8Config failure are unrelated." + }, + { + "left": "issue:44737", + "right": "issue:45859", + "accept": false, + "reason": "XLNet CPU allocation bug and Qwen3.5 MoE rotary embedding CPU-offload incompatibility are different device-placement issues." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer _gpu_broadcast_one NoneType error and Whisper return_language pipeline regression are unrelated." + }, + { + "left": "issue:46285", + "right": "issue:46402", + "accept": false, + "reason": "CLIP structure inconsistency and Sam3Model text encoder weight loading are different model-loading regressions." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "add-new-model-like mapping behavior and torch.export / torch_compilable_check failures are different developer/runtime paths." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "AutoConfig trust_remote_code field loss and PagedAttentionCache linear_attention crashes are unrelated bugs." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Torch dynamo grouped_mm tracing failure in Mixtral and non-template-node compilation errors are different compiler/tracing 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": 23, + "review_unit_index": 10, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:28530", + "issue:34567", + "issue:36032", + "issue:40444", + "issue:40990", + "issue:42175", + "issue:42915", + "issue:43262", + "issue:43317", + "issue:43377", + "issue:43404", + "issue:43421", + "issue:43452", + "issue:43482", + "issue:43582", + "issue:43632", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43742", + "issue:43746", + "issue:43819", + "issue:43883", + "issue:43994", + "issue:44206", + "issue:44220", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44464", + "issue:44466", + "issue:44488", + "issue:44530", + "issue:44534", + "issue:44589", + "issue:44610", + "issue:44811", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45507", + "issue:45584", + "issue:45987", + "issue:46178", + "issue:46326" + ], + "soft_pairs": [ + "issue:43582|issue:43632", + "issue:42915|issue:44530", + "issue:44451|issue:44987", + "issue:40990|issue:43421", + "issue:43644|issue:43646", + "issue:34567|issue:40444", + "issue:42175|issue:46178", + "issue:44811|issue:45584", + "issue:44206|issue:44220", + "issue:43883|issue:44488", + "issue:36032|issue:43645", + "issue:40444|issue:46326", + "issue:45084|issue:45507", + "issue:43452|issue:43482", + "issue:43317|issue:43482", + "issue:44987|issue:44991", + "issue:43262|issue:43377", + "issue:43746|issue:46326", + "issue:42915|issue:45084", + "issue:43883|issue:44991", + "issue:44403|issue:45081", + "issue:44464|issue:45987", + "issue:43646|issue:44534", + "issue:43645|issue:44534", + "issue:44849|issue:45325", + "issue:43404|issue:44466", + "issue:43377|issue:43819", + "issue:42915|issue:44291", + "issue:43994|issue:44610", + "issue:44451|issue:45081", + "issue:44987|issue:45081", + "issue:28530|issue:34567", + "issue:36032|issue:43742", + "issue:42915|issue:44589" + ], + "prepared_review_unit_hash": "b488ed01e7de08cad4f42bd40f0731353b0333175ca6088c0ea8b8f229932a14", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 38005, + "estimated_input_tokens": 9502, + "estimated_eval_tokens": 19260 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous, with mostly unrelated bug reports plus one clear duplicate family around GGUF loading. There is no meaningful PR cluster here; the best issue anchor is the broad v5 regression ticket that has the clearest duplicate relationship in the set.", + "confidence": 0.71, + "canonical_issue_reason": "Issue 43644 is the clearest duplicate anchor among the items: it is a concrete v5 regression and has an almost verbatim twin in 44534, making it the strongest representative of the cluster\u2019s most duplicate-like subthread.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43644 is the best single representative of the set because it is the cleanest, most specific duplicate-worthy report, while most other items are unrelated one-off regressions or model-specific failures.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Different failures: Apple Silicon allocator warmup TypeError vs. _is_hf_initialized flag regression." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "Both mention Qwen, but one is FP8 config failure and the other is PagedAttentionCache linear_attention handling; different code paths." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Both are model-loading failures, but for different models and likely different regressions." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "Perplexity issue vs. tokenizer backend post-processor update bug; unrelated behavior." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 regressions, but one is about non-persistent buffers and the other about custom model initialization; not the same bug." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "Trainer token-count tracking vs. Qwen2.5-VL multi-image iterable dataset failure; unrelated." + }, + { + "left": "issue:42175", + "right": "issue:46178", + "accept": false, + "reason": "Both concern backend packaging/registration, but they describe different backend problems." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Both involve Whisper, but one is batch_decode skip_special_tokens and the other is empty transcription after align_special_tokens; different defects." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": false, + "reason": "Both are audio-feature related, but the reported crashes and code paths differ." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Missing all_tied_weights_keys in Molmo vs. failure loading a specific BERT model; unrelated." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "Tokenizer API conflict vs. custom model initialization breakage; different problems." + }, + { + "left": "issue:40444", + "right": "issue:46326", + "accept": false, + "reason": "Qwen2.5-VL multi-image batching vs. mistral_common continue_final_message behavior; unrelated." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "TypeError compiling template nodes vs. GraniteMoEHybrid invalid method call; unrelated." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": true, + "reason": "Both point to the same GGUF loading regression: generic gguf_file/from_pretrained breakage, with 43482 a specific Qwen2.5-GGUF instance." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "device_map=auto offload/dequantization failure is a different code path from GGUF loading." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Both are model/tokenizer loading failures, but they affect different models and do not show the same underlying bug." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Audio processor sampling-rate defaulting vs. MIMI batch-padding-mask discrepancy; unrelated." + }, + { + "left": "issue:43746", + "right": "issue:46326", + "accept": false, + "reason": "PEFT/local checkpoint loading vs. mistral_common chat-template behavior; unrelated." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "Qwen3Moe FP8 config failure vs. template compilation error; unrelated." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Tied-weights attribute error vs. tokenizer loading failure for a different model; unrelated." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Generic loading noise vs. Mistral regex patch crash; different issues." + }, + { + "left": "issue:44464", + "right": "issue:45987", + "accept": false, + "reason": "Chunked generation inconsistency with compiled forward is not the same as StaticCache seq-length contract bug." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Both are v5 initialization regressions, but one is custom model init and the other is non-persistent buffer corruption; not clearly the same bug." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Custom model initialization in Jupyter is distinct from buffer initialization corruption." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Different Qwen family issues: hidden-states bug vs. rope-index temporal scaling bug." + }, + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Both mention tied lm_head weights, but one is a model-specific tying bug and the other is a serialization inconsistency; not the same defect." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "Different model families and failure modes, only loosely similar because both involve mismatches between expected and actual outputs." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "FP8 Qwen3Moe failure vs. init_empty_weights/_is_hf_initialized TypeError; unrelated." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Different vision-model preprocessing issues; not the same underlying bug." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "General model loading failure vs. Mistral tokenizer regex patch crash; unrelated." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Different model-loading problems with different triggers and code paths." + }, + { + "left": "issue:28530", + "right": "issue:34567", + "accept": false, + "reason": "Early stopping metric selection vs. TrainerState token-count tracking; unrelated." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "Tokenizer method conflict vs. MobileLLM key error; unrelated." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "FP8 config failure vs. missing Float8 storage type; different low-level problems." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet and stays conservative. The chosen anchor around issue 43644/44534 is well supported, and the only accepted soft pair (43452/43482) looks plausibly duplicate-like from the titles. The other pairwise rejections are appropriately cautious and do not overstate similarity." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 11, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:33666", + "issue:35707", + "issue:36032", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42796", + "issue:43089", + "issue:43296", + "issue:43322", + "issue:43381", + "issue:43425", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43540", + "issue:43611", + "issue:43632", + "issue:43746", + "issue:43792", + "issue:44062", + "issue:44079", + "issue:44220", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44479", + "issue:44534", + "issue:44589", + "issue:44610", + "issue:44683", + "issue:44756", + "issue:44811", + "issue:44877", + "issue:44987", + "issue:44991", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45538", + "issue:45663", + "issue:45812", + "issue:45923", + "issue:45987", + "issue:46315", + "issue:46491", + "issue:46710" + ], + "soft_pairs": [ + "issue:43089|issue:45987", + "issue:44062|issue:44987", + "issue:44265|issue:44610", + "issue:44479|issue:45325", + "issue:43322|issue:43540", + "issue:43531|issue:44877", + "issue:36032|issue:41720", + "issue:45216|issue:45357", + "issue:44991|issue:46315", + "issue:43425|issue:44403", + "issue:43296|issue:43531", + "issue:44220|issue:44683", + "issue:43611|issue:44534", + "issue:40444|issue:45325", + "issue:43632|issue:44756", + "issue:45812|issue:46491", + "issue:44403|issue:44991", + "issue:43792|issue:44589", + "issue:43381|issue:45923", + "issue:42796|issue:43746", + "issue:44291|issue:44589", + "issue:33666|issue:40444", + "issue:44079|issue:45663", + "issue:40990|issue:44811", + "issue:42796|issue:46710", + "issue:43482|issue:43525", + "issue:35707|issue:40444", + "issue:43632|issue:44534", + "issue:43482|issue:45538" + ], + "prepared_review_unit_hash": "3babbe25b08339f9d6d9d676626b9a8822d410505837cd7ec5cd1ace86e85c40", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35680, + "estimated_input_tokens": 8920, + "estimated_eval_tokens": 18096 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is heterogeneous: several unrelated model/tokenizer/regression reports are loosely similar, but none of the soft pairs look like true duplicates of the same underlying bug/change. I would not merge any pair.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43089", + "right": "issue:45987", + "accept": false, + "reason": "Performance/profiling issue about per-token GPU syncs vs a cache API contract bug; different code paths and symptoms." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer AddedToken argument conflict vs model loading failure for a specific checkpoint; unrelated failures." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "torch.export/torch_compilable_check bug vs processor image-size mismatch; not the same underlying bug." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Both involve Qwen VL/video-image behavior, but one is a v5.3 video-input regression and the other is a still-image RoPE position-id scaling bug; distinct paths." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Different models and different failures: Llava Next segfault on load vs Qwen3OmniMoe video ValueError." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Sliding-window handling in Qwen3-MoE vs strict config loading for granite_speech; no shared bug." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Tokenizer class method conflict vs auto device mapping CUDA assert; unrelated subsystems." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": false, + "reason": "Both are Qwen3.5 save_pretrained regressions, but they report different incorrect outputs/keys and appear to be separate fixes." + }, + { + "left": "issue:44991", + "right": "issue:46315", + "accept": false, + "reason": "Tokenizer load failure vs slow convert_tokens_to_ids regression; different behaviors and likely different fixes." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Torch 2.10 compatibility failure vs noisy loading output; not the same bug." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "PaddleOCR-VL loading in vLLM vs Qwen3-MoE sliding_window; unrelated issues." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Audio feature extraction bug vs compiled flex_attention failure on newer torch; different code paths." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "Model loading with base_model_prefix vs non-persistent buffer initialization junk; both v5 regressions but not the same issue." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "IterableDataset multi-image finetuning failure vs still-image temporal position-id scaling bug; different multimodal problems." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "_is_hf_initialized flag regression vs mmap/OOM on Strix Halo; unrelated failures." + }, + { + "left": "issue:45812", + "right": "issue:46491", + "accept": false, + "reason": "Both tokenizer regressions, but one is wrong Granite token IDs and the other is a leading-space round-trip issue in CodeLlama; distinct bugs." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Generic noisy-load report vs tokenizer load failure for EMBEDDIA/est-roberta; not duplicates." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Whisper inference failure vs missing Float8 storage object; unrelated." + }, + { + "left": "issue:43381", + "right": "issue:45923", + "accept": false, + "reason": "Eval-mode gradient checkpointing behavior vs missing supports_gradient_checkpointing flag on a trust_remote_code model; related concept, different concrete defect." + }, + { + "left": "issue:42796", + "right": "issue:43746", + "accept": false, + "reason": "Tokenizer BPE-marker handling vs PEFT local-checkpoint loading; unrelated." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "init_empty_weights/_is_hf_initialized constructor issue vs Float8 storage lookup failure; different bugs." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Both are Qwen VL finetuning/training, but the failures are different and not mergeable into one fix." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key assignment bug vs Gemma4 FSDP2 shared_kv_states KeyError; unrelated implementation areas." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "Perplexity regression on gpt-oss-20b vs Whisper batch_decode skip_special_tokens behavior; no shared bug." + }, + { + "left": "issue:42796", + "right": "issue:46710", + "accept": false, + "reason": "Tokenizer BPE-marker issue vs accuracy regression on a model after upgrade; too broad and not the same underlying problem." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "Qwen2.5-GGUF loading with transformers v5 vs Llama4Config missing pad_token_id; different model/config failures." + }, + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Progressive generation with inputs_embeds/past_key_values vs iterable-dataset multi-image finetuning; unrelated." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "Both are v5 regressions, but one is an init-flag breakage and the other is non-persistent buffer corruption; 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": 23, + "review_unit_index": 12, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:40444", + "issue:41553", + "issue:41720", + "issue:42175", + "issue:42796", + "issue:42915", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43421", + "issue:43454", + "issue:43482", + "issue:43493", + "issue:43531", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43701", + "issue:43749", + "issue:43792", + "issue:43818", + "issue:43856", + "issue:43873", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44265", + "issue:44291", + "issue:44462", + "issue:44466", + "issue:44534", + "issue:44568", + "issue:44829", + "issue:44877", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45491", + "issue:45542", + "issue:45701", + "issue:45812", + "issue:45834", + "issue:46402" + ], + "soft_pairs": [ + "issue:41553|issue:43582", + "issue:43531|issue:44291", + "issue:43317|issue:43856", + "issue:42175|issue:45542", + "issue:43421|issue:44568", + "issue:44079|issue:44534", + "issue:41720|issue:45081", + "issue:44829|issue:45479", + "issue:44462|issue:45812", + "issue:43611|issue:43646", + "issue:43377|issue:45491", + "issue:43818|issue:46402", + "issue:42175|issue:43296", + "issue:43644|issue:44079", + "issue:43818|issue:45357", + "issue:43749|issue:44079", + "issue:44265|issue:45072", + "issue:43792|issue:43957", + "issue:45325|issue:45381", + "issue:43493|issue:43994", + "issue:43454|issue:44466", + "issue:45325|issue:45834", + "issue:40444|issue:45381", + "issue:42915|issue:43482", + "issue:44291|issue:44877", + "issue:43632|issue:45072", + "issue:43873|issue:45701", + "issue:43701|issue:45325", + "issue:42796|issue:45381" + ], + "prepared_review_unit_hash": "0b9362b48509b4ebc9e8dcaffbc803795df87ed1150f9b97c4b57d2eb2fd9839", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36578, + "estimated_input_tokens": 9145, + "estimated_eval_tokens": 18546 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft pair is a safe duplicate under strict triage. There are a few thematic overlaps (notably Qwen2.5-VL position-id bugs and several Transformers v5 loading regressions), but each pair appears to describe a different concrete failure mode or code path.", + "confidence": 0.9, + "canonical_issue_reason": "Issue 44291 is the most broadly representative loading regression in the set: it targets a concrete transformers>=5.0.0rc0 instantiation failure (`init_empty_weights` / unexpected `_is_hf_initialized`), which is more central and reusable than the narrower model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44291 is the best single issue to anchor the cluster because it captures a clear upstream v5 loading breakage affecting many models, while the other candidates are either more specialized or only loosely related.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Both are user-facing failures, but one is a tokenizer-loading error for Voxtral and the other is an AppleSilicon allocator TypeError; different subsystems and root causes." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Both mention Transformers v5-era regressions, but one is a Qwen3-MoE sliding-window bug and the other is an `init_empty_weights` loading-API break; not the same code path." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Both involve offload/quantization behavior, but one is a dequantized model load issue with `device_map=auto` and the other is inefficient Qwen3 MoE training memory usage; different bugs." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": false, + "reason": "Both concern TensorFlow-related packaging/backend availability, but one is about `.[torch]` not including TensorFlow and the other is about a missing TF backend when only tensorboard is installed." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Both touch special-token handling, but one is runtime post-processor refresh in TokenizersBackend and the other is `add_special_tokens=True` failing for a specific tokenizer/model." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "These are unrelated v5 regressions: one is `ModelOutput` key assignment semantics, the other is non-persistent buffer initialization." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "One is a Qwen3 auto-device-map CUDA assert, the other is a Mistral tokenizer regex patch crash; different components and failure modes." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "One is flash_attention_3 causing degenerate training, the other is a classification loss edge case with `num_labels=1`; not the same bug." + }, + { + "left": "issue:44462", + "right": "issue:45812", + "accept": false, + "reason": "Both are AutoTokenizer regressions, but one ignores `tokenizer.json` while the other returns wrong token IDs for Granite models; separate tokenizer paths." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Both are loading regressions in Transformers v5, but one is about `base_model_prefix` and the other about custom model initialization; too different to merge." + }, + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "One is a MIMI padding-mask batching discrepancy, the other is Gemma3 sliding-window NaNs; different models and mechanisms." + }, + { + "left": "issue:43818", + "right": "issue:46402", + "accept": false, + "reason": "One is a Video-LLaVA vision-tower/temporal-attention problem, the other is Sam3Model text-encoder weights not loading." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "One is a packaging/backend install issue, the other is a PaddleOCR-VL load failure under a specific transformers/vLLM combination." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "One concerns non-persistent buffers being filled with junk; the other concerns `ModelOutput` key assignment after `None` values." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "These are both multimodal model issues, but one is Video-LLaVA architecture/weights and the other is Qwen3.5 visual-encoder key serialization; different concrete defects." + }, + { + "left": "issue:43749", + "right": "issue:44079", + "accept": false, + "reason": "`FSDP_CPU_RAM_EFFICIENT_LOADING` and `ModelOutput` key assignment are unrelated transformers behaviors." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "One is `torch.export` failing on `torch_compilable_check`, the other is a dtype mismatch in bfloat16 inference; no shared underlying bug." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "One is Whisper model execution failure, the other is `torch.device('meta')` loading breakage across some models; related to loading only at a very high level." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL position-id reports, but one is about still-image temporal scaling and the other about video-input vision position IDs; similar area, not enough evidence of one bug." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": false, + "reason": "Both involve SigLIP2 outputs, but one is an implementation discrepancy against JAX and the other is bad AutoModel/pipeline inference results; not clearly the same defect." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "One is a missing weight-tie bug in AyaVision, the other is inconsistent `lm_head.weight` serialization tied to device in v5." + }, + { + "left": "issue:45325", + "right": "issue:45834", + "accept": false, + "reason": "Both are multimodal/vision-related, but the bugs are different: Qwen2.5-VL rope-index scaling vs Kosmos2.5 OCR index overflow." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Both mention Qwen2.5-VL and image/video inputs, but one is an IterableDataset multi-image training failure and the other is a video position-id regression." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "One is a Qwen3MoE FineGrainedFP8Config failure, the other is a Qwen2.5-GGUF loading regression under transformers v5; distinct model/loading paths." + }, + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "One is an unexpected-argument failure in model init under `init_empty_weights`, the other is strict config validation blocking `granite_speech` config loading." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "One is the `_is_hf_initialized` flag regression, the other is a dtype mismatch in bfloat16 inference; unrelated defects." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "One is an offloading-with-quantization issue, the other is version-dependent tokenization changes; different subsystems entirely." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint resume key mismatch is unrelated to Qwen2.5-VL rope/position-id handling." + }, + { + "left": "issue:42796", + "right": "issue:45381", + "accept": false, + "reason": "A tokenizer BPE-marker bug and a Qwen2.5-VL video position-id bug 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": 23, + "review_unit_index": 13, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:36331", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42796", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43454", + "issue:43526", + "issue:43550", + "issue:43572", + "issue:43606", + "issue:43720", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44186", + "issue:44265", + "issue:44479", + "issue:44488", + "issue:44534", + "issue:44568", + "issue:44610", + "issue:44898", + "issue:45072", + "issue:45584", + "issue:45759", + "issue:45800", + "issue:45823", + "issue:45920", + "issue:45923", + "issue:46050", + "issue:46209", + "issue:46305", + "issue:46315" + ], + "soft_pairs": [ + "issue:43296|issue:45800", + "issue:44479|issue:45823", + "issue:43606|issue:44265", + "issue:40990|issue:45584", + "issue:43262|issue:43746", + "issue:35707|issue:41720", + "issue:43881|issue:44488", + "issue:34634|issue:35707", + "issue:43454|issue:43550", + "issue:43606|issue:43819", + "issue:39692|issue:40444", + "issue:41720|issue:45823", + "issue:44186|issue:44265", + "issue:43377|issue:44610", + "issue:46305|issue:46315", + "issue:43257|issue:45923", + "issue:44568|issue:45920", + "issue:43454|issue:45072", + "issue:42796|issue:46209", + "issue:43317|issue:43873", + "issue:43377|issue:44898", + "issue:43454|issue:43720", + "issue:40990|issue:42947", + "issue:42617|issue:42915", + "issue:43749|issue:44534", + "issue:43572|issue:45759", + "issue:43526|issue:43883", + "issue:43322|issue:46050", + "issue:41720|issue:43322", + "issue:36331|issue:42617" + ], + "prepared_review_unit_hash": "4373fe8c8aa347477fb9c5fe75a31297216038d825a72022496de98a4decc6bf", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36810, + "estimated_input_tokens": 9203, + "estimated_eval_tokens": 18662 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are largely a grab bag of distinct bugs/regressions across loading, offloading, tokenization, multimodal processors, and training/runtime paths. I don\u2019t see any soft pair that looks like the same underlying bug or a mergeable duplicate PR/change.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43873 is the broadest representative of the loading/offloading/quantization problems here and is still active with discussion, so it works best as the cluster anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the most general and active issue in the set, with the clearest umbrella fit for nearby offloading/loading regressions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "Different failures: model loading in vLLM vs a torch/transformers version incompatibility." + }, + { + "left": "issue:44479", + "right": "issue:45823", + "accept": false, + "reason": "Different subsystems and symptoms: video-input regression vs device-mapping/forward device mismatch." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "One is CPU offload device mismatch for Bark; the other is torch.export failing on torch_compilable_check." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "Unrelated bugs: perplexity regression vs Whisper empty-transcription handling." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio template sample-rate default bug vs PEFT adapter checkpoint loading; not the same code path." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Both are generation-related, but one is inputs_embeds/past_key_values behavior and the other is auto device mapping crashing." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "Both are load failures, but for different models and likely different causes." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "BarkProcessor voice_preset issue is unrelated to progressive generation with inputs_embeds/past_key_values." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "Weight-tying/generation bug vs torch.compile+SDPA failure; different concrete problems." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "Device mismatch during CPU offload is unrelated to DAC.from_latents vs forward mismatch." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "Documentation/example errors for SigLIP2 vs finetuning failure on Qwen2.5-VL; not the same issue." + }, + { + "left": "issue:41720", + "right": "issue:45823", + "accept": false, + "reason": "Both involve device placement, but they hit different models and different failure modes." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Tokenizer crash on NER/padding vs export failure from torch_compilable_check; different bugs." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "MIMI batched-vs-single output mismatch is unrelated to OmDet-Turbo processor image-size mismatch." + }, + { + "left": "issue:46305", + "right": "issue:46315", + "accept": false, + "reason": "Sliding-window off-by-one vs tokenizer conversion slowdown; no shared underlying defect." + }, + { + "left": "issue:43257", + "right": "issue:45923", + "accept": false, + "reason": "Weights not converted during accelerate+deepspeed loading vs missing supports_gradient_checkpointing flag; different fixes." + }, + { + "left": "issue:44568", + "right": "issue:45920", + "accept": false, + "reason": "Specific add_special_tokens regression for one tokenizer vs broader wrong-token-id regression across multiple models." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "lm_head tying bug is unrelated to dtype mismatch in bfloat16 inference." + }, + { + "left": "issue:42796", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer BPE marker handling vs FP8 inference on B200; unrelated." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "Both mention offloading/quantization, but one is dequantized model loading on gpu+cpu offload and the other is a broader offloading regression; not clearly the same bug." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Different model components and symptoms: encoder padding-mask mismatch vs image-classification resolution handling." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "Text generation weight-tying bug vs packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "Perplexity regression is unrelated to gradient checkpointing ineffectiveness with PEFT LoRA." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "3d_parallel.py runtime problem vs Qwen3 MoE FP8 config failure; different issues." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "FSDP CPU RAM efficient loading is unrelated to non-persistent buffers being filled with junk." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "Missing pad_token_idx after v5 update vs failing to unwrap text_config in composite multimodal configs." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels bug is unrelated to Molmo missing all_tied_weights_keys." + }, + { + "left": "issue:43322", + "right": "issue:46050", + "accept": false, + "reason": "Llava Next loading segfault vs sinq save/load breakage; different code paths." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Auto device mapping assert on Qwen3 vs segfault loading Llava Next; not the same defect." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "Trainer compute_loss signature breakage vs inability to run 3d_parallel.py; 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": 23, + "review_unit_index": 14, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:26413", + "issue:28530", + "issue:33666", + "issue:36010", + "issue:36331", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:43065", + "issue:43295", + "issue:43317", + "issue:43404", + "issue:43450", + "issue:43531", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43646", + "issue:43708", + "issue:43720", + "issue:43873", + "issue:43881", + "issue:44008", + "issue:44186", + "issue:44387", + "issue:44423", + "issue:44466", + "issue:44610", + "issue:44756", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45161", + "issue:45357", + "issue:45406", + "issue:45632", + "issue:45684", + "issue:45701", + "issue:45704", + "issue:45823", + "issue:46291", + "issue:46402", + "issue:46665" + ], + "soft_pairs": [ + "issue:43881|issue:44991", + "issue:44387|issue:45704", + "issue:41720|issue:44977", + "issue:36331|issue:42915", + "issue:44877|issue:44991", + "issue:44466|issue:45357", + "issue:43450|issue:45701", + "issue:40990|issue:45701", + "issue:41720|issue:42617", + "issue:43065|issue:46402", + "issue:43873|issue:44756", + "issue:44186|issue:45072", + "issue:26413|issue:28530", + "issue:43550|issue:43720", + "issue:43577|issue:43708", + "issue:43550|issue:43606", + "issue:43295|issue:45701", + "issue:43646|issue:46665", + "issue:44186|issue:44610", + "issue:43404|issue:45357", + "issue:42915|issue:43317", + "issue:44610|issue:44898", + "issue:44008|issue:45823", + "issue:36010|issue:42617", + "issue:44423|issue:45406", + "issue:45632|issue:45684", + "issue:41720|issue:45161", + "issue:43531|issue:43572", + "issue:28530|issue:33666", + "issue:42915|issue:46291" + ], + "prepared_review_unit_hash": "8803e6e9097eb1f24f54ebe178e1691e4018da6d1505b92e460b4dcc912737f0", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36178, + "estimated_input_tokens": 9045, + "estimated_eval_tokens": 18346 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: it mixes trainer/checkpointing bugs, tokenizer/processor regressions, quantization/offload issues, multimodal model failures, and serving crashes. None of the soft pairs look like the same underlying bug/change, so I would not merge any of them.", + "confidence": 0.94, + "canonical_issue_reason": "If a single issue must represent the cluster, issue 43708 is the clearest and most concrete bug report; however, the set is too diverse to have a strong true canonical issue.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43708 is the best standalone issue because it is specific, reproducible, and tightly scoped to a real Trainer checkpointing bug, unlike the broader or more model-specific reports in the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "Different problems: GLM-4V loading failure vs tokenizer loading failure for a different model family." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "Different failure modes: int4 quantization OOM vs T5 using apex FusedRMSNorm memory leak." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Different code paths: auto device-map CUDA assert on Qwen3 vs flash-attention generation issue on Qwen3.5." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "Unrelated bugs: custom trainer loss signature change vs Qwen3Moe FineGrainedFP8Config failure." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Different areas: strict config blocking granite_speech load vs tokenizer loading regression for est-roberta." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Both involve model saving, but one is lm_head serialization/tied weights and the other is incorrect visual encoder keys in Qwen3.5." + }, + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Different subsystems: batched video processor shape bug vs tokenization behavior changing across versions." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "Perplexity anomaly on gpt-oss is not the same issue as version-dependent tokenization changes." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Different failures: Qwen3 CUDA assert under auto device mapping vs 3d_parallel.py launch/runtime failure." + }, + { + "left": "issue:43065", + "right": "issue:46402", + "accept": false, + "reason": "Same model family, but different submodules and bugs: dummy Conv2d in pixel decoder vs text encoder weights not loading." + }, + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Quantization offload behavior and Strix Halo mmap OOM are different root causes." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer crash on NER/padding is unrelated to dtype mismatches in inference." + }, + { + "left": "issue:26413", + "right": "issue:28530", + "accept": false, + "reason": "Both are Trainer-related, but checkpoint resume epoch-iterator failure is not the same as early-stopping metric lookup." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "Different code paths: torch.compile with SDPA on Bamba vs BitNet packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "Blip2 dtype propagation and resume_from_checkpoint max_steps miscalculation are unrelated." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Different model-specific bugs: Bamba SDPA compile failure vs bark-small CPU offload device mismatch." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "Processor.tokenizer regression is not the same as version-dependent tokenization output changes." + }, + { + "left": "issue:43646", + "right": "issue:46665", + "accept": false, + "reason": "Both concern custom code/init behavior, but one is Transformers 5.0 breaking model initialization and the other is from_pretrained behavior inside Python classes." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Different processor bugs: LayoutLMv2Tokenizer NER/padding crash vs OmDet-Turbo output-size mismatch." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Different serialization issues: untied lm_head weights in Mistral3 vs incorrect visual encoder keys in Qwen3.5 save_pretrained." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "Qwen3Moe FP8 config failure is unrelated to device_map auto loading failure for a dequantized model with GPU+CPU offload." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Both are image-size related, but they affect different models and different preprocessing/model expectations." + }, + { + "left": "issue:44008", + "right": "issue:45823", + "accept": false, + "reason": "Gemma3n forward variable collision and Gemma4 PLE device mismatch are distinct bugs." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "Importing GenerationMixin and running 3d_parallel.py are unrelated failures." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Both hit serving/multimodal code, but the concrete attribute errors and processors differ." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Cache-path collisions for trust_remote_code and read-only permission propagation in save_pretrained are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Auto device-map CUDA assert is not the same as tensor-parallel not working for GPT-OSS MoE." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Sliding-window behavior for Qwen3-MoE and missing pad_token_idx in StableLmConfig are unrelated." + }, + { + "left": "issue:28530", + "right": "issue:33666", + "accept": false, + "reason": "Early stopping metric resolution and Qwen2-VL multi-GPU training are different areas." + }, + { + "left": "issue:42915", + "right": "issue:46291", + "accept": false, + "reason": "FineGrainedFP8Config failure and kernels import/version requirement failure are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 15, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:28530", + "issue:29159", + "issue:30064", + "issue:33357", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:36032", + "issue:36331", + "issue:41720", + "issue:42490", + "issue:42915", + "issue:43316", + "issue:43322", + "issue:43329", + "issue:43388", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43873", + "issue:43881", + "issue:44186", + "issue:44361", + "issue:44387", + "issue:44451", + "issue:44492", + "issue:44509", + "issue:44534", + "issue:44589", + "issue:44671", + "issue:44743", + "issue:44913", + "issue:44991", + "issue:45059", + "issue:45406", + "issue:45464", + "issue:45701", + "issue:45704", + "issue:45759", + "issue:45859", + "issue:46665" + ], + "soft_pairs": [ + "issue:44671|issue:44743", + "issue:36331|issue:41720", + "issue:44186|issue:44361", + "issue:43526|issue:43881", + "issue:43316|issue:44913", + "issue:28530|issue:30064", + "issue:42915|issue:43322", + "issue:33357|issue:35707", + "issue:43540|issue:43782", + "issue:30064|issue:34634", + "issue:43873|issue:45704", + "issue:43550|issue:43653", + "issue:44387|issue:45701", + "issue:29159|issue:36032", + "issue:43606|issue:43720", + "issue:44492|issue:44509", + "issue:44534|issue:46665", + "issue:43531|issue:45759", + "issue:41720|issue:43746", + "issue:42490|issue:43329", + "issue:43606|issue:43653", + "issue:43388|issue:45059", + "issue:45406|issue:45464", + "issue:34567|issue:35707", + "issue:34634|issue:41720", + "issue:43531|issue:44589", + "issue:45759|issue:45859", + "issue:44451|issue:44991" + ], + "prepared_review_unit_hash": "38463044b00f4772fe35a428d8dacd750c9faf9f79fe3b6c623752c90cf7ce57", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35465, + "estimated_input_tokens": 8867, + "estimated_eval_tokens": 17990 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: the soft pairs share only broad Transformers topical similarity, not the same underlying bug or change. None of the candidate pairs look safe to merge as duplicates, and there are no PRs to select.", + "confidence": 0.97, + "canonical_issue_reason": "If a representative issue is needed, issue 29159 is the broadest and most generally phrased user-facing regression report, but the set does not actually form one duplicate family.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 29159 is the most generic/high-signal problem statement among the set, so it is the best stand-in representative even though it is not a true canonical duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "Different models and failure modes: CamemBERT masked-LM prediction regression vs Qwen recurrent cache state reset." + }, + { + "left": "issue:36331", + "right": "issue:41720", + "accept": false, + "reason": "Trainer API signature mismatch vs Qwen3 device-mapping CUDA assert; unrelated root causes." + }, + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Both tokenizer-related, but one is LayoutLMv2 NER padding/truncation and the other is MLuke task AttributeError." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug vs glm-4v-9b load failure; no shared code-path." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Both are config/reload issues, but Gemma3TextConfig API discrepancy is unrelated to GPTNeoX rotary_pct persistence." + }, + { + "left": "issue:28530", + "right": "issue:30064", + "accept": false, + "reason": "Early stopping metric resolution bug vs void segmentation-map processing bug; different subsystems." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Qwen3Moe FP8 config failure vs Llava Next segmentation fault; different model families and symptoms." + }, + { + "left": "issue:33357", + "right": "issue:35707", + "accept": false, + "reason": "MacOS CLIP bus error vs generation with inputs_embeds/past_key_values; unrelated failures." + }, + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Different Qwen multimodal issues: video input processing error vs weight_only from_pretrained loading error." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Void segmentation maps vs BarkProcessor voice_preset handling; unrelated image vs audio processor bugs." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization/offloading behavior vs T5 RMSNorm memory leak; not the same bug." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "torch.compile SDPA failure vs tokenizer special-token registration bug; completely different paths." + }, + { + "left": "issue:44387", + "right": "issue:45701", + "accept": false, + "reason": "Int4 quantization memory growth vs tokenization changes across versions; different concerns." + }, + { + "left": "issue:29159", + "right": "issue:36032", + "accept": false, + "reason": "Slow/fast tokenizer inconsistency vs T5 add_special_tokens name collision; both tokenizer issues but distinct root causes." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "CPU offload device mismatch vs accelerate loading of packed BitNet weights; related area but different bugs." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "Cache strategy typo vs docs referencing removed pipeline tasks; documentation-only and unrelated." + }, + { + "left": "issue:44534", + "right": "issue:46665", + "accept": false, + "reason": "Non-persistent buffer junk in v5 vs from_pretrained behavior inside Python classes; different code paths." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "Qwen3-MoE sliding_window issue vs composite Qwen multimodal config unwrap bug; same family name only." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Qwen3 auto device mapping CUDA assert vs GraniteSpeech PEFT local checkpoint loading; unrelated." + }, + { + "left": "issue:42490", + "right": "issue:43329", + "accept": false, + "reason": "Trailing-slash local model import warning vs multimodal token counting video-branch bug; different behaviors." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "CPU offload mismatch vs BigBirdTokenizer mask-token registration; unrelated." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "Metric gathering drops tuple labels vs SAM3 text/bounding-box weirdness; different training vs inference problems." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer in serve vs Qwen3.5 streaming chat/completions failure; different models and layers." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState token counter update bug vs generation with inputs_embeds/past_key_values; unrelated." + }, + { + "left": "issue:44991", + "right": "issue:44589", + "accept": false, + "reason": "Tokenizer load failure for a specific model vs missing Float8 storage type; different failure classes." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior vs Float8 storage lookup error; no overlap beyond general Transformers scope." + }, + { + "left": "issue:45759", + "right": "issue:45859", + "accept": false, + "reason": "Composite Qwen config unwrap issue vs Qwen3_5MoeTextRotaryEmbedding CPU offload incompatibility; same ecosystem, different bug." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Model-specific loading regression for ScandiBERT vs tokenizer loading regression for est-roberta; unrelated root causes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 16, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:39692", + "issue:41720", + "issue:42886", + "issue:43381", + "issue:43421", + "issue:43531", + "issue:43644", + "issue:43645", + "issue:43653", + "issue:43746", + "issue:43782", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44062", + "issue:44077", + "issue:44186", + "issue:44368", + "issue:44451", + "issue:44514", + "issue:44545", + "issue:44610", + "issue:44625", + "issue:44671", + "issue:44843", + "issue:44898", + "issue:44918", + "issue:44936", + "issue:44987", + "issue:45042", + "issue:45072", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45704", + "issue:45759", + "issue:45923", + "issue:46032", + "issue:46050", + "issue:46402", + "issue:46463", + "issue:46655", + "issue:46665", + "issue:46693" + ], + "soft_pairs": [ + "issue:45357|issue:46402", + "issue:44077|issue:44625", + "issue:44625|issue:45479", + "issue:43644|issue:46665", + "issue:44898|issue:45072", + "issue:44062|issue:44451", + "issue:44987|issue:45042", + "issue:43381|issue:43844", + "issue:44368|issue:45357", + "issue:39692|issue:44610", + "issue:43531|issue:43645", + "issue:41720|issue:45704", + "issue:43746|issue:46050", + "issue:45704|issue:45923", + "issue:46655|issue:46693", + "issue:43381|issue:45704", + "issue:44918|issue:44936", + "issue:43421|issue:43927", + "issue:42886|issue:44843", + "issue:45759|issue:45923", + "issue:44514|issue:44545", + "issue:46032|issue:46463", + "issue:39692|issue:43873", + "issue:43746|issue:43873", + "issue:43994|issue:44671", + "issue:43421|issue:43653", + "issue:30064|issue:36010", + "issue:43782|issue:45381", + "issue:43421|issue:44186" + ], + "prepared_review_unit_hash": "3afebb1e27fe60d3aa947a116b68eab012472862e784904206d705e58fc477f5", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 36055, + "estimated_input_tokens": 9014, + "estimated_eval_tokens": 18284 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are thematically related but not true duplicates; only the two Qwen2_5_VLProcessor issues appear to describe the same crash on batched `apply_chat_template` with `padding=False`.", + "confidence": 0.94, + "canonical_issue_reason": "Issue 44514 is the best canonical issue: it is the earlier, more detailed report of the exact batched-input `apply_chat_template` crash with `padding=False`.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44514 is the strongest issue to keep as the reference because it has the clearer original report and fuller reproduction context for the shared bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45357", + "right": "issue:46402", + "accept": false, + "reason": "Different models and different failures: incorrect visual encoder keys vs missing text encoder weights." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "Unrelated problems: optional `post_init` policy vs `num_labels` propagation." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "Both involve classification config, but one is missing propagation and the other is a zero-loss regression across models; not the same bug." + }, + { + "left": "issue:43644", + "right": "issue:46665", + "accept": false, + "reason": "Different subsystems and symptoms: buffer initialization vs `from_pretrained` behavior in Python classes." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Different model families and failure modes: image resolution handling vs dtype mismatch in inference." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "The first is a tokenizer constructor error; the second is a specific model load failure. Too broad to treat as the same bug." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "Unrelated: model loading failure for one package vs PIL processor dependency regression." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Different issues: eval-mode gradient checkpointing guard vs anomalous gradient growth under ZeRO-3." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "A warning about `tie_word_embeddings` is unrelated to incorrect visual encoder keys during save/load." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "Different docs/model mismatch issue vs processor image-size mismatch in OmDet-Turbo." + }, + { + "left": "issue:43531", + "right": "issue:43645", + "accept": false, + "reason": "No shared underlying bug: sliding-window behavior in Qwen3-MoE vs notebook custom-model initialization." + }, + { + "left": "issue:41720", + "right": "issue:45704", + "accept": false, + "reason": "CUDA assert on Qwen3 device mapping is unrelated to T5 Apex RMSNorm memory leak." + }, + { + "left": "issue:43746", + "right": "issue:46050", + "accept": false, + "reason": "Both involve checkpoint loading, but one is PEFT adapters and the other is sinq quantization; different code paths." + }, + { + "left": "issue:45704", + "right": "issue:45923", + "accept": false, + "reason": "Completely different concerns: memory leak in T5 RMSNorm vs missing gradient-checkpointing capability flag." + }, + { + "left": "issue:46655", + "right": "issue:46693", + "accept": false, + "reason": "Video processor padding bug is unrelated to FA2 long-context performance regression." + }, + { + "left": "issue:43381", + "right": "issue:45704", + "accept": false, + "reason": "No substantive overlap; one is eval-mode checkpointing, the other is a T5 normalization leak." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "Different trainer failures: embedding unpacking with TRL SFT vs post-train evaluate() failure." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "One is runtime special-token post-processing, the other is save/load loss of custom token IDs; related area but not same bug." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": false, + "reason": "Both mention offline mode, but one is cache loading and the other is an unconditional hub metadata call in a specific regex patch." + }, + { + "left": "issue:45759", + "right": "issue:45923", + "accept": false, + "reason": "Different bugs: unwrapping `text_config` in composite multimodal configs vs a trust_remote_code capability flag." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Same crash, same API, same condition (`apply_chat_template` on batched input with `padding=False`); these are duplicates." + }, + { + "left": "issue:46032", + "right": "issue:46463", + "accept": false, + "reason": "Mamba cache correctness vs T5 decoder-input error messaging are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "Documentation/model mismatch is not the same as quantization/offloading behavior." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "Checkpoint loading with PEFT adapters is unrelated to quantization offloading failures." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "Different models and symptoms: SigLIP2 output quality vs CamemBERT masked-LM predictions." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Related to special tokens, but one is a framework feature gap and the other is a tokenizer-specific registration bug." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Completely unrelated: image processor segmentation-map handling vs import error for `GenerationMixin`." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Both are multimodal Qwen issues, but one is `weight_only` loading and the other is video `vision_position_ids`; different failures." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Different tokenizer issues with different triggers; not the same code-path bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: 44514 is earlier and more detailed than 44545, and the duplicate call is conservative. The soft-edge rejections are generally cautious and align with the issue titles; I don\u2019t see overstatement that would require rejection." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 17, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:30064", + "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:43644", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43749", + "issue:43756", + "issue:43819", + "issue:43927", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44220", + "issue:44442", + "issue:44448", + "issue:44485", + "issue:44488", + "issue:44509", + "issue:44661", + "issue:44671", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45356", + "issue:45381", + "issue:45834", + "issue:45907", + "issue:46294", + "issue:46665", + "issue:46693" + ], + "soft_pairs": [ + "issue:45081|issue:45356", + "issue:45356|issue:45907", + "issue:43720|issue:43819", + "issue:43653|issue:43720", + "issue:43596|issue:44220", + "issue:43550|issue:45072", + "issue:43295|issue:45834", + "issue:43550|issue:43927", + "issue:44062|issue:44488", + "issue:43540|issue:46294", + "issue:45381|issue:46693", + "issue:43618|issue:43697", + "issue:43531|issue:43643", + "issue:43611|issue:46665", + "issue:44220|issue:44661", + "issue:43653|issue:43819", + "issue:44186|issue:44898", + "issue:30064|issue:36331", + "issue:41720|issue:44991", + "issue:43643|issue:44877", + "issue:43550|issue:43819", + "issue:43335|issue:43756", + "issue:44821|issue:45084", + "issue:43540|issue:44008", + "issue:44843|issue:45081", + "issue:39692|issue:45834", + "issue:44448|issue:44671", + "issue:44186|issue:44442", + "issue:44485|issue:44509", + "issue:43644|issue:43749" + ], + "prepared_review_unit_hash": "cb10a8a04965120d1ed0f6c42e2c5fb4287ac8dae13d5cde71c129a658aa6eaa", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36268, + "estimated_input_tokens": 9067, + "estimated_eval_tokens": 18390 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This looks like a loose cluster of mostly unrelated Transformers v5 regressions centered on loading, tokenizers/configs, and a few model-specific behavior changes. I do not see any soft pair that is clearly the same underlying bug/change, so all soft edges are rejected. The broadest representative issue is the generic model-loading regression around `base_model_prefix` / v5 compatibility.", + "confidence": 0.7, + "canonical_issue_reason": "issue:43611 is the broadest and most central regression in the set: it describes a core v5 model-loading breakage, which matches the dominant theme of the cluster better than the model-specific or tokenizer-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43611 is the best standalone representative because it captures a high-level Transformers v5 loading regression rather than a narrow model- or tokenizer-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Both involve tokenizer regex handling, but one is a crash in `_patch_mistral_regex` and the other is a Kimi-K2.5 codec/regression warning issue; different concrete failures." + }, + { + "left": "issue:45356", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer regression vs repo file API forwarding bug; completely different code paths." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "Different models and failures: BitNet accelerate-loading of packed weights vs DAC latent reconstruction/STE mismatch." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "BigBirdTokenizer special-token registration bug is unrelated to BitNet packed-weight loading." + }, + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "Zero3/BertModel indexing failure and fbank feature extraction are unrelated subsystems and symptoms." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Both touch inference/runtime behavior, but one is Bamba torch.compile+SDPA and the other is dtype mismatch in SwitchTransformers/TimmWrapperModel; not the same bug." + }, + { + "left": "issue:43295", + "right": "issue:45834", + "accept": false, + "reason": "Processor/tokenizer regression in v4.57.5 vs Kosmos2.5 OCR long-input index error; different model families and failure modes." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "Torch.compile SDPA issue in Bamba is unrelated to DiaConfig save/load token-id loss." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "Duplicate `special` kwarg in `AddedToken` is unrelated to loading a specific tokenizer model." + }, + { + "left": "issue:43540", + "right": "issue:46294", + "accept": false, + "reason": "Both are video-related Qwen issues, but one is a Qwen3OmniMoe processing ValueError and the other is a chat-template video token mismatch in Qwen3.5-9B; not the same concrete defect." + }, + { + "left": "issue:45381", + "right": "issue:46693", + "accept": false, + "reason": "Video `vision_position_ids` bug and FA2 long-context performance regression are unrelated." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "CLIP output attentions assignment regression and RTDetrV2 output differences are separate model-specific issues." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to `AutoConfig` missing fields with `trust_remote_code=True`." + }, + { + "left": "issue:43611", + "right": "issue:46665", + "accept": false, + "reason": "Both concern loading APIs, but one is a model-loading break due to `base_model_prefix` and the other is inconsistent `from_pretrained` behavior inside Python classes; different bugs." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Audio fbank extraction and `add-new-model-like` mapping behavior are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "Tokenizer special-token registration and DAC latent conversion are unrelated." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "LayoutLMv2 tokenizer padding/truncation crash and Perceiver non-default-resolution classification failure are unrelated." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "ImageProcessor segmentation-map handling and CustomTrainer `compute_loss` signature mismatch are different layers of the stack." + }, + { + "left": "issue:41720", + "right": "issue:44991", + "accept": false, + "reason": "Qwen3 auto device mapping CUDA assert and tokenizer loading for EMBEDDIA/est-roberta are unrelated." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "Both are config-loading regressions, but one is missing custom fields with `trust_remote_code` and the other is strict config blocking `granite_speech`; not the same underlying issue." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Bamba torch.compile/SDPA failure is unrelated to DAC `from_latents` mismatch." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "Sparse layer creation in SwitchTransformers and Smollm3 RoPE layer dropping are different configuration bugs." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "AutoImageProcessor URL loading and a template-node compilation TypeError are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Qwen3OmniMoe video processing failure and Gemma3n variable-name collision are unrelated multimodal bugs." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Offline `model_info()` call in `_patch_mistral_regex` and backend-tokenizer attribute crash are both tokenizer-related but distinct failures." + }, + { + "left": "issue:39692", + "right": "issue:45834", + "accept": false, + "reason": "SigLIP2 doc example errors and Kosmos2.5 OCR index error are unrelated." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Pegasus v4/v5 output drift and CamemBERT masked LM regression are separate model-specific output bugs." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "LayoutLMv2 NER/padding crash and AutoTokenizer failing to load FastSpeech2ConformerTokenizer are unrelated tokenizer issues." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "GLM-5 RoPE implementation discussion and docs about removed pipeline tasks are unrelated." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Non-persistent buffer initialization junk and FSDP CPU RAM efficient loading are different loading/runtime regressions." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 18, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:28218", + "issue:29159", + "issue:34634", + "issue:36010", + "issue:43322", + "issue:43388", + "issue:43421", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43645", + "issue:43716", + "issue:43720", + "issue:43749", + "issue:43782", + "issue:43818", + "issue:43825", + "issue:43827", + "issue:43950", + "issue:43994", + "issue:44008", + "issue:44291", + "issue:44479", + "issue:44561", + "issue:44610", + "issue:44805", + "issue:44811", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45020", + "issue:45072", + "issue:45183", + "issue:45381", + "issue:45479", + "issue:45584", + "issue:45685", + "issue:45834", + "issue:45987", + "issue:46032", + "issue:46192", + "issue:46291", + "issue:46294", + "issue:46463", + "issue:46512", + "issue:46665" + ], + "soft_pairs": [ + "issue:43322|issue:45685", + "issue:34634|issue:36010", + "issue:43606|issue:45072", + "issue:45584|issue:45834", + "issue:45183|issue:45479", + "issue:43716|issue:44008", + "issue:44479|issue:45834", + "issue:43950|issue:45987", + "issue:43388|issue:46032", + "issue:43582|issue:46512", + "issue:44991|issue:46291", + "issue:43825|issue:43994", + "issue:43577|issue:43818", + "issue:46294|issue:46463", + "issue:43782|issue:44977", + "issue:44561|issue:45020", + "issue:43421|issue:44811", + "issue:44805|issue:44936", + "issue:44610|issue:45072", + "issue:43720|issue:45072", + "issue:43540|issue:45381", + "issue:43825|issue:43827", + "issue:28218|issue:29159", + "issue:43749|issue:44291", + "issue:43645|issue:46665", + "issue:43577|issue:43582", + "issue:45987|issue:46192" + ], + "prepared_review_unit_hash": "fa517b950ba29c52414f72a2996a312827da234913a1894e235c840157f05a9d", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35668, + "estimated_input_tokens": 8917, + "estimated_eval_tokens": 18090 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly heterogeneous, with only one plausible duplicate pair around tokenizer added-token spacing versus slow/fast tokenizer inconsistency. Everything else appears to be separate bugs or regressions in different code paths.", + "confidence": 0.84, + "canonical_issue_reason": "issue:29159 is the broader, more canonical tokenizer report: it covers the same slow-vs-fast inconsistency that issue:28218 describes as an extra-space-after-added-token symptom.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:29159 is the best issue anchor because it is the broader, more central description of the tokenizer problem and subsumes the more specific added-token spacing report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Different model/runtime failures: Llava Next loading segfault vs MPS int histogram support error." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "Unrelated components and symptoms: BarkProcessor voice_preset vs GenerationMixin import failure." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "Both are model-loading/inference bugs, but the concrete failures differ: CPU offload device mismatch vs bfloat16 dtype mismatch." + }, + { + "left": "issue:45584", + "right": "issue:45834", + "accept": false, + "reason": "Different pipelines and root causes: Whisper empty transcription handling vs Kosmos2.5 long OCR index error." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "These are distinct: input validation error messaging vs degenerate zero loss in sequence classification." + }, + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Different models and bugs: dtype mismatch in Mistral-3 preprocessing vs Gemma3n variable-name collision." + }, + { + "left": "issue:44479", + "right": "issue:45834", + "accept": false, + "reason": "Separate regressions: Qwen video-input handling vs Kosmos2.5 OCR index error." + }, + { + "left": "issue:43950", + "right": "issue:45987", + "accept": false, + "reason": "Different subsystems: non-persistent buffer corruption in from_pretrained vs StaticCache sequence-length return type." + }, + { + "left": "issue:43388", + "right": "issue:46032", + "accept": false, + "reason": "Unrelated training/eval issues: gather_for_metrics label truncation vs Mamba2Mixer use_cache correctness." + }, + { + "left": "issue:43582", + "right": "issue:46512", + "accept": false, + "reason": "Unrelated dtype/Apple Silicon warmup TypeError vs from_config stale config.dtype." + }, + { + "left": "issue:44991", + "right": "issue:46291", + "accept": false, + "reason": "Tokenizer loading failure for a specific model vs importing pipelines breaking on kernels version requirements." + }, + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "Both mention pipeline, but one is an error message regression and the other is wrong model outputs; not the same bug." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "Different multimodal model issues: Blip2 dtype propagation vs Video-LLaVA architecture/weights problem." + }, + { + "left": "issue:46294", + "right": "issue:46463", + "accept": false, + "reason": "Different validation paths and models: Qwen3.5 video feature/token mismatch vs T5 decoder-input error messaging." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Different failures: from_pretrained weight_only error vs generation failure with flash-attention." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "The second is a broad umbrella over remote_code breakages; the first is one specific removed helper. Not clearly the same concrete bug." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Tokenizer backend special-token update issue vs Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "Different failure modes: mask shape IndexError vs trainer.evaluate after train regression." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Different model families and code paths: image size mismatch vs dtype mismatch in inference." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Packed-weight loading bug vs dtype mismatch bug; not the same underlying issue." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Both involve video inputs, but they target different models and concrete defects." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "One is an error-message regression, the other is stale documentation; related topic but not the same bug/change." + }, + { + "left": "issue:28218", + "right": "issue:29159", + "accept": true, + "reason": "Likely the same tokenizer bug: 28218 is the specific added-token extra-space symptom, while 29159 describes the broader slow/fast tokenizer inconsistency." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "Different loading regressions: FSDP CPU RAM efficient loading vs init_empty_weights unexpected argument handling." + }, + { + "left": "issue:43645", + "right": "issue:46665", + "accept": false, + "reason": "Both involve model initialization, but the reported failures are different enough that they do not look like the same bug." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "Completely different problem areas: dtype propagation in BLIP2 vs Apple Silicon allocator warmup TypeError." + }, + { + "left": "issue:45987", + "right": "issue:46192", + "accept": false, + "reason": "Different API regressions: StaticCache sequence length type vs ignored return_tensors when text_kwargs is passed." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall and appropriately conservative. The only accepted soft-edge pair, 28218 vs 29159, is plausibly the same tokenizer-family bug, while the remaining pairs are credibly separated by model, subsystem, or failure mode." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 19, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:34634", + "issue:36010", + "issue:36331", + "issue:42175", + "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:44671", + "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:45701", + "issue:46209", + "issue:46285", + "issue:46291", + "issue:46326", + "issue:46697", + "issue:46728" + ], + "soft_pairs": [ + "issue:44671|issue:45701", + "issue:42175|issue:45081", + "issue:34634|issue:36331", + "issue:45003|issue:46285", + "issue:46697|issue:46728", + "issue:43653|issue:44912", + "issue:43582|issue:44857", + "issue:43262|issue:43819", + "issue:44589|issue:45042", + "issue:43742|issue:45020", + "issue:43782|issue:45325", + "issue:45081|issue:45406", + "issue:45372|issue:46326", + "issue:45084|issue:45538", + "issue:43550|issue:44912", + "issue:43526|issue:43697", + "issue:43329|issue:45600", + "issue:43526|issue:44554", + "issue:44060|issue:44368", + "issue:43452|issue:44991", + "issue:43257|issue:43329", + "issue:36010|issue:36331", + "issue:36010|issue:43645", + "issue:43906|issue:46209", + "issue:43819|issue:43927", + "issue:43531|issue:45042", + "issue:43638|issue:44936", + "issue:42175|issue:46291" + ], + "prepared_review_unit_hash": "fbaea6d5fa4d145bfd3e6b78d588e8b8acd1e956795db7bfc9e188175f42ca54", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35874, + "estimated_input_tokens": 8969, + "estimated_eval_tokens": 18194 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: tokenizer, processor, training, quantization, and backend/runtime issues that only share broad transformer-adjacent vocabulary. None of the soft pairs look like the same underlying bug or a mergeable duplicate PR.", + "confidence": 0.96, + "canonical_issue_reason": "No single issue is a safe canonical representative for the whole set; the items span unrelated bugs across many subsystems.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45020 is the broadest thematic anchor here, but it is still only a loose umbrella for remote_code regressions and not a true duplicate target for the rest of the items.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "CamemBERT masked-LM prediction bug vs a broad tokenizer-version regression; same library, different failure mode and code path." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Packaging/backend dependency gap vs mistral tokenizer regex patch crash; unrelated symptoms and subsystems." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "Audio voice_preset bug vs trainer.compute_loss signature break; completely different components." + }, + { + "left": "issue:45003", + "right": "issue:46285", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils vs CLIP text model structural inconsistency; different model-loading problems." + }, + { + "left": "issue:46697", + "right": "issue:46728", + "accept": false, + "reason": "Both are image-processor issues, but one is over-padding and the other is patch counting; not the same bug." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "BigBirdTokenizer special-token registration bug vs MXFP4 quantization fallback; unrelated." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "Apple Silicon allocator TypeError vs CUDA AMP loss crash; different platforms and failure points." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio chat-template sampling-rate default vs DAC latent/forward mismatch; different audio paths." + }, + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Float8 storage lookup failure vs PIL backend requiring torchvision; unrelated runtime/dependency issues." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "Specific MobileLLM key error vs broad remote_code breakage; too different to be the same bug." + }, + { + "left": "issue:43782", + "right": "issue:45325", + "accept": false, + "reason": "Qwen3VL weight_only load error vs Qwen2.5-VL rope-index scaling bug; different models and code paths." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Both involve processor loading, but one is mistral regex patching and the other is a missing _tokenizer attribute in Gemma4Processor." + }, + { + "left": "issue:45372", + "right": "issue:46326", + "accept": false, + "reason": "mistral_common import failure for Gemma 4 processor loading vs continue_final_message behavior; shared dependency name only." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "Template compilation TypeError vs tokenizer model_max_length constant; unrelated tokenizer/template behaviors." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "torch.compile SDPA bug in Bamba vs quantization fallback in git-oss-20b; different failure surfaces." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels bug vs RTDetrV2 output mismatch in v5; different image-model issues." + }, + { + "left": "issue:43329", + "right": "issue:45600", + "accept": false, + "reason": "Multimodal token counting video-branch bug vs stale auto_mappings reference to removed configs; unrelated code paths." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "Image processor label reduction bug vs MPS attention correctness issue; no shared underlying defect." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Both mention tied-weights warnings in Qwen models, but one is a modeling_utils bug and the other is a user-facing config warning in fine-tuning; not the same fix." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "gguf_file loading regression vs tokenizer loading failure for est-roberta; different load mechanisms." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Qwen3 MOE weights conversion under accelerate+deepspeed vs multimodal token counting video bug; unrelated model-loading vs token-count logic." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "GenerationMixin import error vs custom trainer loss signature mismatch; both API breaks, but not the same underlying change." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "ImportError from transformers.generation vs Jupyter custom-model initialization breakage; distinct import/runtime failures." + }, + { + "left": "issue:43906", + "right": "issue:46209", + "accept": false, + "reason": "A reproduction of an older issue vs an FP8 inference failure on B200; one is a test case, the other a hardware-specific bug." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "DAC latent/forward mismatch vs DiaConfig losing custom token IDs after save/load; different audio/model configuration defects." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Qwen3-MoE sliding_window issue vs PIL backend torchvision dependency regression; unrelated." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Deepspeed ZeRO-3 index error during training vs evaluate-after-train failure; different trainer lifecycle bugs." + }, + { + "left": "issue:42175", + "right": "issue:46291", + "accept": false, + "reason": "Backend package selection problem vs kernels v0.15 import constraint for pipelines; unrelated dependency 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": 23, + "review_unit_index": 20, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:28282", + "issue:29159", + "issue:29870", + "issue:34567", + "issue:38175", + "issue:39692", + "issue:40990", + "issue:42490", + "issue:42915", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43450", + "issue:43454", + "issue:43525", + "issue:43540", + "issue:43577", + "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:44610", + "issue:44617", + "issue:44756", + "issue:44811", + "issue:44821", + "issue:44912", + "issue:44977", + "issue:45042", + "issue:45459", + "issue:45759", + "issue:45803", + "issue:45859", + "issue:46291", + "issue:46421", + "issue:46424", + "issue:46701" + ], + "soft_pairs": [ + "issue:43720|issue:44610", + "issue:42490|issue:43257", + "issue:39692|issue:44912", + "issue:38175|issue:44610", + "issue:43746|issue:44811", + "issue:43643|issue:45759", + "issue:44387|issue:44617", + "issue:43854|issue:44977", + "issue:43257|issue:43696", + "issue:43262|issue:44811", + "issue:43257|issue:43931", + "issue:42915|issue:43742", + "issue:46421|issue:46424", + "issue:43329|issue:44479", + "issue:43329|issue:43450", + "issue:45459|issue:45803", + "issue:43749|issue:44756", + "issue:44291|issue:45042", + "issue:42915|issue:44821", + "issue:43606|issue:45859", + "issue:34567|issue:40990", + "issue:43650|issue:43906", + "issue:43540|issue:46701", + "issue:43927|issue:44912", + "issue:43454|issue:43577", + "issue:43525|issue:46291", + "issue:29159|issue:34567", + "issue:28282|issue:29870" + ], + "prepared_review_unit_hash": "3fe71c5cd23092718f57f881e47f792c48f368675e5991d5f2dbda50a8308b24", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36249, + "estimated_input_tokens": 9063, + "estimated_eval_tokens": 18382 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated single-bug issues across transformers subsystems; none of the soft-edge pairs look like true duplicates or the same fix target. The cluster is heterogeneous, so the best canonical issue is the broadest/high-signal one rather than a shared root cause.", + "confidence": 0.97, + "canonical_issue_reason": "issue:43749 stands out as the most central representative: it has the highest discussion/inbound activity and concerns a broad loading-path regression with multiple downstream references.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43749 is the best standalone issue in this set because it appears impactful, well-documented, and broadly relevant compared with the many narrow model-specific regressions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "Different failures: packed-weight unpacking during accelerate loading vs processor/model image-size mismatch." + }, + { + "left": "issue:42490", + "right": "issue:43257", + "accept": false, + "reason": "Trailing-slash import warning is unrelated to Qwen MOE weight conversion under accelerate+deepspeed." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "SigLIP2 docs/example errors are not the same bug as gpt-oss-20b MXFP4 quantization fallback." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 is unrelated to the OmDet-Turbo 224px/640px processor-model mismatch." + }, + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "PEFT local-checkpoint loading and Whisper batch_decode skip_special_tokens are separate code paths." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "Different config-loading issues: missing fields with trust_remote_code vs not unwrapping text_config in from_config." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both are memory-related at a high level, but one is int4 reserved-memory growth and the other is a model-specific OOM." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "GLM unit-test loading failure is not the same as Qwen3.5 flash-attention generation corruption." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "Accelerate+deepspeed weight conversion bug is unrelated to plain CUDA OOM during GPT-oss-20b inference." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and Whisper batch_decode special-token handling are distinct defects." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "MOE weight conversion during accelerate loading is not the same as a Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "FineGrainedFP8Config failure and MobileLLM key error are unrelated loading/config problems." + }, + { + "left": "issue:46421", + "right": "issue:46424", + "accept": false, + "reason": "Same broad cache area, but one is lazy-init recompilation with chunked prefill and the other is max_cache_length scoping." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "Undefined video-token counting helpers and the v5.3.0 video-input regression are different concrete bugs." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "One is an undefined-variable bug in multimodal token counting; the other is a batched video-shape bug." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Tokenizer error masking from protobuf handling is not the same as a bare-except bug in Fuyu tensor conversion." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and disabling mmap on Strix Halo are different memory-related issues." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "init_empty_weights argument compatibility is unrelated to PIL backend processors requiring torchvision." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "Qwen MOE FP8 config failure has nothing to do with loading AutoImageProcessor from a URL." + }, + { + "left": "issue:43606", + "right": "issue:45859", + "accept": false, + "reason": "CPU offload device mismatch in bark-small is a different code path from Qwen3_5Moe rotary embedding offload compatibility." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "TrainerState token accounting and GPT-oss-20b perplexity are unrelated issues." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "The placeholder issue and the isolated reproduction of issue 38071 are not the same underlying bug." + }, + { + "left": "issue:43540", + "right": "issue:46701", + "accept": false, + "reason": "Qwen3OmniMoe video-processing ValueError and DiffusionGemma pipeline failure are unrelated model/pipeline bugs." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "DiaConfig save/load token-ID loss is unrelated to MXFP4 quantization fallback." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "AyaVision lm_head tying and Blip2 dtype propagation are different model-initialization issues." + }, + { + "left": "issue:43525", + "right": "issue:46291", + "accept": false, + "reason": "Llama4Config missing pad_token_id is unrelated to kernels import requiring version/revision." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency and TrainerState num_input_tokens_seen not updating do not share the same bug." + }, + { + "left": "issue:28282", + "right": "issue:29870", + "accept": false, + "reason": "AutoModel missing PyTorch and Bart generate decoder_attention_mask initialization are completely separate failures." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 21, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:24643", + "issue:27957", + "issue:28282", + "issue:29127", + "issue:29159", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:36331", + "issue:38175", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43317", + "issue:43404", + "issue:43425", + "issue:43452", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43824", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44560", + "issue:44568", + "issue:44821", + "issue:44849", + "issue:44898", + "issue:44987", + "issue:44991", + "issue:45127", + "issue:45137", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45405", + "issue:45488", + "issue:45538", + "issue:45701", + "issue:45907", + "issue:46291", + "issue:46512", + "issue:46693" + ], + "soft_pairs": [ + "issue:44987|issue:45538", + "issue:44568|issue:45356", + "issue:28282|issue:34689", + "issue:43425|issue:46291", + "issue:42915|issue:45538", + "issue:45137|issue:45405", + "issue:43697|issue:45357", + "issue:43404|issue:45127", + "issue:33357|issue:34567", + "issue:42175|issue:43824", + "issue:43742|issue:44291", + "issue:28282|issue:29127", + "issue:43317|issue:45538", + "issue:44479|issue:46693", + "issue:44062|issue:44991", + "issue:43577|issue:46512", + "issue:43452|issue:43582", + "issue:44849|issue:45701", + "issue:42915|issue:43749", + "issue:41720|issue:43540", + "issue:27957|issue:28282", + "issue:43540|issue:44560", + "issue:44291|issue:44821", + "issue:24643|issue:29159", + "issue:42175|issue:44488", + "issue:43540|issue:44479", + "issue:44987|issue:45907", + "issue:38175|issue:44898", + "issue:44568|issue:45488", + "issue:44479|issue:45381", + "issue:34567|issue:36331" + ], + "prepared_review_unit_hash": "961cb160299ffd17709e4a7b48dbc93e2f962c3b381e18b481f63beacbc7afa0", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37057, + "estimated_input_tokens": 9265, + "estimated_eval_tokens": 18786 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a mix of unrelated closed issue reports across tokenizer loading, model dtype/loading, training, and video/VL regressions. Nearly all soft pairs are false positives; only the Qwen video-input regression pair looks like the same underlying bug thread.", + "confidence": 0.74, + "canonical_issue_reason": "issue:44479 is the broadest and most representative report in the only near-duplicate family here: a Transformers 5.3.0 video-input regression affecting Qwen VL variants.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44479 best represents the cluster\u2019s clearest duplicated bug because it captures the shared Qwen video-input regression at a higher level than the narrower follow-up report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Different bugs: one is a loading regression for a specific model, the other is CLIPTokenizer max_length behavior." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer regressions, but they concern different failure modes and model families." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "Unrelated: PyTorch import detection versus a Llama Vision model-loading regression." + }, + { + "left": "issue:43425", + "right": "issue:46291", + "accept": false, + "reason": "Separate problems: Torch version incompatibility versus a kernels package revision/version requirement." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "Different areas entirely: Qwen3MoE FP8 config handling versus tokenizer max_length." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "DeepSpeed ZeRO3 runtime error and PEFT version pinning are unrelated." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "Different model/code paths: RTDetrV2 output changes versus Qwen3.5 save_pretrained key regression." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both mention weight tying, but one is a model definition bug and the other is LoRA merge collapse with extended vocab." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "MacOS bus error on CLIP loading is unrelated to TrainerState token counting." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "Backend packaging issue versus missing Qwen2.5-VL import; not the same bug." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "Both are loading errors, but one is a key error for a model, the other is init_empty_weights argument handling." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "PyTorch import error and LayoutLMv3 box-validation clarity are unrelated." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "Device-map/offload failure is unrelated to CLIPTokenizer max_length." + }, + { + "left": "issue:44479", + "right": "issue:46693", + "accept": false, + "reason": "Both involve vision/video performance or correctness, but one is a functional video-input regression and the other is a performance regression." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "Tokenizer construction errors, but different causes and different models." + }, + { + "left": "issue:43577", + "right": "issue:46512", + "accept": false, + "reason": "Related to dtype, but one is submodule dtype sticking at float32 and the other is config.dtype being stale after from_config." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "gguf loading/tokenizer issue and Apple Silicon allocator TypeError are unrelated." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Different symptoms: Qwen3.5 hidden-state behavior versus generic tokenization changes." + }, + { + "left": "issue:42915", + "right": "issue:43749", + "accept": false, + "reason": "Qwen3MoE FP8 config failure is unrelated to FSDP CPU RAM efficient loading." + }, + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "Different failures and code paths: auto device mapping cuda assert versus Qwen3OmniMoe video processing ValueError." + }, + { + "left": "issue:27957", + "right": "issue:28282", + "accept": false, + "reason": "Flash Attention 2 model issue and PyTorch import detection are not the same bug." + }, + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "Both are Qwen video-input bugs, but they affect different models and error paths (ValueError vs StopIteration)." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "Unexpected _is_hf_initialized argument is unrelated to loading AutoImageProcessor from a URL." + }, + { + "left": "issue:24643", + "right": "issue:29159", + "accept": false, + "reason": "DeepSpeed weight-dimension runtime error and tokenizer inconsistency are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "Packaging/backend inclusion issue versus model-loading failure for a specific checkpoint." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "These are both Qwen video issues, but the concrete bugs differ; one is a processing ValueError, the other is a broader regression in video position handling." + }, + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "Model loading regression and list_repo_files revision forwarding are unrelated." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Different image-model correctness issues with different failure modes." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Different tokenizer bugs: special-token addition versus LlamaTokenizer overriding tokenizer.json pre-tokenizer behavior." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "Same underlying Qwen VL video-input regression in Transformers 5.3.0; 45381 looks like a narrower manifestation of the broader bug reported in 44479." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "TrainerState token counting and CustomTrainer compute_loss signature changes are unrelated training API issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted soft pair, issue:44479 with issue:45381, is plausibly the same Qwen video-input regression family based on the titles. The rejected soft pairs are generally justified by distinct titles, model families, or failure modes, and none of the verdicts overclaim beyond the available evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 22, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29870", + "issue:32090", + "issue:43257", + "issue:43404", + "issue:43452", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43577", + "issue:43716", + "issue:43720", + "issue:43742", + "issue:43756", + "issue:43906", + "issue:44062", + "issue:44488", + "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:45357", + "issue:45405", + "issue:45479", + "issue:45588", + "issue:45600", + "issue:45823", + "issue:45920", + "issue:45987", + "issue:46121", + "issue:46399", + "issue:46463", + "issue:46468", + "issue:46491", + "issue:46693", + "issue:46701" + ], + "soft_pairs": [ + "issue:44792|issue:44964", + "issue:43257|issue:43756", + "issue:44062|issue:45081", + "issue:43742|issue:44589", + "issue:44737|issue:46463", + "issue:45042|issue:45081", + "issue:43716|issue:46701", + "issue:43756|issue:46399", + "issue:43531|issue:44821", + "issue:43756|issue:45325", + "issue:43577|issue:45072", + "issue:44568|issue:46491", + "issue:45405|issue:45600", + "issue:43906|issue:46468", + "issue:45823|issue:46463", + "issue:24643|issue:29870", + "issue:43404|issue:45987", + "issue:43452|issue:44488", + "issue:44737|issue:45588", + "issue:43906|issue:46693", + "issue:44625|issue:44877", + "issue:43526|issue:45479", + "issue:45357|issue:45920", + "issue:44913|issue:46121", + "issue:43540|issue:46693", + "issue:28282|issue:32090", + "issue:43720|issue:44898" + ], + "prepared_review_unit_hash": "98a8fa7ef0b3d0a1b7e93127b3c54924771be43e271a0e986f3f406aaa2bf7ae", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35635, + "estimated_input_tokens": 8909, + "estimated_eval_tokens": 18074 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: most items are unrelated Transformers bugs/regressions across different models, tokenizers, configs, and loading paths. I found no true duplicate pair among the suggested soft edges.", + "confidence": 0.95, + "canonical_issue_reason": "If a representative issue must be chosen, issue #45920 is the broadest user-facing regression in the set (multi-model AutoTokenizer wrong IDs) and has the strongest cross-cutting impact; however, the cluster does not actually contain a clear duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #45920 is the best single issue here because it describes a widespread regression affecting several models and is more broadly representative than the highly model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Both are multimodal-related, but one is a Janus generation test failure and the other is a Phi-4 model loading regression; different code paths and symptoms." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Qwen3 MOE weight conversion with accelerate/deepspeed is unrelated to Smollm3 RoPE layer dropping; different models and failure modes." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer AddedToken duplication and Mistral regex patching are separate tokenizer bugs with different causes." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "MobileLLM key lookup failure and Float8 storage deserialization are unrelated loading errors." + }, + { + "left": "issue:44737", + "right": "issue:46463", + "accept": false, + "reason": "XLNet CPU device placement bug is unrelated to T5 decoder-input validation messaging." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "Torchvision dependency in PIL image processors is unrelated to Mistral tokenizer regex patching." + }, + { + "left": "issue:43716", + "right": "issue:46701", + "accept": false, + "reason": "Mistral-3 image preprocessor dtype mismatch and DiffusionGemma image-text-to-text pipeline failure are different multimodal issues." + }, + { + "left": "issue:43756", + "right": "issue:46399", + "accept": false, + "reason": "Smollm3 RoPE-layer dropping and Qwen3.5 kernel registration failure are unrelated model-specific bugs." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior has nothing to do with AutoImageProcessor URL loading." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Both mention RoPE/positioning, but Smollm3 layer selection and Qwen2.5-VL temporal position scaling are distinct defects." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Both are dtype-related, but Blip2 loading and SwitchTransformers/TimmWrapperModel bfloat16 inference are different code paths and bugs." + }, + { + "left": "issue:44568", + "right": "issue:46491", + "accept": false, + "reason": "Tokenizer special-token insertion regression and CodeLlama encode/decode whitespace regression are separate tokenizer behaviors." + }, + { + "left": "issue:45405", + "right": "issue:45600", + "accept": false, + "reason": "A version pin to unreleased PEFT and stale auto_mappings references are unrelated maintenance issues." + }, + { + "left": "issue:43906", + "right": "issue:46468", + "accept": false, + "reason": "An isolated reproduction of an older issue and the Nemotron example breakage do not point to the same underlying bug." + }, + { + "left": "issue:45823", + "right": "issue:46463", + "accept": false, + "reason": "Gemma4 device-map mismatch and T5 decoder-input validation messaging are unrelated." + }, + { + "left": "issue:24643", + "right": "issue:29870", + "accept": false, + "reason": "DeepSpeed weight-dimension runtime error and BART decoder_attention_mask UnboundLocalError are different training/generation failures." + }, + { + "left": "issue:43404", + "right": "issue:45987", + "accept": false, + "reason": "Mistral3 lm_head tying and StaticCache return-type mismatch are unrelated implementation bugs." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "gguf_file load failures and cjvt/sleng-bert loading issues are both model-loading related but not the same defect." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "XLNet relative-position CPU allocation and flash_attention s_aux=None crash are unrelated." + }, + { + "left": "issue:43906", + "right": "issue:46693", + "accept": false, + "reason": "The former is a reproduction report for an older bug; the latter is a long-context FA2 performance regression." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and strict granite_speech config loading are different config-handling bugs." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "BEiT Fast label reduction and single-label-classification zero-loss behavior are distinct classification/preprocessing issues." + }, + { + "left": "issue:45357", + "right": "issue:45920", + "accept": false, + "reason": "Incorrect visual encoder keys on save_pretrained and wrong AutoTokenizer token IDs are separate regression classes." + }, + { + "left": "issue:44913", + "right": "issue:46121", + "accept": false, + "reason": "GPTNeoX rotary_pct reload loss and rope-validation ignore-keys type handling are different RoPE/config bugs." + }, + { + "left": "issue:43540", + "right": "issue:46693", + "accept": false, + "reason": "Qwen3OmniMoe video-input processing and FA2 long-context performance regression are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "Missing PyTorch import and NoneType broadcast failure in Trainer are unrelated environment/runtime errors." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "BitNet packed-weight unpacking during accelerate loading and Perceiver non-default resolution handling are different model-loading 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": 23, + "review_unit_index": 23, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:29159", + "issue:33357", + "issue:38175", + "issue:41762", + "issue:42898", + "issue:43295", + "issue:43452", + "issue:43525", + "issue:43531", + "issue:43577", + "issue:43643", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43819", + "issue:43927", + "issue:43994", + "issue:44361", + "issue:44442", + "issue:44464", + "issue:44466", + "issue:44493", + "issue:44568", + "issue:44743", + "issue:44811", + "issue:44843", + "issue:44898", + "issue:44936", + "issue:45005", + "issue:45200", + "issue:45216", + "issue:45356", + "issue:45372", + "issue:45375", + "issue:45542", + "issue:45704", + "issue:45823", + "issue:46050", + "issue:46170", + "issue:46178", + "issue:46402", + "issue:46463", + "issue:46491", + "issue:46701" + ], + "soft_pairs": [ + "issue:43697|issue:45216", + "issue:43643|issue:45375", + "issue:44466|issue:45005", + "issue:45200|issue:46463", + "issue:41762|issue:44936", + "issue:44361|issue:44442", + "issue:44493|issue:46050", + "issue:43653|issue:44361", + "issue:43819|issue:44811", + "issue:43525|issue:46701", + "issue:43688|issue:43749", + "issue:43994|issue:44898", + "issue:43452|issue:44843", + "issue:38175|issue:43295", + "issue:44936|issue:45823", + "issue:43531|issue:43742", + "issue:45372|issue:46170", + "issue:45542|issue:46178", + "issue:42898|issue:46491", + "issue:43452|issue:43577", + "issue:43646|issue:43749", + "issue:43577|issue:43697", + "issue:43295|issue:45356", + "issue:44464|issue:44743", + "issue:43653|issue:44568", + "issue:29159|issue:33357", + "issue:43756|issue:46402", + "issue:43688|issue:45704", + "issue:43653|issue:43927" + ], + "prepared_review_unit_hash": "299c1e56f2ee2e558950a990ac9be4be30ca035f8ed4e6375273e62d3b3b73af", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36740, + "estimated_input_tokens": 9185, + "estimated_eval_tokens": 18626 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous: tokenizer quirks, config/save-load regressions, model-specific loading bugs, training/eval issues, and unrelated backend/dependency errors. None of the soft pairs look like the same underlying bug/change, so all should be rejected. If a single issue must stand in as a canonical anchor, the broadest one is the generic custom-model regression ticket, but it still only loosely represents the set.", + "confidence": 0.88, + "canonical_issue_reason": "Issue 43646 is the broadest umbrella report here: a general Transformers v5 custom model initialization regression, which is closer to the overall \u201cframework regression\u201d theme than the narrow model-specific bugs around it.", + "canonical_pr_reason": null, + "best_issue_reason": "43646 is the most general and reusable issue in the set, but the cluster is too mixed for a truly strong canonical issue; it best serves as a weak anchor rather than a real duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "Different models and failure modes: RTDetrV2 output drift vs Qwen3.5 checkpoint save_pretrained regression." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "Both involve config handling, but one is missing fields from trust_remote_code AutoConfig while the other drops a specific Qwen3_5MoeVisionConfig field." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": false, + "reason": "Both mention tied weights, but one is device-dependent serialization of lm_head.weight and the other is a translation-model-specific tied-weights issue; not the same concrete bug." + }, + { + "left": "issue:45200", + "right": "issue:46463", + "accept": false, + "reason": "Unrelated areas: Gemma 4 mm_token_type_ids defaulting vs a T5 decoder-input validation error message." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "Gemma3 ZeRO-3 loading IndexError is unrelated to trainer.evaluate() failing after trainer.train()." + }, + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Different tokenizer families and errors: MLukeTokenizer AttributeError vs AutoTokenizer failing to load FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:44493", + "right": "issue:46050", + "accept": false, + "reason": "Position-id key mismatch is unrelated to quantized-model save/load breakage after sinq quantization." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "Both are tokenizer bugs, but BigBirdTokenizer special-token registration and MLukeTokenizer task AttributeError are different root causes." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "DAC.from_latents forward mismatch and Whisper batch_decode skip_special_tokens handling are unrelated audio-model behaviors." + }, + { + "left": "issue:43525", + "right": "issue:46701", + "accept": false, + "reason": "Llama4Config pad_token_id attribute error is unrelated to the DiffusionGemmaGenerationOutput pipeline failure." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Aux-loss normalization and FSDP CPU RAM efficient loading are separate training/distributed issues." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-model bugs: SigLIP2 nonsensical outputs vs Perceiver failing with non-default resolution/interpolate_pos_encoding." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "Both affect AutoTokenizer paths, but gguf_file loading breakage and offline model_info calls in _patch_mistral_regex are distinct bugs." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Siglip2 zero-probabilities and a v4.57 custom-model regression involving processor.tokenizer/images are unrelated." + }, + { + "left": "issue:44936", + "right": "issue:45823", + "accept": false, + "reason": "Trainer evaluate/train lifecycle issue is unrelated to Gemma4 PLE device placement with device_map=\"auto\"." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and MobileLLM-125M loading KeyError are different model/config bugs." + }, + { + "left": "issue:45372", + "right": "issue:46170", + "accept": false, + "reason": "Gemma 4 processor loading import failure from mistral_common is unrelated to the Windows ROCm torch.distributed dependency problem." + }, + { + "left": "issue:45542", + "right": "issue:46178", + "accept": false, + "reason": "Both are backend/dependency related, but tensorboard-without-TF undefined backend and tensorflow_text missing from BACKENDS_MAPPING are not the same bug." + }, + { + "left": "issue:42898", + "right": "issue:46491", + "accept": false, + "reason": "Generic v5 clean_up_tokenization_spaces behavior change is not the same as CodeLlama's leading-space round-trip regression." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "gguf_file AutoTokenizer/AutoModel loading breakage is unrelated to Blip2 dtype remaining float32." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "A broad custom-model initialization regression is unrelated to the FSDP CPU RAM efficient loading failure." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "Blip2 dtype propagation and RTDetrV2 output inconsistency are separate model-specific issues." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Both are regressions, but one is custom processor/model access and the other is Kimi-K2.5 tokenizer codec handling; different code paths." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Chunked generation with compiled forward and Qwen3_5MoE recurrent-state reset are distinct generation/caching bugs." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Special-token registration on BigBirdTokenizer and missing BOS/EOS insertion on mdeberta-v3 tokenizer are different tokenizer-specific defects." + }, + { + "left": "issue:29159", + "right": "issue:33357", + "accept": false, + "reason": "Inconsistent slow vs fast tokenizer behavior is unrelated to a MacOS bus error on a community CLIP model." + }, + { + "left": "issue:43756", + "right": "issue:46402", + "accept": false, + "reason": "Smollm3 RoPE-layer mismatch and Sam3Model text encoder weights not loading are unrelated model-architecture/loading issues." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "Aux-loss normalization and T5 using apex.FusedRMSNorm with a memory leak are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and DiaConfig losing custom token IDs after save/load are both token-related, but 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": 23, + "review_unit_index": 24, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:32090", + "issue:33666", + "issue:36010", + "issue:36331", + "issue:39692", + "issue:41720", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43493", + "issue:43504", + "issue:43525", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43643", + "issue:43688", + "issue:43720", + "issue:43742", + "issue:43782", + "issue:43825", + "issue:43927", + "issue:44361", + "issue:44464", + "issue:44466", + "issue:44479", + "issue:44484", + "issue:44554", + "issue:44561", + "issue:44610", + "issue:44625", + "issue:44704", + "issue:44734", + "issue:44821", + "issue:44843", + "issue:44964", + "issue:44977", + "issue:44991", + "issue:45405", + "issue:45440", + "issue:45491", + "issue:45701", + "issue:46002", + "issue:46050", + "issue:46365", + "issue:46612", + "issue:46693" + ], + "soft_pairs": [ + "issue:43688|issue:45701", + "issue:44843|issue:44991", + "issue:44479|issue:44625", + "issue:45491|issue:46002", + "issue:43550|issue:44361", + "issue:44484|issue:44704", + "issue:43742|issue:44821", + "issue:33666|issue:36010", + "issue:43688|issue:45440", + "issue:44734|issue:46612", + "issue:44466|issue:46050", + "issue:32090|issue:33666", + "issue:41720|issue:44464", + "issue:44464|issue:44977", + "issue:43504|issue:43550", + "issue:43504|issue:43720", + "issue:44977|issue:46693", + "issue:43504|issue:44610", + "issue:43577|issue:44554", + "issue:43720|issue:43927", + "issue:33666|issue:36331", + "issue:43643|issue:44561", + "issue:43606|issue:43927", + "issue:43278|issue:43688", + "issue:43782|issue:46693", + "issue:43262|issue:43295", + "issue:39692|issue:43493", + "issue:43720|issue:44361", + "issue:43525|issue:43572", + "issue:44964|issue:45405", + "issue:44554|issue:46365", + "issue:43493|issue:43825" + ], + "prepared_review_unit_hash": "7e49f7842cae90152f838dcdc7db42833e0ffa3cff28aa772b4fc6d32604556b", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37389, + "estimated_input_tokens": 9348, + "estimated_eval_tokens": 18952 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the soft-linked pairs cover unrelated bugs across trainers, configs, tokenization, generation, processors, and model loading. No pair looks like the same underlying fix, so all soft edges are rejected. As a cluster anchor, the broadest and most referenced issue is the tokenization/version-regression report.", + "confidence": 0.96, + "canonical_issue_reason": "issue:45701 is the broadest, most cross-cutting regression in the set and has the most evidence of impact (multiple inbound references), so it is the best anchor issue if one must be chosen.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45701 best represents the cluster because it is a generic version-induced behavior change, not a narrow model-specific edge case, and it has strong discussion/impact signals.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "Different bugs: auxiliary-loss normalization vs tokenization changes." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Both involve tokenizer loading, but the failure modes and affected paths are different." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Unrelated regressions: video-input handling vs num_labels propagation." + }, + { + "left": "issue:45491", + "right": "issue:46002", + "accept": false, + "reason": "Both mention NaNs, but the root causes and code paths are unrelated." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "Different subsystems entirely: torch.compile/SDPA vs tokenizer/task AttributeError." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "save_pretrained shard sizing vs AutoProcessor/cached_file kwargs are different issues." + }, + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "Different loading problems: model key error vs loading AutoImageProcessor from URL." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Multi-GPU training issue vs GenerationMixin import failure." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Both concern model correctness, but these are distinct implementation discrepancies." + }, + { + "left": "issue:44734", + "right": "issue:46612", + "accept": false, + "reason": "Different generation-cache bugs in different execution paths; not one fix." + }, + { + "left": "issue:44466", + "right": "issue:46050", + "accept": false, + "reason": "Serialization inconsistency vs quantized save/load breakage are separate problems." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "Trainer broadcast TypeError is unrelated to Qwen2-VL multi-GPU training." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "Auto device mapping crash vs compiled-forward chunked generation mismatch." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Both affect generation output, but one is compiled forward and the other is flash-attention-specific." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Pretrained preset loading bug vs Bamba SDPA compile failure are unrelated." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Legacy-field load failure vs BitNet packed-weight loading are different bugs." + }, + { + "left": "issue:44977", + "right": "issue:46693", + "accept": false, + "reason": "Correctness bug in Qwen3.5 generation vs performance regression in FA2 long-context generation." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "Model preset loading and processor/model size mismatch are different issues." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "Blip2 dtype loading regression vs MPS attention correctness issue." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "Packed-weight accelerate loading vs config token-ID persistence are unrelated." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Multi-GPU training problem vs unexpected trainer kwarg regression." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Both touch trust_remote_code, but one is missing config fields and the other is removed helper incompatibility." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "CPU offload device mismatch vs DiaConfig token-ID loss are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43688", + "accept": false, + "reason": "Embedding dtype drift vs auxiliary-loss normalization are different model behaviors." + }, + { + "left": "issue:43782", + "right": "issue:46693", + "accept": false, + "reason": "from_pretrained weight_only failure vs FA2 performance regression are different failure modes." + }, + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Audio sampling-rate default bug vs processor.tokenizer regression are separate processor issues." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "Docs/example errors and implementation discrepancy in SigLIP2 are not the same concrete bug." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "BitNet weight-loading bug vs MLukeTokenizer AttributeError are unrelated." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Similar missing-pad-field symptoms, but different config classes and code paths." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Model loading regression vs dependency version bump are unrelated." + }, + { + "left": "issue:44554", + "right": "issue:46365", + "accept": false, + "reason": "MPS attention correctness issue vs Trainer import conflict with datasets are unrelated." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 implementation discrepancy vs pipeline error-message regression 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": 23, + "review_unit_index": 25, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:28218", + "issue:33357", + "issue:33666", + "issue:39692", + "issue:42796", + "issue:43257", + "issue:43278", + "issue:43295", + "issue:43299", + "issue:43334", + "issue:43452", + "issue:43493", + "issue:43504", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43701", + "issue:43824", + "issue:43844", + "issue:43927", + "issue:43994", + "issue:44112", + "issue:44361", + "issue:44451", + "issue:44466", + "issue:44488", + "issue:44514", + "issue:44568", + "issue:44589", + "issue:44625", + "issue:44792", + "issue:44843", + "issue:44849", + "issue:44977", + "issue:45290", + "issue:45357", + "issue:45405", + "issue:45520", + "issue:45584", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:46050", + "issue:46192", + "issue:46209", + "issue:46489", + "issue:46710" + ], + "soft_pairs": [ + "issue:44514|issue:45290", + "issue:44625|issue:45584", + "issue:44977|issue:46209", + "issue:43452|issue:44451", + "issue:45823|issue:46050", + "issue:28218|issue:33357", + "issue:43994|issue:46192", + "issue:42796|issue:44977", + "issue:43493|issue:43844", + "issue:43278|issue:43844", + "issue:39692|issue:43295", + "issue:44849|issue:46209", + "issue:33357|issue:33666", + "issue:43646|issue:44466", + "issue:44568|issue:46489", + "issue:46209|issue:46710", + "issue:43278|issue:43701", + "issue:43575|issue:45405", + "issue:44792|issue:45405", + "issue:43927|issue:44361", + "issue:44589|issue:45520", + "issue:44849|issue:45357", + "issue:43824|issue:44589", + "issue:44568|issue:45812", + "issue:44488|issue:44843", + "issue:43504|issue:43606", + "issue:43257|issue:43299", + "issue:43334|issue:43572", + "issue:43334|issue:45759", + "issue:43606|issue:44112" + ], + "prepared_review_unit_hash": "eb5b984f8accdcb77ea82be8ab4580e4172123f604997f8172a2ae01ae22abb6", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36754, + "estimated_input_tokens": 9189, + "estimated_eval_tokens": 18634 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a loose collection of tokenizer, config, and model-loading regression reports in Transformers v5, but the soft pairs are too behaviorally different to treat as true duplicates. No PRs are present.", + "confidence": 0.67, + "canonical_issue_reason": "issue:45812 is the most central and broadly scoped tokenizer regression here: it describes a silent v4\u2192v5 break affecting all Granite models, which fits the cluster\u2019s strongest shared theme better than the narrower or unrelated reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45812 is the best representative issue overall because it is concrete, high-impact, and broadly applicable; it also aligns with the most repeated topic in the cluster (tokenizer behavior regressions).", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Both mention apply_chat_template crashes, but the failure modes differ: batched VL input with padding=False vs assistant tool-call messages with empty content. Not the same concrete bug." + }, + { + "left": "issue:44625", + "right": "issue:45584", + "accept": false, + "reason": "One is about num_labels propagation in Qwen3.5 config; the other is Whisper generation failing after align_special_tokens. Different subsystems and root causes." + }, + { + "left": "issue:44977", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3.5 flash-attention generation problems and FP8 inference failures on B200 are unrelated code paths." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "Both are loading-related, but one is about gguf_file breaking from_pretrained APIs and the other is a specific model failing to load. Too broad to merge." + }, + { + "left": "issue:45823", + "right": "issue:46050", + "accept": false, + "reason": "Gemma4 device-map forward mismatch vs sinq quantization save/load breakage are distinct bugs." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer extra-space behavior and a MacOS bus error in a CLIP model are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:46192", + "accept": false, + "reason": "SigLIP2 wrong outputs and return_tensors being ignored when text_kwargs is passed do not share the same concrete failure." + }, + { + "left": "issue:42796", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer BPE marker handling and Qwen3.5 flash-attention generation are different issues." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and a DeepSpeed ZeRO-3 gradient increase issue are not the same bug." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Embedding dtype drift during evaluate and gradient blow-up under ZeRO-3 are different training/eval problems." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "SigLIP2 doc example errors and a regression in processor.tokenizer/custom image passing are unrelated." + }, + { + "left": "issue:44849", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3.5 output_hidden_states behavior and FP8 inference failure on B200 are separate issues." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "A bus error on MacOS and Qwen2-VL multi-GPU training are unrelated." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Custom model initialization breakage and lm_head serialization inconsistency affect different code paths." + }, + { + "left": "issue:44568", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer regressions, but one is missing BOS/EOS insertion for mdeberta and the other is DeepSeek-Coder output corruption tied to a specific prior fix gap; not enough evidence of one underlying bug." + }, + { + "left": "issue:46209", + "right": "issue:46710", + "accept": false, + "reason": "FP8 inference on B200 and an accuracy regression after upgrading Transformers are different symptom sets and likely different causes." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "Embedding dtype change in evaluation and resume_from_checkpoint key mismatch are unrelated." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "TP OOM when loading Qwen2-57B and an unmet PEFT version bump are not the same issue." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "Janus image-generation test failure and a dependency version bump are unrelated." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "DiaConfig custom token IDs loss and MLukeTokenizer AttributeError are different tokenizer/config bugs." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Float8 storage lookup failure and a Python 3.13 flash_attn import key error are separate compatibility issues." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Qwen3.5 hidden-states behavior and incorrect visual encoder keys in save_pretrained are different regressions." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "ImportError for Qwen2_5_VLForConditionalGeneration and Float8 storage lookup failure are unrelated." + }, + { + "left": "issue:44568", + "right": "issue:45812", + "accept": false, + "reason": "Both concern tokenization regressions, but they affect different tokenizers and different mechanics (special token addition vs Granite token IDs). Not the same concrete bug." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "Failure to load a specific BERT model and offline-mode hub access in _patch_mistral_regex are unrelated." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "BEiT legacy-field loading and CPU offload device mismatch are different model-loading failures." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Both involve Qwen3 MoE loading, but one is accelerate+deepspeed weight conversion and the other is a dev-version model-loading break in Qwen3VL. Not clearly the same code-path bug." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Qwen3-VL pad_token_id missing and StableLm pad_token_idx missing are separate config regressions." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "Missing pad_token_id in Qwen3-VL text config and not unwrapping text_config in AutoModelForCausalLM.from_config are related only loosely via composite configs, but not the same bug." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "These are both CI/device-mismatch flavored, but one is CPU offload for bark-small and the other is a stale GraniteSpeech test. Different defects." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 26, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:15354", + "issue:28218", + "issue:29159", + "issue:29942", + "issue:30333", + "issue:32090", + "issue:34567", + "issue:38175", + "issue:41720", + "issue:42490", + "issue:43065", + "issue:43493", + "issue:43504", + "issue:43572", + "issue:43653", + "issue:43696", + "issue:43701", + "issue:43844", + "issue:43873", + "issue:43937", + "issue:43994", + "issue:44291", + "issue:44423", + "issue:44464", + "issue:44493", + "issue:44554", + "issue:44734", + "issue:44869", + "issue:44964", + "issue:44977", + "issue:45084", + "issue:45200", + "issue:45229", + "issue:45245", + "issue:45406", + "issue:45446", + "issue:45520", + "issue:45584", + "issue:45588", + "issue:45823", + "issue:46050", + "issue:46121", + "issue:46285", + "issue:46369", + "issue:46424", + "issue:46701", + "issue:46710", + "issue:46752" + ], + "soft_pairs": [ + "issue:43572|issue:44291", + "issue:45245|issue:46710", + "issue:44493|issue:46285", + "issue:43504|issue:43653", + "issue:44464|issue:46424", + "issue:43937|issue:46701", + "issue:44464|issue:45584", + "issue:45200|issue:45229", + "issue:45229|issue:45823", + "issue:44869|issue:46752", + "issue:43844|issue:44964", + "issue:43873|issue:46050", + "issue:29942|issue:30333", + "issue:43994|issue:44493", + "issue:45588|issue:46050", + "issue:29159|issue:32090", + "issue:42490|issue:43696", + "issue:45084|issue:45520", + "issue:44977|issue:46701", + "issue:28218|issue:34567", + "issue:43065|issue:46369", + "issue:44554|issue:45446", + "issue:15354|issue:38175", + "issue:44423|issue:44734", + "issue:43493|issue:43701", + "issue:41720|issue:46710", + "issue:45406|issue:46121" + ], + "prepared_review_unit_hash": "48a32b1c77cd6a521bbe0c462ab21b5f6c09ce48cbfd243f2b7bc7c1de44b101", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35813, + "estimated_input_tokens": 8954, + "estimated_eval_tokens": 18164 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: tokenizer quirks, model-loading issues, serving crashes, quantization/offloading bugs, and training/runtime errors. None of the suggested soft pairs look like the same underlying bug or change.", + "confidence": 0.97, + "canonical_issue_reason": "No strong canonical issue; the listed issues span unrelated bugs, so any single representative would be arbitrary.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45229 is the most discussed and concrete report, with a clear reproduction and impact, so it is the best representative if one must be picked.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Config field regression vs an unexpected init_empty_weights argument; different code paths and symptoms." + }, + { + "left": "issue:45245", + "right": "issue:46710", + "accept": false, + "reason": "Category-cardinality runtime error vs an accuracy regression after upgrade; not the same bug." + }, + { + "left": "issue:44493", + "right": "issue:46285", + "accept": false, + "reason": "Position-id key handling vs CLIP text model structure drift; unrelated model internals." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "Legacy field loading failure in BEiT vs BigBird tokenizer special-token registration; different components." + }, + { + "left": "issue:44464", + "right": "issue:46424", + "accept": false, + "reason": "Compiled-forward generation mismatch vs cache-length config semantics; not one concrete defect." + }, + { + "left": "issue:43937", + "right": "issue:46701", + "accept": false, + "reason": "Invalid GenerationConfig vs a broken image-text pipeline; different failure surfaces." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Chunked generation with compiled forward is unrelated to Whisper empty-transcription decode failure." + }, + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Missing mm_token_type_ids defaults vs multi-GPU CUDA OOM; separate Gemma4 issues." + }, + { + "left": "issue:45229", + "right": "issue:45823", + "accept": false, + "reason": "OOM during inference vs device-mismatch during forward; not the same root cause." + }, + { + "left": "issue:44869", + "right": "issue:46752", + "accept": false, + "reason": "Whisper timestamp decode crash vs chat-template empty-list IndexError; different API paths." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "ZeRO-3 gradient anomaly vs model loading failure for Phi-4 multimodal; unrelated." + }, + { + "left": "issue:43873", + "right": "issue:46050", + "accept": false, + "reason": "Offloading/quantization behavior vs sinq save/load breakage; same broad area but distinct bugs." + }, + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention test failures vs MLflow reporting on job failure; no shared underlying defect." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 output nonsense vs unexpected position-id key handling; different problems." + }, + { + "left": "issue:45588", + "right": "issue:46050", + "accept": false, + "reason": "Flash-attention sink-less model crash vs quantized-model persistence bug; unrelated." + }, + { + "left": "issue:29159", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer inconsistency vs Trainer broadcast NoneType error; different layers entirely." + }, + { + "left": "issue:42490", + "right": "issue:43696", + "accept": false, + "reason": "Path trailing-slash import warning vs CUDA OOM; no duplicate signal." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Template-node compilation error vs flash_attn import KeyError on Python 3.13; unrelated." + }, + { + "left": "issue:44977", + "right": "issue:46701", + "accept": false, + "reason": "Qwen3.5 flash-attention generation issue vs image-text pipeline failure; different models and code paths." + }, + { + "left": "issue:28218", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer spacing bug vs TrainerState token counter not updating; not the same bug." + }, + { + "left": "issue:43065", + "right": "issue:46369", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder vs memory overflow on sam3-litetext; different SAM3 issues." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "MPS attention correctness issue vs incorrect PyTorch version check for flex_attention import; unrelated." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "TorchScript GeneratorExp export failure vs SigLIP2 zero-probability output; different functionality." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Serve crash from multimodal input handling vs KV-cache continuation indexing bug; both in serve, but distinct bugs." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 HF/JAX discrepancy vs resume_from_checkpoint key mismatch; unrelated training/model issues." + }, + { + "left": "issue:41720", + "right": "issue:46710", + "accept": false, + "reason": "Device-mapping CUDA assert on A800 vs an accuracy regression after upgrade; not the same defect." + }, + { + "left": "issue:45406", + "right": "issue:46121", + "accept": false, + "reason": "Gemma4 processor missing _tokenizer vs rope-parameter conversion type error; 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": 23, + "review_unit_index": 27, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:28218", + "issue:29870", + "issue:29942", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:42490", + "issue:42915", + "issue:43232", + "issue:43278", + "issue:43295", + "issue:43404", + "issue:43450", + "issue:43493", + "issue:43540", + "issue:43550", + "issue:43572", + "issue:43575", + "issue:43653", + "issue:43824", + "issue:43844", + "issue:43856", + "issue:43873", + "issue:43957", + "issue:44112", + "issue:44186", + "issue:44466", + "issue:44521", + "issue:44554", + "issue:44610", + "issue:44849", + "issue:44987", + "issue:45072", + "issue:45127", + "issue:45198", + "issue:45200", + "issue:45290", + "issue:45356", + "issue:45381", + "issue:45520", + "issue:45588", + "issue:45759", + "issue:45823", + "issue:46178", + "issue:46192", + "issue:46209", + "issue:46710" + ], + "soft_pairs": [ + "issue:43404|issue:46710", + "issue:45072|issue:45198", + "issue:37428|issue:43957", + "issue:45588|issue:45759", + "issue:43295|issue:43450", + "issue:43540|issue:45200", + "issue:44186|issue:46192", + "issue:45200|issue:45823", + "issue:43295|issue:43873", + "issue:45823|issue:46192", + "issue:43295|issue:44610", + "issue:29870|issue:29942", + "issue:45520|issue:46178", + "issue:43572|issue:44987", + "issue:43450|issue:43575", + "issue:43550|issue:43844", + "issue:43232|issue:44849", + "issue:43572|issue:43957", + "issue:37428|issue:43824", + "issue:44554|issue:45356", + "issue:42490|issue:43493", + "issue:38175|issue:43493", + "issue:39692|issue:46192", + "issue:43653|issue:44112", + "issue:44466|issue:45127", + "issue:43278|issue:43856", + "issue:45381|issue:46209", + "issue:44521|issue:45290", + "issue:28218|issue:35141", + "issue:42915|issue:43824" + ], + "prepared_review_unit_hash": "26a22a119cc76027a63c2acfbb1d765214d432029b55502c470c878ac87c4b2f", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36969, + "estimated_input_tokens": 9243, + "estimated_eval_tokens": 18742 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The items are a loose mix of unrelated Transformers issues. None of the soft pairs look like the same concrete bug/change, so I would not merge any of them. If forced to pick a single issue as the most representative umbrella report, the multimodal processor/tokenizer regression in #43295 is the closest fit.", + "confidence": 0.39, + "canonical_issue_reason": "#43295 is the most central of the weakly related reports: it describes a broad multimodal processor/tokenizer regression and has the most soft links to other processor-facing issues.", + "canonical_pr_reason": null, + "best_issue_reason": "#43295 is the best umbrella issue among this set because it is a reproducible user-facing regression with the widest apparent overlap to other multimodal processor/tokenizer complaints.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43404", + "right": "issue:46710", + "accept": false, + "reason": "Different bugs: one is tied-weight handling in a specific multimodal generation model, the other is an accuracy regression in DeepSeek after an upgrade." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "Bfloat16 dtype mismatches during inference are unrelated to save_pretrained/tokenization failures." + }, + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "An import error for a flash-attention helper is not the same as a meta-device model-loading regression." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention sink handling and AutoModelForCausalLM text_config unwrapping are separate code paths and separate bugs." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Both are multimodal, but one is a processor/tokenizer regression and the other is a batched video-shape bug." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Qwen3OmniMoe video processing failure and Gemma4 mm_token_type_ids defaults are different model-specific issues." + }, + { + "left": "issue:44186", + "right": "issue:46192", + "accept": false, + "reason": "LayoutLMv2 tokenizer crashes on NER/padding are unrelated to return_tensors being ignored when text_kwargs is passed." + }, + { + "left": "issue:45200", + "right": "issue:45823", + "accept": false, + "reason": "One is about missing/default mm_token_type_ids, the other is a device placement mismatch in Gemma4 forward." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor/tokenizer regression and quantization offloading behavior are not the same underlying defect." + }, + { + "left": "issue:45823", + "right": "issue:46192", + "accept": false, + "reason": "A Gemma4 device_map bug does not match a text_kwargs/return_tensors precedence issue." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Custom model code breaking on processor.tokenizer/images is different from OmDet-Turbo's 640px vs 224px preprocessing mismatch." + }, + { + "left": "issue:29870", + "right": "issue:29942", + "accept": false, + "reason": "Bart generate decoder_attention_mask initialization and Flash Attention 2 test failures are distinct problems." + }, + { + "left": "issue:45520", + "right": "issue:46178", + "accept": false, + "reason": "Python 3.13 flash_attn import metadata and tensorflow_text backend mapping are unrelated import-system errors." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig is not the same as failing to load a specific model checkpoint." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "A batched video processor shape bug is unrelated to a tensor-parallel OOM when loading Qwen2-57B." + }, + { + "left": "issue:43550", + "right": "issue:43844", + "accept": false, + "reason": "torch.compile SDPA failure and ZeRO-3 gradient growth are different runtime/training issues." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation kwargs handling after sync_gpus is unrelated to an output_hidden_states bug in Qwen3.5." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "StableLmConfig missing pad_token_idx and meta-device loading regressions are not the same bug." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "These are both import errors, but they involve different missing symbols and different models." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "An MPS attention correctness issue is unrelated to Kimi-K2.5 tokenizer codec/warning regressions." + }, + { + "left": "issue:42490", + "right": "issue:43493", + "accept": false, + "reason": "A local-path import warning and a SigLIP2 HF-vs-JAX discrepancy are different issues." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": false, + "reason": "Unexpected zero probabilities in one SigLIP2 model are not the same as a broader implementation discrepancy report." + }, + { + "left": "issue:39692", + "right": "issue:46192", + "accept": false, + "reason": "Documentation example errors for SigLIP2 are unrelated to return_tensors being ignored with text_kwargs." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and GraniteSpeech CI device-override failure are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "Both touch tied weights, but one is serialization inconsistency and the other is LoRA merge collapse with extended vocab." + }, + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Embedding dtype drift during eval and Qwen3 MoE memory inefficiency are different failures." + }, + { + "left": "issue:45381", + "right": "issue:46209", + "accept": false, + "reason": "Qwen2.5-VL video position ids and FP8 inference failure on B200 are unrelated." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "All-zero assistant masks for multimodal inputs and tool-call tokenize crashes are different apply_chat_template bugs." + }, + { + "left": "issue:28218", + "right": "issue:35141", + "accept": false, + "reason": "Tokenizer extra spacing after added tokens is unrelated to embedding reinitialization after resizing." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "A FineGrainedFP8Config failure in Qwen3Moe is unrelated to the Qwen2_5_VL import error." + } + ] + }, + "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": 23, + "review_unit_index": 28, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:29942", + "issue:32090", + "issue:37428", + "issue:38175", + "issue:41720", + "issue:43262", + "issue:43334", + "issue:43335", + "issue:43388", + "issue:43425", + "issue:43450", + "issue:43452", + "issue:43525", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43819", + "issue:43824", + "issue:43827", + "issue:43901", + "issue:43950", + "issue:43975", + "issue:44112", + "issue:44206", + "issue:44442", + "issue:44479", + "issue:44561", + "issue:44610", + "issue:44792", + "issue:44857", + "issue:44898", + "issue:45083", + "issue:45198", + "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:46512" + ], + "soft_pairs": [ + "issue:43452|issue:46512", + "issue:43950|issue:46285", + "issue:43824|issue:43975", + "issue:43334|issue:43525", + "issue:43262|issue:43450", + "issue:44442|issue:45198", + "issue:44792|issue:45563", + "issue:44479|issue:46192", + "issue:43819|issue:44112", + "issue:44898|issue:45412", + "issue:43827|issue:43901", + "issue:45200|issue:45325", + "issue:44610|issue:44857", + "issue:43388|issue:43526", + "issue:43335|issue:43550", + "issue:44561|issue:46234", + "issue:44857|issue:45593", + "issue:41720|issue:46209", + "issue:43819|issue:45083", + "issue:44206|issue:45584", + "issue:45759|issue:46285", + "issue:43606|issue:45083", + "issue:45520|issue:46170", + "issue:38175|issue:45412", + "issue:37428|issue:43425", + "issue:43335|issue:43653", + "issue:43550|issue:44112", + "issue:29942|issue:32090" + ], + "prepared_review_unit_hash": "7f2a7a73079cde2b004ffa4a4c23a9340e7de3d5dcd47a56ca67862a58f085ba", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36431, + "estimated_input_tokens": 9108, + "estimated_eval_tokens": 18472 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated issues with only superficial title overlap (e.g. missing config attrs, inference regressions, docs/API drift). None of the soft pairs look like the same underlying bug or change, so this cluster should not be collapsed into duplicates.", + "confidence": 0.81, + "canonical_issue_reason": "issue:43425 is the broadest, most discussion-heavy compatibility/regression report in the set, so it is the best umbrella representative even though the cluster is heterogeneous.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43425 is the most central representative for triage: broad impact, high discussion, and a general runtime compatibility framing rather than a narrow one-off symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43452", + "right": "issue:46512", + "accept": false, + "reason": "Different failures: tokenizer/model loading for gguf vs stale dtype on from_config; no shared code path." + }, + { + "left": "issue:43950", + "right": "issue:46285", + "accept": false, + "reason": "One is a from_pretrained non-persistent buffer regression; the other is a CLIP class structure inconsistency. Different bugs." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "ImportError for a missing model class vs incorrect detokenization for a specific model; unrelated symptoms and fixes." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": false, + "reason": "Both mention missing pad_token_id, but they affect different model families/config classes and are not the same bug." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sampling-rate default vs batched video processor shape bug; different subsystems and root causes." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Tokenizer loading failure for one model vs save_pretrained/tokenization failure for another; not the same concrete issue." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "A failing model_generate_images test is unrelated to a stale paged generate warning." + }, + { + "left": "issue:44479", + "right": "issue:46192", + "accept": false, + "reason": "Video-input regression in specific Qwen VL variants vs return_tensors being ignored when text_kwargs is passed; different behaviors." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "DAC forward/from_latents mismatch is a model computation bug; GraniteSpeech stale device override is a CI test issue." + }, + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Perceiver resolution handling vs RT-DETR memory release on deletion; no shared underlying defect." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs issues, but they cover different pipeline APIs and deprecations, so not duplicates." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "MM token type IDs defaulting to zeros is unrelated to Qwen2.5-VL rope index scaling." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Processor input-size mismatch vs AMP/float16 loss crash; different code paths." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Both concern label handling, but one drops tuple elements in gather_for_metrics while the other reduces labels incorrectly in an image processor." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "SwitchTransformers sparse-layer creation and Bamba torch.compile/SDPA failure are unrelated model/config issues." + }, + { + "left": "issue:44561", + "right": "issue:46234", + "accept": false, + "reason": "Two separate missing utility-symbol regressions affecting different optional dependencies and callers." + }, + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "One is a CUDA AMP crash in a loss; the other is auxiliary losses not used when denoising is off. Distinct bugs." + }, + { + "left": "issue:41720", + "right": "issue:46209", + "accept": false, + "reason": "Both are device/hardware failures, but one is an A800 cudaErrorAssert in Qwen3 auto-mapping and the other is an FP8 inference issue on B200." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "Different helper/functionality: DAC latent conversion mismatch vs qwen3_omni_moe feature-length computation." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Unsupported center arg in a feature extractor vs empty-transcription generation failure after token alignment; unrelated." + }, + { + "left": "issue:45759", + "right": "issue:46285", + "accept": false, + "reason": "Composite Qwen config unwrapping bug vs CLIP text model structural inconsistency; different model families and fixes." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "CPU offload device mismatch in bark-small vs feature-length helper behavior in qwen3_omni_moe; no overlap." + }, + { + "left": "issue:45520", + "right": "issue:46170", + "accept": false, + "reason": "Flash-attn import lookup on Python 3.13 vs Windows ROCm/torch.distributed dependency failure; different environment breakages." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP vs RT-DETR memory not released; unrelated model behaviors." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Missing flash support import vs a broad Torch 2.10 compatibility report; not the same underlying regression." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "Sparse layer creation in SwitchTransformers vs BigBird tokenizer special-token registration; different components." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "Bamba SDPA/torch.compile failure vs stale device override test in GraniteSpeech; different issues." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention 2 test failures are unrelated to a Trainer _gpu_broadcast_one NoneType error." + } + ] + }, + "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": 23, + "review_unit_index": 29, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:26413", + "issue:28218", + "issue:28282", + "issue:29127", + "issue:31501", + "issue:32090", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:39692", + "issue:42490", + "issue:43278", + "issue:43335", + "issue:43504", + "issue:43526", + "issue:43606", + "issue:43611", + "issue:43720", + "issue:43819", + "issue:43825", + "issue:43854", + "issue:43931", + "issue:44112", + "issue:44351", + "issue:44448", + "issue:44493", + "issue:44514", + "issue:44521", + "issue:44743", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:45083", + "issue:45127", + "issue:45200", + "issue:45245", + "issue:45290", + "issue:45310", + "issue:45478", + "issue:45588", + "issue:45593", + "issue:45759", + "issue:45859", + "issue:46326", + "issue:46431", + "issue:46512", + "issue:46566" + ], + "soft_pairs": [ + "issue:28282|issue:33357", + "issue:43611|issue:43931", + "issue:44448|issue:44743", + "issue:45759|issue:46512", + "issue:43335|issue:43504", + "issue:44855|issue:45083", + "issue:43278|issue:43825", + "issue:43720|issue:44112", + "issue:29127|issue:32090", + "issue:43335|issue:45083", + "issue:43335|issue:44112", + "issue:43335|issue:45593", + "issue:37428|issue:44351", + "issue:44514|issue:44521", + "issue:45588|issue:45859", + "issue:34689|issue:35141", + "issue:46431|issue:46566", + "issue:45290|issue:46326", + "issue:45127|issue:45245", + "issue:42490|issue:43854", + "issue:45200|issue:45588", + "issue:43606|issue:44857", + "issue:43819|issue:44493", + "issue:45310|issue:45478", + "issue:44448|issue:44779", + "issue:39692|issue:43526", + "issue:26413|issue:31501", + "issue:28218|issue:32090" + ], + "prepared_review_unit_hash": "8ef91cbe64e8e58d439a0648c5459e865a5b6a7a756e315444c83e89b68ae7b1", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36455, + "estimated_input_tokens": 9114, + "estimated_eval_tokens": 18484 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a grab-bag of unrelated Transformers bug reports. The only clear duplicate pair is the Qwen3.5 MoE `from_pretrained` failure (issues 45310 and 45478); the rest are broad subsystem similarities, not the same underlying bug.", + "confidence": 0.94, + "canonical_issue_reason": "issue:45310 is the earlier, more detailed report, and issue:45478 is an obvious duplicate with the same title/root failure on transformers>=5.4.0.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is the best canonical issue because it cleanly represents the one real duplicate pair and has the fuller original report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "Both are import/runtime failures, but one is missing PyTorch for AutoModel and the other is a MacOS CLIP bus error; different root causes." + }, + { + "left": "issue:43611", + "right": "issue:43931", + "accept": false, + "reason": "Both concern model loading, but one is about `base_model_prefix` handling and the other is a weight shape mismatch for a different model." + }, + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Pegasus output regression vs Qwen3 recurrent-state/cache bug; these are different code paths and failures." + }, + { + "left": "issue:45759", + "right": "issue:46512", + "accept": false, + "reason": "One is about unwrapping `text_config` in composite configs; the other is about stale `model.config.dtype` after `from_config(dtype=...)`." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "SwitchTransformers sparse-layer construction and BEiT legacy preset loading are unrelated model-specific bugs." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Python 3.13 `@torch.jit.script` parsing/import issue vs a helper length-calculation bug in `qwen3_omni_moe`; not the same problem." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Training/eval dtype drift and a pipeline translation error message are unrelated issues." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "BitNet packed-weight unpacking and GraniteSpeech stale device-override test failures are different models and different mechanisms." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "LayoutLMv3 box validation and Trainer `_gpu_broadcast_one` NoneType errors do not share the same underlying bug." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "SwitchTransformers config bug vs qwen3_omni_moe feature-length helper bug; unrelated." + }, + { + "left": "issue:43335", + "right": "issue:44112", + "accept": false, + "reason": "SwitchTransformers sparse-layer issue vs GraniteSpeech CI device-override issue; unrelated." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "SwitchTransformers config bug vs D-FINE auxiliary-loss issue; different subsystem and failure mode." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they concern different missing symbols (`_flash_supports_window_size` vs `HybridCache`) and different breakages." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Batched `apply_chat_template(..., padding=False)` crash and zero assistant masks for multimodal inputs are distinct chat-template bugs." + }, + { + "left": "issue:45588", + "right": "issue:45859", + "accept": false, + "reason": "Flash-attention `s_aux=None` crash and Qwen3.5 MoE rotary embedding CPU-offload incompatibility are separate code paths." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Llama 3.2 Vision model-loading regression vs output embedding reinitialization after resizing embeddings; different defects." + }, + { + "left": "issue:46431", + "right": "issue:46566", + "accept": false, + "reason": "Missing chat template error vs DiffusionGemma staying on `meta` with `device_map='auto'`; unrelated loading failures." + }, + { + "left": "issue:45290", + "right": "issue:46326", + "accept": false, + "reason": "Tool-call assistant message crash and `continue_final_message` behavior for mistral_common tokenizers are related only at a high level, not the same bug." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA merge with extended vocab causing collapse and a `2^24` category limit runtime error are not the same issue." + }, + { + "left": "issue:42490", + "right": "issue:43854", + "accept": false, + "reason": "Local-path trailing-slash import warning and GLM-4.7-Flash unit-test loading failure are different model-loading problems." + }, + { + "left": "issue:45200", + "right": "issue:45588", + "accept": false, + "reason": "Gemma 4 `mm_token_type_ids` defaulting and flash_attention `s_aux=None` crashes are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "Bark CPU-offload device mismatch and LwDetrImageLoss float16 AMP crash are different failures in different models." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "DAC `from_latents` mismatch with missing STE and unexpected position-id keys are not the same underlying bug." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "These are duplicate reports of the same Qwen3.5 MoE `from_pretrained` error on transformers>=5.4.0, with matching symptom and root cause." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Pegasus output regression and Deepseek tokenizer correctness regression are both v4/v5 comparison issues, but not the same bug." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "SigLIP2 documentation example failures and BEiT image-processor `reduce_labels` behavior are unrelated." + }, + { + "left": "issue:26413", + "right": "issue:31501", + "accept": false, + "reason": "Checkpoint resume epoch-iterator failure and FineWeb SLM training not starting are different training issues." + }, + { + "left": "issue:28218", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer added-space behavior and Trainer `_gpu_broadcast_one` NoneType error are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall. The packet clearly supports 45310/45478 as the only exact duplicate pair (same title and same transformers>=5.4.0 Qwen3.5 MoE from_pretrained failure), and the other pairwise rejections stay conservative given the distinct titles. The canonical issue choice is reasonable." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 30, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:10000", + "issue:15354", + "issue:27957", + "issue:29870", + "issue:29942", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:40444", + "issue:42175", + "issue:42490", + "issue:43278", + "issue:43329", + "issue:43381", + "issue:43526", + "issue:43550", + "issue:43575", + "issue:43701", + "issue:43927", + "issue:44206", + "issue:44360", + "issue:44448", + "issue:44493", + "issue:44561", + "issue:44617", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45083", + "issue:45229", + "issue:45290", + "issue:45381", + "issue:45459", + "issue:45588", + "issue:45698", + "issue:45823", + "issue:45920", + "issue:46192", + "issue:46234", + "issue:46399", + "issue:46655", + "issue:46697" + ], + "soft_pairs": [ + "issue:44617|issue:46697", + "issue:39692|issue:44493", + "issue:44206|issue:45823", + "issue:45459|issue:46234", + "issue:44857|issue:45072", + "issue:43701|issue:45083", + "issue:32090|issue:33357", + "issue:42490|issue:43575", + "issue:43550|issue:45083", + "issue:44360|issue:44561", + "issue:38175|issue:44493", + "issue:29942|issue:35141", + "issue:27957|issue:34689", + "issue:44448|issue:44977", + "issue:29870|issue:32090", + "issue:43278|issue:43329", + "issue:45920|issue:46192", + "issue:40444|issue:43526", + "issue:10000|issue:15354", + "issue:44977|issue:45229", + "issue:43278|issue:43381", + "issue:45290|issue:45381", + "issue:44855|issue:46399", + "issue:44857|issue:45588", + "issue:29870|issue:33290", + "issue:43927|issue:45698", + "issue:37428|issue:42175", + "issue:45290|issue:46655", + "issue:32090|issue:33290" + ], + "prepared_review_unit_hash": "c93a63be15489855ec7879470e9a8270c77bc9f0b79b64d9197e9980dee36cb1", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36535, + "estimated_input_tokens": 9134, + "estimated_eval_tokens": 18524 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a grab-bag of unrelated Transformer issues spanning different models, features, and failure modes. The soft edges look like superficial text similarity rather than true duplicates, so none should be merged.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue fits the cluster: the items cover unrelated bugs across loading, generation, flash-attention, dtype, save/load, and model-specific processor behavior.", + "canonical_pr_reason": null, + "best_issue_reason": "No issue is a strong global representative because there is no shared underlying bug or change across the cluster; the safest triage is to keep them separate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44617", + "right": "issue:46697", + "accept": false, + "reason": "Both involve OOM/padding-adjacent vision workflows, but one is Sam3Video CUDA memory pressure and the other is Swin2SR over-padding. Different code paths and symptoms." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 doc example errors are about example/model-processor mismatch and quantization, while the other is a position-id key warning. Not the same bug." + }, + { + "left": "issue:44206", + "right": "issue:45823", + "accept": false, + "reason": "One is a LasrFeatureExtractor regression from an unsupported `center` arg; the other is a Gemma4 device-map mismatch during forward. Different subsystems." + }, + { + "left": "issue:45459", + "right": "issue:46234", + "accept": false, + "reason": "Tokenizer/protobuf error handling is unrelated to the missing `is_falcon_mamba_ssm_available` import_utils helper. No shared failure path." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "Float16 AMP crash in LwDetrImageLoss is not the same as dtype mismatch issues in SwitchTransformers/TimmWrapperModel. Different model paths." + }, + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Checkpoint key mismatch and `_get_feat_extract_output_lengths` behavior in qwen3_omni_moe are unrelated bugs." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "Trainer broadcast of `NoneType` and MacOS bus errors on CLIP loading are distinct runtime failures." + }, + { + "left": "issue:42490", + "right": "issue:43575", + "accept": false, + "reason": "Trailing-slash import warning on local paths is unrelated to Qwen2-57B tensor-parallel OOM." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "Bamba-9B-v2 torch.compile/SDPA failure and qwen3_omni_moe feature-length logic are different concrete bugs." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "Missing ReLU in DSA indexer is unrelated to removal of `is_torch_fx_available` breaking trust_remote_code models." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 and unexpected position-id keys are not the same underlying issue." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention 2 test failures do not match the embedding reinitialization bug in `post_init`." + }, + { + "left": "issue:27957", + "right": "issue:34689", + "accept": false, + "reason": "Both mention model loading/attention, but one is XLMRoberta with Flash Attention 2 and the other is Llama 3.2 Vision loading regression. Different models and failures." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Pegasus v4/v5 output differences and Qwen3.5 flash-attention generation problems are unrelated model-specific regressions." + }, + { + "left": "issue:29870", + "right": "issue:32090", + "accept": false, + "reason": "BART `generate()` decoder mask bug and Trainer broadcast `NoneType` error are separate execution paths." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Embedding dtype drift in eval and undefined variables in qwen multimodal token counting are different bugs." + }, + { + "left": "issue:45920", + "right": "issue:46192", + "accept": false, + "reason": "Wrong token IDs across many tokenizers is not the same as `return_tensors` being ignored when `text_kwargs` is passed." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "Qwen2.5-VL multi-image dataset finetuning failure and BeitImageProcessorFast label reduction are unrelated processor/training issues." + }, + { + "left": "issue:10000", + "right": "issue:15354", + "accept": false, + "reason": "German DistilBertModel issue and GeneratorExp unsupported by `torch.jit.script` are unrelated old bugs." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3.5 flash-attention generation problems and Gemma4 multi-GPU CUDA OOM are different runtime issues." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Eval/train dtype mismatch is unrelated to gradient checkpointing being disallowed in eval mode." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Chat template crashes on tool-call messages and qwen2.5-vl video position-id bugs are different code paths." + }, + { + "left": "issue:44855", + "right": "issue:46399", + "accept": false, + "reason": "Python parsing/indentation failure around decorators is unrelated to a missing decorator registration bug for `apply_rotary_pos_emb`." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "LwDetr float16 AMP crash and `flash_attention.py` crashing on `s_aux=None` are distinct failures." + }, + { + "left": "issue:29870", + "right": "issue:33290", + "accept": false, + "reason": "BART generation mask bug and DeepSpeed Adafactor OOM do not share a concrete root cause." + }, + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "DiaConfig losing custom token IDs after save/load and loading the wrong custom module after `save_pretrained` are both serialization-related, but they are different bugs with different failure modes." + }, + { + "left": "issue:37428", + "right": "issue:42175", + "accept": false, + "reason": "Missing flash-attention helper import and packaging missing TensorFlow backend are unrelated dependency/import issues." + }, + { + "left": "issue:45290", + "right": "issue:46655", + "accept": false, + "reason": "Chat template tool-call handling and SmolVLM video padding calculation are separate feature bugs." + }, + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Trainer `NoneType` broadcast and Adafactor/DeepSpeed OOM are different problems despite both occurring in training." + } + ] + }, + "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": 23, + "review_unit_index": 31, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:25251", + "issue:29127", + "issue:29870", + "issue:29942", + "issue:30064", + "issue:32090", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:39692", + "issue:41720", + "issue:42490", + "issue:42915", + "issue:42994", + "issue:43278", + "issue:43381", + "issue:43412", + "issue:43575", + "issue:43611", + "issue:43653", + "issue:43701", + "issue:43756", + "issue:43873", + "issue:43976", + "issue:44186", + "issue:44206", + "issue:44361", + "issue:44442", + "issue:44448", + "issue:44466", + "issue:44479", + "issue:44561", + "issue:44568", + "issue:44743", + "issue:44779", + "issue:44871", + "issue:44977", + "issue:45198", + "issue:45229", + "issue:45372", + "issue:45406", + "issue:45440", + "issue:45491", + "issue:45563", + "issue:45588", + "issue:45803", + "issue:45823", + "issue:46121" + ], + "soft_pairs": [ + "issue:44561|issue:44779", + "issue:43873|issue:45229", + "issue:25251|issue:32090", + "issue:44448|issue:44466", + "issue:42490|issue:43873", + "issue:43976|issue:45440", + "issue:43278|issue:43575", + "issue:45588|issue:45823", + "issue:41720|issue:45229", + "issue:44568|issue:45491", + "issue:43381|issue:43701", + "issue:33357|issue:39692", + "issue:44977|issue:45563", + "issue:29127|issue:34567", + "issue:44871|issue:44977", + "issue:45198|issue:45803", + "issue:43653|issue:44743", + "issue:44361|issue:45198", + "issue:45372|issue:45588", + "issue:44186|issue:45198", + "issue:43611|issue:46121", + "issue:42915|issue:42994", + "issue:29870|issue:34689", + "issue:43412|issue:45588", + "issue:44206|issue:44479", + "issue:45372|issue:45406", + "issue:43653|issue:44442", + "issue:29942|issue:30064", + "issue:43756|issue:45440" + ], + "prepared_review_unit_hash": "7bbff46ca77ff682346e4c3064f526a8bfb893993a43b44c31b714bfb20a3f12", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36539, + "estimated_input_tokens": 9135, + "estimated_eval_tokens": 18526 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is a loose semantic grouping of unrelated issues, mostly spanning model loading, tokenizer/processor bugs, generation regressions, and memory/offloading problems. None of the soft pairs look like true duplicates or the same concrete code-path change.", + "confidence": 0.91, + "canonical_issue_reason": "Issue #43873 is the broadest and most central representative of the loose memory/device/quantization subset, with active discussion and inbound references; the rest are narrower, model-specific regressions or unrelated bugs.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #43873 best represents the cluster because it is a prominent, ongoing infrastructure-level bug around offloading/quantization, whereas most other items are isolated model- or feature-specific failures.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "Different failures: a removed utility breaking trust_remote_code models vs a DeepSeek tokenizer regression." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Both are memory-related, but one is quantized offloading behavior and the other is a Gemma4 multi-GPU OOM; not the same bug." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "Pipeline top_k nesting and Trainer _gpu_broadcast_one TypeError are unrelated code paths." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "One is a generation-output regression, the other is device-dependent serialization of lm_head.weight." + }, + { + "left": "issue:42490", + "right": "issue:43873", + "accept": false, + "reason": "Local-path import warning is unrelated to quantization/offloading behavior." + }, + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Python-version compatibility break and DeepSeekV3MoE divergence are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43575", + "accept": false, + "reason": "Embedding dtype drift in eval is not the same as tp-induced OOM on a large Qwen model." + }, + { + "left": "issue:45588", + "right": "issue:45823", + "accept": false, + "reason": "Flash-attention s_aux=None crash and Gemma4 PLE device mismatch affect different subsystems and failure modes." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Different models and errors: Qwen3 auto device mapping assert vs Gemma4 OOM." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Tokenizer special-token handling is unrelated to Gemma3 NaN embeddings from sliding-window attention." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Eval-mode gradient-checkpointing restriction is not the same as a resume_from_checkpoint key mismatch." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "A MacOS bus error on a community CLIP model and a broken SigLIP2 docs example are different issues." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Qwen3.5 flash-attention generation regression vs a stale warning in paged generate are different." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and TrainerState token counter updates are unrelated." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "Gemma eos_token_id config mismatch is not the same as Qwen3.5 flash-attention generation failure." + }, + { + "left": "issue:45198", + "right": "issue:45803", + "accept": false, + "reason": "Wav2Vec2 save/tokenization failures are unrelated to a bare-except bug in FuyuBatchFeature.convert_to_tensors()." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and Qwen3.5 recurrent-state reset are different bugs." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "Different tokenizers and failure modes: MLuke AttributeError vs Wav2Vec2 save/tokenization issues." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "Gemma4 processor import/dependency failure and flash_attention s_aux=None crash are separate code paths." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "LayoutLMv2 NER/padding crashes and Wav2Vec2 save/tokenization failures are not the same underlying bug." + }, + { + "left": "issue:43611", + "right": "issue:46121", + "accept": false, + "reason": "Model loading with base_model_prefix and rope-validation JSON list handling are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and quantized model saving failure are different quantization problems." + }, + { + "left": "issue:29870", + "right": "issue:34689", + "accept": false, + "reason": "BART generate-time UnboundLocalError and Llama 3.2 model-loading regression are unrelated." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Executorch export failures and flash-attention runtime crashes do not share the same code path." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Feature-extractor center-arg crash and Qwen video-input regression are separate regressions." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Both involve Gemma4, but one is an import-time dependency problem and the other is a serve-time missing _tokenizer attribute." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "BigBird special-token registration and AutoTokenizer loading FastSpeech2ConformerTokenizer are unrelated tokenizer bugs." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "Flash Attention 2 CI failures and void segmentation-map processing are different subsystems." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Smollm3 RoPE-layer mismatch and DeepSeekV3MoE divergence are different model-specific 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": 23, + "review_unit_index": 32, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:27957", + "issue:29127", + "issue:33290", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:42490", + "issue:42915", + "issue:43065", + "issue:43066", + "issue:43334", + "issue:43381", + "issue:43421", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43701", + "issue:43761", + "issue:43824", + "issue:43931", + "issue:43994", + "issue:44186", + "issue:44265", + "issue:44279", + "issue:44387", + "issue:44442", + "issue:44462", + "issue:45072", + "issue:45198", + "issue:45245", + "issue:45250", + "issue:45372", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45563", + "issue:45588", + "issue:45593", + "issue:45600", + "issue:45712", + "issue:45834", + "issue:45859", + "issue:46121", + "issue:46369", + "issue:46399", + "issue:46640", + "issue:46693" + ], + "soft_pairs": [ + "issue:45491|issue:46693", + "issue:45459|issue:45712", + "issue:29127|issue:33357", + "issue:43065|issue:45412", + "issue:45245|issue:45563", + "issue:43065|issue:43994", + "issue:44265|issue:45198", + "issue:43575|issue:43701", + "issue:45372|issue:45459", + "issue:43066|issue:44462", + "issue:45440|issue:45593", + "issue:29127|issue:35141", + "issue:42490|issue:43381", + "issue:43994|issue:45412", + "issue:45412|issue:46369", + "issue:43421|issue:45491", + "issue:43575|issue:44387", + "issue:44442|issue:45072", + "issue:43526|issue:45834", + "issue:42915|issue:43931", + "issue:43526|issue:46693", + "issue:43526|issue:44186", + "issue:27957|issue:29127", + "issue:43824|issue:44279", + "issue:46121|issue:46399", + "issue:27957|issue:33290", + "issue:43994|issue:46369", + "issue:43334|issue:43531", + "issue:33290|issue:34689", + "issue:33290|issue:35141", + "issue:45250|issue:46640", + "issue:45588|issue:45600", + "issue:43761|issue:45859" + ], + "prepared_review_unit_hash": "b8e7cd293991e7608a3b079736a91875f69f76ae0c8a50d063f0e4207c08c5e0", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37449, + "estimated_input_tokens": 9363, + "estimated_eval_tokens": 18982 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: most pairs only share superficial wording or a broad subsystem, not the same concrete bug/change. I would not merge any of the soft pairs as duplicates, and I do not see a safe single canonical issue for the whole set.", + "confidence": 0.78, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45491", + "right": "issue:46693", + "accept": false, + "reason": "Both mention Flash Attention 2, but one is a NaN/attention correctness bug and the other is a long-context performance regression; different failures and fixes." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "Tokenizer error handling with missing protobuf is unrelated to leftover dummy classes and repo-check failures." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "LayoutLMv3 box-validation messaging and a CLIP MacOS bus error are unrelated issues." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "A dummy Conv2d placeholder in Sam3PixelDecoder is not the same bug as RT-DETR memory not being released." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Category-cardinality runtime error and a stale generate() warning are different code paths and symptoms." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "Sam3PixelDecoder dummy layer handling is unrelated to SigLIP2 producing nonsensical outputs." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export exportability with torch_compilable_check is unrelated to Wav2Vec2 save/tokenization failure." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "OOM during tensor-parallel model load and a resume_from_checkpoint key mismatch are different problems." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "Gemma 4 processor import failure from mistral_common is unrelated to tokenizer error suppression when protobuf is absent." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Wrong tokenizer decoder type in v5 and ignoring tokenizer.json are both tokenizer-related, but not the same underlying bug." + }, + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "DeepSeekV3MoE implementation divergence and D-FINE auxiliary-loss omission are unrelated model-specific bugs." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "LayoutLMv3 error clarity and embedding reinitialization after resize are unrelated." + }, + { + "left": "issue:42490", + "right": "issue:43381", + "accept": false, + "reason": "Local-path import warning and gradient checkpointing in eval mode are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 bad outputs and RT-DETR memory leaks do not share the same code-path bug." + }, + { + "left": "issue:45412", + "right": "issue:46369", + "accept": false, + "reason": "RT-DETR GC/memory leak and Sam3 LiteText memory overflow are different memory issues in different models." + }, + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Runtime post-processor updates for special tokens are unrelated to Gemma3 NaNs from sliding-window attention." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both involve OOM, but one is tensor-parallel loading and the other is int4 quantization reserved-memory growth." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "FastSpeech2Conformer tokenizer loading and dtype mismatches in bfloat16 inference are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:45834", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and Kosmos2.5 OCR index errors are different bugs." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "FineGrainedFP8Config failure in Qwen3Moe and Qwen3-VL weight-shape mismatch are unrelated model-loading problems." + }, + { + "left": "issue:43526", + "right": "issue:46693", + "accept": false, + "reason": "BeitImageProcessorFast label handling is unrelated to FA2 long-context performance regression." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "Label reduction in an image processor is not the same as LayoutLMv2Tokenizer crashing on NER padding/truncation." + }, + { + "left": "issue:27957", + "right": "issue:29127", + "accept": false, + "reason": "Flash Attention 2 support for XLMRoberta and LayoutLMv3 error messaging are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Missing Qwen2_5_VL import and a generic dependency issue are not the same concrete problem." + }, + { + "left": "issue:46121", + "right": "issue:46399", + "accept": false, + "reason": "Rope-parameter validation TypeError and a missing decorator for apply_rotary_pos_emb registration are different issues." + }, + { + "left": "issue:27957", + "right": "issue:33290", + "accept": false, + "reason": "FA2 support for XLMRoberta and Adafactor OOM in deepspeed are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:46369", + "accept": false, + "reason": "SigLIP2 bad outputs and Sam3 LiteText memory overflow are different model-specific failures." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Missing pad_token_id for Qwen3-VL and a sliding_window issue for Qwen3-MoE are different config/behavior bugs." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "Adafactor OOM and Llama 3.2 Vision loading breakage are unrelated." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "OOM in deepspeed with Adafactor is unrelated to output embedding reinitialization after resize." + }, + { + "left": "issue:45250", + "right": "issue:46640", + "accept": false, + "reason": "These are both Flash Attention 2 themed, but one is a generic request and the other is an open issue; no evidence they describe the same concrete bug/change." + }, + { + "left": "issue:45588", + "right": "issue:45600", + "accept": false, + "reason": "FA integration crash on s_aux=None and stale auto_mappings references to removed configs are unrelated." + }, + { + "left": "issue:43761", + "right": "issue:45859", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression and Qwen3.5 CPU-offload incompatibility 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": 23, + "review_unit_index": 33, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:27957", + "issue:29870", + "issue:42175", + "issue:42490", + "issue:43262", + "issue:43316", + "issue:43381", + "issue:43493", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43720", + "issue:43723", + "issue:43749", + "issue:43761", + "issue:43827", + "issue:43873", + "issue:43931", + "issue:44062", + "issue:44077", + "issue:44079", + "issue:44188", + "issue:44206", + "issue:44442", + "issue:44479", + "issue:44514", + "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:45381", + "issue:45440", + "issue:45479", + "issue:45759", + "issue:45907", + "issue:46178", + "issue:46234", + "issue:46697" + ], + "soft_pairs": [ + "issue:43381|issue:43873", + "issue:44514|issue:45381", + "issue:43646|issue:44077", + "issue:43262|issue:44206", + "issue:44977|issue:45245", + "issue:44442|issue:44898", + "issue:44743|issue:44811", + "issue:42490|issue:44743", + "issue:43720|issue:44442", + "issue:45003|issue:45440", + "issue:42175|issue:46234", + "issue:27957|issue:29870", + "issue:43316|issue:43525", + "issue:43493|issue:43827", + "issue:43723|issue:45479", + "issue:44077|issue:44479", + "issue:45325|issue:46697", + "issue:43873|issue:44871", + "issue:44062|issue:45907", + "issue:43761|issue:44079", + "issue:43749|issue:43931", + "issue:43643|issue:44913", + "issue:43316|issue:45759", + "issue:44987|issue:46178", + "issue:44855|issue:45137", + "issue:43761|issue:44188", + "issue:43931|issue:45216" + ], + "prepared_review_unit_hash": "78a18e03751e77b5a62e416781fb6387251a0787f7e37e06f6c1d9e4ab7fa87f", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35459, + "estimated_input_tokens": 8865, + "estimated_eval_tokens": 17986 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is not a duplicate set; it contains many distinct bugs/regressions across unrelated areas (generation, audio, configs, loading, quantization, docs, and model-specific logic). The soft pairs are mostly only superficially similar by subsystem or model family.", + "confidence": 0.97, + "canonical_issue_reason": "No single issue is a clear canonical representative; the items span unrelated bug reports rather than one underlying defect.", + "canonical_pr_reason": null, + "best_issue_reason": "If one issue must anchor the cluster, issue 43749 has the strongest activity/signal (discussion and inbound references), but it is still just one distinct report among many unrelated ones.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Both involve training/runtime behavior, but one is gradient-checkpointing in eval mode and the other is quantization offloading; different code paths and fixes." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Same model family, but one is batched chat-template padding behavior and the other is video RoPE/position-id computation." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "Missing fields from trust_remote_code config loading is unrelated to patchtsmixer post-init validation." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Both are audio-related, but one is sampling-rate defaulting and the other is an unsupported feature-extractor argument." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Completely unrelated: flash-attention generation bug versus a categorical-cardinality runtime error." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer loading and image-classification positional interpolation are different subsystems with different failure modes." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Recurrent cache reset in modular Qwen code is unrelated to Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:42490", + "right": "issue:44743", + "accept": false, + "reason": "Path-trailing-slash import warning is unrelated to Qwen recurrent-state cache behavior." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "BitNet packed-weight loading and FastSpeech2Conformer tokenizer loading are different code paths." + }, + { + "left": "issue:45003", + "right": "issue:45440", + "accept": false, + "reason": "Generic sys.modules access safety issue is not the same as DeepSeekV3MoE implementation divergence." + }, + { + "left": "issue:42175", + "right": "issue:46234", + "accept": false, + "reason": "Both touch backend/import utilities, but they are about different missing backends and unrelated symbols." + }, + { + "left": "issue:27957", + "right": "issue:29870", + "accept": false, + "reason": "Flash Attention 2 support for XLM-Roberta is unrelated to BART generate() decoder_attention_mask assignment." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Both are config/API discrepancies, but they concern different models and different missing fields." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "SigLIP2 implementation mismatch is unrelated to documentation still referencing removed pipeline() APIs." + }, + { + "left": "issue:43723", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer-loading regression and degenerate classification loss are unrelated." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "patchtsmixer post_init validation is unrelated to Qwen video-input regression." + }, + { + "left": "issue:45325", + "right": "issue:46697", + "accept": false, + "reason": "Both are preprocessing issues, but one is video temporal-position scaling and the other is image padding behavior." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization/offloading behavior and Gemma-3 eos_token_id configuration are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:45907", + "accept": false, + "reason": "AddedToken constructor argument duplication is unrelated to composite Qwen text_config unwrapping." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "A model failing to return hidden_states is not the same as ModelOutput key assignment when a prior key was None." + }, + { + "left": "issue:43749", + "right": "issue:43931", + "accept": false, + "reason": "FSDP CPU RAM-efficient loading and Qwen3-VL shape mismatch are different loading problems." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "trust_remote_code missing fields is unrelated to GPTNeoX rotary_pct not persisting on reload." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Both involve config handling for multimodal/text configs, but they are distinct API regressions with different fixes." + }, + { + "left": "issue:44987", + "right": "issue:46178", + "accept": false, + "reason": "One is a model-loading regression, the other is a backend-mapping/import error for tensorflow_text." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 torch.jit parsing issue is unrelated to DeepSpeed ZeRO3 deque underflow." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "Hidden-states output regression is unrelated to attention-kernel divergence under torch.compile." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "A loading shape mismatch for Qwen3-VL is not the same as a save_pretrained regression for Qwen3.5." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 34, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:25251", + "issue:27957", + "issue:28282", + "issue:32090", + "issue:42175", + "issue:42617", + "issue:42907", + "issue:43232", + "issue:43295", + "issue:43316", + "issue:43388", + "issue:43404", + "issue:43441", + "issue:43452", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43701", + "issue:43716", + "issue:43742", + "issue:43825", + "issue:44077", + "issue:44162", + "issue:44190", + "issue:44442", + "issue:44464", + "issue:44496", + "issue:44534", + "issue:44568", + "issue:44792", + "issue:44871", + "issue:44913", + "issue:45479", + "issue:45584", + "issue:45632", + "issue:45698", + "issue:45803", + "issue:46050", + "issue:46170", + "issue:46234", + "issue:46291", + "issue:46365", + "issue:46421", + "issue:46424", + "issue:46506" + ], + "soft_pairs": [ + "issue:44190|issue:44496", + "issue:43232|issue:43388", + "issue:44442|issue:45803", + "issue:42907|issue:45698", + "issue:43701|issue:46506", + "issue:42617|issue:43575", + "issue:43575|issue:43742", + "issue:27957|issue:32090", + "issue:43550|issue:43701", + "issue:25251|issue:28282", + "issue:44568|issue:45479", + "issue:43606|issue:43701", + "issue:44464|issue:46421", + "issue:46234|issue:46291", + "issue:43295|issue:43825", + "issue:16998|issue:22355", + "issue:43295|issue:44871", + "issue:44871|issue:45584", + "issue:43316|issue:43716", + "issue:42175|issue:43316", + "issue:44077|issue:44534", + "issue:44871|issue:46050", + "issue:43646|issue:44913", + "issue:43452|issue:44496", + "issue:44162|issue:44792", + "issue:45803|issue:46365", + "issue:45803|issue:46170", + "issue:43404|issue:43441", + "issue:45632|issue:46424" + ], + "prepared_review_unit_hash": "3ce12af171ae7918a501359909574545aab1c40b10882fbea896024ce183b587", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36307, + "estimated_input_tokens": 9077, + "estimated_eval_tokens": 18410 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: most items are unrelated Transformers issues around loading, generation, config, cache, or environment-specific failures. I found no true duplicate pair to merge.", + "confidence": 0.94, + "canonical_issue_reason": "No clear canonical issue: the cluster spans many distinct bugs and regressions across different models and subsystems. If forced to anchor the set, issue 45698 is the broadest active save/load regression, but it is not a good representative duplicate target for the rest.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45698 is the strongest single issue to keep as the representative anchor because it is open, active, and broadly about incorrect custom-module loading after save_pretrained/from_pretrained. It is still not a duplicate of the other items.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Local dataset loading failure and unrecognized-model config error are different code paths and different failure modes." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "One is a generation kwargs/sync_gpus issue; the other is a metric-gathering bug dropping tuple labels in the last batch." + }, + { + "left": "issue:44442", + "right": "issue:45803", + "accept": false, + "reason": "Tokenizer loading failure for FastSpeech2Conformer is unrelated to Fuyu's bare-except error swallowing exceptions in tensor conversion." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "Both involve save/load, but one is about dequantized model saving and the other is wrong custom-module resolution after reload; different underlying bugs." + }, + { + "left": "issue:43701", + "right": "issue:46506", + "accept": false, + "reason": "Checkpoint key mismatch and CPU/CUDA output divergence are unrelated." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "3d_parallel startup failure and tp OOM on Qwen2 are different problems." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "Tensor-parallel OOM is not the same as a key error when loading MobileLLM-125M." + }, + { + "left": "issue:27957", + "right": "issue:32090", + "accept": false, + "reason": "FlashAttention support for XLMRoberta and a Trainer broadcast TypeError are separate issues." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "torch.compile/SDPA failure for Bamba is unrelated to resume_from_checkpoint key mismatch." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline top_k output nesting bug and missing PyTorch ImportError are unrelated." + }, + { + "left": "issue:44568", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer special-token insertion regression and degenerate sequence-classification loss are different subsystems." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU offload device mismatch and checkpoint key mismatch are not the same bug." + }, + { + "left": "issue:44464", + "right": "issue:46421", + "accept": false, + "reason": "Both mention chunked generation/prefill, but one reports wrong outputs while the other reports spurious recompilation from cache lazy init." + }, + { + "left": "issue:46234", + "right": "issue:46291", + "accept": false, + "reason": "Missing import_utils helper and kernels version/revision import failure are distinct package/API issues." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Custom model/processor regression is unrelated to the pipeline translation error message change." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "A model_max_length question for DeBERTa-V3 and a missing transformers.onnx module are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Processor/tokenizer regression and Gemma3 eos_token_id inconsistency are different model/API problems." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Gemma3 eos_token_id configuration and Whisper empty-transcription handling after align_special_tokens are not the same bug." + }, + { + "left": "issue:43316", + "right": "issue:43716", + "accept": false, + "reason": "Gemma3TextConfig API mismatch and Mistral-3 image-preprocessor dtype mismatch are separate issues." + }, + { + "left": "issue:42175", + "right": "issue:43316", + "accept": false, + "reason": "Backend installation packaging issue and Gemma3 config API discrepancy are unrelated." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "Optional post_init validation and non-persistent buffer junk are different code paths." + }, + { + "left": "issue:44871", + "right": "issue:46050", + "accept": false, + "reason": "Gemma3 eos_token_id mismatch and quantized-model save/load breakage are not the same defect." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "Custom model initialization breakage in v5 and GPTNeoX rotary_pct persistence/reload bug are distinct." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "gguf_file/from_pretrained breakage and unrecognized model config errors are related to loading, but they are different failure mechanisms." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 runtime breakage and a janus test failure are not the same issue." + }, + { + "left": "issue:45803", + "right": "issue:46365", + "accept": false, + "reason": "Bare except in Fuyu tensor conversion and Trainer import failure caused by datasets are unrelated." + }, + { + "left": "issue:45803", + "right": "issue:46170", + "accept": false, + "reason": "Fuyu exception handling bug and Windows ROCm torch.distributed import dependency issue are different." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Untied lm_head weights in Mistral3ForConditionalGeneration and FlashAttention failure for Ministral-3 are separate bugs." + }, + { + "left": "issue:45632", + "right": "issue:46424", + "accept": false, + "reason": "Cache path collision for trust_remote_code and max_cache_length being honored only for Quantized Cache are unrelated cache 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": 23, + "review_unit_index": 35, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "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:43650", + "issue:43673", + "issue:43688", + "issue:43761", + "issue:43827", + "issue:43994", + "issue:44190", + "issue:44479", + "issue:44514", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44734", + "issue:44743", + "issue:44792", + "issue:44855", + "issue:44913", + "issue:44964", + "issue:44977", + "issue:45127", + "issue:45200", + "issue:45406", + "issue:45542", + "issue:45588", + "issue:45600", + "issue:45632", + "issue:45698", + "issue:46082", + "issue:46178", + "issue:46202", + "issue:46234", + "issue:46369", + "issue:46421", + "issue:46617", + "issue:46640", + "issue:46697" + ], + "soft_pairs": [ + "issue:44610|issue:46369", + "issue:43650|issue:46202", + "issue:43761|issue:44479", + "issue:43572|issue:46178", + "issue:43540|issue:44514", + "issue:45542|issue:46234", + "issue:43065|issue:44617", + "issue:45127|issue:45200", + "issue:43504|issue:45698", + "issue:44792|issue:44977", + "issue:43388|issue:43688", + "issue:44792|issue:45600", + "issue:43352|issue:46640", + "issue:43673|issue:44743", + "issue:41950|issue:43475", + "issue:43525|issue:43531", + "issue:44190|issue:44964", + "issue:44534|issue:44913", + "issue:35141|issue:42490", + "issue:44734|issue:45406", + "issue:46082|issue:46697", + "issue:45632|issue:46617", + "issue:44855|issue:45588", + "issue:43827|issue:43994", + "issue:43673|issue:46421" + ], + "prepared_review_unit_hash": "5490088610f13c3752f3323dd70a516c6e1f5ce143ff60b0f4eaa56249786635", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 25, + "serialized_chars": 35227, + "estimated_input_tokens": 8807, + "estimated_eval_tokens": 17870 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is not a true duplicate set; it contains many unrelated issues across different models, processors, configs, and cache paths. I rejected all soft edges because the paired items only share broad subsystem vocabulary, not the same concrete bug.", + "confidence": 0.95, + "canonical_issue_reason": "No clear canonical issue: the items are heterogeneous and the similarity links look like keyword-based false positives rather than one underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "No single best issue to represent the cluster; picking one would be arbitrary because the linked issues do not share a common code-path failure.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44610", + "right": "issue:46369", + "accept": false, + "reason": "Both involve SAM-related models, but one is an input-size/processor mismatch and the other is a memory overflow; different failure modes." + }, + { + "left": "issue:43650", + "right": "issue:46202", + "accept": false, + "reason": "Unrelated: one is a vague data request, the other is a Gemma pipeline roundtrip/tokenization issue." + }, + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "Different models and different bugs: CLIP hidden states regression vs Qwen video-input regression." + }, + { + "left": "issue:43572", + "right": "issue:46178", + "accept": false, + "reason": "Config attribute mismatch vs backend registration error; no shared code-path." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Qwen video-processing ValueError and chat-template batching crash are separate bugs." + }, + { + "left": "issue:45542", + "right": "issue:46234", + "accept": false, + "reason": "TensorBoard/TF backend detection issue is unrelated to missing Falcon Mamba availability helper." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "SAM3 dummy Conv2d implementation and SAM3 video OOM are not the same bug." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "LoRA merge with extended vocab and Gemma4 mm_token_type_ids defaulting are different model behaviors." + }, + { + "left": "issue:43504", + "right": "issue:45698", + "accept": false, + "reason": "Legacy field loading failure vs wrong custom module resolution after save/load are distinct persistence bugs." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Janus image-generation test failure and Qwen3.5 flash-attention generation issue do not match." + }, + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "Metric gathering drops tuple labels; the other is auxiliary-loss normalization in MoE models." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "Janus generation test and removed Sam3LiteText auto-mapping references are unrelated." + }, + { + "left": "issue:43352", + "right": "issue:46640", + "accept": false, + "reason": "Both mention Flash Attention, but one is an explicit unsupported-model guard and the other is an unspecified open issue." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Chunked-prefill cache initialization and recurrent-state reset with cache are different implementation bugs." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Video-classification pipeline processor selection and missing fpn_position_embeddings on SAM3 output are unrelated." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Llama4Config pad_token_id and Qwen3-MoE sliding_window handling are separate config issues." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Local dataset loading in a training script and loading Phi-4 multimodal are different model/input loading failures." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer initialization bug vs GPTNeoX rotary_pct reload defaulting are not the same issue." + }, + { + "left": "issue:35141", + "right": "issue:42490", + "accept": false, + "reason": "Token embedding resize reinitialization and trailing-slash local model import warning have no concrete overlap." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "KV-cache continuation indexing crash in serve and Gemma4Processor missing _tokenizer are separate failures." + }, + { + "left": "issue:46082", + "right": "issue:46697", + "accept": false, + "reason": "LlamaConfig head_dim validation and Swin2SR over-padding are unrelated bugs." + }, + { + "left": "issue:45632", + "right": "issue:46617", + "accept": false, + "reason": "Both touch trust_remote_code caching, but one is a leaf-directory collision and the other is a FileNotFoundError; not enough to merge." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Python 3.13 decorator parsing error in DebertaV2Model is unrelated to sink-less flash-attention s_aux handling." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs referencing removed pipeline() APIs and SigLIP2 nonsensical outputs are different problems." + }, + { + "left": "issue:43673", + "right": "issue:46421", + "accept": false, + "reason": "Both mention cache/chunked prefill, but one is missing GenerationMixin cache and the other is lazy cache init causing recompilation." + } + ] + }, + "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": 23, + "review_unit_index": 36, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:25251", + "issue:27957", + "issue:29870", + "issue:35141", + "issue:39692", + "issue:41720", + "issue:43065", + "issue:43295", + "issue:43335", + "issue:43526", + "issue:43540", + "issue:43550", + "issue:43644", + "issue:43761", + "issue:43931", + "issue:44060", + "issue:44077", + "issue:44112", + "issue:44190", + "issue:44242", + "issue:44464", + "issue:44479", + "issue:44610", + "issue:44617", + "issue:44734", + "issue:44737", + "issue:44779", + "issue:44855", + "issue:44877", + "issue:44913", + "issue:44933", + "issue:45127", + "issue:45245", + "issue:45276", + "issue:45290", + "issue:45357", + "issue:45362", + "issue:45381", + "issue:45563", + "issue:45584", + "issue:45600", + "issue:45803", + "issue:46369", + "issue:46489", + "issue:46655", + "issue:46697", + "issue:46752" + ], + "soft_pairs": [ + "issue:43931|issue:44060", + "issue:43761|issue:44242", + "issue:29870|issue:35141", + "issue:44617|issue:45584", + "issue:45563|issue:45584", + "issue:43644|issue:44913", + "issue:25251|issue:27957", + "issue:44933|issue:45362", + "issue:43335|issue:44855", + "issue:44479|issue:44734", + "issue:44077|issue:44877", + "issue:43550|issue:44190", + "issue:45381|issue:46369", + "issue:41720|issue:45245", + "issue:43335|issue:44190", + "issue:43065|issue:44610", + "issue:44464|issue:45563", + "issue:39692|issue:44190", + "issue:45127|issue:45357", + "issue:43526|issue:46697", + "issue:44112|issue:45600", + "issue:45381|issue:46655", + "issue:44617|issue:46369", + "issue:44779|issue:46489", + "issue:43295|issue:45276", + "issue:45290|issue:46752", + "issue:43540|issue:44734", + "issue:44737|issue:45803", + "issue:45276|issue:45584" + ], + "prepared_review_unit_hash": "5e53cd958f6483b242cad829496071c2a07eecf0c01a063ffa9cc20eedbef265", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35961, + "estimated_input_tokens": 8991, + "estimated_eval_tokens": 18238 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most soft pairs are not true duplicates. The only strong matches are the apply_chat_template empty-input crash pair and the DeepSeek tokenizer regression pair.", + "confidence": 0.89, + "canonical_issue_reason": "issue:46752 is the broadest and clearest representative of the apply_chat_template empty-input crash, and it is still open; its counterpart is a narrower variant.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:46752 is the best issue to anchor this cluster because it captures the general failure mode with a simple repro and likely subsumes the narrower tool-call/no-content report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Different bugs: model load weight-shape mismatch vs an incorrect tied-weights warning." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "Different models and failures: missing hidden_states vs missing load-balancing loss." + }, + { + "left": "issue:29870", + "right": "issue:35141", + "accept": false, + "reason": "Both touch generation/embeddings, but the concrete bugs differ: decoder attention mask init error vs embedding reinitialization on resize." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "Unrelated symptoms and code paths: Sam3Video OOM vs Whisper empty transcription crash." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "Stale warning in paged generate is not the same bug as Whisper failing after align_special_tokens." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Buffer initialization issue vs config reload reverting rotary_pct; related area, different underlying defects." + }, + { + "left": "issue:25251", + "right": "issue:27957", + "accept": false, + "reason": "Different features entirely: pipeline top_k nesting vs Flash Attention 2 support." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "Import error in image_utils is unrelated to Qwen3.5 chat crashes." + }, + { + "left": "issue:43335", + "right": "issue:44855", + "accept": false, + "reason": "Configuration/model-construction bug vs Python 3.13 parsing error in a scripted function." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Both are regressions, but one is about video input handling and the other is KV-cache continuation indexing in serve." + }, + { + "left": "issue:44077", + "right": "issue:44877", + "accept": false, + "reason": "Different config-policy problems with no shared concrete code path." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "torch.compile/SDPA failure in Bamba is unrelated to local dataset loading in the training example." + }, + { + "left": "issue:45381", + "right": "issue:46369", + "accept": false, + "reason": "Both mention model/resource problems, but one is a Qwen video-position regression and the other is Sam3-LiteText memory overflow." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "GPU device-mapping assert and category-count limit are unrelated." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "Single-layer sparse config bug is unrelated to dataset loading." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "Sam3PixelDecoder dummy layer issue is not the same as OmDet-Turbo processor size mismatch." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency is a different failure from a stale warning." + }, + { + "left": "issue:39692", + "right": "issue:44190", + "accept": false, + "reason": "SigLIP2 doc-example errors are unrelated to local dataset loading." + }, + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "LoRA/extended-vocab collapse is a different bug from incorrect visual encoder keys on save_pretrained." + }, + { + "left": "issue:43526", + "right": "issue:46697", + "accept": false, + "reason": "Label reduction bug and image over-padding bug are different preprocessing defects." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "CI test staleness and auto_mappings referencing removed configs are different issues." + }, + { + "left": "issue:45381", + "right": "issue:46655", + "accept": false, + "reason": "Both are video-related, but they affect different processors and failure mechanisms." + }, + { + "left": "issue:44617", + "right": "issue:46369", + "accept": false, + "reason": "Both are Sam3 memory problems, but the affected models and symptoms differ." + }, + { + "left": "issue:44779", + "right": "issue:46489", + "accept": true, + "reason": "Same underlying DeepSeek tokenizer regression in v5; 46489 is a more specific instance of the broader issue reported in 44779." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Regression around processor.tokenizer access is unrelated to gemma4 resize_token_embeddings behavior." + }, + { + "left": "issue:45290", + "right": "issue:46752", + "accept": true, + "reason": "Both describe apply_chat_template crashing on empty/missing assistant content; 46752 is the broader empty-list form of the same bug." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Qwen3OmniMoe video-input validation error is unrelated to serve-side KV-cache continuation indexing." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "XLNet CPU arange issue and bare-except swallowing errors in FuyuBatchFeature are unrelated." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Gemma4 resize-token-embeddings/output-embeddings bug is unrelated to Whisper empty-transcription handling." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet, and the only accepted soft pairs are conservative matches supported by the issue titles. The rejected pairs also read as appropriately cautious rather than overstated. " + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 37, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "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:43479", + "issue:43531", + "issue:43643", + "issue:43650", + "issue:43673", + "issue:43854", + "issue:43976", + "issue:44016", + "issue:44112", + "issue:44186", + "issue:44220", + "issue:44246", + "issue:44279", + "issue:44295", + "issue:44462", + "issue:44496", + "issue:44521", + "issue:44734", + "issue:44938", + "issue:44964", + "issue:44987", + "issue:45003", + "issue:45216", + "issue:45245", + "issue:45375", + "issue:45381", + "issue:45542", + "issue:45584", + "issue:45600", + "issue:45632", + "issue:45907", + "issue:46082", + "issue:46121", + "issue:46234", + "issue:46424", + "issue:46506" + ], + "soft_pairs": [ + "issue:44938|issue:45542", + "issue:44521|issue:45381", + "issue:44462|issue:45907", + "issue:25251|issue:29870", + "issue:45632|issue:46234", + "issue:44246|issue:45003", + "issue:43854|issue:44964", + "issue:44964|issue:45600", + "issue:43643|issue:44496", + "issue:44112|issue:44964", + "issue:27957|issue:35141", + "issue:46082|issue:46121", + "issue:42175|issue:44279", + "issue:45245|issue:45584", + "issue:42796|issue:44521", + "issue:44186|issue:44734", + "issue:36246|issue:41628", + "issue:43404|issue:45216", + "issue:43650|issue:44220", + "issue:32090|issue:35141", + "issue:36246|issue:38617", + "issue:43479|issue:45375", + "issue:41950|issue:44016", + "issue:44295|issue:46506", + "issue:43531|issue:45600", + "issue:43673|issue:46424", + "issue:44016|issue:44220", + "issue:43976|issue:44987" + ], + "prepared_review_unit_hash": "7c70a8dff1d4c029790eff49fe995399b6b19734ef4bf96b2c6ff8590b8217f6", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36112, + "estimated_input_tokens": 9028, + "estimated_eval_tokens": 18312 + }, + "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. Several pairs share only broad subsystem similarity (imports, loading, multimodal, cache, tokenizer) but not the same concrete defect, so almost all soft edges should be rejected.", + "confidence": 0.42, + "canonical_issue_reason": "If a single issue must represent the cluster, issue 44964 is the most concrete and broadly user-facing model-loading regression, and it sits nearest to other load/config-related reports. Still, the cluster is not a true duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "44964 is the best overall issue candidate because it is a clear failure mode with a specific reproduction and broad impact, making it the most representative anchor for triage among these heterogeneous reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Different failures: Python 3.14 import/load problem vs missing TensorFlow backend when only tensorboard is installed." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "Both are multimodal-related, but one is about all-zero assistant masks and the other about wrong vision position ids; not the same bug." + }, + { + "left": "issue:44462", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer.json handling and list_repo_files revision forwarding are unrelated code paths." + }, + { + "left": "issue:25251", + "right": "issue:29870", + "accept": false, + "reason": "Pipeline output nesting and BART decoder_attention_mask assignment are different issues." + }, + { + "left": "issue:45632", + "right": "issue:46234", + "accept": false, + "reason": "Cache-path collision for trust_remote_code is unrelated to a missing import_utils symbol." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Both touch import behavior, but one is slow import performance and the other is unsafe sys.modules access; not the same defect." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Different model-loading failures for different models and root causes." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "One is a model load regression; the other is stale auto_mappings references breaking CI." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "Both involve config/model loading, but one loses fields under trust_remote_code while the other lacks model_type for an unrecognized model." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "GraniteSpeech CI device-override failure is unrelated to Phi-4 multimodal loading." + }, + { + "left": "issue:27957", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention support and embedding reinitialization after resize are different code paths." + }, + { + "left": "issue:46082", + "right": "issue:46121", + "accept": false, + "reason": "Both are validation edge cases, but they concern different configs and different failure modes." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "Generic dependency/backend issues, but not the same package or symptom." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "Category-count overflow and empty-transcription generation are unrelated." + }, + { + "left": "issue:42796", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer BPE marker handling is unrelated to multimodal assistant mask generation." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "LayoutLMv2 tokenizer crashes and serve KV-cache continuation indexing are different bugs." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but for different symbols and modules." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Weight tying in Mistral3 and save_pretrained checkpoint correctness in Qwen3.5 are distinct regressions." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "A placeholder/data request is unrelated to _torch_extract_fbank_features." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer broadcast NoneType and embedding reinitialization are unrelated failures." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Both are missing-import reports, but they concern different missing names and code paths." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Defaulting None configs and missing deepstack_visual_indexes under strict serialization are different issues." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "Video-classification pipeline image-processor lookup and a notebook syntax error are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:46506", + "accept": false, + "reason": "Reading position_ids after buffer registration is unrelated to CPU/CUDA output divergence." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Sliding_window behavior in Qwen3-MoE and stale auto_mappings references are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:46424", + "accept": false, + "reason": "Both mention cache, but one is missing GenerationMixin cache in chunked_prefill while the other is about max_cache_length being honored only for quantized cache." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error and fbank feature extraction are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "Python version compatibility and a specific model-loading failure are not the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 38, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:39692", + "issue:41950", + "issue:42907", + "issue:43295", + "issue:43299", + "issue:43381", + "issue:43441", + "issue:43526", + "issue:43577", + "issue:43673", + "issue:43824", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44075", + "issue:44279", + "issue:44327", + "issue:44336", + "issue:44368", + "issue:44496", + "issue:44521", + "issue:44683", + "issue:44779", + "issue:44792", + "issue:44933", + "issue:45020", + "issue:45042", + "issue:45083", + "issue:45216", + "issue:45335", + "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:46512", + "issue:46515", + "issue:46693", + "issue:46752" + ], + "soft_pairs": [ + "issue:43295|issue:45335", + "issue:44683|issue:45446", + "issue:43526|issue:43901", + "issue:45083|issue:46506", + "issue:43381|issue:45593", + "issue:43901|issue:44496", + "issue:44496|issue:45042", + "issue:42907|issue:43441", + "issue:43673|issue:44336", + "issue:43976|issue:44279", + "issue:44496|issue:45759", + "issue:43441|issue:45588", + "issue:44016|issue:44062", + "issue:44792|issue:46424", + "issue:43824|issue:44933", + "issue:44327|issue:46752", + "issue:46123|issue:46369", + "issue:43299|issue:45216", + "issue:44521|issue:46693", + "issue:44279|issue:46515", + "issue:45335|issue:45584", + "issue:46082|issue:46399", + "issue:44279|issue:45405", + "issue:43295|issue:44779", + "issue:39692|issue:43901", + "issue:43577|issue:45216", + "issue:44075|issue:44368", + "issue:43976|issue:45405", + "issue:46082|issue:46512", + "issue:41950|issue:45020" + ], + "prepared_review_unit_hash": "ede50ccf5a465220a969c7551979fc61eee9ce8e9a2ebb7990c6f31f028c31f9", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36154, + "estimated_input_tokens": 9039, + "estimated_eval_tokens": 18334 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is heterogeneous: the soft pairs mostly share only broad subsystem or version-regression wording, not the same concrete bug. I would not merge any of the candidate pairs; the closest thing to an umbrella issue is the broad remote-code/version-regression report, but it still does not subsume the rest.", + "confidence": 0.44, + "canonical_issue_reason": "issue:45020 is the broadest umbrella-style regression report in the set (recent versions breaking `remote_code` models) and is the closest available cluster anchor, though it does not actually duplicate most others.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the best global representative because it is broad, user-facing, and spans the kind of recent-version breakage that shows up in several other reports here; still, it is not a true duplicate hub.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "Different bugs: processor/tokenizer regression vs `resize_token_embeddings` not updating `decoder.embed_tokens`." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Both mention flex_attention, but one is a torch>=2.9 compiled-flex-attention failure and the other is an import/version-check bug." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "Unrelated APIs: BEiT fast processor label reduction vs text-classification docs for `return_all_scores`." + }, + { + "left": "issue:45083", + "right": "issue:46506", + "accept": false, + "reason": "Different model paths and symptoms; helper-length logic in qwen3_omni_moe is not the same as CPU/CUDA output mismatch in PPDocLayoutV3." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Distinct failures: gradient checkpointing in eval mode vs D-FINE auxiliary loss handling when denoising is off." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "Docs mismatch vs model-loading/config rejection; same broad theme of API breakage, but not the same bug." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "One is an unrecognized model/config issue, the other is a PIL image-processor torchvision dependency regression." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Different Ministral-related problems: save/dequantize failure vs FlashAttention crash." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Unrelated concerns: GenerationMixin cache during chunked prefill vs ANSI codes in loading_report." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "One is a Python-version compatibility bug; the other is a generic dependency issue with transformers." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Different composite-model loading issues: unrecognized model config vs `from_config` not unwrapping `text_config`." + }, + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Both are flash-attention related, but they hit different code paths and failure modes." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error and tokenizer keyword conflict are unrelated." + }, + { + "left": "issue:44792", + "right": "issue:46424", + "accept": false, + "reason": "A Janus test failure is not the same bug as cache config max length handling." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Different import problems in different modules (`Qwen2_5_VLForConditionalGeneration` vs `image_utils`)." + }, + { + "left": "issue:44327", + "right": "issue:46752", + "accept": false, + "reason": "Different edge-case crashes: QA span decoding top-k bounds vs empty chat-template input." + }, + { + "left": "issue:46123", + "right": "issue:46369", + "accept": false, + "reason": "Silently dropping final batch results is unrelated to sam3-litetext memory overflow." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Both are Qwen regressions, but one is model loading and the other is incorrect save_pretrained output; separate code paths." + }, + { + "left": "issue:44521", + "right": "issue:46693", + "accept": false, + "reason": "apply_chat_template assistant masks vs FA2 long-context performance regression are unrelated." + }, + { + "left": "issue:44279", + "right": "issue:46515", + "accept": false, + "reason": "Generic dependency issue and torch 2.4.1 incompatibility are not the same defect." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "Embedding resize bug and Whisper empty-transcription generation failure are different issues." + }, + { + "left": "issue:46082", + "right": "issue:46399", + "accept": false, + "reason": "LlamaConfig head_dim validation is unrelated to the Qwen3.5 kernel registration decorator bug." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "Dependency complaints, but one is about transformers compatibility and the other is a PEFT minimum-version bump." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Both mention tokenizer behavior, but the affected models and failures are different (processor/tokenizer API regression vs Deepseek tokenization correctness)." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "SigLIP2 doc-example errors are not the same as the text-classification pipeline docs mismatch." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "Blip2 dtype propagation and Qwen3.5 save_pretrained regression are separate dtype/state bugs." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "Optimizer argument handling is unrelated to the Qwen3.5 `tie_word_embeddings` warning." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Python support failure and PEFT version bump are different release/compatibility issues." + }, + { + "left": "issue:46082", + "right": "issue:46512", + "accept": false, + "reason": "Both involve config fields, but one is head_dim validation and the other is stale dtype metadata after from_config." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "Video-classification pipeline processor lookup is not the same issue as remote_code model breakage." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 39, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:33290", + "issue:41950", + "issue:42175", + "issue:42796", + "issue:42994", + "issue:43299", + "issue:43317", + "issue:43441", + "issue:43479", + "issue:43525", + "issue:43650", + "issue:43653", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44242", + "issue:44246", + "issue:44261", + "issue:44263", + "issue:44291", + "issue:44351", + "issue:44361", + "issue:44462", + "issue:44466", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:45081", + "issue:45103", + "issue:45200", + "issue:45276", + "issue:45278", + "issue:45357", + "issue:45372", + "issue:45397", + "issue:45542", + "issue:45593", + "issue:45600", + "issue:45656", + "issue:45712", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:45907", + "issue:46209" + ], + "soft_pairs": [ + "issue:44351|issue:45278", + "issue:43299|issue:45357", + "issue:42796|issue:43650", + "issue:44779|issue:45812", + "issue:44242|issue:45200", + "issue:43479|issue:43525", + "issue:44466|issue:45276", + "issue:43299|issue:43854", + "issue:41950|issue:45823", + "issue:43653|issue:45593", + "issue:44683|issue:45542", + "issue:43650|issue:46209", + "issue:44016|issue:44291", + "issue:44462|issue:45081", + "issue:44361|issue:45593", + "issue:43441|issue:45834", + "issue:45103|issue:45372", + "issue:42994|issue:43317", + "issue:33290|issue:45656", + "issue:45081|issue:45907", + "issue:45600|issue:45712", + "issue:44623|issue:44749", + "issue:44261|issue:44263", + "issue:42175|issue:43976", + "issue:44671|issue:44779", + "issue:43901|issue:44246", + "issue:44623|issue:45397", + "issue:43441|issue:44683" + ], + "prepared_review_unit_hash": "ee8ca5abccc1a7e04685da2809074dad5162d0db12c557c482396639eba340ce", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36017, + "estimated_input_tokens": 9005, + "estimated_eval_tokens": 18266 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is not a true duplicate cluster: the items span unrelated tokenizer, model-loading, config, training, CI, and documentation bugs. All proposed soft edges look like false positives, so none should be merged.", + "confidence": 0.94, + "canonical_issue_reason": "No clear canonical issue; the set is heterogeneous and the soft edges do not point to one shared underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "No single issue is a strong global representative for this cluster because the reports cover many different code paths and regressions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "Missing import of one symbol vs broad import errors after upgrade; not the same bug." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Model loading failure vs save_pretrained key serialization regression; different code paths." + }, + { + "left": "issue:42796", + "right": "issue:43650", + "accept": false, + "reason": "Tokenizer BPE marker handling vs unrelated placeholder issue; no shared bug." + }, + { + "left": "issue:44779", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and have different failure modes." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "MoE load-balancing loss bug vs Gemma4 mm_token_type_ids defaulting; unrelated." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Different config classes and errors: default-init bug vs missing attribute." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Serialization inconsistency vs embedding-resize propagation; different functionality." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Different models and symptoms; only broadly both are loading issues." + }, + { + "left": "issue:41950", + "right": "issue:45823", + "accept": false, + "reason": "Pipeline processor selection bug vs device mismatch during forward; unrelated." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "Tokenizer special-token registration vs auxiliary-loss omission; unrelated." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "torch flex_attention compile failure vs missing TensorFlow backend package; unrelated." + }, + { + "left": "issue:43650", + "right": "issue:46209", + "accept": false, + "reason": "No substantive overlap: placeholder issue vs FP8 inference failure on B200." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Notebook syntax error vs init_empty_weights TypeError; unrelated." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer loading preference vs regex patch crash; different loader paths." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "Tokenizer AttributeError vs D-FINE loss bug; unrelated." + }, + { + "left": "issue:43441", + "right": "issue:45834", + "accept": false, + "reason": "FlashAttention failure for Ministral vs OCR index error in Kosmos2.5; unrelated." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "auto_docstring annotation bug vs Gemma4 processor import error; unrelated." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Quantized save failure vs dequantized load/offload failure; related area but different bugs." + }, + { + "left": "issue:33290", + "right": "issue:45656", + "accept": false, + "reason": "Adafactor OOM vs DeepSpeed double optimizer step; different optimizer problems." + }, + { + "left": "issue:45081", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer regex crash vs repo file listing missing revision forwarding; unrelated." + }, + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Stale auto_mappings references vs leftover dummy classes; different CI failures." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "processor.save_pretrained missing files vs slow filtering after upgrade; unrelated." + }, + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "MLA layernorm epsilon bug vs GlmMoeDsaIndexer split return issue; unrelated." + }, + { + "left": "issue:42175", + "right": "issue:43976", + "accept": false, + "reason": "Missing TensorFlow backend in torch install vs Python version compatibility regression; unrelated." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "Different model regressions and different code paths; not the same bug." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "Docs mismatch vs import performance regression; unrelated." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "Processor save-pretrained file omission vs Gemma4 zero3 from_pretrained; different load/save paths." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "Both are tokenizer-related, but one ignores tokenizer.json and the other crashes while patching Mistral regex." + }, + { + "left": "issue:44779", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and have 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": 23, + "review_unit_index": 40, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:41628", + "issue:42491", + "issue:43299", + "issue:43525", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43643", + "issue:43673", + "issue:43723", + "issue:43761", + "issue:43784", + "issue:43824", + "issue:43881", + "issue:43976", + "issue:44016", + "issue:44162", + "issue:44188", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44297", + "issue:44336", + "issue:44373", + "issue:44451", + "issue:44485", + "issue:44496", + "issue:44589", + "issue:44871", + "issue:45003", + "issue:45103", + "issue:45362", + "issue:45397", + "issue:45542", + "issue:45561", + "issue:45563", + "issue:45593", + "issue:45800", + "issue:45823", + "issue:46291", + "issue:46399", + "issue:46424", + "issue:46614", + "issue:46655", + "issue:46665", + "issue:46726", + "issue:46752" + ], + "soft_pairs": [ + "issue:45542|issue:45800", + "issue:43723|issue:44162", + "issue:43673|issue:43761", + "issue:43824|issue:43976", + "issue:43525|issue:43881", + "issue:44016|issue:45823", + "issue:46291|issue:46399", + "issue:43525|issue:43643", + "issue:44589|issue:45362", + "issue:44188|issue:44336", + "issue:46655|issue:46726", + "issue:44336|issue:44373", + "issue:41628|issue:43976", + "issue:44246|issue:44496", + "issue:43784|issue:45542", + "issue:44297|issue:45397", + "issue:45561|issue:45563", + "issue:46614|issue:46665", + "issue:45103|issue:46399", + "issue:43526|issue:45003", + "issue:44242|issue:45593", + "issue:42491|issue:43299", + "issue:43531|issue:45362", + "issue:43540|issue:46752", + "issue:44246|issue:44871", + "issue:44295|issue:45397", + "issue:44336|issue:44485", + "issue:44451|issue:46752", + "issue:44336|issue:46424" + ], + "prepared_review_unit_hash": "754709a988ba0c7f053dd6454a7b7da8e5c4f5640c2d58908079c16b77c83d53", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35717, + "estimated_input_tokens": 8930, + "estimated_eval_tokens": 18116 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: most items are separate regressions, import errors, model-specific load failures, or unrelated tooling/doc issues. I would not merge any of the soft pairs as duplicates; the only reasonable anchor is the broad config-loading regression in issue 43643, but even that is not a true duplicate of the others.", + "confidence": 0.22, + "canonical_issue_reason": "Issue 43643 is the broadest and most central loading/config regression in the set, with a generic code path and clearer triage value than the highly model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43643 is the best representative issue because it targets a common upstream loading path and is more generally actionable than the other narrowly scoped bug reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "Different failure modes: TensorFlow backend handling vs a torch/transformers version incompatibility." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "Tokenizer loading regression and ESM2 model breakage are unrelated code paths." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Generation cache issue during chunked prefill is not the same bug as CLIP hidden-states being dropped." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Missing Qwen2.5-VL import and Python-version support problems are different issues." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Llama4Config missing pad_token_id and glm-4v-9b loading failure do not share the same root cause." + }, + { + "left": "issue:44016", + "right": "issue:45823", + "accept": false, + "reason": "Notebook syntax error is unrelated to Gemma4 device placement during forward." + }, + { + "left": "issue:46291", + "right": "issue:46399", + "accept": false, + "reason": "Kernel dependency version/revision validation and a missing decorator on rotary-pos-emb registration are different problems." + }, + { + "left": "issue:43525", + "right": "issue:43643", + "accept": false, + "reason": "Both are config-related, but one is a missing attribute on Llama4Config while the other is trust_remote_code returning incomplete AutoConfig fields." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Float8 storage lookup failure and Qwen3.5 chat crash are unrelated." + }, + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "torch.compile attention divergence is not the same as ANSI codes emitted by loading_report." + }, + { + "left": "issue:46655", + "right": "issue:46726", + "accept": false, + "reason": "Video padding calculation in SmolVLM is unrelated to AutoProcessor's missing-vision-backend error." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "A terminal-color output bug and a docstring issue are not duplicates." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "AutoImageProcessor import failure is not the same as Python 3.9/3.10 compatibility breakage." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Load-balancing loss omission and unrecognized model/config loading are different code paths." + }, + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "sentence-transformers import NameError and undefined tf backend initialization are unrelated." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer save metadata mismatch is different from Gemma4 zero3 from_pretrained loading." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "A pytest-xdist file race and a stale generation warning are not the same bug." + }, + { + "left": "issue:46614", + "right": "issue:46665", + "accept": false, + "reason": "Stale phonemizer backend state and from_pretrained behavior inside Python classes are unrelated." + }, + { + "left": "issue:45103", + "right": "issue:46399", + "accept": false, + "reason": "auto_docstring annotation handling and Qwen3.5 kernel registration are different subsystems and failures." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and modeling_utils sys.modules access are unrelated." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "Missing load-balancing loss and missing auxiliary losses are similar at a high level but not the same concrete bug." + }, + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both mention Qwen3 MoE, but they describe different loading failures and model variants, so they are not safe duplicates." + }, + { + "left": "issue:43531", + "right": "issue:45362", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and Qwen3.5 chat crashes are different issues." + }, + { + "left": "issue:43540", + "right": "issue:46752", + "accept": false, + "reason": "Video-input processing failure in Qwen3OmniMoe is unrelated to empty chat-template input handling." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Import performance and Gemma-3 eos_token_id configuration are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "position_ids buffer access and Gemma4 zero3 loading are different failure modes." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "ANSI output formatting and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:44451", + "right": "issue:46752", + "accept": false, + "reason": "A specific model loading failure and empty list handling in apply_chat_template are not duplicates." + }, + { + "left": "issue:44336", + "right": "issue:46424", + "accept": false, + "reason": "Loading-report terminal coloring and cache length semantics 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": 23, + "review_unit_index": 41, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:36246", + "issue:41950", + "issue:42757", + "issue:42886", + "issue:43257", + "issue:43262", + "issue:43295", + "issue:43475", + "issue:43540", + "issue:43550", + "issue:43596", + "issue:43673", + "issue:43901", + "issue:44038", + "issue:44162", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44351", + "issue:44466", + "issue:44479", + "issue:44496", + "issue:44561", + "issue:44623", + "issue:44734", + "issue:44749", + "issue:44779", + "issue:44857", + "issue:44877", + "issue:45003", + "issue:45137", + "issue:45200", + "issue:45292", + "issue:45335", + "issue:45375", + "issue:45561", + "issue:45584", + "issue:45600", + "issue:45759", + "issue:45859", + "issue:46132", + "issue:46369", + "issue:46399", + "issue:46424", + "issue:46506", + "issue:46655", + "issue:46726" + ], + "soft_pairs": [ + "issue:42886|issue:46132", + "issue:43550|issue:46506", + "issue:41950|issue:44220", + "issue:44779|issue:45292", + "issue:43901|issue:45003", + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:43257|issue:45137", + "issue:45003|issue:45561", + "issue:44734|issue:46424", + "issue:44496|issue:45003", + "issue:43540|issue:46655", + "issue:43673|issue:44561", + "issue:44162|issue:45200", + "issue:44242|issue:44857", + "issue:45859|issue:46399", + "issue:43262|issue:44246", + "issue:45375|issue:45759", + "issue:44479|issue:46655", + "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:44561|issue:46399", + "issue:45292|issue:45584", + "issue:44466|issue:45335", + "issue:44623|issue:46726", + "issue:44038|issue:46506" + ], + "prepared_review_unit_hash": "0278dd12ce2a4290d2bc340cf3c7bc5b2efe881b8d0d72c15be4fc5c0d750c7f", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36713, + "estimated_input_tokens": 9179, + "estimated_eval_tokens": 18614 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose topical grouping of unrelated issues: import/load failures, cache/processor regressions, DeepSpeed problems, and several model-specific video/text bugs. The soft pairs are similar in wording or subsystem, but they do not look like the same underlying defect.", + "confidence": 0.85, + "canonical_issue_reason": "issue:44479 is the cleanest concrete regression in the set: it names the affected Qwen VL families, version, and symptom, and is more actionable than the broader or more ambiguous reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44479 is the best representative issue because it is a well-scoped, user-facing bug report with a clear failure mode and model/version context.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42886", + "right": "issue:46132", + "accept": false, + "reason": "Offline tokenizer cache loading failure and AttentionInterface.register behavior change are different code paths and symptoms." + }, + { + "left": "issue:43550", + "right": "issue:46506", + "accept": false, + "reason": "torch.compile/SDPA crash in Bamba is unrelated to CPU-vs-CUDA output differences in PPDocLayoutV3." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification pipeline image-processor lookup and _torch_extract_fbank_features are separate bugs in different subsystems." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "Deepseek tokenizer corruption and resize_token_embeddings not updating output embeddings are different failures." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "A docs/behavior mismatch for return_all_scores is not the same as unsafe sys.modules access in modeling_utils." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they concern different missing symbols and different loading paths." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Different missing imports: Qwen2_5_VLImageProcessor versus HybridCache." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Qwen3 MOE weight conversion under accelerate/deepspeed is unrelated to a ZeRO3 empty-deque IndexError." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "modeling_utils sys.modules access and xdist race on captured_info.txt are unrelated bugs." + }, + { + "left": "issue:44734", + "right": "issue:46424", + "accept": false, + "reason": "Serve KV-cache continuation indexing and cache max_length behavior are distinct cache issues with different fixes." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model/config loading and unsafe sys.modules access are different root causes." + }, + { + "left": "issue:43540", + "right": "issue:46655", + "accept": false, + "reason": "Both involve video inputs, but they are different model families and different processor bugs." + }, + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Missing GenerationMixin cache in chunked_prefill and removal of is_torch_fx_available break different features." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "ESM2 failure and Gemma 4 mm_token_type_ids defaulting are unrelated model-specific bugs." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Missing load-balancing loss and an AMP/CUDA crash in LwDetrImageLoss are different defects." + }, + { + "left": "issue:45859", + "right": "issue:46399", + "accept": false, + "reason": "Both target Qwen3.5 MOE, but CPU-offload incompatibility and missing decorator registration are different concrete problems." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio sampling-rate defaulting and slow import time do not share the same underlying bug." + }, + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Missing config field in Qwen3.5MoeVisionConfig and text_config unwrapping in AutoModelForCausalLM are separate composite-config issues." + }, + { + "left": "issue:44479", + "right": "issue:46655", + "accept": false, + "reason": "Both are video-related regressions, but they affect different models and stem from different processor logic." + }, + { + "left": "issue:43475", + "right": "issue:46369", + "accept": false, + "reason": "SAM 3 Video missing attribute and sam3-litetext memory overflow are not the same bug." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "Buffer access IndexError and performance slowdown after upgrade are unrelated." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque failure and stale auto_mappings config references are different issues." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "Qwen2_5_VLImageProcessor import failure and BertModel ZeRO3 IndexError are unrelated." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Strict granite_speech config loading and missing deepstack_visual_indexes are both config issues, but for different models and fields." + }, + { + "left": "issue:43295", + "right": "issue:45292", + "accept": false, + "reason": "Processor.tokenizer/image handling regression and resize_token_embeddings not updating output_embeddings are different API regressions." + }, + { + "left": "issue:44561", + "right": "issue:46399", + "accept": false, + "reason": "Broken trust_remote_code from removed is_torch_fx_available is unrelated to the Qwen3.5 decorator problem." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "Embedding resize behavior and Whisper empty-transcription generation are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Both concern serialization/resizing of embeddings, but they describe different model components and failure modes." + }, + { + "left": "issue:44623", + "right": "issue:46726", + "accept": false, + "reason": "Missing save_pretrained files and unrecognized image processor on AutoProcessor load are different processor-path bugs." + }, + { + "left": "issue:44038", + "right": "issue:46506", + "accept": false, + "reason": "Qwen3-VL-Moe version bug and CPU/CUDA divergence in PPDocLayoutV3 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": 23, + "review_unit_index": 42, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:29870", + "issue:30333", + "issue:42886", + "issue:43065", + "issue:43352", + "issue:43408", + "issue:43475", + "issue:43540", + "issue:43653", + "issue:43723", + "issue:43906", + "issue:43931", + "issue:43994", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44188", + "issue:44297", + "issue:44361", + "issue:44368", + "issue:44479", + "issue:44483", + "issue:44623", + "issue:44749", + "issue:44877", + "issue:44977", + "issue:44987", + "issue:44995", + "issue:45200", + "issue:45229", + "issue:45250", + "issue:45397", + "issue:45412", + "issue:45459", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45656", + "issue:45685", + "issue:45987", + "issue:46082", + "issue:46123", + "issue:46170", + "issue:46202", + "issue:46421", + "issue:46431", + "issue:46710", + "issue:46726" + ], + "soft_pairs": [ + "issue:29870|issue:30333", + "issue:44479|issue:46170", + "issue:44987|issue:45459", + "issue:43352|issue:45250", + "issue:45412|issue:46123", + "issue:44060|issue:44077", + "issue:44368|issue:44977", + "issue:43653|issue:44162", + "issue:44877|issue:46170", + "issue:44995|issue:45200", + "issue:43475|issue:43540", + "issue:42886|issue:45229", + "issue:45397|issue:45656", + "issue:43065|issue:43408", + "issue:44368|issue:45563", + "issue:44995|issue:45588", + "issue:43931|issue:44368", + "issue:43723|issue:46710", + "issue:45987|issue:46082", + "issue:44188|issue:46202", + "issue:43906|issue:46202", + "issue:44483|issue:46202", + "issue:44188|issue:46421", + "issue:45685|issue:46431", + "issue:45584|issue:46170", + "issue:43065|issue:46123", + "issue:44361|issue:44749", + "issue:44297|issue:44623", + "issue:43994|issue:46726" + ], + "prepared_review_unit_hash": "b79c8ff4355596e2069f79c15fb1469b3a65f8d0480ce0af49405c2a593684f3", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36481, + "estimated_input_tokens": 9121, + "estimated_eval_tokens": 18498 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the items span unrelated tokenizer, generation, config, vision, memory, and infrastructure problems. None of the soft-edge pairs look like the same underlying bug or change, so all should be rejected.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue is appropriate here because the issues are not duplicates of one another; they describe distinct bugs and regressions across different models, APIs, and subsystems.", + "canonical_pr_reason": "No pull requests were provided in this cluster.", + "best_issue_reason": "No single issue is a good representative for the cluster, since there is no shared underlying defect tying these reports together.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:29870", + "right": "issue:30333", + "accept": false, + "reason": "Bart generate() UnboundLocalError and MLflow job-status reporting are unrelated failures." + }, + { + "left": "issue:44479", + "right": "issue:46170", + "accept": false, + "reason": "Video input regression in Qwen/VL models is unrelated to Windows ROCm import/dependency errors." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "A model-loading failure for physical-intelligence/fast is not the same as protobuf masking tokenizer errors." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "NemotronHForCausalLM not supporting Flash Attention 2.0 is not the same as a generic Flash Attention 2.0 issue." + }, + { + "left": "issue:45412", + "right": "issue:46123", + "accept": false, + "reason": "RT-DETR memory release and MaskGenerationPipeline batch handling are different code paths and bugs." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "A Qwen3-Next tied-weights warning and patchtsmixer post_init validation are unrelated model/config issues." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "A tie_word_embeddings warning during LoRA fine-tuning is not the same as a flash-attention generation bug." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and ESM2 breakage are distinct tokenizer/model issues." + }, + { + "left": "issue:44877", + "right": "issue:46170", + "accept": false, + "reason": "Strict config loading for granite_speech is unrelated to a Windows ROCm torch.distributed import failure." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "GlmMoeDsa stale cache crash and Gemma4 mm_token_type_ids defaulting are different defects." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "Missing fpn_position_embeddings in SAM 3 video and Qwen3OmniMoe video ValueError are separate model implementation problems." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Offline tokenizer cache loading and Gemma4 multi-GPU CUDA OOM are unrelated." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Gemma-4 zero3 from_pretrained and a deepspeed double optimizer step are different training/runtime bugs." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder and sam3_video/sam3_tracker type mismatch are separate SAM 3 issues." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "A tie_word_embeddings warning and a stale num_return_sequences warning in paged generate() are not the same bug." + }, + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "GlmMoeDsa cache corruption and flash_attention.py failing on s_aux=None are different failures in different subsystems." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch is unrelated to Qwen3.5 tie_word_embeddings warnings." + }, + { + "left": "issue:43723", + "right": "issue:46710", + "accept": false, + "reason": "Tokenizer loading in v5 and DeepSeek-R1 accuracy regression are not the same underlying problem." + }, + { + "left": "issue:45987", + "right": "issue:46082", + "accept": false, + "reason": "StaticCache return-type contract and LlamaConfig head_dim validation are unrelated API/config issues." + }, + { + "left": "issue:44188", + "right": "issue:46202", + "accept": false, + "reason": "torch.compile attention-kernel divergence is unrelated to Gemma pipeline roundtrip tokenization failure." + }, + { + "left": "issue:43906", + "right": "issue:46202", + "accept": false, + "reason": "An isolated reproduction of another issue and a Gemma pipeline roundtrip failure are not duplicates." + }, + { + "left": "issue:44483", + "right": "issue:46202", + "accept": false, + "reason": "chat/completions request rejection in v5.3 and Gemma roundtrip token handling are different problems." + }, + { + "left": "issue:44188", + "right": "issue:46421", + "accept": false, + "reason": "Attention-kernel divergence and cache lazy-init recompilation are different performance/correctness bugs." + }, + { + "left": "issue:45685", + "right": "issue:46431", + "accept": false, + "reason": "MPS histogram support for Int and apply_chat_template missing chat template are unrelated errors." + }, + { + "left": "issue:45584", + "right": "issue:46170", + "accept": false, + "reason": "Whisper empty transcription handling and Windows ROCm import issues have no common code path." + }, + { + "left": "issue:43065", + "right": "issue:46123", + "accept": false, + "reason": "SAM 3 Conv2d placeholder and MaskGenerationPipeline partial-batch dropping are unrelated." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "MLukeTokenizer task AttributeError and a Transformers upgrade slowdown are not the same issue." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "tokenizer.save_pretrained class metadata mismatch and processor.save_pretrained missing files are different serialization bugs." + }, + { + "left": "issue:43994", + "right": "issue:46726", + "accept": false, + "reason": "SigLIP2 nonsensical outputs and AutoProcessor missing-vision-backend handling are distinct vision-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": 23, + "review_unit_index": 43, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:39692", + "issue:41628", + "issue:42371", + "issue:42994", + "issue:43232", + "issue:43441", + "issue:43475", + "issue:43482", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43701", + "issue:43723", + "issue:43784", + "issue:43873", + "issue:43986", + "issue:44038", + "issue:44162", + "issue:44188", + "issue:44351", + "issue:44623", + "issue:44749", + "issue:44869", + "issue:44871", + "issue:44933", + "issue:44945", + "issue:44964", + "issue:45092", + "issue:45137", + "issue:45230", + "issue:45310", + "issue:45356", + "issue:45375", + "issue:45381", + "issue:45397", + "issue:45406", + "issue:45478", + "issue:45656", + "issue:45753", + "issue:46132", + "issue:46202", + "issue:46326", + "issue:46399", + "issue:46726", + "issue:46752" + ], + "soft_pairs": [ + "issue:43646|issue:44162", + "issue:44749|issue:45656", + "issue:43701|issue:44749", + "issue:44871|issue:46132", + "issue:42994|issue:43482", + "issue:44623|issue:45656", + "issue:43441|issue:46752", + "issue:43232|issue:43673", + "issue:43475|issue:43643", + "issue:45375|issue:45478", + "issue:28282|issue:30990", + "issue:44964|issue:45137", + "issue:43986|issue:46726", + "issue:41628|issue:44933", + "issue:45310|issue:45375", + "issue:44869|issue:45356", + "issue:42994|issue:43525", + "issue:44038|issue:44623", + "issue:45230|issue:45397", + "issue:39692|issue:46726", + "issue:43723|issue:46326", + "issue:43475|issue:45381", + "issue:42371|issue:45092", + "issue:43784|issue:46399", + "issue:43232|issue:44188", + "issue:43873|issue:44945", + "issue:45406|issue:46202", + "issue:44945|issue:45753", + "issue:41628|issue:44351" + ], + "prepared_review_unit_hash": "4513848cce2fccb68e7b7bc0ba0cdf0d731d70e376f566a01b0759fc8128abfd", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36075, + "estimated_input_tokens": 9019, + "estimated_eval_tokens": 18294 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are superficially similar but refer to different underlying bugs. The only strong duplicate-like match is the AutoProcessor/vision-backend missing-dependency crash between issue:43986 and issue:46726.", + "confidence": 0.81, + "canonical_issue_reason": "issue:43986 is the better canonical issue because it states the broader AutoProcessor crash pattern more generally and was reported earlier, making it a better umbrella for the duplicate pair.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43986 is the strongest issue anchor overall because it captures the shared AutoProcessor failure in the broadest, most reusable wording for the duplicate pair.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "Both are Transformers v5 regressions, but one is a generic custom-model-init breakage and the other is an ESM2-specific failure; different code paths and fixes." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "A filtering slowdown is unrelated to DeepSpeed causing optimizer.step to run twice." + }, + { + "left": "issue:43701", + "right": "issue:44749", + "accept": false, + "reason": "Checkpoint resume key mismatch is not the same bug as a data filtering slowdown." + }, + { + "left": "issue:44871", + "right": "issue:46132", + "accept": false, + "reason": "EOS token config inconsistency is unrelated to AttentionInterface.register changing registered-function behavior." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Quantized model saving failure and GGUF loading failure are different operations with different fixes." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "Missing files in processor.save_pretrained is unrelated to a DeepSpeed optimizer-step double-call." + }, + { + "left": "issue:43441", + "right": "issue:46752", + "accept": false, + "reason": "FlashAttention failure for Ministral-3 is a different path from apply_chat_template crashing on an empty list." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "Both touch generation internals, but sync_gpus keyword handling and chunked_prefill cache absence are distinct failures." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "SAM 3 missing an output attribute is not the same as trust_remote_code dropping fields from AutoConfig results." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Same model family, but one is a strict/config-field omission and the other is a from_pretrained error; too little evidence of one underlying bug." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "ImportError for missing PyTorch and Sentence Transformers loading hang are unrelated." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Phi-4 multimodal loading and DeepSpeed ZeRO3 deque underflow are unrelated." + }, + { + "left": "issue:43986", + "right": "issue:46726", + "accept": true, + "reason": "Both describe the same underlying AutoProcessor failure when a required vision backend is missing; one is the specific video-model/torchvision case and the other is the generic error-message variant." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Different missing-import symbols and different import locations; not clearly the same bug." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "Both involve Qwen3.5 Moe, but one is a from_pretrained error and the other is a missing config field silently dropped; likely separate issues." + }, + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Whisper timestamp decoding crash and Kimi tokenizer codec regression are different tokenizer problems." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized model saving and Llama4Config missing pad_token_id are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe loading and processor.save_pretrained file omission are different code paths." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "A placeholder bug report does not establish the same problem as Gemma-4 zero3 from_pretrained." + }, + { + "left": "issue:39692", + "right": "issue:46726", + "accept": false, + "reason": "Documentation/example errors for SigLIP2 are not the same as missing-vision-backend processor loading behavior." + }, + { + "left": "issue:43723", + "right": "issue:46326", + "accept": false, + "reason": "AutoTokenizer loading in v5 and continue_final_message for mistral_common tokenizers are separate tokenizer bugs." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "Both are video/vision-related, but one is a missing attribute in SAM 3 and the other is wrong vision_position_ids in Qwen2.5-VL." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 settings guidance and remote-code checkpoint meta-initialization incompatibility are unrelated." + }, + { + "left": "issue:43784", + "right": "issue:46399", + "accept": false, + "reason": "Import-time NameError in sentence-transformers is unrelated to Qwen3.5 kernel registration failure." + }, + { + "left": "issue:43232", + "right": "issue:44188", + "accept": false, + "reason": "Generation kwargs update after sync_gpus and attention-kernel divergence under torch.compile are different regressions." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Quantization offloading behavior and pipeline-parallelism output correctness are distinct problems." + }, + { + "left": "issue:45406", + "right": "issue:46202", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer and Gemma pipeline roundtrip failure are not the same bug." + }, + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Pipeline parallelism incorrect output and Qwen3_5 infinite loop on a specific image are unrelated symptoms." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors from transformers, but they concern different symbols and likely different missing exports." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The pair-level judgment for issue:43986 vs issue:46726 is conservative and grounded, but the canonical-issue rationale overstates the evidence: issue:43986 does not clearly have \"more discussion/inbound context\" than issue:46726 (46726 has more inbound references, while 43986 has more discussion activity). If you want 43986 as canonical, justify it by its broader wording / earlier report instead of a comparative context claim. The rest of the summary looks fine." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 44, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:33290", + "issue:35141", + "issue:36010", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42673", + "issue:42831", + "issue:42898", + "issue:42994", + "issue:43232", + "issue:43278", + "issue:43377", + "issue:43479", + "issue:43504", + "issue:43704", + "issue:43723", + "issue:43824", + "issue:43976", + "issue:44016", + "issue:44038", + "issue:44079", + "issue:44222", + "issue:44242", + "issue:44496", + "issue:44568", + "issue:44623", + "issue:44779", + "issue:44792", + "issue:44869", + "issue:44933", + "issue:45081", + "issue:45103", + "issue:45137", + "issue:45200", + "issue:45310", + "issue:45561", + "issue:45600", + "issue:45712", + "issue:45812", + "issue:46077", + "issue:46123", + "issue:46202", + "issue:46231", + "issue:46693", + "issue:46701" + ], + "soft_pairs": [ + "issue:42994|issue:43824", + "issue:45081|issue:45103", + "issue:42673|issue:43704", + "issue:46202|issue:46701", + "issue:43278|issue:45137", + "issue:44038|issue:45310", + "issue:36010|issue:41628", + "issue:44222|issue:45310", + "issue:42831|issue:44222", + "issue:43232|issue:44792", + "issue:30333|issue:33290", + "issue:43504|issue:45561", + "issue:44496|issue:45561", + "issue:42175|issue:44933", + "issue:45561|issue:45600", + "issue:43976|issue:44933", + "issue:30990|issue:35141", + "issue:42175|issue:45712", + "issue:42898|issue:44568", + "issue:43377|issue:44242", + "issue:46077|issue:46123", + "issue:44623|issue:45310", + "issue:43479|issue:44869", + "issue:42673|issue:46693", + "issue:44016|issue:46231", + "issue:42898|issue:45812", + "issue:43723|issue:44568", + "issue:42222|issue:45200", + "issue:42898|issue:44779", + "issue:43479|issue:44079" + ], + "prepared_review_unit_hash": "58620e0f520aa13cc52473d2ca24b8952e3a67356f15e69a3bfdb4edbbc16fc7", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "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": "One clear duplicate pair appears: the two Qwen3ForCausalLM VRAM-leak reports. The rest are similar only at a broad theme level (imports, tokenizers, saving/loading, or model-specific regressions) and should stay separate.", + "confidence": 0.92, + "canonical_issue_reason": "issue:42673 is the stronger canonical issue: it is the earlier, fuller report of the exact same Qwen3ForCausalLM multi-dataloader-thread VRAM leak, with more discussion and inbound references than the later duplicate.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42673 is the best representative issue in this cluster because it is the clearest confirmed duplicate target and the most fully developed report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Different failures: quantized model saving vs a specific missing import for Qwen2_5_VLForConditionalGeneration. Same general release-era breakage, but not the same bug." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "Both are AttributeError crashes, but they occur in different subsystems and code paths: Mistral tokenizer regex patching vs auto-docstring processing under future annotations." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Same exact title and same underlying bug report: Qwen3ForCausalLM VRAM leak when used in multiple dataloader threads." + }, + { + "left": "issue:46202", + "right": "issue:46701", + "accept": false, + "reason": "Both mention Gemma-related generation/pipeline behavior, but one is a tokenizer roundtrip issue and the other is an image-text-to-text pipeline failure." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Completely different problems: dtype drift during evaluation vs a DeepSpeed ZeRO3 deque IndexError." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Both involve Qwen3 MoE models, but the concrete failures and affected versions differ; not the same code-path problem." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors from transformers, but they concern different symbols and likely different export/regression paths." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "One is FP8 save_pretrained behavior for MoE models; the other is a from_pretrained import/load error. Related area, different bug." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "FineGrainedFP8 accuracy regression is not the same as FP8 save_pretrained failure; different symptoms and code paths." + }, + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Generation-kwargs sync logic and a janus image-generation test failure are unrelated concrete issues." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "MLflow reporting on failed training jobs is unrelated to an AdaFactor/DeepSpeed OOM." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Pretrained model loading with a legacy field has nothing to do with the xdist captured_info.txt race." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Unrecognized model/config loading error and a pytest-xdist file race are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Dependency/backend packaging issue vs missing image_utils import; not the same underlying bug." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python-version compatibility breakage and a removed import are different problems." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "SentenceTransformer loading hang and token-embedding reinitialization are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Backend dependency omission is distinct from leftover dummy classes leaking into dir(transformers)." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Both are tokenizer behavior regressions in v5, but one is cleanup-space handling and the other is add_special_tokens/BOS-EOS insertion; different bugs." + }, + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "Padding-mask correctness in MIMI encoder is unrelated to router-logits load-balancing loss calculation." + }, + { + "left": "issue:46077", + "right": "issue:46123", + "accept": false, + "reason": "Bidirectional attention mask construction and final-partial-batch bookkeeping in MaskGenerationPipeline are separate issues." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Missing processor save files is not the same as Qwen3.5 MoE from_pretrained import/load failure." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Multimodal config default initialization and Whisper timestamp decoding crash are different code paths." + }, + { + "left": "issue:42673", + "right": "issue:46693", + "accept": false, + "reason": "Both mention performance/memory, but one is a VRAM leak in dataloader threads and the other is a long-context FA2 slowdown regression." + }, + { + "left": "issue:44016", + "right": "issue:46231", + "accept": false, + "reason": "Notebook syntax error and documentation parameter-name mismatch are not the same bug." + }, + { + "left": "issue:42898", + "right": "issue:45812", + "accept": false, + "reason": "Different tokenizer regressions affecting different model families and behaviors." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer loading failure in v5 and special-token insertion behavior are different issues." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "VitPose model breakage and Gemma4 mm_token_type_ids defaulting are unrelated." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and different failure modes; not one concrete bug." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Multimodal config defaults and ModelOutput key assignment are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded and conservative overall. The only clear duplicate is issue:42673 / issue:43704, which share the exact title and same VRAM-leak report; the choice of issue:42673 as canonical is supported by earlier date and more discussion. The remaining soft-edge rejections stay appropriately cautious and do not overstate equivalence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 45, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:28282", + "issue:30333", + "issue:34689", + "issue:35141", + "issue:41628", + "issue:42757", + "issue:42796", + "issue:42831", + "issue:42898", + "issue:43116", + "issue:43295", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:44060", + "issue:44075", + "issue:44190", + "issue:44222", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44448", + "issue:44485", + "issue:44623", + "issue:44734", + "issue:44933", + "issue:45071", + "issue:45092", + "issue:45103", + "issue:45125", + "issue:45245", + "issue:45310", + "issue:45356", + "issue:45405", + "issue:45656", + "issue:45712", + "issue:46082", + "issue:46132", + "issue:46231", + "issue:46665", + "issue:46710", + "issue:46762" + ], + "soft_pairs": [ + "issue:43784|issue:43824", + "issue:44734|issue:45103", + "issue:27957|issue:30333", + "issue:30333|issue:34689", + "issue:43874|issue:45092", + "issue:42898|issue:43295", + "issue:42898|issue:45356", + "issue:44060|issue:45125", + "issue:43116|issue:44190", + "issue:44295|issue:45656", + "issue:45405|issue:45712", + "issue:30333|issue:35141", + "issue:22355|issue:28282", + "issue:44933|issue:45712", + "issue:44222|issue:44297", + "issue:44075|issue:46231", + "issue:44360|issue:46762", + "issue:44448|issue:46132", + "issue:42831|issue:44623", + "issue:44297|issue:45656", + "issue:42796|issue:43723", + "issue:43479|issue:43644", + "issue:44222|issue:44623", + "issue:45245|issue:46132", + "issue:43479|issue:46665", + "issue:41628|issue:42757", + "issue:45071|issue:46082", + "issue:44373|issue:44485", + "issue:44297|issue:45310", + "issue:46132|issue:46710" + ], + "prepared_review_unit_hash": "436a98453fc98f5250464960406f97169efb3d0aa1e562ed1fc70747c73db137", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36612, + "estimated_input_tokens": 9153, + "estimated_eval_tokens": 18562 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly heterogeneous Transformers issues with only superficial similarity; none of the soft pairs look like true duplicates or mergeable PR-equivalents. The nearest overlap is in tokenizer/import/regression themes, but the concrete bugs and code paths differ.", + "confidence": 0.92, + "canonical_issue_reason": "No clear canonical issue: the cluster is heterogeneous, and the soft pairs are mostly unrelated regressions/import errors rather than duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44297 is the most concrete, self-contained bug here (tokenizer save_pretrained writes the wrong tokenizer_class), so it's the best single representative, though it doesn't unify the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Different import failures: sentence-transformers hits a missing nn symbol, while the other is a missing Qwen2_5_VL class in transformers." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Server KV-cache indexing crash and auto-docstring future-annotations AttributeError are unrelated code paths." + }, + { + "left": "issue:27957", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention 2 support for XLMRoberta is unrelated to MLFlow reporting on failed training jobs." + }, + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "MLFlow status handling and Llama 3.2 Vision model loading are different bugs with no shared failing path." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Both are multimodal/model-loading issues, but for different models and different failures; not the same bug." + }, + { + "left": "issue:42898", + "right": "issue:43295", + "accept": false, + "reason": "One is a clean_up_tokenization_spaces v5 behavior change; the other is a processor.tokenizer/images regression." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Tokenizer cleanup behavior and Kimi codec/BPE regression are separate tokenizer bugs." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Qwen3-Next tied-weights warning and Qwen3.5 MoE tensor-parallel plan are distinct model-specific issues." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "Different example scripts and different failures: multi-label classification output vs local dataset loading." + }, + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "position_ids buffer access and deepspeed double optimizer-step are unrelated runtime bugs." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "A PEFT version pin problem and leftover dummy class leakage are unrelated." + }, + { + "left": "issue:30333", + "right": "issue:35141", + "accept": false, + "reason": "MLFlow reporting on failed jobs and embedding resize/post_init reinitialization are different problems." + }, + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Both are import errors, but one is a missing submodule and the other is a missing PyTorch dependency; not the same bug." + }, + { + "left": "issue:44933", + "right": "issue:45712", + "accept": false, + "reason": "Both concern public namespace/import surface, but one is a missing import from image_utils and the other is dummy PT classes leaking into dir(transformers)." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "Save_pretrained is involved in both, but one is FP8 MoE model saving and the other is tokenizer config metadata; different artifacts and code paths." + }, + { + "left": "issue:44075", + "right": "issue:46231", + "accept": false, + "reason": "Optimizer SGD args being ignored is a code bug; the set_training parameter-name mismatch is a documentation issue." + }, + { + "left": "issue:44360", + "right": "issue:46762", + "accept": false, + "reason": "DSA indexer ReLU discussion and Minimax attention numerical differences are unrelated mathematical/model bugs." + }, + { + "left": "issue:44448", + "right": "issue:46132", + "accept": false, + "reason": "Pegasus output regression across versions and AttentionInterface.register behavior change are different underlying causes." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FineGrainedFP8 accuracy and processor.save_pretrained missing files are unrelated save/model-quality issues." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch and deepspeed double-step are different bugs." + }, + { + "left": "issue:42796", + "right": "issue:43723", + "accept": false, + "reason": "Both mention tokenizers, but one is model-specific BPE marker handling and the other is a generic AutoTokenizer loading regression." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Default config initialization with None and non-persistent buffer junk filling are unrelated initialization bugs." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "Model save_pretrained and processor save_pretrained failures are different persistence paths." + }, + { + "left": "issue:45245", + "right": "issue:46132", + "accept": false, + "reason": "Category-count runtime error and AttentionInterface.register behavior change are unrelated." + }, + { + "left": "issue:43479", + "right": "issue:46665", + "accept": false, + "reason": "Incorrect multimodal config defaults and from_pretrained behavior inside Python classes are not the same failure." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Missing AutoImageProcessor in transformers and missing is_offline_mode in huggingface_hub are separate dependency/export issues." + }, + { + "left": "issue:45071", + "right": "issue:46082", + "accept": false, + "reason": "Both involve config validation, but one is PretrainedConfig type checking and the other is LlamaConfig head_dim validation." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "A wrong docstring and a GLM-5 RoPE implementation issue are not duplicates." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer save metadata mismatch and Qwen3.5 MoE from_pretrained loading error are different bugs." + }, + { + "left": "issue:46132", + "right": "issue:46710", + "accept": false, + "reason": "AttentionInterface.register behavior change and DeepSeek accuracy regression are 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": 23, + "review_unit_index": 46, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:30990", + "issue:38175", + "issue:41628", + "issue:42831", + "issue:42907", + "issue:43122", + "issue:43232", + "issue:43408", + "issue:43475", + "issue:43644", + "issue:43723", + "issue:43824", + "issue:44038", + "issue:44190", + "issue:44206", + "issue:44222", + "issue:44261", + "issue:44263", + "issue:44279", + "issue:44295", + "issue:44360", + "issue:44373", + "issue:44483", + "issue:44484", + "issue:44623", + "issue:44734", + "issue:44829", + "issue:44977", + "issue:45092", + "issue:45103", + "issue:45310", + "issue:45357", + "issue:45381", + "issue:45406", + "issue:45834", + "issue:46002", + "issue:46050", + "issue:46082", + "issue:46209", + "issue:46305", + "issue:46328", + "issue:46402", + "issue:46421", + "issue:46424", + "issue:46431", + "issue:46655" + ], + "soft_pairs": [ + "issue:43723|issue:44977", + "issue:44222|issue:44484", + "issue:44190|issue:46082", + "issue:43232|issue:46424", + "issue:44206|issue:45103", + "issue:43122|issue:46209", + "issue:43232|issue:45834", + "issue:44263|issue:45092", + "issue:43475|issue:46305", + "issue:44734|issue:46421", + "issue:46209|issue:46431", + "issue:46305|issue:46655", + "issue:44263|issue:44360", + "issue:44295|issue:44623", + "issue:30990|issue:38175", + "issue:44263|issue:44373", + "issue:45381|issue:46305", + "issue:43644|issue:46305", + "issue:42907|issue:46050", + "issue:42831|issue:45310", + "issue:44279|issue:46328", + "issue:44829|issue:46002", + "issue:45092|issue:45357", + "issue:22355|issue:27957", + "issue:44483|issue:45406", + "issue:44261|issue:45092", + "issue:42831|issue:44038", + "issue:41628|issue:43824", + "issue:43408|issue:46402" + ], + "prepared_review_unit_hash": "bc4db0223e0925843a6bf96e73585bf5f3655e0f72566ef97a2f9141a5476552", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36067, + "estimated_input_tokens": 9017, + "estimated_eval_tokens": 18290 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a loose collection of unrelated Transformers issues, with a few broad v5-regression and save/load reports but no clear duplicate pairings. I found no soft-edge pair that looks like the same underlying bug or change.", + "confidence": 0.88, + "canonical_issue_reason": "Issue 43723 is the best representative because it is a core v5 loading regression and sits near the center of the cluster\u2019s repeated import/load/save complaints, while still being concrete enough to anchor triage.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43723 is the most broadly representative issue in this set: it concerns a general Transformers v5 loading problem rather than a narrow model-specific failure.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Both mention model loading/generation behavior, but one is tokenizer loading in v5 and the other is Qwen3.5 flash-attention generation; different code paths and symptoms." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "FP8 save_pretrained issues and max_shard_size defaults are different save/export concerns, not the same bug." + }, + { + "left": "issue:44190", + "right": "issue:46082", + "accept": false, + "reason": "Local dataset loading in an example script is unrelated to LlamaConfig head_dim validation." + }, + { + "left": "issue:43232", + "right": "issue:46424", + "accept": false, + "reason": "Generation kwargs after sync_gpus and cache max length honoring are distinct cache/generation behaviors." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "One is a feature-extractor argument crash; the other is an auto-docstring annotation parsing crash." + }, + { + "left": "issue:43122", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer behavior across versions and FP8 inference on B200 are unrelated." + }, + { + "left": "issue:43232", + "right": "issue:45834", + "accept": false, + "reason": "Generation kwarg handling is not the same as a Kosmos2.5 OCR index error." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "GlmMoeDsaIndexer torch.split behavior and InternVL2 meta-init compatibility are different model-specific failures." + }, + { + "left": "issue:43475", + "right": "issue:46305", + "accept": false, + "reason": "SAM3 vision encoder missing attributes and Moonshine streaming window off-by-one are unrelated code paths." + }, + { + "left": "issue:44734", + "right": "issue:46421", + "accept": false, + "reason": "Serve KV-cache continuation crashes and cache lazy initialization recompilation issues are different cache problems with different failure modes." + }, + { + "left": "issue:46209", + "right": "issue:46431", + "accept": false, + "reason": "FP8 inference failure on B200 is unrelated to processor.save_pretrained missing files." + }, + { + "left": "issue:46305", + "right": "issue:46655", + "accept": false, + "reason": "Moonshine sliding-window off-by-one and SmolVLM video padding calculation are separate video-processing bugs." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "Both reference DSA indexer internals, but one is about torch.split outputs and the other about a missing ReLU; not the same bug." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "A position_ids buffer read error and missing files from processor.save_pretrained are different save/load issues." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "Sentence Transformers loading stalls and SigLIP2 zero probabilities are unrelated symptoms and subsystems." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "A runtime tensor-splitting issue is not the same as a docstring mismatch for position_ids." + }, + { + "left": "issue:45381", + "right": "issue:46305", + "accept": false, + "reason": "Qwen2.5-VL video position ids and Moonshine streaming windowing are different video logic bugs." + }, + { + "left": "issue:43644", + "right": "issue:46305", + "accept": false, + "reason": "Non-persistent buffer initialization junk and streaming mask off-by-one are unrelated regressions." + }, + { + "left": "issue:42907", + "right": "issue:46050", + "accept": false, + "reason": "Dequantized model saving and sinq quantized save/from_pretrained breakage are both serialization-related but appear to be different mechanisms and model types." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "FineGrainedFP8 accuracy issues and Qwen3.5 Moe from_pretrained errors are not the same underlying defect." + }, + { + "left": "issue:44279", + "right": "issue:46328", + "accept": false, + "reason": "A generic dependency issue and a SequenceFeatureExtractor pad performance regression are unrelated." + }, + { + "left": "issue:44829", + "right": "issue:46002", + "accept": false, + "reason": "Flash-attention-3 degenerate training and NaNs after checkout/init are different training instability reports." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "Old InternVL2 remote-code/meta-init incompatibility and Qwen3.5 visual encoder key save regression are separate model-family bugs." + }, + { + "left": "issue:22355", + "right": "issue:27957", + "accept": false, + "reason": "Missing transformers.onnx import and XLMRoberta Flash Attention 2 are unrelated issues." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "Chat/completions request rejection in serve and Gemma4Processor missing _tokenizer are different server/processor failures." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "A missing rms_norm_eps precision issue in MLA q_a_layernorm is unrelated to InternVL2 meta initialization compatibility." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "FineGrainedFP8 accuracy and Qwen3-VL-Moe v5 compatibility are not the same bug." + }, + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "AutoImageProcessor import failure and Qwen2_5_VLForConditionalGeneration import failure are separate missing-export issues." + }, + { + "left": "issue:43408", + "right": "issue:46402", + "accept": false, + "reason": "A model-type warning for sam3_tracker and missing Sam3Model text encoder weights are distinct SAM3 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": 23, + "review_unit_index": 47, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:28218", + "issue:30990", + "issue:33357", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42757", + "issue:42796", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43066", + "issue:43122", + "issue:43232", + "issue:43408", + "issue:43576", + "issue:43646", + "issue:43756", + "issue:43784", + "issue:43937", + "issue:43976", + "issue:44060", + "issue:44117", + "issue:44188", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44373", + "issue:44488", + "issue:44568", + "issue:44625", + "issue:44704", + "issue:44829", + "issue:44945", + "issue:44964", + "issue:45278", + "issue:45290", + "issue:45593", + "issue:45600", + "issue:45753", + "issue:46002", + "issue:46172", + "issue:46369", + "issue:46762" + ], + "soft_pairs": [ + "issue:22355|issue:28218", + "issue:42831|issue:45753", + "issue:39401|issue:44568", + "issue:27957|issue:30990", + "issue:42886|issue:43066", + "issue:42757|issue:45278", + "issue:42886|issue:42994", + "issue:44188|issue:44373", + "issue:43232|issue:45593", + "issue:43408|issue:46369", + "issue:44261|issue:44373", + "issue:43646|issue:46002", + "issue:42796|issue:43122", + "issue:42907|issue:44945", + "issue:30990|issue:33357", + "issue:43576|issue:43784", + "issue:43756|issue:44060", + "issue:39401|issue:44488", + "issue:42371|issue:44263", + "issue:43976|issue:44945", + "issue:44295|issue:44297", + "issue:42907|issue:44964", + "issue:43784|issue:45600", + "issue:46172|issue:46762", + "issue:44117|issue:44704", + "issue:42907|issue:43937", + "issue:44625|issue:44829", + "issue:41628|issue:45278", + "issue:44704|issue:45290" + ], + "prepared_review_unit_hash": "32611f77416276366494f60498fa8e8adc09264ab84e7fd632c88513d1c09b24", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 35863, + "estimated_input_tokens": 8966, + "estimated_eval_tokens": 18188 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: most items are unrelated model, tokenizer, import, or runtime-regression issues, so I would not merge any soft pair as duplicates. If forced to pick a representative issue, the broad Transformers v5 import-breakage report is the best anchor, but only weakly.", + "confidence": 0.29, + "canonical_issue_reason": "issue:45278 is the broadest, most general regression report in the set (multiple import errors after a version upgrade), so it is the least narrow representative of this noisy cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best single anchor because it captures a broad upgrade-related failure mode rather than a niche model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:22355", + "right": "issue:28218", + "accept": false, + "reason": "One is a missing-module import error; the other is a tokenizer spacing bug. Different symptoms and code paths." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "FineGrainedFP8 accuracy loss and Qwen3_5 image-specific infinite loop are unrelated bugs." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Both are tokenizer regressions, but one is offset mapping and the other is special-token insertion; different concrete behavior." + }, + { + "left": "issue:27957", + "right": "issue:30990", + "accept": false, + "reason": "Flash Attention 2 on XLM-Roberta and sentence-transformers loading hang are different problems." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Offline cache loading failure and wrong tokenizer decoder type are distinct tokenizer issues." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "A specific missing hub symbol is not the same underlying bug as a broad set of import errors after upgrade." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Tokenizer cache/offline loading and quantized model saving failure do not share the same code path." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "Torch.compile attention divergence and a docstring mismatch are not duplicates." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation-state update logic and D-FINE auxiliary loss behavior are unrelated." + }, + { + "left": "issue:43408", + "right": "issue:46369", + "accept": false, + "reason": "Model-type warning for sam3_tracker versus memory overflow in sam3-litetext are different SAM3 issues." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "Missing rms_norm_eps precision bug and wrong docstring for position_ids are unrelated." + }, + { + "left": "issue:43646", + "right": "issue:46002", + "accept": false, + "reason": "Custom model initialization breakage and NaNs in classification heads are different failure modes." + }, + { + "left": "issue:42796", + "right": "issue:43122", + "accept": false, + "reason": "A specific tokenizer BPE-marker bug is not the same as a general tokenization change across versions." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Quantized model saving failure and incorrect pipeline-parallel output are separate problems." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Loading hang for sentence-transformers and Mac bus error on a community CLIP model are different symptoms." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "Broken CLI env command and NameError during sentence-transformers import are unrelated." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "RoPE layer dropping in Smollm3 and incorrect tied-weights warning in Qwen3-Next are different model bugs." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Wrong offset mapping and failure to load a specific model are not the same underlying tokenizer issue." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 settings guidance and torch.split return values in a GLM indexer are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python version compatibility failure and pipeline-parallel output corruption are distinct issues." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "Reading position_ids after buffer registration and tokenizer_class mismatch on save_pretrained are separate bugs." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Dequantized Ministral/Devstral saving failure and Phi-4 multimodal load failure are unrelated." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "Import NameError in sentence-transformers and stale auto_mappings references are different breakages." + }, + { + "left": "issue:46172", + "right": "issue:46762", + "accept": false, + "reason": "A specific DeepSeek tensor operation bug and numerical differences in Minimax MSA attention are not the same code-path problem." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "Tokenizer mapping None-handling and AutoProcessor kwargs forwarding are different API bugs." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Quantized saving failure and invalid GenerationConfig for GLM-5 are unrelated." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "Config propagation for num_labels and degenerate training with flash_attention_3 are separate issues." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "A single missing import name is not the same as a broader multi-import failure after upgrade." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "Cached-file kwargs forwarding and chat-template crashes on tool-call messages 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": 23, + "review_unit_index": 48, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:29159", + "issue:36296", + "issue:38617", + "issue:39401", + "issue:42175", + "issue:42673", + "issue:42757", + "issue:42898", + "issue:42907", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43278", + "issue:43408", + "issue:43452", + "issue:43475", + "issue:43493", + "issue:43611", + "issue:43824", + "issue:43937", + "issue:43976", + "issue:44038", + "issue:44188", + "issue:44222", + "issue:44295", + "issue:44297", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44485", + "issue:44521", + "issue:44779", + "issue:44938", + "issue:44945", + "issue:45072", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45446", + "issue:45478", + "issue:45753", + "issue:46002", + "issue:46231", + "issue:46326", + "issue:46328", + "issue:46614" + ], + "soft_pairs": [ + "issue:39401|issue:44521", + "issue:44360|issue:44485", + "issue:43408|issue:43475", + "issue:43122|issue:44521", + "issue:43122|issue:46326", + "issue:44038|issue:44297", + "issue:42907|issue:43452", + "issue:22355|issue:29159", + "issue:43116|issue:45200", + "issue:45216|issue:45478", + "issue:39401|issue:46614", + "issue:44038|issue:44222", + "issue:44038|issue:44295", + "issue:43824|issue:44351", + "issue:43122|issue:44779", + "issue:43278|issue:44945", + "issue:42757|issue:44351", + "issue:36296|issue:43116", + "issue:22355|issue:29127", + "issue:42175|issue:44938", + "issue:43493|issue:45753", + "issue:44188|issue:46328", + "issue:42898|issue:46614", + "issue:43937|issue:44485", + "issue:43611|issue:46002", + "issue:38617|issue:45278", + "issue:43116|issue:43493", + "issue:43976|issue:45446", + "issue:44355|issue:46231", + "issue:43066|issue:44779", + "issue:42673|issue:45072", + "issue:44038|issue:45478" + ], + "prepared_review_unit_hash": "f6b10657ccbb05b37bbaaf1ec0ed8c0667392a00cb351596e10140b9d8d6403a", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 36852, + "estimated_input_tokens": 9213, + "estimated_eval_tokens": 18682 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: most items are unrelated bugs across tokenization, imports, model-specific regressions, training/runtime issues, and docs. I would not merge any of the proposed soft pairs as duplicates.", + "confidence": 0.34, + "canonical_issue_reason": "No clear canonical issue: the set spans multiple unrelated problem areas, and none of the soft pairs is close enough to be a safe duplicate merge.", + "canonical_pr_reason": null, + "best_issue_reason": "If a representative is needed, issue 44521 is the most active/open and has the richest discussion, but it is not an umbrella for the rest of the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Both involve tokenization-related output, but one is offset_mapping for Qwen3 and the other is multimodal assistant masks in apply_chat_template; different code paths." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer missing ReLU and GLM-5 RoPE implementation are unrelated model internals." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Both are SAM3 Video adjacent, but one is a model-type warning and the other is a missing attribute in encoder output; different bugs." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer version regression vs multimodal assistant mask generation; not the same underlying defect." + }, + { + "left": "issue:43122", + "right": "issue:46326", + "accept": false, + "reason": "Both mention tokenizer behavior, but one is general tokenization drift and the other is continue_final_message for mistral_common tokenizers." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe loading bug and tokenizer save_pretrained metadata mismatch are different failures." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Dequantized model save failure and gguf_file from_pretrained breakage are separate loading/saving issues." + }, + { + "left": "issue:22355", + "right": "issue:29159", + "accept": false, + "reason": "onnx import failure and slow/fast tokenizer inconsistency are unrelated." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Example script multi-label output bug and Gemma 4 mm_token_type_ids defaulting are different code paths." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Both are Qwen3.5 regression reports, but one is save_pretrained checkpoint corruption and the other is from_pretrained loading error; not the same concrete bug." + }, + { + "left": "issue:39401", + "right": "issue:46614", + "accept": false, + "reason": "Different tokenizer features: offset mapping vs phonemizer language state leakage." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Both mention MoE/FP8 context, but one is Qwen3-VL-Moe loading and the other is FP8 save_pretrained behavior." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Qwen3-VL-Moe loading and position_ids buffer reading are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Two separate import errors for different symbols; same broad symptom, different underlying breakage." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and failure modes." + }, + { + "left": "issue:43278", + "right": "issue:44945", + "accept": false, + "reason": "Embedding dtype drift in evaluate vs incorrect output under pipeline parallelism are distinct runtime bugs." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Different missing imports from different modules; not the same issue." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor parallel training bug and classification example script output bug are unrelated." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "onnx import error and LayoutLMv3 box error-message clarity are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Package extras/backend dependency issue and Python 3.14 import/load failure are different environment problems." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and Qwen3_5 infinite-loop-on-image are unrelated model bugs." + }, + { + "left": "issue:44188", + "right": "issue:46328", + "accept": false, + "reason": "torch.compile attention-kernel divergence and SequenceFeatureExtractor.pad performance are unrelated." + }, + { + "left": "issue:42898", + "right": "issue:46614", + "accept": false, + "reason": "v5 tokenization-space cleanup behavior and stale phonemizer backend state are different tokenizer concerns." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "GLM-5 GenerationConfig validation and GLM-5 RoPE implementation are different GLM-5 issues." + }, + { + "left": "issue:43611", + "right": "issue:46002", + "accept": false, + "reason": "base_model_prefix loading regression and NaNs in classification heads are separate problems." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "Specific missing layer_type_validation import vs broad import failures after upgrade; not the same bug." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Example script classification bug and SigLIP2 model discrepancy are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python-version compatibility failure and flex_attention version-check bug are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:46231", + "accept": false, + "reason": "Compiled Python file errors and a docs parameter-name mismatch are unrelated." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Wrong tokenizer decoder type and Deepseek tokenizer regression are different tokenizer issues." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "VRAM leak in dataloader threads and dtype mismatches in bfloat16 inference are different runtime bugs." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5/Qwen3-VL-Moe loading regressions, but the model variants and symptoms differ; too far apart to merge safely." + } + ] + }, + "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": 23, + "review_unit_index": 49, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:28348", + "issue:33453", + "issue:39401", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:42796", + "issue:43064", + "issue:43066", + "issue:43116", + "issue:43441", + "issue:43673", + "issue:43825", + "issue:43827", + "issue:43874", + "issue:43937", + "issue:43992", + "issue:44075", + "issue:44117", + "issue:44263", + "issue:44291", + "issue:44380", + "issue:44466", + "issue:44625", + "issue:44704", + "issue:44936", + "issue:44995", + "issue:45071", + "issue:45083", + "issue:45092", + "issue:45356", + "issue:45357", + "issue:45405", + "issue:45446", + "issue:45478", + "issue:45563", + "issue:45588", + "issue:46002", + "issue:46077", + "issue:46328", + "issue:46614", + "issue:46640", + "issue:46655", + "issue:46697", + "issue:46710", + "issue:46752", + "issue:46762" + ], + "soft_pairs": [ + "issue:44625|issue:46002", + "issue:45083|issue:45446", + "issue:45357|issue:45478", + "issue:42371|issue:43874", + "issue:44075|issue:44936", + "issue:42222|issue:43441", + "issue:43116|issue:43825", + "issue:43992|issue:44704", + "issue:44995|issue:46697", + "issue:43066|issue:46710", + "issue:28348|issue:46640", + "issue:46614|issue:46752", + "issue:44263|issue:46762", + "issue:43827|issue:45446", + "issue:45092|issue:46762", + "issue:43064|issue:43673", + "issue:42673|issue:42796", + "issue:43937|issue:45563", + "issue:33453|issue:39401", + "issue:44380|issue:46077", + "issue:45356|issue:46614", + "issue:43673|issue:46328", + "issue:42673|issue:46655", + "issue:42222|issue:46752", + "issue:45405|issue:45446", + "issue:44291|issue:45071", + "issue:44117|issue:45588", + "issue:44466|issue:46002" + ], + "prepared_review_unit_hash": "3f4a0d13f1858b67cb35cb490b412ef99c206d434e473b87042eda5bf00f12d8", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35441, + "estimated_input_tokens": 8861, + "estimated_eval_tokens": 17978 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated issues spanning tokenizers, generation, training, vision, and attention backends. No soft pair looks like the same underlying bug/change closely enough to merge.", + "confidence": 0.96, + "canonical_issue_reason": "No clear duplicate cluster is present; if a representative issue is needed, issue:33453 is the broadest and most generally framed report in the set (tokenizer loading regression).", + "canonical_pr_reason": null, + "best_issue_reason": "issue:33453 is the best single representative because it is broad, well-scoped, and has substantial discussion, unlike the more niche model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44625", + "right": "issue:46002", + "accept": false, + "reason": "Different failures: config propagation for Qwen3.5 vs NaNs in classification heads. No shared code path." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "Unrelated subsystems: qwen3_omni_moe feature-length helper vs a flex_attention PyTorch version check." + }, + { + "left": "issue:45357", + "right": "issue:45478", + "accept": false, + "reason": "Both are Qwen3.5-related, but one is save_pretrained key serialization and the other is from_pretrained load failure; not the same bug." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 API settings warning is unrelated to a missing image-patch method in GLM46V." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "Optimizer SGD args not used is unrelated to trainer.evaluate() failing after train()." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "VitPose model breakage and Ministral FlashAttention failure affect different models and code paths." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Example script multi-label output bug is unrelated to a pipeline translation error message regression." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Missing encoder weights on model load is unrelated to AutoProcessor kwargs forwarding." + }, + { + "left": "issue:44995", + "right": "issue:46697", + "accept": false, + "reason": "Stale GlmMoeDsa indexer cache and Swin2SR over-padding are different image/model bugs." + }, + { + "left": "issue:43066", + "right": "issue:46710", + "accept": false, + "reason": "Tokenizer decoder-type issue is unrelated to a DeepSeek accuracy regression." + }, + { + "left": "issue:28348", + "right": "issue:46640", + "accept": false, + "reason": "Generic Flash Attention 2 issue is too vague and does not clearly match GPT2 FA2 support." + }, + { + "left": "issue:46614", + "right": "issue:46752", + "accept": false, + "reason": "Stale phonemizer backend state is unrelated to apply_chat_template crashing on empty input." + }, + { + "left": "issue:44263", + "right": "issue:46762", + "accept": false, + "reason": "GlmMoeDsa split return values bug is unrelated to Minimax attention numerical differences." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs referencing pipeline() after removals is unrelated to a flex_attention import/version check." + }, + { + "left": "issue:45092", + "right": "issue:46762", + "accept": false, + "reason": "Remote-code meta-init incompatibility and attention numerical drift are different problems." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "FSDP2/PEFT optimizer-state corruption is unrelated to missing GenerationMixin cache in chunked prefill." + }, + { + "left": "issue:42673", + "right": "issue:42796", + "accept": false, + "reason": "VRAM leak in multi-dataloader-thread Qwen3ForCausalLM is unrelated to tokenizer BPE marker handling." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "Invalid GenerationConfig and stale num_return_sequences warning are different generation issues." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Both are tokenizer reports, but one is a generic loading regression and the other is a specific offset_mapping bug." + }, + { + "left": "issue:44380", + "right": "issue:46077", + "accept": false, + "reason": "GPT2 attention scaling under SDPA/FlashAttention and Gemma4 bidirectional mask construction are distinct attention-path bugs." + }, + { + "left": "issue:45356", + "right": "issue:46614", + "accept": false, + "reason": "Kimi tokenizer codec/regression issue is unrelated to Wav2Vec2 phonemizer backend state." + }, + { + "left": "issue:43673", + "right": "issue:46328", + "accept": false, + "reason": "Generation cache missing and feature-extractor padding inefficiency are unrelated." + }, + { + "left": "issue:42673", + "right": "issue:46655", + "accept": false, + "reason": "VRAM leak in dataloader threads is unrelated to SmolVLM video padding logic." + }, + { + "left": "issue:42222", + "right": "issue:46752", + "accept": false, + "reason": "VitPose model breakage and apply_chat_template empty-list crash do not share a concrete code path." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "PEFT minimum-version bump and AuxRequest import version check are unrelated maintenance issues." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "Both involve v5-era compatibility, but one is init_empty_weights/_is_hf_initialized and the other is PretrainedConfig type checking; not the same bug." + }, + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "Tokenizer mapping None handling and flash_attention sink-less AttributeError are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:46002", + "accept": false, + "reason": "lm_head serialization inconsistency is unrelated to NaNs in classification heads." + } + ] + }, + "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": 23, + "review_unit_index": 50, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42222", + "issue:42371", + "issue:42617", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43408", + "issue:43450", + "issue:43673", + "issue:43742", + "issue:43937", + "issue:43986", + "issue:44075", + "issue:44117", + "issue:44220", + "issue:44261", + "issue:44263", + "issue:44410", + "issue:44484", + "issue:44485", + "issue:44493", + "issue:44496", + "issue:44779", + "issue:44829", + "issue:44877", + "issue:44964", + "issue:44977", + "issue:45071", + "issue:45081", + "issue:45216", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45362", + "issue:45478", + "issue:45561", + "issue:45656", + "issue:46178", + "issue:46209", + "issue:46231", + "issue:46298", + "issue:46328", + "issue:46512", + "issue:46617", + "issue:46762" + ], + "soft_pairs": [ + "issue:44410|issue:45310", + "issue:39401|issue:44779", + "issue:43937|issue:44977", + "issue:45071|issue:46512", + "issue:45341|issue:45561", + "issue:44829|issue:45216", + "issue:45341|issue:46328", + "issue:43450|issue:43986", + "issue:44117|issue:45081", + "issue:42617|issue:45341", + "issue:38617|issue:41628", + "issue:43408|issue:43742", + "issue:45362|issue:46328", + "issue:43122|issue:46298", + "issue:42222|issue:46178", + "issue:44877|issue:46231", + "issue:44484|issue:45310", + "issue:42371|issue:44261", + "issue:43937|issue:44496", + "issue:39401|issue:46209", + "issue:44410|issue:45478", + "issue:42371|issue:46617", + "issue:44261|issue:46762", + "issue:38617|issue:44220", + "issue:44075|issue:45656", + "issue:44964|issue:45278", + "issue:44263|issue:44485", + "issue:43066|issue:43122", + "issue:43097|issue:44493", + "issue:43673|issue:45341" + ], + "prepared_review_unit_hash": "39ecb9978cf063aba9bc990e44c647e5293eedb4a371904e16d0a44c9e0e1758", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36285, + "estimated_input_tokens": 9072, + "estimated_eval_tokens": 18400 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "All soft pairs are false positives: they share Transformers/tokenizer/model-loading themes, but each reports a different bug, model family, or code path. No duplicate cluster emerges here.", + "confidence": 0.92, + "canonical_issue_reason": "No single canonical issue is appropriate; the set is heterogeneous across unrelated regressions and utilities, with only superficial topical overlap.", + "canonical_pr_reason": "No PR canonical candidate in this cluster.", + "best_issue_reason": "No single issue is a safe global representative because the reports span distinct subsystems (tokenization, config/loading, training, generation, testing, and docs).", + "best_pr_reason": "No PRs provided.", + "soft_edge_verdicts": [ + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Different symptoms and paths: missing Qwen3.5 layer projections vs MoE from_pretrained loading failure." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Both tokenizer-related, but one is offset_mapping alignment and the other is incorrect DeepSeek tokenization outputs." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "GLM-5 GenerationConfig validation is unrelated to Qwen3.5 flash-attention generation behavior." + }, + { + "left": "issue:45071", + "right": "issue:46512", + "accept": false, + "reason": "One is a typing regression in PretrainedConfig; the other is stale config.dtype after from_config(dtype=...)." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "A small testing_utils bug is distinct from an xdist race on captured_info.txt." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "Degenerate training with flash_attention_3 is unrelated to a save_pretrained checkpoint regression." + }, + { + "left": "issue:45341", + "right": "issue:46328", + "accept": false, + "reason": "testing_utils behavior bug vs a SequenceFeatureExtractor.pad performance issue." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Incorrect batched video processor shapes differ from AutoProcessor failing without torchvision." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None is unrelated to _patch_mistral_regex crashing on backend_tokenizer." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "3d_parallel.py execution failure is not the same as the testing_utils bug." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Separate import regressions: missing layer_type_validation vs missing AutoImageProcessor export." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "A model-type warning for sam3 is unrelated to MobileLLM key errors during load." + }, + { + "left": "issue:45362", + "right": "issue:46328", + "accept": false, + "reason": "Qwen3.5 chat crash and a padding performance issue are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:46298", + "accept": false, + "reason": "Tokenizer output changes across versions are not the same as a noisy logging message." + }, + { + "left": "issue:42222", + "right": "issue:46178", + "accept": false, + "reason": "Vitpose breakage is a model-specific issue, not the backend mapping error for tensorflow_text." + }, + { + "left": "issue:44877", + "right": "issue:46231", + "accept": false, + "reason": "Strict granite_speech config loading and a docs parameter mismatch are unrelated." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "save_pretrained shard-size question is not the same as Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 settings guidance is unrelated to MLA q_a_layernorm missing rms_norm_eps." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "GenerationConfig invalidity is a different bug from unrecognized model config / missing model_type." + }, + { + "left": "issue:39401", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer offset mapping and FP8 inference failure on B200 are different problems." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "Same model family, but distinct failures: missing projections vs Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:42371", + "right": "issue:46617", + "accept": false, + "reason": "TF32 API messaging is unrelated to trust_remote_code cache FileNotFoundError." + }, + { + "left": "issue:44261", + "right": "issue:46762", + "accept": false, + "reason": "Both numerical, but one is rms_norm_eps precision mismatch and the other is Minimax attention reference divergence." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "ImportError for layer_type_validation is unrelated to _torch_extract_fbank_features()." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "SGD args not used and optimizer step called twice under deepspeed are different optimizer bugs." + }, + { + "left": "issue:44964", + "right": "issue:45278", + "accept": false, + "reason": "A specific Phi-4 multimodal loading failure is not the same as broad post-upgrade import errors." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "torch.split return-value handling in GlmMoeDsaIndexer is unrelated to GLM-5 RoPE implementation." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Wrong tokenizer decoder type and changed tokenization across versions are related broadly, but not the same concrete bug." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "Removal of tie_embeddings_and_encoder_decoder and unexpected position-id keys are different regressions." + }, + { + "left": "issue:43673", + "right": "issue:45341", + "accept": false, + "reason": "Generation cache missing in chunked_prefill is unrelated to the testing_utils 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": 23, + "review_unit_index": 51, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:27957", + "issue:38617", + "issue:41628", + "issue:42222", + "issue:42757", + "issue:43012", + "issue:43064", + "issue:43097", + "issue:43352", + "issue:43441", + "issue:43452", + "issue:43531", + "issue:43577", + "issue:43673", + "issue:43874", + "issue:43931", + "issue:43986", + "issue:44038", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44315", + "issue:44360", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44829", + "issue:44933", + "issue:44987", + "issue:45003", + "issue:45071", + "issue:45341", + "issue:45372", + "issue:45375", + "issue:45440", + "issue:45464", + "issue:45478", + "issue:45542", + "issue:45588", + "issue:45753", + "issue:46050", + "issue:46123", + "issue:46172", + "issue:46231", + "issue:46369", + "issue:46655", + "issue:46664" + ], + "soft_pairs": [ + "issue:42222|issue:45542", + "issue:41628|issue:45071", + "issue:43931|issue:44410", + "issue:45071|issue:45542", + "issue:44484|issue:45478", + "issue:44483|issue:45464", + "issue:43012|issue:43577", + "issue:44263|issue:46172", + "issue:44360|issue:46172", + "issue:43874|issue:44261", + "issue:27957|issue:43352", + "issue:43931|issue:44038", + "issue:43441|issue:44829", + "issue:43097|issue:43452", + "issue:45003|issue:45341", + "issue:43986|issue:45341", + "issue:44261|issue:44485", + "issue:44933|issue:46231", + "issue:45753|issue:46369", + "issue:44117|issue:44987", + "issue:43531|issue:45341", + "issue:43673|issue:46123", + "issue:43874|issue:44263", + "issue:45440|issue:46172", + "issue:43097|issue:45588", + "issue:43986|issue:45372", + "issue:44410|issue:45375", + "issue:43064|issue:44315", + "issue:43097|issue:46050", + "issue:44038|issue:44410", + "issue:38617|issue:42757", + "issue:43986|issue:44484", + "issue:46655|issue:46664" + ], + "prepared_review_unit_hash": "a56870ac1c2790470273ff929896062b12823c921155b82a46c3e385d53b756b", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37379, + "estimated_input_tokens": 9345, + "estimated_eval_tokens": 18946 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous, with mostly unrelated import, model-specific, training, and processor bugs. None of the proposed soft pairs looks like a true duplicate; they mostly share only a broad subsystem or keyword.", + "confidence": 0.92, + "canonical_issue_reason": "No single true canonical duplicate stands out because the cluster spans many unrelated bugs. If a representative must be chosen, issue:45071 is the broadest core-regression item and the least model-specific.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45071 is the best overall representative because it concerns a core API regression in PretrainedConfig and is broader than the mostly model-specific or peripheral reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "Different problems: vitpose model breakage vs TensorBoard-only TF backend handling." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "One is an import failure for AutoImageProcessor; the other is a type-checking regression in PretrainedConfig." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Both are Qwen-family, but they describe different model failures and different code paths." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Core type-checking regression vs missing/undefined TensorFlow backend error; not the same bug." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "Unrelated: shard-size/save_pretrained behavior vs Qwen3.5 MoE loading error." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Both involve chat/completions, but one is a general API rejection and the other is a streaming-specific Qwen3.5-0.8B failure." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "Both mention dtype/precision, but one is a PyTorch warning during compile and the other is a BLIP2 loading dtype mismatch." + }, + { + "left": "issue:44263", + "right": "issue:46172", + "accept": false, + "reason": "Different model components and different failure modes; no shared code-path bug." + }, + { + "left": "issue:44360", + "right": "issue:46172", + "accept": false, + "reason": "DSA indexer ReLU issue vs DeepSeek V4 dtype/transpose issue are unrelated." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Different multimodal/model bugs: image patch counting vs MLA q_a_layernorm epsilon precision." + }, + { + "left": "issue:27957", + "right": "issue:43352", + "accept": false, + "reason": "Same Flash Attention 2 keyword, but different models and different support issues." + }, + { + "left": "issue:43986", + "right": "issue:44038", + "accept": false, + "reason": "One is AutoProcessor/torchvision handling; the other is a Qwen3-VL-Moe bug." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "Both mention FlashAttention, but one is a model compatibility failure and the other is degenerate training with FA3." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "Version-5 breaking change vs gguf_file loading regression; no concrete shared bug." + }, + { + "left": "issue:45003", + "right": "issue:45341", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to a small testing_utils bug." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "Video-model/AutoProcessor crash vs testing utility issue are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "Both are precision/architecture discussions, but they affect different models and different implementation paths." + }, + { + "left": "issue:44933", + "right": "issue:46231", + "accept": false, + "reason": "Import/export issue from image_utils vs documentation parameter mismatch; not the same change." + }, + { + "left": "issue:45753", + "right": "issue:46369", + "accept": false, + "reason": "Different symptoms and subsystems: image-specific infinite loop vs memory overflow in a different model." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer mapping None handling and a specific model load failure are not the same underlying bug." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to a testing_utils bug." + }, + { + "left": "issue:43673", + "right": "issue:46123", + "accept": false, + "reason": "Cache handling in generation vs partial-batch result dropping in MaskGenerationPipeline are different code paths." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Different GLM-related bugs with different APIs and failure modes." + }, + { + "left": "issue:45440", + "right": "issue:46172", + "accept": false, + "reason": "Both touch DeepSeek-family models, but one is a remote/native implementation divergence and the other is a dtype/transpose bug." + }, + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "Removal of a config helper is unrelated to a flash_attention integration crash on sink-less models." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "AutoProcessor/torchvision crash vs mistral_common ReasoningEffort import error are unrelated." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Different Qwen variants and different missing fields/projections; not one duplicate bug." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "FSDP2/PEFT optimizer-state corruption is unrelated to Liger Kernel not being applied with model_init." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Both are Qwen-related, but one is a general transformers-v5 bug and the other is missing projections in a specific layer." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Two separate import errors from different modules/dependencies." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "AutoProcessor video-model crash vs shard-size question are unrelated." + }, + { + "left": "issue:46655", + "right": "issue:46664", + "accept": false, + "reason": "Both are SmolVLM video-processor issues, but one is padding calculation and the other is num_frames semantics; not clearly the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 52, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:36683", + "issue:38617", + "issue:42175", + "issue:42890", + "issue:42907", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43352", + "issue:43502", + "issue:43550", + "issue:43582", + "issue:43825", + "issue:43873", + "issue:43906", + "issue:43976", + "issue:43986", + "issue:43992", + "issue:44062", + "issue:44077", + "issue:44117", + "issue:44162", + "issue:44222", + "issue:44351", + "issue:44368", + "issue:44380", + "issue:44484", + "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", + "issue:46701", + "issue:46710", + "issue:46752" + ], + "soft_pairs": [ + "issue:46298|issue:46752", + "issue:42890|issue:46298", + "issue:46172|issue:46710", + "issue:38617|issue:44351", + "issue:43097|issue:43873", + "issue:44062|issue:44117", + "issue:44117|issue:46701", + "issue:44936|issue:46365", + "issue:43502|issue:43992", + "issue:44704|issue:45103", + "issue:44380|issue:46421", + "issue:43986|issue:44117", + "issue:44938|issue:46178", + "issue:42907|issue:44655", + "issue:42907|issue:45753", + "issue:43825|issue:44655", + "issue:43208|issue:43550", + "issue:43012|issue:43550", + "issue:45712|issue:46298", + "issue:43012|issue:44368", + "issue:43992|issue:44484", + "issue:43066|issue:43906", + "issue:43352|issue:43976", + "issue:43352|issue:44380", + "issue:43097|issue:46298", + "issue:44162|issue:46172", + "issue:44380|issue:46132", + "issue:44077|issue:46365", + "issue:45584|issue:46365", + "issue:44222|issue:45478", + "issue:36683|issue:43582", + "issue:44380|issue:45479", + "issue:42175|issue:46365", + "issue:43064|issue:44380" + ], + "prepared_review_unit_hash": "391e600863c3bde8a91e56732f6acb4405f57c6e7588196531c7ec7746eb2ed8", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37770, + "estimated_input_tokens": 9443, + "estimated_eval_tokens": 19142 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are a loose set of unrelated Transformers issues rather than a duplicate cluster. I found no soft-edge pair that clearly describes the same underlying bug/change, so all candidate links should be rejected.", + "confidence": 0.41, + "canonical_issue_reason": "issue:43873 has the strongest standalone signal in this set: active discussion, inbound references, and a concrete user-facing runtime bug with a clear reproduction around quantization/offloading.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best representative issue here because it is specific, actionable, and broadly impactful compared with the mostly unrelated import/test/config complaints in the rest of the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46298", + "right": "issue:46752", + "accept": false, + "reason": "A noisy logging message is unrelated to an IndexError in apply_chat_template on an empty list." + }, + { + "left": "issue:42890", + "right": "issue:46298", + "accept": false, + "reason": "Test flakiness from missing set_seed is unrelated to logging output." + }, + { + "left": "issue:46172", + "right": "issue:46710", + "accept": false, + "reason": "These involve different failure modes: a model-specific tensor operation bug versus an accuracy regression after upgrade." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and different code paths." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "Removed API warning behavior and quantization offloading failure are different issues." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "Tokenizer AddedToken construction error is unrelated to TOKENIZER_MAPPING_NAMES returning None." + }, + { + "left": "issue:44117", + "right": "issue:46701", + "accept": false, + "reason": "A tokenizer mapping assumption bug is unrelated to a pipeline failure for DiffusionGemma output." + }, + { + "left": "issue:44936", + "right": "issue:46365", + "accept": false, + "reason": "trainer.evaluate() after train() and failing to import Trainer due to datasets are different problems." + }, + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "local_files_only network leakage is unrelated to a missing embed_tokens.weight during UMT5Encoder loading." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "Passing kwargs to cached_file and auto_docstring crashes from future annotations are unrelated code paths." + }, + { + "left": "issue:44380", + "right": "issue:46421", + "accept": false, + "reason": "Attention scaling being ignored and cache lazy initialization causing recompilation are different bugs." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "AutoProcessor/torchvision crash is unrelated to tokenizer mapping returning None." + }, + { + "left": "issue:44938", + "right": "issue:46178", + "accept": false, + "reason": "Python 3.14 import failure and missing BACKENDS_MAPPING entry for tensorflow_text are separate compatibility issues." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Saving dequantized models and saving Pipeline objects both involve serialization, but they are not the same defect." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Model save failure and an infinite-loop generation bug are unrelated." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "A v5 pipeline error message and save_pretrained on Pipeline objects are different behaviors." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "xLSTM training bugs and Bamba torch.compile SDPA failure are model-specific and distinct." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "A precision warning during bf16 compilation is not the same as a compile failure with SDPA." + }, + { + "left": "issue:45712", + "right": "issue:46298", + "accept": false, + "reason": "Dummy class leakage in no-torch environments is unrelated to a logging message." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "These are different warnings with different causes and remediation." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Missing weights on load is unrelated to the default 50GB shard size in save_pretrained." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Wrong tokenizer decoder type and an isolated reproduction of another issue are not the same bug." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "Flash Attention support for a model and Python version compatibility are separate problems." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "They both mention attention backends, but one is missing FA2 support for a model while the other ignores attention scaling config." + }, + { + "left": "issue:43097", + "right": "issue:46298", + "accept": false, + "reason": "Removed tie_embeddings_and_encoder_decoder and noisy logging are unrelated." + }, + { + "left": "issue:44162", + "right": "issue:46172", + "accept": false, + "reason": "ESM2 breakage and a DeepSeek tensor operation issue are unrelated model-specific bugs." + }, + { + "left": "issue:44380", + "right": "issue:46132", + "accept": false, + "reason": "Ignoring GPT2 attention scaling and AttentionInterface.register changing function behavior are different code-path issues." + }, + { + "left": "issue:44077", + "right": "issue:46365", + "accept": false, + "reason": "A patchtsmixer post_init policy issue is unrelated to importing Trainer failing because of datasets." + }, + { + "left": "issue:45584", + "right": "issue:46365", + "accept": false, + "reason": "Whisper empty-transcription generation failure is unrelated to a Trainer import problem." + }, + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "FP8 save_pretrained and Qwen3.5 MoE from_pretrained errors are different save/load bugs." + }, + { + "left": "issue:36683", + "right": "issue:43582", + "accept": false, + "reason": "A missing config attribute and an Apple Silicon TypeError are unrelated." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Attention backend config being ignored is unrelated to a degenerate zero-loss classification setup." + }, + { + "left": "issue:42175", + "right": "issue:46365", + "accept": false, + "reason": "Backend installation packaging and Trainer import interference from datasets are different problems." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "Distributed optimizer state corruption and attention scaling backend behavior are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 53, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:29870", + "issue:36010", + "issue:36296", + "issue:36683", + "issue:38617", + "issue:42490", + "issue:42757", + "issue:42994", + "issue:43010", + "issue:43012", + "issue:43064", + "issue:43208", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43502", + "issue:43525", + "issue:43526", + "issue:43611", + "issue:43618", + "issue:43756", + "issue:43761", + "issue:43824", + "issue:43867", + "issue:43901", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44077", + "issue:44230", + "issue:44380", + "issue:44483", + "issue:44485", + "issue:44492", + "issue:44514", + "issue:44625", + "issue:44655", + "issue:44683", + "issue:44938", + "issue:45216", + "issue:45276", + "issue:45278", + "issue:45310", + "issue:46077", + "issue:46082", + "issue:46365", + "issue:46369" + ], + "soft_pairs": [ + "issue:42757|issue:43824", + "issue:42490|issue:43012", + "issue:43012|issue:44683", + "issue:43208|issue:43526", + "issue:44483|issue:44514", + "issue:44038|issue:44230", + "issue:43618|issue:43761", + "issue:43618|issue:44380", + "issue:38617|issue:43502", + "issue:43756|issue:44485", + "issue:42994|issue:44655", + "issue:43299|issue:43867", + "issue:44625|issue:46365", + "issue:45276|issue:46077", + "issue:36683|issue:43475", + "issue:36010|issue:42757", + "issue:43352|issue:43937", + "issue:44483|issue:46369", + "issue:43611|issue:44655", + "issue:43901|issue:44655", + "issue:36683|issue:44077", + "issue:44625|issue:46077", + "issue:43976|issue:44938", + "issue:43867|issue:45216", + "issue:46077|issue:46082", + "issue:43618|issue:45216", + "issue:36683|issue:43525", + "issue:36683|issue:44625", + "issue:29870|issue:36296", + "issue:43992|issue:45310", + "issue:43064|issue:44492", + "issue:43010|issue:45278" + ], + "prepared_review_unit_hash": "3529b1991b2797db8d4b6671a863072ec280c158a3b3d5aa559a75efb46398a4", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 36940, + "estimated_input_tokens": 9235, + "estimated_eval_tokens": 18726 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These issues are largely heterogeneous: import errors, model-specific config/load regressions, saving bugs, attention/mask regressions, and unrelated training/runtime problems. None of the soft pairs look like the same concrete underlying bug or change, so all soft edges should be rejected.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45278 is the broadest, most representative upgrade-regression report in this set; it best matches the generic import-breakage theme, whereas most others are narrower model- or feature-specific tickets.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is also the best standalone triage candidate because it has a clear user-facing symptom and broad scope; the rest are more specialized or orthogonal.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different symbols, different modules, and likely different root causes." + }, + { + "left": "issue:42490", + "right": "issue:43012", + "accept": false, + "reason": "One is about a trailing-slash local model path warning; the other is a PyTorch warning during bfloat16 compilation." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "Both involve compilation/warnings, but one is a warning on a bfloat16 model while the other is a flex_attention failure on newer torch." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "Completely unrelated subsystems: xLSTM training bugs vs BeitImageProcessorFast label reduction." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "A chat-completions API request failure is not the same bug as a batched processor/chat-template crash." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Same model family, but one is a loading regression and the other is fp8 inference support; not the same fix." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": false, + "reason": "Both are CLIP regressions, but they affect different outputs and different code paths." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "CLIP attentions assignment and GPT2 attention scaling under SDPA/FlashAttention are unrelated bugs." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "ImportError for a missing utility and unintended network requests with local_files_only=True are different problems." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Both mention RoPE, but they concern different models and different implementation issues." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Saving quantized models and saving Pipeline objects are different save_pretrained code paths." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Both are load failures, but one is a Qwen3VL-Moe regression and the other is a state_dict ordering error." + }, + { + "left": "issue:44625", + "right": "issue:46365", + "accept": false, + "reason": "Qwen3.5 label propagation and Trainer import failure are unrelated." + }, + { + "left": "issue:45276", + "right": "issue:46077", + "accept": false, + "reason": "Both are Gemma4 bugs, but one is token embedding resize propagation and the other is bidirectional-attention mask construction." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Different model families and different missing attributes; no shared code path is evident." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they involve different removed/moved symbols in different packages." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "FlashAttention support for a model and an invalid GenerationConfig error are not the same bug." + }, + { + "left": "issue:44483", + "right": "issue:46369", + "accept": false, + "reason": "One is an API endpoint request failure; the other is a Trainer import conflict with datasets." + }, + { + "left": "issue:43611", + "right": "issue:44655", + "accept": false, + "reason": "Base_model_prefix loading regression and Pipeline save_pretrained failure are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:44655", + "accept": false, + "reason": "Documentation mismatch for return_all_scores is not the same as pipeline serialization failing." + }, + { + "left": "issue:36683", + "right": "issue:44077", + "accept": false, + "reason": "Gemma3 config missing vocab_size and patchtsmixer post_init validation are unrelated." + }, + { + "left": "issue:44625", + "right": "issue:46077", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and Gemma4 bidirectional attention are different model-specific regressions." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both mention Python compatibility, but they refer to different versions and different failure modes." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "State_dict-sorting load error and incorrect Qwen3.5 checkpoint saving are different save/load bugs." + }, + { + "left": "issue:46077", + "right": "issue:46082", + "accept": false, + "reason": "Gemma4 attention-mask behavior and LlamaConfig head_dim validation are unrelated." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "CLIP attention output regression and Qwen3.5 save_pretrained checkpoint regression are unrelated." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Different attribute errors in different model configs; no common underlying bug." + }, + { + "left": "issue:36683", + "right": "issue:44625", + "accept": false, + "reason": "Gemma3 vocab_size missing and Qwen3.5 num_labels propagation are separate config issues." + }, + { + "left": "issue:29870", + "right": "issue:36296", + "accept": false, + "reason": "Bart generate() decoder_attention_mask bug and tensor-parallel training bug do not share a concrete code path." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "Both are Qwen3.5-related loading issues, but one is missing embed_tokens weights and the other is a broader from_pretrained regression." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Optimizer-state corruption under FSDP2/PEFT is unrelated to a typo in cache strategy naming." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "A no_grad decoration request for cache updates is unrelated to broad import failures after version upgrade." + } + ] + }, + "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": 23, + "review_unit_index": 54, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:28282", + "issue:29870", + "issue:33453", + "issue:36296", + "issue:36683", + "issue:41669", + "issue:42890", + "issue:42898", + "issue:43010", + "issue:43064", + "issue:43065", + "issue:43502", + "issue:43575", + "issue:43618", + "issue:43723", + "issue:43824", + "issue:43867", + "issue:43976", + "issue:43992", + "issue:44062", + "issue:44075", + "issue:44077", + "issue:44112", + "issue:44222", + "issue:44246", + "issue:44355", + "issue:44360", + "issue:44387", + "issue:44556", + "issue:44734", + "issue:44855", + "issue:44857", + "issue:44938", + "issue:44964", + "issue:45071", + "issue:45125", + "issue:45357", + "issue:45375", + "issue:45440", + "issue:45478", + "issue:45561", + "issue:45600", + "issue:45850", + "issue:46202", + "issue:46285", + "issue:46365", + "issue:46617", + "issue:46664" + ], + "soft_pairs": [ + "issue:44075|issue:46365", + "issue:29870|issue:33453", + "issue:44355|issue:44855", + "issue:44938|issue:45071", + "issue:43867|issue:46202", + "issue:36683|issue:43064", + "issue:43867|issue:45375", + "issue:42890|issue:45561", + "issue:43010|issue:45850", + "issue:41669|issue:44246", + "issue:36296|issue:43575", + "issue:43618|issue:44077", + "issue:43010|issue:43976", + "issue:43992|issue:45478", + "issue:43867|issue:45357", + "issue:28282|issue:33453", + "issue:43618|issue:46285", + "issue:43618|issue:44857", + "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:44222|issue:45125", + "issue:43723|issue:44556", + "issue:44387|issue:44556", + "issue:43824|issue:46617", + "issue:43010|issue:43065" + ], + "prepared_review_unit_hash": "2c7b6ddff1b17aef4bd1eadae5c6a2fde77b395b58e385f1c7b47a589c1f8822", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36363, + "estimated_input_tokens": 9091, + "estimated_eval_tokens": 18438 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a grab-bag of unrelated Transformers issues across import/loading, tokenizer, tensor-parallel, quantization, and CI bugs. The soft pairs share broad keywords or subsystems, but they do not look like the same underlying bug/change, so none should be merged as duplicates.", + "confidence": 0.84, + "canonical_issue_reason": "issue:43824 is the strongest representative issue here: it has clear user impact, multiple inbound references, and a concrete import failure with a specific model/component.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43824 is the best overall issue candidate because it is concrete, widely impactful, and has more evidence/discussion than most others in the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44075", + "right": "issue:46365", + "accept": false, + "reason": "Optimizer arguments not being used is a different bug from importing Trainer failing due to datasets interaction." + }, + { + "left": "issue:29870", + "right": "issue:33453", + "accept": false, + "reason": "Both involve modeling/tokenizer code, but one is a BART generate-time mask bug and the other is a tokenizer-loading regression." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Different failure modes: running compiled Python files versus a Python 3.13 @torch.jit.script parsing/IndentationError issue." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Python 3.14 load failure and PretrainedConfig type-checking regression are unrelated changes." + }, + { + "left": "issue:43867", + "right": "issue:46202", + "accept": false, + "reason": "State-dict ordering load error and Gemma pipeline roundtrip on \u2581 are different model-loading problems." + }, + { + "left": "issue:36683", + "right": "issue:43064", + "accept": false, + "reason": "Missing Gemma3Config.vocab_size is unrelated to the FSDP2+PEFT optimizer-state mismatch." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "A generic load-model state_dict issue is not the same as a missing Qwen3.5 vision config field." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "Both touch testing, but one is missing set_seed() in a model integration test and the other is an xdist file race." + }, + { + "left": "issue:43010", + "right": "issue:45850", + "accept": false, + "reason": "A no_grad decoration request for cache updates is not the same as a broad report that v5 breaks many models." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": false, + "reason": "Removing import * to speed import time is different from intermittent slow import behavior." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Tensor-parallel training bug and tp-related OOM on a huge model are related at a high level but not the same concrete defect." + }, + { + "left": "issue:43618", + "right": "issue:44077", + "accept": false, + "reason": "CLIPOutput attentions assignment and patchtsmixer post_init restriction are unrelated model issues." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Cache/Layer no_grad decoration has no relation to the Python 3.9+ / 3.10+ compatibility issue." + }, + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Missing embed_tokens.weight in UMT5Encoder and Qwen3.5 MoE from_pretrained failure are different loading bugs." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "They both concern loading/saving, but one is a sorted state_dict load error and the other is incorrect visual encoder keys on save_pretrained." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "Missing PyTorch import error is unrelated to a tokenizer-loading regression." + }, + { + "left": "issue:43618", + "right": "issue:46285", + "accept": false, + "reason": "CLIPOutput attentions missing and CLIP text-model structure inconsistency are distinct CLIP-related regressions." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "A CLIPOutput field issue is unrelated to an AMP/CUDA crash in LwDetrImageLoss." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "DSA indexer missing ReLU and DeepSeekV3 implementation divergence do not describe the same bug." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "A flaky test due to missing seed is unrelated to a latest-transformers model-loading failure." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "Seed-related test flakiness and a stale device override failure are different CI issues." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "Test seeding flakiness is unrelated to auto_mappings referencing removed Sam3LiteText configs." + }, + { + "left": "issue:42898", + "right": "issue:46664", + "accept": false, + "reason": "Tokenization_spaces behavior in v5 and SmolVLM num_frames semantics are unrelated API/processor issues." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "local_files_only still making network requests is not the same as AddedToken receiving duplicate special arguments." + }, + { + "left": "issue:44734", + "right": "issue:46617", + "accept": false, + "reason": "KV-cache continuation indexing in serve and trust_remote_code cache FileNotFoundError are different code paths." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 save_pretrained for MoE and missing _tp_plan for Qwen3_5MoeForConditionalGeneration are separate tensor-parallel/serialization bugs." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Tokenizer loading in v5 and checkpoint reload incompatibility after upgrading versions are related only broadly, not the same defect." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Int4 quantization memory growth and checkpoint reload incompatibility are different regressions." + }, + { + "left": "issue:43824", + "right": "issue:46617", + "accept": false, + "reason": "Qwen2_5_VL import failure and a trust_remote_code cache FileNotFoundError are both loading issues but not the same bug." + }, + { + "left": "issue:43010", + "right": "issue:43065", + "accept": false, + "reason": "Cache/Layer no_grad decoration and a dummy Conv2d in Sam3PixelDecoder are unrelated changes." + } + ] + }, + "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": 23, + "review_unit_index": 55, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "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:43824", + "issue:43827", + "issue:43874", + "issue:43901", + "issue:44222", + "issue:44230", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44485", + "issue:44556", + "issue:44655", + "issue:44749", + "issue:44908", + "issue:44961", + "issue:45071", + "issue:45125", + "issue:45362", + "issue:45375", + "issue:45468", + "issue:45478", + "issue:45850", + "issue:46132", + "issue:46143", + "issue:46144", + "issue:46209", + "issue:46617", + "issue:46664", + "issue:46726" + ], + "soft_pairs": [ + "issue:43502|issue:44485", + "issue:43874|issue:46617", + "issue:43576|issue:46132", + "issue:45125|issue:45478", + "issue:30333|issue:33453", + "issue:44355|issue:44749", + "issue:42548|issue:44351", + "issue:45125|issue:45375", + "issue:43010|issue:43232", + "issue:46664|issue:46726", + "issue:30990|issue:31795", + "issue:43502|issue:44961", + "issue:43502|issue:44222", + "issue:43827|issue:44961", + "issue:30333|issue:36296", + "issue:29870|issue:31795", + "issue:43299|issue:44556", + "issue:44355|issue:45071", + "issue:33453|issue:34689", + "issue:44355|issue:44655", + "issue:42757|issue:43576", + "issue:43576|issue:45071", + "issue:44315|issue:44360", + "issue:31795|issue:38617", + "issue:43576|issue:43824", + "issue:44961|issue:45468", + "issue:44351|issue:46617", + "issue:44485|issue:44961", + "issue:44908|issue:44961", + "issue:36296|issue:43901", + "issue:38617|issue:43576", + "issue:46143|issue:46144", + "issue:43475|issue:46617", + "issue:44222|issue:44230", + "issue:41628|issue:42548", + "issue:44315|issue:46209", + "issue:44485|issue:46664", + "issue:45362|issue:45850" + ], + "prepared_review_unit_hash": "d71dc2d3b16e01f7a9c09815a72b8f2d97f14feccd5aa941f9a359a2a195f3db", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 38, + "serialized_chars": 37959, + "estimated_input_tokens": 9490, + "estimated_eval_tokens": 19236 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous: a mix of v5 regressions, import/export failures, model-loading bugs, docs issues, and unrelated runtime problems. The only near-alignment is the Qwen3.5-Moe loading/tensor-parallel pair; the rest do not look like the same underlying bug.", + "confidence": 0.58, + "canonical_issue_reason": "issue:45478 is the most concrete, version-scoped regression ticket in the set and best represents the Qwen3.5-Moe loading failures; the other issues are largely separate import, docs, or runtime problems.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45478 is the strongest standalone issue for triage because it names the affected model family and the regression window clearly, making it the best anchor for this cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "Different bugs: local_files_only causing network access vs a GLM-5 RoPE implementation problem." + }, + { + "left": "issue:43874", + "right": "issue:46617", + "accept": false, + "reason": "Different failures: a missing image-processor method vs a cache loading FileNotFoundError." + }, + { + "left": "issue:43576", + "right": "issue:46132", + "accept": false, + "reason": "Unrelated: a broken env command in v5 vs AttentionInterface.register changing function behavior." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": true, + "reason": "Both point to the same Qwen3.5-Moe loading/tensor-parallel regression on v5.4+, and a single fix could plausibly address both." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "Different areas: MLflow job status reporting vs tokenizer loading regression." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Unrelated: compiled Python-file errors vs a slowdown while filtering data after upgrade." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but the missing symbols are different and there is no clear evidence of the same underlying export bug." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Different Qwen3.5-Moe issues: missing tensor-parallel plan vs a config field being silently dropped." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "Related generation/cache themes, but the specific bugs differ: no_grad decoration vs post-sync_gpus generation kwargs handling." + }, + { + "left": "issue:46664", + "right": "issue:46726", + "accept": false, + "reason": "Different vision/video processor issues: num_frames semantics vs missing vision backend handling in AutoProcessor." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "One is a Sentence Transformers load hang; the other is confusing forward-argument documentation." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "No relation: remote API requests despite local_files_only vs an unrelated one-word issue title." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "Different bugs: local_files_only network access vs FP8 save_pretrained on MoE models." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs about pipeline() removal vs an unrelated one-word issue." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLflow job reporting bug is unrelated to tensor-parallel training." + }, + { + "left": "issue:29870", + "right": "issue:31795", + "accept": false, + "reason": "Different topics: BART generation UnboundLocalError vs docs confusion around model.forward arguments." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Both mention version regressions, but one is Qwen3.5-VL MoE loading and the other is checkpoint reload compatibility; not the same bug." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled-file execution errors are unrelated to PretrainedConfig type-checking regressions." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Both are loading regressions, but they affect different models and likely different code paths." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Different failure modes: compiled Python files vs inability to save Pipeline objects." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Importing is_offline_mode from huggingface_hub is unrelated to the transformers env command bug." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "A broken env command and PretrainedConfig type checking are separate issues." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger Kernel model_init behavior and a DSA indexer discussion are not the same bug." + }, + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Docs confusion in model.forward is unrelated to an import error for layer_type_validation." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "A broken env command and a missing Qwen2_5_VLForConditionalGeneration export are different failures." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "No evidence of relation: placeholder title vs Gemma-4 audio positional encoding bug." + }, + { + "left": "issue:44351", + "right": "issue:46617", + "accept": false, + "reason": "Missing HybridCache export and trust_remote_code cache FileNotFoundError are different problems." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "GLM-5 RoPE implementation issue is unrelated to the placeholder racoon issue." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs handling is unrelated to the placeholder racoon issue." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor-parallel training bug is not the same as TextClassificationPipeline docs mentioning return_all_scores." + }, + { + "left": "issue:38617", + "right": "issue:43576", + "accept": false, + "reason": "Missing layer_type_validation import is unrelated to the env command regression." + }, + { + "left": "issue:46143", + "right": "issue:46144", + "accept": false, + "reason": "Both concern RoFormer, but one is missing kwargs passthrough and the other is attention-interface integration; not the same concrete bug." + }, + { + "left": "issue:43475", + "right": "issue:46617", + "accept": false, + "reason": "SAM 3 video attribute error and trust_remote_code cache loading failure are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "FP8 save_pretrained on MoE and FP8 inference support for Qwen3-VL/Qwen3.5 are related only at a broad subsystem level, not the same bug." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Different import errors: AutoImageProcessor vs PreTrainedModel." + }, + { + "left": "issue:44315", + "right": "issue:46209", + "accept": false, + "reason": "Liger Kernel model_init behavior is unrelated to FP8 inference on B200." + }, + { + "left": "issue:44485", + "right": "issue:46664", + "accept": false, + "reason": "GLM-5 RoPE implementation and SmolVLM num_frames semantics are different bugs." + }, + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "Different scope: Qwen3.5-35B chat crash vs a broad complaint that v5 breaks many models." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded and suitably conservative. The packet supports treating issue:45478 as the best anchor for the Qwen3.5-Moe regression cluster, and the summary correctly avoids claiming stronger links than the titles/evidence show. The only plausible same-bug pairing is 45125/45478; the other soft-edge rejects look 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": 23, + "review_unit_index": 56, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:31795", + "issue:33453", + "issue:41669", + "issue:42371", + "issue:42890", + "issue:43352", + "issue:43408", + "issue:43525", + "issue:43572", + "issue:43576", + "issue:43673", + "issue:43827", + "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:44492", + "issue:44556", + "issue:44623", + "issue:44704", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:45003", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45278", + "issue:45341", + "issue:45468", + "issue:45478", + "issue:45542", + "issue:45753", + "issue:45850", + "issue:46095", + "issue:46218", + "issue:46328", + "issue:46664" + ], + "soft_pairs": [ + "issue:43673|issue:44492", + "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: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:41669|issue:46218", + "issue:44230|issue:46095", + "issue:43408|issue:44315", + "issue:42890|issue:44908", + "issue:44246|issue:44273", + "issue:43576|issue:44908", + "issue:43352|issue:44393", + "issue:43976|issue:44393", + "issue:44704|issue:44908", + "issue:44908|issue:45071", + "issue:44297|issue:45230", + "issue:44295|issue:45230", + "issue:41669|issue:44351", + "issue:44492|issue:46328", + "issue:41669|issue:45341", + "issue:43976|issue:45850", + "issue:44393|issue:45478", + "issue:43976|issue:46218", + "issue:44077|issue:44861", + "issue:43827|issue:44908", + "issue:31795|issue:33453", + "issue:43576|issue:44273", + "issue:44315|issue:44829", + "issue:44360|issue:45468" + ], + "prepared_review_unit_hash": "c584e256b073708eb389f301c6d6c608b97f96443feb0b033104114646432520", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 40, + "serialized_chars": 38554, + "estimated_input_tokens": 9639, + "estimated_eval_tokens": 19534 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a grab-bag of distinct Transformers issues: docs, import/lazy-loading, v5 migration regressions, and model-specific bugs. The soft-similarity pairs mostly share broad keywords or subsystems, but not the same concrete defect, so none should be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": "No true canonical issue stands out: the items are mostly unrelated bug reports, and the similarity pairs do not share a single concrete root cause. Issue 45850 is the broadest regression report, but it is too generic to serve as a reliable canonical target.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45850 is the broadest umbrella-style report, but it is still too vague to be a good deduplication anchor for this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Different problems: chunked_prefill cache missing vs a cache-strategy typo; only the word \"cache\" overlaps." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Unrelated subsystems: lazy loading behavior vs TensorFlow backend detection with tensorboard installed." + }, + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "No concrete overlap; one is missing processor save files, the other is an unspecified generic bug report." + }, + { + "left": "issue:44360", + "right": "issue:46664", + "accept": false, + "reason": "Different model/code-path issues: DSA ReLU discussion vs video processor num_frames semantics." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Same family name, but different defects: fp8 inference support vs missing tensor-parallel plan." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": false, + "reason": "Lazy loading/performance problem vs unsafe sys.modules access; not the same bug." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "Qwen3-VL bbox output issue vs PretrainedConfig type-checking regression; unrelated root causes." + }, + { + "left": "issue:44393", + "right": "issue:45753", + "accept": false, + "reason": "Both involve Qwen3-VL, but one is bbox output errors and the other is an infinite-loop case on a specific image." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import-star cleanup/performance issue vs missing GenerationMixin cache in chunked prefill; not the same defect." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Both are v5 regressions, but one is the env command and the other is checkpoint reload compatibility." + }, + { + "left": "issue:43525", + "right": "issue:45850", + "accept": false, + "reason": "Specific Llama4Config attribute bug vs a very broad \"many models break\" complaint; not the same issue." + }, + { + "left": "issue:43572", + "right": "issue:45850", + "accept": false, + "reason": "StableLm pad_token_idx regression is model-specific and distinct from the broad v5 breakage report." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Both concern upgrade incompatibility, but one is checkpoint reload after v5.2/v5.3 and the other is remote-code meta-init incompatibility." + }, + { + "left": "issue:44492", + "right": "issue:45850", + "accept": false, + "reason": "A typo in cache strategies is not the same as a broad claim that many models break in v5." + }, + { + "left": "issue:45278", + "right": "issue:45850", + "accept": false, + "reason": "Many import errors after upgrading is broader, but still a different report from the generic \"too many models break\" complaint." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 API guidance vs cache-strategy typo; different topics entirely." + }, + { + "left": "issue:41669", + "right": "issue:46218", + "accept": false, + "reason": "Import-time slowdown from import * removal is different from a module import failure for Qwen3_5ForCausalLM." + }, + { + "left": "issue:44230", + "right": "issue:46095", + "accept": false, + "reason": "FP8 inference support request vs weight-conversion tooling question; different changes." + }, + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "Sam model type warning vs Liger Kernel not applied with model_init; unrelated bugs." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "Test flakiness from missing set_seed() vs scheduler kwargs being ignored; different failure modes." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Import slowdown/timing issue vs lazy loading not working properly; related area, but not the same concrete bug." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "Env command regression and inverse_sqrt scheduler kwargs issue are unrelated." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Nemotron Flash Attention 2 support gap vs Qwen3-VL bbox hallucination; no shared root cause." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Python version compatibility problem vs Qwen3-VL output bug; unrelated." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "AutoProcessor cached_file kwargs forwarding vs inverse_sqrt scheduler kwargs; different APIs and symptoms." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwarg handling vs PretrainedConfig type checking; no real overlap." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "Tokenizer save_pretrained mismatch vs generic bug report; nothing concrete matches." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "position_ids buffer access error vs generic bug report; not a duplicate pair." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-star cleanup/performance issue vs missing HybridCache import; different import-path bugs." + }, + { + "left": "issue:44492", + "right": "issue:46328", + "accept": false, + "reason": "Cache-strategy typo vs SequenceFeatureExtractor.pad inefficiency; unrelated." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Import-star removal discussion vs a bug in testing_utils; no shared defect." + }, + { + "left": "issue:43976", + "right": "issue:45850", + "accept": false, + "reason": "Python support mismatch vs broad v5 breakage complaint; different problem classes." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3-VL hallucination/error vs Qwen3.5 Moe from_pretrained failure; same family but distinct issues." + }, + { + "left": "issue:43976", + "right": "issue:46218", + "accept": false, + "reason": "Python compatibility issue vs module import failure; unrelated." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "patchtsmixer post_init policy issue vs _tied_weight_keys AttributeError; different code paths." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs still referencing pipeline() vs scheduler kwarg handling; docs cleanup is not the same as runtime scheduler bug." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Model.forward docs confusion vs tokenizer loading regression; unrelated." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Env command regression vs lazy loading not working; same broad v5 area, but not the same bug." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel not applied with model_init vs FlashAttention 3 degenerate training; different implementations and symptoms." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "DSA/ReLU discussion vs Gemma-4 audio positional encoding bug; 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": 23, + "review_unit_index": 57, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "nodes": [ + "issue:28348", + "issue:29870", + "issue:30333", + "issue:31515", + "issue:31795", + "issue:34689", + "issue:36296", + "issue:38617", + "issue:41669", + "issue:42371", + "issue:42913", + "issue:43066", + "issue:43298", + "issue:43352", + "issue:43502", + "issue:43519", + "issue:43525", + "issue:43572", + "issue:43576", + "issue:43606", + "issue:43618", + "issue:43704", + "issue:43717", + "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", + "issue:45468", + "issue:45850", + "issue:46018", + "issue:46095", + "issue:46144", + "issue:46172", + "issue:46218", + "issue:46655" + ], + "soft_pairs": [ + "issue:43519|issue:46655", + "issue:44273|issue:44908", + "issue:44655|issue:44861", + "issue:43576|issue:46218", + "issue:45230|issue:45310", + "issue:43519|issue:44485", + "issue:43519|issue:44908", + "issue:41669|issue:43704", + "issue:31795|issue:34689", + "issue:46018|issue:46172", + "issue:43704|issue:44485", + "issue:44485|issue:45468", + "issue:43502|issue:43519", + "issue:45850|issue:46218", + "issue:42371|issue:43704", + "issue:43519|issue:44623", + "issue:43606|issue:45468", + "issue:44263|issue:45468", + "issue:44261|issue:45468", + "issue:43298|issue:46095", + "issue:43519|issue:43704", + "issue:44351|issue:46218", + "issue:43352|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:43525|issue:44861", + "issue:43572|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" + ], + "prepared_review_unit_hash": "39d7b8829c222439a3a96883c418ddf542a14a6475685a60251e594fcb5f2d89", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 43, + "serialized_chars": 39687, + "estimated_input_tokens": 9922, + "estimated_eval_tokens": 20100 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous, with one clear duplicate pair around the tokenizer v5 regression. Most other soft pairs share broad subsystem vocabulary but do not describe the same concrete bug or change.", + "confidence": 0.81, + "canonical_issue_reason": "issue:43066 is the more specific, actionable wording of the tokenizer v5 regression and best represents the duplicate pair with issue:42913.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43066 is the strongest representative issue overall because it states the concrete failure mode directly, rather than a broad behavioral complaint.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43519", + "right": "issue:46655", + "accept": false, + "reason": "Both are processor-related calculation bugs, but they affect different models and different code paths (timestamp vs padding)." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading and inverse_sqrt scheduler kwargs are unrelated failures." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Saving Pipeline objects and _get_tied_weight_keys are separate save/load internals, not the same bug." + }, + { + "left": "issue:43576", + "right": "issue:46218", + "accept": false, + "reason": "One is the env CLI command, the other is an import failure for Qwen3.5; no shared underlying defect." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "Generic bug report vs a specific Qwen3.5 MoE from_pretrained regression are not the same issue." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Different model families and different bugs: Qwen3VL timestamp math vs GLM-5 RoPE implementation." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Processor timestamp calculation and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import-star cleanup/performance and a VRAM leak in Qwen3ForCausalLM are different changes." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Documentation confusion about forward inputs does not match a Llama 3.2 model-loading breakage." + }, + { + "left": "issue:46018", + "right": "issue:46172", + "accept": false, + "reason": "DeepSeek-V4 shared-expert gating and a comb.to(dtype).transpose bug are distinct implementation problems." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "A VRAM leak in dataloader threads is unrelated to GLM-5 RoPE behavior." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "Different models and different numerical/positional encoding issues." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "local_files_only network behavior is unrelated to Qwen3VL timestamp calculation." + }, + { + "left": "issue:45850", + "right": "issue:46218", + "accept": false, + "reason": "A broad v5 compatibility complaint is not the same as a specific Qwen3.5 import failure." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 configuration guidance and a VRAM leak are unrelated." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Processor timestamp math and processor.save_pretrained file emission are different bugs." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch for bark-small is unrelated to Gemma-4 audio positional encoding." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "torch.split return handling in GlmMoeDsaIndexer does not match Gemma-4 audio positional encoding." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "Missing rms_norm_eps in MLA config is a different precision issue than Gemma-4 audio positional encoding." + }, + { + "left": "issue:43298", + "right": "issue:46095", + "accept": false, + "reason": "Continuous batching for audio models and DeepSeek-V4 weight conversion are not the same change." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Different subsystems and failure modes: processor timestamp math vs VRAM leak." + }, + { + "left": "issue:44351", + "right": "issue:46218", + "accept": false, + "reason": "Missing HybridCache import and missing Qwen3_5ForCausalLM import are different import regressions." + }, + { + "left": "issue:43352", + "right": "issue:46218", + "accept": false, + "reason": "Flash Attention 2.0 support for NemotronHForCausalLM is unrelated to Qwen3.5 import errors." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": true, + "reason": "Both describe the same tokenizer v5 regression; 43066 is the more specific statement of the bug." + }, + { + "left": "issue:28348", + "right": "issue:45250", + "accept": false, + "reason": "Both mention Flash Attention 2.0, but one is GPT2-specific support and the other is a generic topic; not clearly the same concrete change." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwargs handling and an unrelated placeholder bug report do not match." + }, + { + "left": "issue:44492", + "right": "issue:44998", + "accept": false, + "reason": "A typo in cache strategies is not the same as the unrelated placeholder issue." + }, + { + "left": "issue:43618", + "right": "issue:44998", + "accept": false, + "reason": "CLIPOutput attentions assignment is unrelated to the placeholder issue." + }, + { + "left": "issue:44998", + "right": "issue:45850", + "accept": false, + "reason": "A placeholder issue is not the same as a broad v5 regression complaint." + }, + { + "left": "issue:44938", + "right": "issue:44998", + "accept": false, + "reason": "Python 3.14 load failure and the placeholder issue are unrelated." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Tied-weight key handling and chat template crashes are separate code paths." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "Slow from_pretrained checkpoint loading and an ImportError for layer_type_validation are different problems." + }, + { + "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:43572", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig is a different config issue than tied-weight key handling." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 model support and processor.save_pretrained file output are unrelated." + }, + { + "left": "issue:28348", + "right": "issue:43352", + "accept": false, + "reason": "Both mention Flash Attention 2.0, but they target different models and different support requests." + }, + { + "left": "issue:29870", + "right": "issue:31515", + "accept": false, + "reason": "BART decoder_attention_mask generate crash and slow checkpoint loading are unrelated." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLFlow reporting on failed training jobs is unrelated to from_pretrained load latency." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "Slow loading and tensor parallel training bugs are different issues." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Load performance and documentation confusion are not the same bug." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 inference support and Qwen3.5 MoE from_pretrained errors are different requests." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Audio continuous batching and Qwen3-VL/Qwen3.5 FP8 support are unrelated." + }, + { + "left": "issue:43717", + "right": "issue:46144", + "accept": false, + "reason": "Mamba init_weights usage and RoFormer attention interface usage are unrelated implementation issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet: it correctly centers the tokenizer v5 duplicate pair and avoids claiming stronger relationships for the other soft pairs. The soft-edge verdicts are conservative overall, with no obvious overstatement of evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 23, + "review_unit_index": 58, + "review_unit_count": 58, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-484", + "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": "ffb51e8be7b4b67329940bc1647e4b8eae0eb44ce3d2a60e1236e34992b11ce5", + "component_budget": { + "node_count": 484, + "item_count": 484, + "soft_pair_count": 1779, + "serialized_chars": 922739, + "estimated_input_tokens": 230685, + "estimated_eval_tokens": 461626 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 11, + "serialized_chars": 7772, + "estimated_input_tokens": 1943, + "estimated_eval_tokens": 4142 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: the issues describe different bugs/features across unrelated models and components (normalization, Mask2Former loss handling, config typing, tokenizer warnings, DeepSeek routing, RoFormer attention interface, etc.). The text similarity is superficial and does not indicate shared underlying defects.", + "confidence": 0.96, + "canonical_issue_reason": "No safe canonical issue: the items span distinct code paths and model families, so there is no clear duplicate center.", + "canonical_pr_reason": "No PRs in the cluster.", + "best_issue_reason": "No single issue is a strong global representative; the cluster does not form one coherent bug family.", + "best_pr_reason": "No PRs in the cluster.", + "soft_edge_verdicts": [ + { + "left": "issue:42936", + "right": "issue:46018", + "accept": false, + "reason": "Different bugs in different models: 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 issue; the other is a tokenizer regex warning regression. No shared code path." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Mask2Former ignore_value handling is unrelated to kwargs not being forwarded to a loss function." + }, + { + "left": "issue:43240", + "right": "issue:46144", + "accept": false, + "reason": "Loss-function kwarg plumbing and RoFormer attention-interface usage are different problems." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "Mask2Former ignore_value bug and tokenizer warning regression are unrelated subsystems." + }, + { + "left": "issue:42936", + "right": "issue:46144", + "accept": false, + "reason": "Mask2Former loss behavior and RoFormer attention implementation do not describe the same bug." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "Model loss/ignore_value handling is unrelated to a typo in a quantization attribute name." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Passing kwargs into loss calculation is unrelated to ModernBERTConfig norm_eps typing." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "Config type annotation and quantization-attribute typo are distinct documentation/code issues." + }, + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "Possible normalization wording overlap only; one asks about Llama4TextL2Norm semantics, the other is a config type hint." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "Mask2Former ignore_value usage and ModernBERTConfig type hint 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": 23, + "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-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": "122812cd3ecf49fce08fe0b93f6bbba0e12a994e77083104f9d49a61a28953fc", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2877, + "estimated_input_tokens": 720, + "estimated_eval_tokens": 1696 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2877, + "estimated_input_tokens": 720, + "estimated_eval_tokens": 1696 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These three issues are not duplicates: they cover unrelated feature requests in different parts of the product (tokenizer truncation, learning-rate scheduling, and Wav2Vec2 word offsets). The soft similarities are only at a generic \"request for capability\" level.", + "confidence": 0.95, + "canonical_issue_reason": "Issue #10519 is the most concrete and specific of the set, with a clear user-facing behavior change request and the most actionable title/body among the three.", + "canonical_pr_reason": null, + "best_issue_reason": "#10519 is the best representative issue only in the sense that it is the clearest standalone request; however, the cluster itself does not form a true duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:10519", + "right": "issue:11307", + "accept": false, + "reason": "Different features and code paths: tokenizer truncation behavior vs. Wav2Vec2 word boundary time offsets." + }, + { + "left": "issue:10519", + "right": "issue:11005", + "accept": false, + "reason": "Unrelated requests: truncation strategy in tokenization vs. ReduceLROnPlateau-like scheduler functionality." + }, + { + "left": "issue:11005", + "right": "issue:11307", + "accept": false, + "reason": "No shared underlying bug/change; one is about training schedule behavior, the other about speech model alignment metadata." + } + ] + }, + "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": 23, + "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-13244-3", + "nodes": [ + "issue:13244", + "issue:15980", + "issue:16618" + ], + "soft_pairs": [ + "issue:13244|issue:16618", + "issue:15980|issue:16618" + ], + "prepared_review_unit_hash": "2ffdc7934073480c46fe5585c89b54f77a2e9c18ea2d3dc273e72d8e4d6c1afe", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2476, + "estimated_input_tokens": 619, + "estimated_eval_tokens": 1494 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2476, + "estimated_input_tokens": 619, + "estimated_eval_tokens": 1494 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The three issues are only loosely text-similar and do not describe the same bug. #16618 is a generic model-loading failure, #15980 is specifically about the error message for unauthorized private-model downloads, and #13244 concerns TAPAS tokenization mismatch vs TensorFlow.", + "confidence": 0.95, + "canonical_issue_reason": "#16618 is the broadest and most central user-facing failure in this set: it targets model loading itself, whereas the other two are unrelated edge cases in tokenization and auth/error messaging.", + "canonical_pr_reason": null, + "best_issue_reason": "#16618 is the best representative issue because it describes the core symptom most likely to map to a real code-path problem in the loader, while the others are narrower and unrelated.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Tokenization mismatch for TAPAS and inability to load bert-base-uncased are different subsystems and different bugs." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "Private-model auth/download error messaging is not the same underlying issue as a general failure to load bert-base-uncased." + } + ] + }, + "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": 23, + "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-26350-2", + "nodes": [ + "issue:26350", + "issue:33260" + ], + "soft_pairs": [ + "issue:26350|issue:33260" + ], + "prepared_review_unit_hash": "2f6e595ea72d61d7a6f557babd8558777582282c610e8ea1efae8be5a31b3030", + "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 not duplicates: one requests Flash Attention 2 support across more architectures, the other requests GGUF support across more architectures. The overlap is only the generic phrasing about architectures.", + "confidence": 0.98, + "canonical_issue_reason": "No single canonical issue: the two issues target different features/backends and do not describe the same underlying work.", + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a strong global representative for the pair because they are unrelated feature requests rather than duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Different integration targets: Flash Attention 2 vs GGUF. Shared wording about supporting more architectures is superficial." + } + ] + }, + "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": 23, + "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-27088-8", + "nodes": [ + "issue:27088", + "issue:28936", + "issue:38435", + "issue:43834", + "issue:44365", + "issue:44370", + "issue:46222", + "issue:46226" + ], + "soft_pairs": [ + "issue:27088|issue:44370", + "issue:27088|issue:44365", + "issue:27088|issue:43834", + "issue:27088|issue:38435", + "issue:27088|issue:28936", + "issue:28936|issue:43834", + "issue:28936|issue:44365", + "issue:28936|issue:38435", + "issue:28936|issue:44370", + "issue:38435|issue:44370", + "issue:38435|issue:44365", + "issue:38435|issue:43834", + "issue:38435|issue:46222", + "issue:38435|issue:46226", + "issue:43834|issue:44370", + "issue:43834|issue:44365", + "issue:43834|issue:46222", + "issue:43834|issue:46226", + "issue:44365|issue:44370", + "issue:44365|issue:46222", + "issue:44365|issue:46226", + "issue:44370|issue:46222", + "issue:44370|issue:46226", + "issue:46222|issue:46226" + ], + "prepared_review_unit_hash": "6bddc2b6df4b2776a3b5a3dc3172e23d059b3261fa0566a242c93ce8a81f2b95", + "component_budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 24, + "serialized_chars": 13526, + "estimated_input_tokens": 3382, + "estimated_eval_tokens": 7020 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 24, + "serialized_chars": 13526, + "estimated_input_tokens": 3382, + "estimated_eval_tokens": 7020 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a mixed set of i18n translation-request issues in different languages, so most high-similarity pairs are not true duplicates. The only clear duplicate groups are the identical generic-template issues (43834/44365/44370) and the identical Thai-title pair (46222/46226).", + "confidence": 0.88, + "canonical_issue_reason": "No single canonical for the full cluster because the items span different language requests. If consolidating the exact-title duplicate subgroup, issue:43834 is the best anchor because it is the earliest of the generic-template duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43834 is the oldest and most representative of the exact-match generic i18n template subgroup, but it does not represent the whole mixed cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Different language request (Turkish vs generic placeholder); shared i18n wording is just template overlap." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "Different language request; similarity is from the common translation template, not the same issue." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "Turkish request vs generic template issue; not the same underlying request." + }, + { + "left": "issue:27088", + "right": "issue:38435", + "accept": false, + "reason": "Distinct translation targets (Turkish vs Romanian); same broad docs-i18n theme only." + }, + { + "left": "issue:27088", + "right": "issue:28936", + "accept": false, + "reason": "Turkish and Spanish are separate translation requests, not duplicates." + }, + { + "left": "issue:28936", + "right": "issue:43834", + "accept": false, + "reason": "Spanish request vs generic template issue; similarity is superficial." + }, + { + "left": "issue:28936", + "right": "issue:44365", + "accept": false, + "reason": "Spanish request and generic-template issue are different underlying requests." + }, + { + "left": "issue:28936", + "right": "issue:38435", + "accept": false, + "reason": "Spanish vs Romanian translation requests are distinct." + }, + { + "left": "issue:28936", + "right": "issue:44370", + "accept": false, + "reason": "Spanish request vs generic template issue; not a true duplicate." + }, + { + "left": "issue:38435", + "right": "issue:44370", + "accept": false, + "reason": "Romanian request vs generic template issue; shared template language does not mean duplicate." + }, + { + "left": "issue:38435", + "right": "issue:44365", + "accept": false, + "reason": "Romanian request vs generic template issue; different underlying language target." + }, + { + "left": "issue:38435", + "right": "issue:43834", + "accept": false, + "reason": "Romanian request and generic-template issue are not the same request." + }, + { + "left": "issue:38435", + "right": "issue:46222", + "accept": false, + "reason": "Different language requests; the high similarity is from the issue template, not a shared bug/change." + }, + { + "left": "issue:38435", + "right": "issue:46226", + "accept": false, + "reason": "Romanian vs Thai translation request; not the same underlying issue." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": true, + "reason": "Same generic placeholder title and near-identical issue text; looks like the same translation request duplicated." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": true, + "reason": "Same generic placeholder title/content; these appear to be duplicate submissions of the same issue." + }, + { + "left": "issue:43834", + "right": "issue:46222", + "accept": false, + "reason": "Titles differ completely; this looks like a different translation request, despite template similarity." + }, + { + "left": "issue:43834", + "right": "issue:46226", + "accept": false, + "reason": "Different title and likely different language request; not the same issue." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": true, + "reason": "Exact same generic placeholder title and very similar body, so this is a clear duplicate pair." + }, + { + "left": "issue:44365", + "right": "issue:46222", + "accept": false, + "reason": "Different titles and different request language; template overlap only." + }, + { + "left": "issue:44365", + "right": "issue:46226", + "accept": false, + "reason": "Not the same underlying translation request." + }, + { + "left": "issue:44370", + "right": "issue:46222", + "accept": false, + "reason": "Different title/language target; the similarity is only from the shared i18n template." + }, + { + "left": "issue:44370", + "right": "issue:46226", + "accept": false, + "reason": "Different translation requests; not a true duplicate." + }, + { + "left": "issue:46222", + "right": "issue:46226", + "accept": true, + "reason": "Identical Thai title and same-size content; this looks like a direct duplicate submission." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: the language-specific issues are clearly distinct by title, while the two duplicate subgroups are supported by identical titles and very high similarity/near-identical lengths. The soft-edge verdicts are conservative, and the canonical choice is limited appropriately to the exact generic-template subgroup rather than the whole mixed cluster." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 6, + "component_count": 23, + "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-36817-18", + "nodes": [ + "issue:36817", + "issue:36879", + "issue:37015", + "issue:41115", + "issue:41211", + "issue:42503", + "issue:42738", + "issue:43671", + "issue:43837", + "issue:43909", + "issue:44121", + "issue:44205", + "issue:45027", + "issue:45177", + "issue:45322", + "issue:46133", + "issue:46645", + "issue:46653" + ], + "soft_pairs": [ + "issue:36817|issue:41211", + "issue:36817|issue:43671", + "issue:36817|issue:36879", + "issue:36879|issue:41211", + "issue:36879|issue:43909", + "issue:36879|issue:44121", + "issue:36879|issue:43837", + "issue:36879|issue:43671", + "issue:37015|issue:43837", + "issue:37015|issue:44121", + "issue:37015|issue:43909", + "issue:36879|issue:37015", + "issue:37015|issue:42503", + "issue:41115|issue:43837", + "issue:41115|issue:44121", + "issue:41115|issue:42503", + "issue:41115|issue:45027", + "issue:41211|issue:43671", + "issue:42503|issue:43837", + "issue:42503|issue:43671", + "issue:42503|issue:42738", + "issue:42738|issue:45027", + "issue:42738|issue:46653", + "issue:43671|issue:43837", + "issue:43671|issue:45027", + "issue:43671|issue:45177", + "issue:43671|issue:44121", + "issue:43671|issue:43909", + "issue:43837|issue:43909", + "issue:43837|issue:45322", + "issue:43837|issue:44121", + "issue:43909|issue:44205", + "issue: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": "3f9b100a06f06a21f6f19e83e3f4a07c287cebee42f8a8e2f866f183815d2efb", + "component_budget": { + "node_count": 18, + "item_count": 18, + "soft_pair_count": 40, + "serialized_chars": 24145, + "estimated_input_tokens": 6037, + "estimated_eval_tokens": 12330 + }, + "budget": { + "node_count": 18, + "item_count": 18, + "soft_pair_count": 40, + "serialized_chars": 24145, + "estimated_input_tokens": 6037, + "estimated_eval_tokens": 12330 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is not a duplicate set; it mixes many unrelated model-request issues across different model families and modalities. All soft edges should be rejected.", + "confidence": 0.72, + "canonical_issue_reason": "No true canonical duplicate stands out. If one issue must anchor the cluster, 42738 is the broadest umbrella request ('BERT-like models with RoPE'), but the items are still separate model requests rather than duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "42738 is the closest representative of the general model-support theme, but only as an umbrella issue; the rest are distinct requests and should not be merged as duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "EuroBert config support vs DEIMv2: different model families and different requests." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "EuroBert config support vs Qwen3-TTS: unrelated model families and modalities." + }, + { + "left": "issue:36817", + "right": "issue:36879", + "accept": false, + "reason": "EuroBert vs RF-DETR are different model additions, not the same underlying change." + }, + { + "left": "issue:36879", + "right": "issue:41211", + "accept": false, + "reason": "RF-DETR and DEIMv2 are distinct vision-model requests." + }, + { + "left": "issue:36879", + "right": "issue:43909", + "accept": false, + "reason": "RF-DETR vs LFM2.5 Audio is a vision-model request versus an audio-model request." + }, + { + "left": "issue:36879", + "right": "issue:44121", + "accept": false, + "reason": "RF-DETR and weight-sparse transformer are different code paths and model families." + }, + { + "left": "issue:36879", + "right": "issue:43837", + "accept": false, + "reason": "RF-DETR vs Qwen3-ASR: different models and different modality support." + }, + { + "left": "issue:36879", + "right": "issue:43671", + "accept": false, + "reason": "RF-DETR vs Qwen3-TTS: different model families and different modality." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT and Qwen3-ASR are unrelated requests." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT vs weight-sparse transformer: different model changes, not one fix." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "NeoBERT and LFM2.5 Audio are different model families/modalities." + }, + { + "left": "issue:36879", + "right": "issue:37015", + "accept": false, + "reason": "RF-DETR and NeoBERT are separate model additions." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT vs ModernVBERT are different BERT-like models, not the same request." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 architecture support vs Qwen3-ASR support are unrelated changes." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 architecture support and sparse-transformer support are different requests." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "MiniCPM3 and ModernVBERT are distinct model additions." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "MiniCPM3 architecture work vs Voxtral TTS support are different model changes." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 and Qwen3-TTS are different model families and modalities." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT and Qwen3-ASR are unrelated model-support requests." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT vs Qwen3-TTS: different models and different modality support." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "ModernVBERT is a specific model request; 42738 is a broad architectural feature request, so they are not the same change." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "BERT-like RoPE support vs Voxtral TTS support are unrelated." + }, + { + "left": "issue:42738", + "right": "issue:46653", + "accept": false, + "reason": "Generic model-architecture request vs an unrelated name-only issue." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "Qwen3-TTS and Qwen3-ASR are related in area but are still separate support requests for different capabilities." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-TTS and Voxtral TTS are different model families, so not duplicates." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "TTS support and OCR2 support are different modalities and code paths." + }, + { + "left": "issue:43671", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-TTS support vs sparse transformer support are unrelated changes." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS support and LFM2.5 Audio are different audio-model requests." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-ASR and LFM2.5 Audio are distinct model requests." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "ASR support vs EUPE are unrelated model additions." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-ASR support and sparse-transformer support are different changes." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "LFM2.5 Audio and SAM3-LiteText are different model families and tasks." + }, + { + "left": "issue:43909", + "right": "issue:46645", + "accept": false, + "reason": "Audio-model support vs Granite SWA+sink option are different code changes." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Weight-sparse transformer and EUPE are separate model requests." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral TTS are different model families; similar domain, not the same fix." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT vs Voxtral TTS are unrelated." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "ASR support vs OCR support are different modalities and code paths." + }, + { + "left": "issue:45322", + "right": "issue:46133", + "accept": false, + "reason": "EUPE and TIPSv2 are separate model requests with no shared underlying bug or change." + }, + { + "left": "issue:42503", + "right": "issue:46653", + "accept": false, + "reason": "ModernVBERT and the name-only issue are unrelated." + }, + { + "left": "issue:45027", + "right": "issue:46653", + "accept": false, + "reason": "Voxtral TTS support and the name-only issue 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": 7, + "component_count": 23, + "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-38740-2", + "nodes": [ + "issue:38740", + "issue:46443" + ], + "soft_pairs": [ + "issue:38740|issue:46443" + ], + "prepared_review_unit_hash": "bc3aaaab09893a71d032b933ea941b55330b1ba76e152f3ef31475c87e7c6ee1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1505, + "estimated_input_tokens": 377, + "estimated_eval_tokens": 1010 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1505, + "estimated_input_tokens": 377, + "estimated_eval_tokens": 1010 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are not duplicates: one is a docs request about adding Pruna as an optimization framework, and the other is a vision RoPE refactor. They target different areas and different kinds of work.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue for this pair; the issues describe unrelated topics and do not share a concrete bug or change.", + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a good canonical representative for the other because they are about different domains and different requested work.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:38740", + "right": "issue:46443", + "accept": false, + "reason": "Titles and descriptions point to unrelated work: documentation for an optimization framework vs a vision RoPE refactor. No shared code path or same underlying issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 8, + "component_count": 23, + "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-41084-2", + "nodes": [ + "issue:41084", + "issue:43037" + ], + "soft_pairs": [ + "issue:41084|issue:43037" + ], + "prepared_review_unit_hash": "ee9dfe6a4aa28cf63bff9a3334ebe7b981cd2197f791ea88c9ba8f77dbf782f9", + "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 appear unrelated: one is about block decoding behavior, the other is a feature request for DeepSeek v3.2 support.", + "confidence": 0.97, + "canonical_issue_reason": "No clear canonical issue: the titles describe different problems and there is no evidence they are duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "If a representative issue must be chosen, #43037 is more concrete and actionable than the vague #41084 title.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "Different underlying topics: block decoding configuration vs model/version 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": 9, + "component_count": 23, + "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-41306-2", + "nodes": [ + "issue:41306", + "issue:41826" + ], + "soft_pairs": [ + "issue:41306|issue:41826" + ], + "prepared_review_unit_hash": "2be15634d418d78dbfc28ddaabcffca42a8a8da457b2584f49f4f6879ccf1d95", + "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 at a broad level (Tiled/Liger model efficiency) but they request different features: one is about adding a `from_pretrained()` flag for FLCE, the other about integrating TiledMLP for lower memory use. They are not duplicates.", + "confidence": 0.94, + "canonical_issue_reason": "No canonical issue: these are distinct feature requests with different goals and implementation scopes.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 41306 is the more concrete and self-contained request, so it is the better representative issue for this small cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41306", + "right": "issue:41826", + "accept": false, + "reason": "Reject: they concern different enhancements in the same broad subsystem, not the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 10, + "component_count": 23, + "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-43598-2", + "nodes": [ + "issue:43598", + "issue:43600" + ], + "soft_pairs": [ + "issue:43598|issue:43600" + ], + "prepared_review_unit_hash": "014c0512c4bc857d59894043898c6084cf121ac05e4bb6f7a742d73159bc368e", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are only superficially related Trainer issues; they describe different bugs and should not be treated as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": "No canonical issue: #43598 is about `remove_unused_column` customizability, while #43600 is about FLOPS estimation for custom trainers; they target different code paths.", + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a clean cluster representative because the underlying problems differ; if forced, #43598 is the broader Trainer customization report, but it does not subsume #43600.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "Same subsystem, but different underlying problems: one concerns column filtering customizability, the other FLOPS counting for custom trainers." + } + ] + }, + "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": 23, + "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-43684-2", + "nodes": [ + "issue:43684", + "issue:44405" + ], + "soft_pairs": [ + "issue:43684|issue:44405" + ], + "prepared_review_unit_hash": "79637e79c982f841187360382a4a6367491be9d61a05acdffd0e6c1bd2176d71", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are not duplicates: one asks for Qwen3-Omni registration in AutoModel/AutoModelForConditionalGeneration, while the other requests AutoModelForSequenceClassification support for Qwen3.5 (Qwen3_5Config). They target different model families and different auto-class paths.", + "confidence": 0.96, + "canonical_issue_reason": "Issue #43684 is the broader, more central request because it concerns adding model registration to core auto-model classes, making it the better representative of the cluster even though it is not a duplicate of #44405.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #43684 is the best issue to keep as the cluster representative because it describes the broader framework-level registration gap rather than a task-specific classifier mapping.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Different underlying changes: Qwen3-Omni auto-model registration versus Qwen3.5 sequence-classification support. They involve different configs, different auto classes, and different code paths, so they should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 12, + "component_count": 23, + "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-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": "b1424a237b7372630aa32d40284fd5e7740d101ecf16e635e53d1f4aa856951d", + "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": "This cluster is mostly low-information issue spam: three issues share the same generic title \"huggingface \" and one unrelated issue is titled \"Fett\". There is not enough content to confirm a real underlying bug, so only a thin duplicate representative can be chosen.", + "confidence": 0.53, + "canonical_issue_reason": "issue:43845 is the earliest of the three identical \"huggingface \" issues and is the best available representative for that repeated low-signal thread.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43845 is the most suitable cluster representative because it is earliest among the near-identical issues and has the same sparse metadata as the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "Titles do not match and there is no body-level evidence of the same underlying issue." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "Titles do not match and the available metadata is too sparse to prove a shared bug." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "Different title and no concrete issue details tie these together." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": false, + "reason": "The titles are identical, but there is no actual problem description to confirm they report the same bug rather than duplicate noise." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": false, + "reason": "Same generic title, but the record is too underspecified to safely merge as the same underlying issue." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": false, + "reason": "Same generic title, but insufficient content exists to verify a shared 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": 13, + "component_count": 23, + "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-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": "4603826e4036a51b124b785e3028838e1f98010da769d1c9d05f37d44d36f7d3", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5794, + "estimated_input_tokens": 1449, + "estimated_eval_tokens": 3154 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5794, + "estimated_input_tokens": 1449, + "estimated_eval_tokens": 3154 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All five issues are extremely low-signal and appear spammy or generic, with no evidence they describe the same underlying problem. None of the soft duplicate links should be accepted.", + "confidence": 0.94, + "canonical_issue_reason": "Issue 44418 is the most neutral and descriptively titled of the set, but the cluster lacks enough substance to support a strong canonical issue choice.", + "canonical_pr_reason": null, + "best_issue_reason": "44418 is the least obviously spam-oriented and the only one with a somewhat meaningful label, so it is the best weak representative among otherwise non-informative issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44371", + "right": "issue:44572", + "accept": false, + "reason": "Both are just '' with no body evidence; identical low-quality titles are not enough to treat them as the same issue." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Same spam placeholder title and no content; no concrete bug or topic can be established." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "One is spam-like and the other is a vague documentation suggestion; too little overlap to indicate the same underlying issue." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "'Check out \"Google LLC\"' does not match the spam placeholder issue in a way that suggests a shared bug or request." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "A generic documentation request and a spam placeholder are not enough to infer duplicate intent." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "The titles are both generic, but they do not point to the same concrete problem or change." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Identical spam placeholders, but no actual issue description or target exists to confirm duplicate status." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "No shared subject matter beyond being low-signal issue titles; not a reliable duplicate match." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "The titles are unrelated in content and lack any concrete bug/topic alignment." + } + ] + }, + "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": 23, + "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-45145-2", + "nodes": [ + "issue:45145", + "issue:45151" + ], + "soft_pairs": [ + "issue:45145|issue:45151" + ], + "prepared_review_unit_hash": "34a11714027441596cacd48c63b7bf8002ff7efe2b83fc43eafc6cd699f5ee54", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1646, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1646, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two closed issues are highly similar and likely duplicates: they share the exact title, have very similar body sizes, and there are no distinguishing targets or filenames. The evidence is strong enough to treat them as the same reported topic, but the packet does not expose body text, so the duplicate judgment should stay conservative. #45145 is the better canonical representative because it was opened earlier and has an inbound reference.", + "confidence": 0.94, + "canonical_issue_reason": "#45145 is the better canonical issue because it is earlier and has an inbound reference, while the available signals strongly suggest #45151 is a near-duplicate or highly similar report.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the best issue representative for the cluster because it is the earlier, slightly more established copy of the same likely-duplicate report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": true, + "reason": "Same title, similar body length, and no distinguishing metadata strongly indicate the same underlying issue, though the evidence is not strong enough to claim identical body content." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The duplicate call is a bit too strong for the evidence shown. The packet supports a strong similarity case (same title, similar body lengths, no distinguishing metadata, and soft_similarity), but it does not expose issue body content, so saying they are the \u201csame issue content\u201d or fully duplicate is slightly overstated. A more conservative verdict would be that they are likely duplicates or highly similar." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 15, + "component_count": 23, + "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": 23, + "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-45672-2", + "nodes": [ + "issue:45672", + "issue:45674" + ], + "soft_pairs": [ + "issue:45672|issue:45674" + ], + "prepared_review_unit_hash": "d32306b150aad393600d5cdb76f0286cd1a0655fd6c6fcf2dad8a725690d12db", + "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": "These are related only at a broad level (4-bit/BitsAndBytes inference), but they describe different failure modes and likely different fixes: one is a uint8/NF4 audio-module torch.finfo() TypeError, the other is a BitsAndBytesConfig regression that drops the default lm_head exclusion and triggers an AssertionError.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:45674 is the better representative issue because it states a clearer, more general configuration regression with an explicit failure mechanism and likely code-path to fix.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45672", + "right": "issue:45674", + "accept": false, + "reason": "Do not merge: both are 4-bit inference problems, but one concerns torch.finfo() on uint8 weights in audio modules while the other concerns llm_int8_skip_modules clearing the default lm_head exclusion. Different root causes and fixes." + } + ] + }, + "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": 23, + "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-45762-2", + "nodes": [ + "issue:45762", + "issue:45763" + ], + "soft_pairs": [ + "issue:45762|issue:45763" + ], + "prepared_review_unit_hash": "d57bcfcfd84e4aae899ebbc7716cf785f68a3ca4b817c175aa2e5b39306f263c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two closed issues are unrelated personal-name/date entries with no body, links, or technical context. They do not appear to describe the same underlying problem, so this is not a duplicate cluster.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a good canonical representative: both are sparse, unrelated, and lack any shared bug report or actionable technical context.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45762", + "right": "issue:45763", + "accept": false, + "reason": "Different names and dates with no shared subject, code path, or problem description; they do not look like 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": 18, + "component_count": 23, + "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-45764-2", + "nodes": [ + "issue:45764", + "issue:45765" + ], + "soft_pairs": [ + "issue:45764|issue:45765" + ], + "prepared_review_unit_hash": "67ce8b9a2b36e0550091ea526c0b33ad7194bcc702b07c4504e834b695d76d33", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two closed issues are not duplicates; they appear to be unrelated entries that only share a surname and similarly formatted dates in the titles.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45764", + "right": "issue:45765", + "accept": false, + "reason": "Titles reference different people and different dates; there is no shared bug report, code path, or explicit linkage. The similarity is superficial only." + } + ] + }, + "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": 23, + "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-45810-2", + "nodes": [ + "issue:45810", + "issue:46295" + ], + "soft_pairs": [ + "issue:45810|issue:46295" + ], + "prepared_review_unit_hash": "05e19b770374146b09d8e291a887d63d67d2165c656c61bb3881b0aebb373382", + "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 two issues are unrelated: one requests a new Qwen token-classification head, the other requests FP8 training support. No duplicate/merge relationship is present.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue: the requests concern different areas (model head support vs. FP8 training).", + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a better canonical representative because they do not describe the same underlying bug or feature.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45810", + "right": "issue:46295", + "accept": false, + "reason": "Different feature areas with no shared code-path problem; one is about adding Qwen3_5ForTokenClassification, the other about FP8 training support." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 20, + "component_count": 23, + "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-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": "091e1c7463a06dfbcdb5799fc506509a14d8c47729a861a6c4ea90faf7913fe8", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5726, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5726, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster splits into two exact-duplicate issue pairs: 45870/45871 and 45872/45873. Issue 45869 is a different dataset listing and does not look like the same underlying request as the others.", + "confidence": 0.93, + "canonical_issue_reason": "45870 is the earliest clear duplicate representative in the main pair, with an identical title and body length to 45871.", + "canonical_pr_reason": null, + "best_issue_reason": "45870 is the safest cluster representative: it is an exact duplicate of 45871 and predates it, while the 45872/45873 pair appears to be a separate duplicate topic.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45869", + "right": "issue:45870", + "accept": false, + "reason": "Different topic: a dataset bundle listing versus a DataForge example issue; not the same underlying request." + }, + { + "left": "issue:45869", + "right": "issue:45871", + "accept": false, + "reason": "Different topic: dataset bundle listing versus DataForge example issue." + }, + { + "left": "issue:45870", + "right": "issue:45871", + "accept": true, + "reason": "Same title and same body length, created minutes apart; they look like duplicate submissions of the same issue." + }, + { + "left": "issue:45870", + "right": "issue:45872", + "accept": false, + "reason": "Both are financial/DataForge themed, but the titles describe different requests (example prediction vs scraping service)." + }, + { + "left": "issue:45870", + "right": "issue:45873", + "accept": false, + "reason": "Different concrete request: example prediction issue versus scraping-service issue." + }, + { + "left": "issue:45871", + "right": "issue:45872", + "accept": false, + "reason": "Different concrete request with no evidence they describe the same issue." + }, + { + "left": "issue:45871", + "right": "issue:45873", + "accept": false, + "reason": "Different concrete request with no evidence they describe the same issue." + }, + { + "left": "issue:45872", + "right": "issue:45873", + "accept": true, + "reason": "Same title and same body length, created close together; they appear to be duplicate submissions of the same issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: 45870/45871 share the same title, body length, and close timestamps, as do 45872/45873, while 45869 is clearly a different dataset-bundle topic. The soft-edge verdicts are conservative and do not overstate weaker pairs. No PR-mergeability issue applies here." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 21, + "component_count": 23, + "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-46439-2", + "nodes": [ + "issue:46439", + "issue:46441" + ], + "soft_pairs": [ + "issue:46439|issue:46441" + ], + "prepared_review_unit_hash": "f26398c5150b4a8fb828667c6ddecfda684ae219c59ee1844fc9d9fa39dd12d3", + "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 concern static-cache failures around linear-attention/Qwen3.5, but they describe different concrete failure modes: one is a shape-corruption RuntimeError during early initialization, the other is a KeyError for `linear_attention` in Qwen3.5. Similar subsystem, but not clearly the same bug.", + "confidence": 0.78, + "canonical_issue_reason": "Issue 46439 is the better canonical issue because it describes the broader underlying static-cache/linear-attention corruption with a concrete runtime failure and likely encompasses more variants than the Qwen3.5-specific KeyError report.", + "canonical_pr_reason": null, + "best_issue_reason": "46439 is the strongest representative of the cluster: it is more general, more diagnostic, and names the likely root area (`early_initialization` + linear-attention shape corruption) rather than a single model-specific crash.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46439", + "right": "issue:46441", + "accept": false, + "reason": "Related but not the same concrete bug: one reports wrong-shape corruption leading to RuntimeError, the other a KeyError for `linear_attention` on Qwen3.5." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 22, + "component_count": 23, + "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-46551-2", + "nodes": [ + "issue:46551", + "issue:46552" + ], + "soft_pairs": [ + "issue:46551|issue:46552" + ], + "prepared_review_unit_hash": "e0458dbb1ba652d0c3e2b0fc5c66fe75e7aed0f90bb4ac2741393d3a74ebfefb", + "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": "The two issues are near-perfect duplicates with identical titles and matching bodies describing the same Wav2Vec2CTCTokenizer v5 regression. The open issue is the better canonical target; there are no PRs in this cluster.", + "confidence": 0.99, + "canonical_issue_reason": "Issue 46552 is the open report with the same text and has inbound references, making it the better live tracking issue.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 46552 is the best issue to keep because it is open, current, and already referenced by others; 46551 appears to be the duplicate/closed copy.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46551", + "right": "issue:46552", + "accept": true, + "reason": "Identical title and matching description indicate the same underlying bug; this is a clear duplicate pair." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both issues have identical titles and bodies, issue 46552 is open with inbound references, and 46551 is closed. The duplicate verdict is conservative given the exact-match similarity evidence, and there are no PRs to evaluate." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 23, + "component_count": 23, + "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": "8bbb237adefcde7eb3a297ee3c8ae2d434a85e42df093b5291675d8632ca7e66", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1568, + "estimated_input_tokens": 392, + "estimated_eval_tokens": 1040 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1568, + "estimated_input_tokens": 392, + "estimated_eval_tokens": 1040 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are not duplicates: one reports a tensor shape mismatch during object-detection training, while the other reports out-of-memory after several epochs in RT-DETR v2. They may be broadly related to training instability, but they describe different failure modes and likely different fixes.", + "confidence": 0.95, + "canonical_issue_reason": "Issue #46657 is the more concrete, narrowly scoped bug report (tensor size mismatch during object detection training) and is the better representative if a single issue must be chosen.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #46657 is the best issue anchor because it describes a specific reproducible error rather than a broader resource problem.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46657", + "right": "issue:46692", + "accept": false, + "reason": "Different underlying problems: #46657 is a tensor dimension mismatch during training, while #46692 is an out-of-memory failure after some epochs in RT-DETR v2. Shared training context is not enough to treat them as the same 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": 37, + "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": "ede7dd2766c805e2ddc64a1845b151f97a7b2ed5a18c48486a14a9cb49c89387", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7224, + "estimated_input_tokens": 1806, + "estimated_eval_tokens": 3868 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7224, + "estimated_input_tokens": 1806, + "estimated_eval_tokens": 3868 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs are related to VibeVoice but are not the same change: #40546 is the base VibeVoice implementation, while #42781 adds VibeVoice Realtime plus acoustic-tokenizer/feature-extraction and generation-path changes. They overlap on shared model integration files, but they target different concrete functionality and should not be deduplicated.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "#40546 is the more canonical base PR: it introduces the core VibeVoice model/docs integration and is the broader foundational implementation, whereas #42781 is a separate realtime extension built on top of that work.", + "best_issue_reason": null, + "best_pr_reason": "#40546 is the best representative of the cluster because it is the foundational VibeVoice implementation and has the clearest standalone scope; #42781 is a later, narrower realtime follow-up with additional tokenizer/feature-extraction work.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Reject: these are related VibeVoice PRs, but not the same underlying change. #40546 implements VibeVoice itself; #42781 adds a separate realtime variant and extra acoustic-tokenizer/feature-extraction/generation changes, so 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 + }, + { + "label": "pull_request", + "component_index": 2, + "component_count": 37, + "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": "8438154b2b81ce848874db2f56ef18b21b67707af0a789569ca2b5b2466a353a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4676, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4676, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two pull requests are unrelated feature additions in different model families: TimesFM covariate forecasting versus DINOv3 image classification support. They do not fix the same bug or implement the same change path, so they should not be merged as duplicates.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": "41224 is the more focused and self-contained PR, adding a concrete DINOv3 image-classification path; 40755 is a broader TimesFM forecasting feature in a different subsystem.", + "best_issue_reason": null, + "best_pr_reason": "41224 is the better representative PR because it is narrower, more concrete, and easier to review/merge as a single feature change than the broader TimesFM covariate work.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Different model families and different code paths: TimesFM covariate forecasting vs DINOv3 image classification support. They are not the same underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 3, + "component_count": 37, + "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": 37, + "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-41855-14", + "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", + "pull_request:46744" + ], + "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:46744", + "pull_request:44827|pull_request:46048", + "pull_request:44827|pull_request:45213", + "pull_request:43823|pull_request:45213", + "pull_request:44827|pull_request:45655", + "pull_request:43251|pull_request:43254" + ], + "prepared_review_unit_hash": "f78b61df4dd42171716fd28ae526d107ddc36b33b007730e62600aa3ba73d1b9", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 19, + "serialized_chars": 36844, + "estimated_input_tokens": 9211, + "estimated_eval_tokens": 18678 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 19, + "serialized_chars": 36844, + "estimated_input_tokens": 9211, + "estimated_eval_tokens": 18678 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Only one soft edge looks like a true duplicate: the two cross-entropy loss PRs both change the same helper in the same way. The rest are merely in the same broad Transformers codebase or thematically similar, but they fix unrelated models, docs, tests, or subsystems.", + "confidence": 0.78, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:43251 is the better canonical PR for the only true duplicate pair: it has the explicit issue target, a concrete loss-utils fix, and a slightly fuller implementation of the same cross-entropy kwargs change.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:43251 is the strongest standalone PR in this set because it addresses a specific bug in a single code path (`fixed_cross_entropy`), has an explicit issue link, and is tightly scoped compared with the more speculative similarity-only matches.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Different areas and goals: Mistral tokenizer API additions vs adding a new MobileLLM model. No shared bug or code path." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Tokenizer method completion vs Executorch dynamic-shape export testing; unrelated changes." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor-loading robustness across multiple models vs adding a brand-new model. Broad framework overlap only." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "Loss-function kwargs fix vs new model addition; entirely different functionality." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "Both touch training/modeling code, but one fixes `cross_entropy` kwargs and the other is a Mistral4 model/test update. Not the same change." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "Executorch export testing and Mistral4 test fixes are unrelated test coverage in different subsystems." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "Dynamic export test vs new model implementation; no shared underlying bug." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "Adding MobileLLM vs fixing config annotation resolution in `modeling_utils.py`; unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "New model addition vs Falcon output-interface refactor; different models and different code paths." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon model refactor vs model-creation skill/docs; no shared bug or mergeable overlap." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Falcon output-capture refactor and Mistral4 test fixes are distinct model-specific changes." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 test fixes vs a CLIP-like `layrnorm` rename/conversion mapping fix; unrelated models and bugs." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46267", + "accept": false, + "reason": "Mistral4 tests vs chat-response parsing overhaul; completely different functionality." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46744", + "accept": false, + "reason": "Mistral4 tests vs Apple Silicon/safetensors docs update; not the same underlying change." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46048", + "accept": false, + "reason": "Mistral4 tests vs TDT loss kernel integration; unrelated code paths." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 tests/modeling vs model-creation skill/dataset/doc tooling. The shared auto-model file is incidental, not a duplicate bug fix." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "New MobileLLM model implementation vs a repository skill/docs PR. Similar 'model creation' theme only." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "Mistral4 test/model changes vs config-resolution logic in the base modeling utility; different layers and problems." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46048", + "accept": false, + "reason": "Mistral4 test fixes vs kernel-loading for TDT loss; unrelated subsystems." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Model test/modeling changes vs model-creation tooling; no shared underlying code change." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Same file, same helper (`fixed_cross_entropy`), same added kwargs (`weight`, `label_smoothing`), and same issue target 43240. These are the same underlying fix." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary stays within the packet evidence, and the non-duplicate verdicts are conservative: they treat broad thematic overlap as non-duplication. The only accepted soft edge, pull_request:43251 vs pull_request:43254, is well-supported by the shared file, shared issue target, and same kwargs-to-cross_entropy fix." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 5, + "component_count": 37, + "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": "575e438ebec92de7d9851f707d94123fb53d190790fa7e699f0f615dddb85214", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests are broad GitHub Actions dependency bumps, but they target different underlying changes: PR 42975 is a Node 24 compatibility sweep across many workflow actions, while PR 42976 specifically upgrades docker/build-push-action and related workflow versions. They overlap in files and theme, but not in the same concrete fix.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 42975 is the better canonical representative because it has the clearest unifying purpose (Node 24 compatibility) and covers the wider workflow-action upgrade set.", + "best_issue_reason": null, + "best_pr_reason": "PR 42975 is the best single PR to represent the cluster since it expresses a concrete compatibility goal rather than a narrower action-version refresh.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "They are related GitHub Actions maintenance PRs, but they fix different things: one is a Node 24 compatibility upgrade across workflow actions, the other is a docker/build-push-action version bump. Shared workflow files are 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": 6, + "component_count": 37, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-42978-7", + "nodes": [ + "pull_request:42978", + "pull_request:43448", + "pull_request:43451", + "pull_request:45490", + "pull_request:46562", + "pull_request:46658", + "pull_request:46705" + ], + "soft_pairs": [ + "pull_request:42978|pull_request:43448", + "pull_request:42978|pull_request:43451", + "pull_request:42978|pull_request:45490", + "pull_request:42978|pull_request:46705", + "pull_request:43448|pull_request:43451", + "pull_request:43448|pull_request:45490", + "pull_request:43448|pull_request:46705", + "pull_request:43451|pull_request:45490", + "pull_request:43451|pull_request:46705", + "pull_request:45490|pull_request:46562", + "pull_request:45490|pull_request:46705", + "pull_request:46562|pull_request:46658" + ], + "prepared_review_unit_hash": "8701d24d4d5f226878fde60ba7db3806947945b9e6c1bb61da50997d3b49683f", + "component_budget": { + "node_count": 7, + "item_count": 7, + "soft_pair_count": 12, + "serialized_chars": 24621, + "estimated_input_tokens": 6156, + "estimated_eval_tokens": 12568 + }, + "budget": { + "node_count": 7, + "item_count": 7, + "soft_pair_count": 12, + "serialized_chars": 24621, + "estimated_input_tokens": 6156, + "estimated_eval_tokens": 12568 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Most PRs are separate model-addition or infrastructure changes that only overlap on common registration/docs files. The only likely duplicate-like pair is the Step 3.7 Flash feature PR and its WIP companion script/model package; everything else should stay separate.", + "confidence": 0.78, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46562 is the best canonical PR because it is the main library-side Step 3.7 Flash model integration with docs/tests; #46658 looks like supporting WIP conversion/build tooling for the same feature, not the primary implementation.", + "best_issue_reason": null, + "best_pr_reason": "PR #46562 is the strongest representative of the only accepted duplicate-like pair and is the most self-contained concrete model change in the set.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42978", + "right": "pull_request:43448", + "accept": false, + "reason": "Both are unrelated new model additions; shared docs/auto-registration files are boilerplate and do not indicate the same underlying change." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "ViT NEPA and Molmo2 are different models with different code paths; overlap is only in shared framework registration/docs." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "Different model families entirely (ViT NEPA vs CTSM); shared auto/docs edits are generic integration work." + }, + { + "left": "pull_request:42978", + "right": "pull_request:46705", + "accept": false, + "reason": "ViT NEPA model addition is unrelated to the distributed config PR; no shared bug or concrete change." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are distinct model additions with different architectures/files; overlap is just model-registration boilerplate." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM are different models; common files like docs and auto mappings are not enough to treat them as the same change." + }, + { + "left": "pull_request:43448", + "right": "pull_request:46705", + "accept": false, + "reason": "Molmo2 model work is unrelated to distributed configuration support." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM are separate model integrations with different implementations and tests." + }, + { + "left": "pull_request:43451", + "right": "pull_request:46705", + "accept": false, + "reason": "Molmo2 model addition and distributed config are distinct changes with no shared code-path problem." + }, + { + "left": "pull_request:45490", + "right": "pull_request:46562", + "accept": false, + "reason": "CTSM and Step 3.7 Flash are unrelated model additions; shared auto/docs files are normal repo integration overlap." + }, + { + "left": "pull_request:45490", + "right": "pull_request:46705", + "accept": false, + "reason": "CTSM model addition does not match the distributed config PR; no same underlying bug/change." + }, + { + "left": "pull_request:46562", + "right": "pull_request:46658", + "accept": true, + "reason": "Both are Step 3.7 Flash work and overlap on the same model family and registration plumbing; #46562 is the core model implementation while #46658 is WIP conversion/build scaffolding for the same feature, so they could plausibly be consolidated." + } + ] + }, + "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": 37, + "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-43291-8", + "nodes": [ + "pull_request:43291", + "pull_request:43498", + "pull_request:43665", + "pull_request:44815", + "pull_request:45114", + "pull_request:46380", + "pull_request:46539", + "pull_request:46579" + ], + "soft_pairs": [ + "pull_request:43291|pull_request:43665", + "pull_request:43291|pull_request:44815", + "pull_request:43291|pull_request:46380", + "pull_request:43498|pull_request:46579", + "pull_request:43665|pull_request:46539", + "pull_request:43665|pull_request:44815", + "pull_request:43665|pull_request:46380", + "pull_request:43665|pull_request:46579", + "pull_request:44815|pull_request:46539", + "pull_request:44815|pull_request:46380", + "pull_request:44815|pull_request:46579", + "pull_request:45114|pull_request:46579", + "pull_request:46380|pull_request:46539", + "pull_request:46380|pull_request:46579" + ], + "prepared_review_unit_hash": "c75cd616b33a8c70760607a4c297e0085816ead8dbc2d77b3131568dc5476db8", + "component_budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 14, + "serialized_chars": 23167, + "estimated_input_tokens": 5792, + "estimated_eval_tokens": 11840 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 14, + "serialized_chars": 23167, + "estimated_input_tokens": 5792, + "estimated_eval_tokens": 11840 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is a mix of unrelated open PRs: Whisper test fixes, tie_weights backward-compatibility, vision test cleanup, FP8 dequant changes, doctest updates, generation watermarking validation, CI infra changes, and typo/documentation cleanup. None of the soft edges look like true duplicates of the same underlying 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:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Both are test-related, but they target different models and bugs: Whisper tokenizer decode behavior vs CLIP/ViT test edits. No shared code path." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper decoding tests are unrelated to FP8 dequant/loading logic. Different subsystem and fix scope." + }, + { + "left": "pull_request:43291", + "right": "pull_request:46380", + "accept": false, + "reason": "Whisper test fix vs generation watermarking/validation and broad model test changes. Only superficial similarity as test fixes." + }, + { + "left": "pull_request:43498", + "right": "pull_request:46579", + "accept": false, + "reason": "Backward-compatibility for tie_weights is unrelated to typo cleanup across many files. No common bug or mergeable change." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46539", + "accept": false, + "reason": "Vision model test tweaks are unrelated to Phimoe CI/integration test infrastructure changes." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "These touch different areas: test assertions for CLIP/ViT versus dequant/fp8 model loading and quantization code." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46380", + "accept": false, + "reason": "Both mention tests, but the underlying changes are different: CLIP/ViT test files versus generation config/utils and Whisper behavior." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46579", + "accept": false, + "reason": "Vision test file edits and code-typo cleanup are not the same bug or change." + }, + { + "left": "pull_request:44815", + "right": "pull_request:46539", + "accept": false, + "reason": "FP8/dequant logic changes are unrelated to Phimoe integration-test CI adjustments." + }, + { + "left": "pull_request:44815", + "right": "pull_request:46380", + "accept": false, + "reason": "Different code paths: finegrained FP8 loading/quantization versus generation configuration and Whisper-related test fixes." + }, + { + "left": "pull_request:44815", + "right": "pull_request:46579", + "accept": false, + "reason": "FP8/model-loading changes do not match a broad typo-fix cleanup PR." + }, + { + "left": "pull_request:45114", + "right": "pull_request:46579", + "accept": false, + "reason": "Doctest documentation updates are not the same underlying change as code typo fixes across multiple modules." + }, + { + "left": "pull_request:46380", + "right": "pull_request:46539", + "accept": false, + "reason": "Generation/watermarking and Whisper test fixes are unrelated to Phimoe CI environment changes." + }, + { + "left": "pull_request:46380", + "right": "pull_request:46579", + "accept": false, + "reason": "Broad generation/Whisper bugfixes are not the same as a multi-file typo cleanup." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 8, + "component_count": 37, + "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": "466decdcca13ae52aaf9bd3c927219e036c85c2fdaee60e8729322f44fd2daf2", + "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": "Three unrelated \"don't merge\"/demo PRs that only overlap at a very broad level (repo-bot, workflow, and test-related changes). They do not look like the same underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43532 is the most representative of the cluster as a generic diff/demo PR: it touches several workflow and test files and best matches the overall placeholder 'show diff' theme.", + "best_issue_reason": null, + "best_pr_reason": "PR 43532 is the best global representative because it has the broadest, most illustrative change set and the clearest 'show diff' intent among the three.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:46747", + "accept": false, + "reason": "Different concrete changes: 43488 tweaks setup/dependency/test files, while 46747 changes a security-gate workflow trigger. Same broad repo-maintenance theme, but not the same bug or change." + }, + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "They both touch testing-related files, but 43488 is about setup/version/dependency and a Vit test, whereas 43532 mainly rewires comment/notification workflows and test utilities. Not one fix." + }, + { + "left": "pull_request:43532", + "right": "pull_request:46747", + "accept": false, + "reason": "43532 is a multi-file workflow/test utility demo, while 46747 is a single workflow security-gate trigger change. Overlap is only superficial at the workflow level." + } + ] + }, + "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": 37, + "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": "e94c39db0b6341f74b5bfafb4f46a06c9eebabfaf86c6cb7c339bd1903f335a6", + "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": "Both PRs add new, similarly named segmentation pipelines, but they target different features and codepaths: promptable concept segmentation vs promptable visual segmentation. They share docs/registry scaffolding, yet the underlying implementations, tests, and model integrations are distinct, so they should not be merged as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR #43613 is the stronger representative of this cluster because it is broader and more integrated, covering multiple model test suites and a dedicated pipeline implementation; #43612 is a narrower, separate feature.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Similar naming and shared docs/registry changes, but they implement different pipelines with different filenames, model integrations, and test targets. They are related feature additions, not the same concrete change or bug 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": 10, + "component_count": 37, + "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": "f55c1cb5e5eb64505a908e61a242c4a1369b423c2a62d8e6ec92f140b48b3f83", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two PRs both modify `trainer.py` and both mention custom trainer extensibility, but they address different problems: one changes gradient-accumulation loss scaling, the other refactors signature-column filtering. They are not the same underlying fix and should not be merged as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43651 is the more concrete bug fix: it changes `training_step` loss normalization for gradient accumulation and introduces `_loss_is_scaled_for_ga`, which is a specific behavior change rather than a broad refactor.", + "best_issue_reason": null, + "best_pr_reason": "PR 43651 is the better representative of the cluster because it is a narrowly scoped correctness fix in the training path, whereas PR 43663 is a separate extensibility refactor for input-column filtering.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Different code paths and different behaviors: 43651 fixes loss scaling during gradient accumulation, while 43663 adds an overridable helper for signature-column filtering. Shared file/subsystem alone 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": 11, + "component_count": 37, + "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": "3192996f63d910653cf95ac283b226f8ca5f1a01dfc34d1b1beedb79ab21c9ee", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3663, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3663, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs touch `src/transformers/cli/serve.py` and are about a TypeAdapter-related NameError in the CLI, but they use different fixes: PR 43656 is a broader postponed-annotations/type-hint cleanup, while PR 43842 specifically guards the optional `pydantic.TypeAdapter` import with `TYPE_CHECKING` and quotes the annotation. They are related, but not clearly the same concrete change.", + "confidence": 0.78, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43842 is the tighter, more explicit fix for the concrete failure mode: it directly addresses `TypeAdapter` being unavailable when `pydantic` is not installed. That makes it a better representative of the underlying bug than the broader annotation cleanup in PR 43656.", + "best_issue_reason": null, + "best_pr_reason": "PR 43842 is the best single PR in the cluster because it targets the specific import-time NameError path and is narrowly scoped to the optional dependency problem.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": false, + "reason": "They overlap in file and theme, but the fixes are not the same concrete change: 43656 applies a broader `from __future__ import annotations`/typing cleanup, while 43842 specifically makes `TypeAdapter` safe when `pydantic` is absent. They could coexist, but this is not strong enough to treat as a duplicate pair." + } + ] + }, + "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": 37, + "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": 37, + "review_unit_index": 1, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43996", + "pull_request:43997", + "pull_request:44007", + "pull_request:44010", + "pull_request:44013", + "pull_request:44017", + "pull_request:44018", + "pull_request:44019", + "pull_request:44044", + "pull_request:44056", + "pull_request:44059", + "pull_request:44066", + "pull_request:44068", + "pull_request:44071", + "pull_request:44072", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44085", + "pull_request:44086", + "pull_request:44098", + "pull_request:44101", + "pull_request:44116", + "pull_request:44129", + "pull_request:44154", + "pull_request:44161", + "pull_request:44722" + ], + "soft_pairs": [ + "pull_request:43996|pull_request:44085", + "pull_request:43996|pull_request:44044", + "pull_request:44066|pull_request:44085", + "pull_request:44007|pull_request:44072", + "pull_request:44072|pull_request:44722", + "pull_request:44066|pull_request:44072", + "pull_request:44013|pull_request:44044", + "pull_request:44066|pull_request:44086", + "pull_request:44066|pull_request:44071", + "pull_request:44018|pull_request:44068", + "pull_request:44066|pull_request:44068", + "pull_request:44018|pull_request:44066", + "pull_request:44019|pull_request:44722", + "pull_request:44086|pull_request:44722", + "pull_request:44019|pull_request:44085", + "pull_request:44018|pull_request:44019", + "pull_request:44019|pull_request:44071", + "pull_request:44068|pull_request:44722", + "pull_request:44068|pull_request:44116", + "pull_request:44018|pull_request:44116", + "pull_request:44071|pull_request:44722", + "pull_request:44018|pull_request:44085", + "pull_request:44018|pull_request:44086", + "pull_request:44068|pull_request:44086", + "pull_request:44071|pull_request:44085", + "pull_request:44085|pull_request:44086", + "pull_request:44059|pull_request:44086", + "pull_request:44098|pull_request:44101", + "pull_request:44018|pull_request:44071", + "pull_request:44059|pull_request:44722", + "pull_request:44017|pull_request:44161", + "pull_request:44017|pull_request:44722", + "pull_request:44017|pull_request:44073", + "pull_request:44068|pull_request:44071", + "pull_request:43997|pull_request:44073", + "pull_request:44071|pull_request:44086", + "pull_request:44017|pull_request:44071", + "pull_request:44073|pull_request:44101", + "pull_request:44073|pull_request:44161", + "pull_request:44059|pull_request:44071", + "pull_request:44056|pull_request:44101", + "pull_request:44018|pull_request:44059", + "pull_request:44059|pull_request:44068", + "pull_request:44101|pull_request:44161", + "pull_request:43997|pull_request:44017", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:44056|pull_request:44161", + "pull_request:43997|pull_request:44018", + "pull_request:43997|pull_request:44161", + "pull_request:44074|pull_request:44161", + "pull_request:44018|pull_request:44161", + "pull_request:43997|pull_request:44071", + "pull_request:44010|pull_request:44071", + "pull_request:44010|pull_request:44073", + "pull_request:44018|pull_request:44129", + "pull_request:43997|pull_request:44074", + "pull_request:44010|pull_request:44018", + "pull_request:44071|pull_request:44129", + "pull_request:43997|pull_request:44129", + "pull_request:43996|pull_request:44161", + "pull_request:43996|pull_request:44017", + "pull_request:44010|pull_request:44044", + "pull_request:44044|pull_request:44074", + "pull_request:44066|pull_request:44722", + "pull_request:44085|pull_request:44722", + "pull_request:44085|pull_request:44129", + "pull_request:43996|pull_request:44007", + "pull_request:43996|pull_request:44129", + "pull_request:44044|pull_request:44129", + "pull_request:44044|pull_request:44154" + ], + "prepared_review_unit_hash": "30001924bb5826d219991b7c9f09996b9981a2196cee95192a0fa263a7ebc8d8", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84884, + "estimated_input_tokens": 21221, + "estimated_eval_tokens": 42698 + }, + "budget": { + "node_count": 27, + "item_count": 27, + "soft_pair_count": 72, + "serialized_chars": 75081, + "estimated_input_tokens": 18771, + "estimated_eval_tokens": 37798 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of model-specific refactors to standardized output tracing. The only clear duplicate subclusters are GPT-J/CodeGen and GPT-Neo variants; the rest are similar but separate model PRs, not the same code change.", + "confidence": 0.78, + "canonical_issue_reason": null, + "canonical_pr_reason": "No issue artifact to anchor on; PR 44722 is the broadest/latest representative of the GPT-J/CodeGen output-tracing refactor subcluster.", + "best_issue_reason": null, + "best_pr_reason": "PR 44722 best represents the duplicate subcluster because it covers the same GPT-J code path as earlier GPT-J-only PRs and also includes the related CodeGen refactor.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Different model families (CVT/FNet vs GPT-J); same issue target is not enough to call these duplicates." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "Unrelated model refactors (CVT/FNet vs DeBERTa); similar output-tracing theme only." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": true, + "reason": "Same GPT-J output-tracing change on the same file; 44066 is the broader version and 44085 is GPT-J-only." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "Different model code paths (RegNet/ResNet vs EfficientNet); both are output-tracing refactors but not the same change." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet is a separate model from GPT-J/CodeGen; shared refactor style only." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen vs EfficientNet are distinct model-specific changes." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "MobileNetV2 and DeBERTa are unrelated code paths." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen vs MGP-STR are unrelated model implementations." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen and MPT are different model refactors." + }, + { + "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 target the same concrete change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J/CodeGen and GPT-Neo are different model code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "GPT-Neo vs GPT-J/CodeGen are distinct model-specific PRs." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet is unrelated to GPT-J/CodeGen." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR and GPT-J/CodeGen are unrelated model changes." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44085", + "accept": false, + "reason": "ResNet and GPT-J are different model refactors." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet do not share the same concrete code path." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet and MPT are separate model implementations." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo and GPT-J/CodeGen are different models; similar decorator migration only." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are unrelated model code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are unrelated model code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT and GPT-J/CodeGen are distinct model changes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo vs GPT-J are different model-specific refactors." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are unrelated." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are unrelated model code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44085", + "accept": false, + "reason": "MPT and GPT-J are unrelated model implementations." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J and MGP-STR are different models; same issue target does not make them duplicates." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT2 and MGP-STR are unrelated model-specific changes." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT and XLM/Flaubert are separate code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are different models." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT2 and GPT-J/CodeGen are separate model refactors." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "SegFormer and LongT5 are unrelated model changes." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44101", + "accept": false, + "reason": "VisualBERT and XLM/Flaubert are different model code paths." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT2 and MPT are unrelated model refactors." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44101", + "accept": false, + "reason": "MPNet and XLM/Flaubert are different model implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet/ResNet and GPT-Neo are unrelated." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet/ResNet and LongT5 are unrelated model changes." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44076", + "accept": false, + "reason": "MPT and ImageGPT are different model code paths." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet and XLM/Flaubert are unrelated model changes." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 are separate model refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44017", + "accept": false, + "reason": "RegNet/ResNet and SegFormer are unrelated." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet/ResNet and MPNet are distinct code paths." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44017", + "accept": false, + "reason": "CVT/FNet and SegFormer are different models." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44161", + "accept": false, + "reason": "CVT/FNet and LongT5 are unrelated changes." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBERT and MPT are unrelated model implementations." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBERT and VisualBERT are different code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo and SpeechT5 are unrelated model refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet/ResNet and TextNet are separate model changes." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44018", + "accept": false, + "reason": "SqueezeBERT and GPT-Neo are unrelated model code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are different model implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44071", + "accept": false, + "reason": "RegNet/ResNet and MPT are unrelated." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBERT and DeBERTa are different model refactors." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa and TextNet are unrelated code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44071", + "accept": false, + "reason": "SegFormer and MPT are unrelated model changes." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44161", + "accept": false, + "reason": "VisualBERT and LongT5 are different models." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT2 and GPT-Neo are different model refactors." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary stays within the packet: it identifies the GPT-J/CodeGen and GPT-Neo clusters as the only clearly duplicate-looking groups and treats the rest conservatively as separate model-specific refactors. The two accepted soft edges are plausibly mergeable subset/broader pairs on the same code path, and the rejected pairs are appropriately cautious." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 13, + "component_count": 37, + "review_unit_index": 2, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43997", + "pull_request:44010", + "pull_request:44024", + "pull_request:44056", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44129" + ], + "soft_pairs": [ + "pull_request:44024|pull_request:44076", + "pull_request:44073|pull_request:44074", + "pull_request:44056|pull_request:44073", + "pull_request:44056|pull_request:44074", + "pull_request:43997|pull_request:44010", + "pull_request:44010|pull_request:44129" + ], + "prepared_review_unit_hash": "fdb40461182ac596bcdadd2ed17d2c61520a08c418b7beea7845c0a270658962", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84884, + "estimated_input_tokens": 21221, + "estimated_eval_tokens": 42698 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 6, + "serialized_chars": 17124, + "estimated_input_tokens": 4281, + "estimated_eval_tokens": 8818 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These PRs are all model-specific refactors around standardized output tracing/capture_outputs, but they touch different architectures and code paths. They are related by theme, not duplicates of the same change.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44073 is a good representative canonical PR: it clearly shows the standardized output-tracing migration pattern, includes the modern decorators/imports, and is scoped to a single model implementation.", + "best_issue_reason": null, + "best_pr_reason": "PR 44073 is the best exemplar in the set because its title and diff directly express the core refactor pattern (capture_outputs + can_return_tuple) without conflating it with unrelated model changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": false, + "reason": "Both are output-tracing refactors, but for different models (FocalNet vs ImageGPT) and different internals. They are not the same bug/change and would not be merged as one PR." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "VisualBert and TextNet are separate model implementations. The changes are similar refactors, but not the same underlying code-path problem." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "Both modernize output recording, but MPNet and VisualBert are unrelated model code paths. Similar pattern, not duplicate change." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "MPNet vs TextNet are distinct model refactors with no shared concrete bug or patch to combine." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are different architectures; these are separate standardized-output migrations, not one duplicate fix." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "SqueezeBert and SpeechT5 touch different model implementations and the SpeechT5 PR is tied to a specific issue. Same broad theme, but not the same underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 37, + "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": "af1df07749f8b683c6f32f8baf6741f585fffede22b4c73a4c18ff5906f734c6", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40174, + "estimated_input_tokens": 10044, + "estimated_eval_tokens": 20344 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40174, + "estimated_input_tokens": 10044, + "estimated_eval_tokens": 20344 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is a set of similar-looking but mostly unrelated refactor PRs across different model families, all centered on output tracing / `can_return_tuple` / `capture_outputs`. They share theme, not the same underlying change, so they should stay separate.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #44003 is the strongest representative of the cluster: it is the broadest single change here, spanning closely related Mamba/FalconMamba code and adding the output-capture machinery used by the refactor.", + "best_issue_reason": null, + "best_pr_reason": "PR #44003 best fits the cluster\u2019s overall theme and has the most substantial, cross-file implementation among the candidates.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model families (`timm_backbone` vs `upernet`); same refactor theme, but not the same bug or mergeable change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "Different code paths and models (`timm_backbone` vs `vision_text_dual_encoder`); only a broad output-tracing refactor in common." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Unrelated model implementations (`timm_backbone` vs `univnet`); no shared concrete fix." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "Different architectures and files; both are output-refactor work, but not the same underlying change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "`timm_backbone` and `codegen` are unrelated code paths; similar refactoring pattern only." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families (`timm_backbone` vs `rwkv`); no evidence of one shared fix." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "`mobilenet_v1` and `univnet` are unrelated; they only share the generic output-tracing refactor motif." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "Different subsystems and files (`mobilenet_v1` vs `upernet`); not the same concrete change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "`mobilenet_v1` and `depth_anything` are separate model families; same style of refactor, not duplicate work." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "`mobilenet_v1` vs `vision_encoder_decoder`; no shared code path or bug." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "`mobilenet_v1` and `speech_encoder_decoder` are different models; only superficial similarity." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "`vision_text_dual_encoder` and `univnet` are unrelated implementations; same refactor pattern is not enough." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model families (`vision_text_dual_encoder` vs `upernet`); no shared concrete fix." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "Different code paths and architectures; these are separate refactors, not one mergeable PR." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "`vision_text_dual_encoder` and `depth_anything` do not share the same underlying bug/change." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "`univnet` vs `upernet` are unrelated model code paths; only the generic output-tracing theme matches." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Different model families (`univnet` vs `mamba`); not the same fix." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "`univnet` and `depth_anything` are separate changes in different model stacks." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44029", + "accept": false, + "reason": "`upernet` and `rwkv` are unrelated code paths; no evidence of the same underlying issue." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "Different architectures and files; similar refactor intent only." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44028", + "accept": false, + "reason": "`mamba`/`falcon_mamba` vs `superpoint` are different model families; not one shared bug or change." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "`mamba` and `codegen` touch different model implementations; same refactor category, not a duplicate." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "Different code paths (`timm_backbone` vs `codegen`); the overlap is only in output-tracing cleanup." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "`univnet` and `codegen` are unrelated models; not mergeable as one fix." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "`upernet` and `codegen` do not address the same underlying issue." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "`depth_anything` and `vision_encoder_decoder` are different model stacks; broad theme match only." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "`depth_anything` and `speech_encoder_decoder` are separate refactors, not the same bug fix." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "`depth_anything` vs `superpoint`; different files and model logic." + }, + { + "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": "`depth_anything` and `dpr` do not share a concrete code-path problem." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "Encoder-decoder wrappers for different modalities; same style of cleanup, not the same change." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "`vision_encoder_decoder` and `superpoint` are different subsystems; no duplicate evidence." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "`vision_encoder_decoder` vs `rwkv` are unrelated model implementations." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "`vision_encoder_decoder` and `dpr` are separate refactors in different code paths." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "`speech_encoder_decoder` and `superpoint` are unrelated model families; not one underlying fix." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "`speech_encoder_decoder` vs `rwkv`; same output-tracing motif only." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "`speech_encoder_decoder` and `dpr` do not appear to fix the same code-path issue." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "`superpoint` and `rwkv` are separate model implementations; no shared concrete bug." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "`superpoint` and `dpr` touch different subsystems; not duplicates." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "`rwkv` and `dpr` are unrelated model families; only broad refactor similarity." + } + ] + }, + "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": 37, + "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": "d05a02297d867ef2099f202c2917cf70d822d346ef43a6afe813753edbd712aa", + "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-addition PRs: PenguinVL and Voxtral TTS. They overlap mainly in transformer registry/docs scaffolding, but they implement different models and code paths.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44662 is the better representative PR here: it is non-draft, has extensive review activity, and contains the more complete model integration.", + "best_issue_reason": null, + "best_pr_reason": "PR 44662 is the strongest standalone PR in the cluster because it is a fully fleshed-out implementation with substantial review history, whereas 45401 is still a draft tied to a specific issue.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Reject: these add different models (PenguinVL vs Voxtral TTS) with different architectures and tasks. Shared registry/docs changes are incidental and not the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 16, + "component_count": 37, + "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": "c857ad2b9689ac484e634d46ec50b70bb32fe6e80694aeb8c2ff2808bf738625", + "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 feature PRs around module/kernel fusion, but they target different layers and code paths: one introduces a generic Module Fusion API, the other implements n-to-1 kernel fusion via `KernelConfig` and integration plumbing. I do not treat them as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR stands out as a duplicate target; the PRs are adjacent but distinct feature work rather than the same change.", + "best_issue_reason": null, + "best_pr_reason": "PR 45363 is the more concrete end-user feature implementation, but it is still separate from the generic API work in PR 44979.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "Related fusion efforts, but not the same underlying change: 44979 adds a generic module fusion API, while 45363 wires in n-to-1 kernel fusion through `KernelConfig` and integration/modeling changes. They touch different abstractions and would not plausibly be merged into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 17, + "component_count": 37, + "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": "d48f8ba93b0b37e26d1f1a88357dd2c8808eb2673912f80351586fb25ab62150", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests target the same Mistral regex AttributeError in `tokenization_utils_tokenizers.py` and make the same code-path fix; `45317` is the more complete version because it adds regression tests.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "`45317` is the stronger representative: it fixes the same `_patch_mistral_regex` AttributeError as `45086` but also includes a regression test, making it the more complete patch for the same underlying bug.", + "best_issue_reason": null, + "best_pr_reason": "`45317` is the best overall candidate because it addresses the same concrete bug, keeps the same fix, and adds tests that validate the behavior change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same underlying bug and same code-path fix in `_patch_mistral_regex` (`backend_tokenizer.pre_tokenizer` -> `tokenizer.pre_tokenizer`); the second PR is just a more complete version with tests, so they are plausibly mergeable into one change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 18, + "component_count": 37, + "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-45115-3", + "nodes": [ + "pull_request:45115", + "pull_request:45569", + "pull_request:46565" + ], + "soft_pairs": [ + "pull_request:45115|pull_request:45569", + "pull_request:45569|pull_request:46565" + ], + "prepared_review_unit_hash": "bc2adbb4b9608ca4346a28ae079cf797a3436322dd2778cb0d5aeabf7169a3ed", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 9123, + "estimated_input_tokens": 2281, + "estimated_eval_tokens": 4818 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 9123, + "estimated_input_tokens": 2281, + "estimated_eval_tokens": 4818 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three items are pull requests in the Nemotron family, but they target different changes: one is a conversion-mapping/refactor PR, one is a larger NemotronH split/architecture PR, and one is an unrelated Nemotron3.5 ASR addition. No soft pair looks like the same concrete change.", + "confidence": 0.86, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45569 is the most central and substantial NemotronH implementation/refactor item in the set, with the broadest code and docs/test coverage.", + "best_issue_reason": null, + "best_pr_reason": "PR 45569 is the best representative of the cluster because it covers the main NemotronH rework, while 45115 is narrower conversion-mapping work and 46565 is a different model family entirely.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "Related NemotronH work, but not the same underlying change: 45115 is a conversion-mapping and inheritance refactor, while 45569 is a broader model split/addition with docs and tests." + }, + { + "left": "pull_request:45569", + "right": "pull_request:46565", + "accept": false, + "reason": "Different model families and code paths: NemotronH architecture work versus Nemotron3.5 ASR streaming. They are 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": 19, + "component_count": 37, + "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": "638bf2b526d636f0827c07b1c1fab086b014bdc8a478fcf1f3f499d7e184b1a8", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5466, + "estimated_input_tokens": 1367, + "estimated_eval_tokens": 2990 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5466, + "estimated_input_tokens": 1367, + "estimated_eval_tokens": 2990 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related but not duplicates: one PR introduces per-layer heterogeneous config support, while the other adds broader heterogeneous model/runtime changes that build on that foundation.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45333 is the cleaner canonical representative because it focuses on the foundational per-layer configuration layer; #45332 is a broader follow-on that includes modeling/cache integration.", + "best_issue_reason": null, + "best_pr_reason": "PR #45333 is the best single representative for the cluster if one must be chosen, since it is the more narrowly scoped core change and is likely prerequisite infrastructure for the larger modeling PR.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "They are adjacent pieces of the heterogeneous-model effort, but not the same concrete change: #45333 adds per-layer config plumbing, while #45332 adds modeling/cache support on top of that. They could be separate, sequential PRs rather than 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": 20, + "component_count": 37, + "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": "f6cf2316e6a74ed0ecf8ada272a4f03b07835d93e4b8e48f4bea45f81c01d508", + "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": "These are both typing-related PRs, but they address different work: one is a broad type-checking sweep across many transformer modules, while the other adds specific typing protocols/helpers and touches a smaller set of modeling/PEFT-related code paths. They should not be treated as duplicates.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: 45415 is a broad type-checking cleanup across many files, while 45425 is a separate typing-helper change focused on protocol definitions and related model utilities.", + "best_issue_reason": null, + "best_pr_reason": "PR 45415 is the better cluster representative because it is the broader umbrella typing sweep and subsumes more of the generic type-checking cleanup work.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Both are about typing, but they fix different concrete changes: 45415 is a wide annotation/type-checking pass, while 45425 adds typing protocol aliases and helper interfaces for specific modeling/PEFT code paths. They would not plausibly merge into one PR as the same underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 21, + "component_count": 37, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45513-2", + "nodes": [ + "pull_request:45513", + "pull_request:46084" + ], + "soft_pairs": [ + "pull_request:45513|pull_request:46084" + ], + "prepared_review_unit_hash": "e4614d0b70699a9da9507774ba28318ba97497761e4d43e082b5c8ce9536c273", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4948, + "estimated_input_tokens": 1237, + "estimated_eval_tokens": 2730 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4948, + "estimated_input_tokens": 1237, + "estimated_eval_tokens": 2730 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two unrelated pull requests about cached/multi-token forward behavior in different model families: one for Qwen3.5/OLMo hybrid/Qwen3-next linear attention, and one for Mamba2 use_cache with seq_len > 1. They do not look like the same underlying bug or a PR that could plausibly be merged with the other.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46084 is the cleaner canonical representative: it has an explicit issue target (#46032), a narrowly scoped Mamba2 fix, and a focused code-path change around cached forward with seq_len > 1.", + "best_issue_reason": null, + "best_pr_reason": "PR #46084 is the strongest standalone representative because it is tightly scoped, explicitly tied to a specific bug report, and its diff clearly addresses one concrete cached-forward edge case.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45513", + "right": "pull_request:46084", + "accept": false, + "reason": "Different model families and different bugs: #45513 fixes multi-token cached forward for GDN linear attention in Qwen3.5/OLMo/Qwen3-next, while #46084 fixes Mamba2 use_cache behavior when seq_len > 1. They are not the same underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 22, + "component_count": 37, + "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": "b6a70245f6ca0e42e45dc632971ec2874053148b579ebfc89f39f31f1eba1291", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs touch expert-parallel loading, but they fix different bugs: one removes `distributed_config` from serialization and adds `has_ep` for Accelerate/ZeRO-3 loading, while the other prevents expert parameters from being overwritten during FSDP2/rank-0 broadcast by post-wrapping sharded tensors. Related area, not the same change.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45662 is the better representative because it targets a concrete distributed-loading overwrite bug with broader code-path coverage, whereas #45548 is a narrower serialization/detection fix.", + "best_issue_reason": null, + "best_pr_reason": "PR #45662 is the strongest standalone PR in this cluster: it addresses a specific expert-parameter corruption path during distributed loading and includes the main fix in the loading stack.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Do not merge: they address different underlying failures. #45548 fixes EP/ZeRO-3 loading via Accelerate by excluding `distributed_config` from serialization and exposing `has_ep`; #45662 fixes expert parameters being overwritten after sharding/broadcast in FSDP2. Same subsystem, different bug and code path." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 23, + "component_count": 37, + "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": "92c690e818ecabaffa04db1f810812356454f46b0948ee84d348d87b90d07bb6", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4589, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4589, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs address the same xdist collision around captured_info test logs; one is a narrow fix in testing_utils, the other extends it with CI handling for multiple log files.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45645 is the more complete representative: it fixes the xdist-safe output filename in testing_utils and updates CI/log collection to handle the resulting per-worker files.", + "best_issue_reason": null, + "best_pr_reason": "PR #45645 best matches the cluster because it covers the core bug fix plus the downstream artifact collection changes needed for the new behavior.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45639", + "right": "pull_request:45645", + "accept": true, + "reason": "Both PRs fix the same concrete problem: patched testing logs colliding under pytest-xdist. #45645 appears to subsume #45639 with the same core code-path change plus related CI adjustments, so they could plausibly be merged into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 24, + "component_count": 37, + "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": "ee95f657f424021aa7e8a4a034eaba3b34dc6afb88ee7b5b6d4b557bfaccb570", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7249, + "estimated_input_tokens": 1813, + "estimated_eval_tokens": 3882 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7249, + "estimated_input_tokens": 1813, + "estimated_eval_tokens": 3882 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs are both broad linter-driven changes across many Transformer model files, but they target different underlying problems: one updates forward signatures/tuple returns, the other refactors weight initialization rules. They are not the same fix and should not be merged together.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45720 is the better canonical PR because it is a more concrete, narrowly scoped change: implementing part of TRF003 by adding `can_return_tuple` and adjusting forward signatures across affected models.", + "best_issue_reason": null, + "best_pr_reason": "PR #45720 is the best representative of the cluster since it has the clearest single code-path change, whereas PR #46259 is a separate initialization-rule refactor with different behavior.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45720", + "right": "pull_request:46259", + "accept": false, + "reason": "Reject: they touch overlapping model files, but the underlying changes are unrelated. #45720 changes forward/return-tuple behavior; #46259 changes `_init_weights`/initialization logic. Shared filenames and linter theme are not enough to treat them as the same PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 25, + "component_count": 37, + "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": 37, + "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": "600bb0c74e7cb2ab5ce235f159e9a9a854e5af41e16240c68886d3ae651e1dce", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5160, + "estimated_input_tokens": 1290, + "estimated_eval_tokens": 2836 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5160, + "estimated_input_tokens": 1290, + "estimated_eval_tokens": 2836 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs, but they are not duplicates: one is a CI/graphana-style testing and model-code PR, the other is a test-migration PR for kernel/DDP and vision model tests. They touch different code paths and aim at different changes.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45932 is the better representative of the cluster only because it has broader scope and more discussion activity; it is not a duplicate of #46100.", + "best_issue_reason": null, + "best_pr_reason": "PR #45932 is the strongest PR to keep as the cluster representative since it has more activity and a more central code-change footprint, while #46100 is a narrower test migration.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45932", + "right": "pull_request:46100", + "accept": false, + "reason": "Different underlying work: #45932 updates CI/secrets and several model implementations, while #46100 adds DDP kernel test fixtures and migrates AltCLIP/ViT tests. They do not fix the same concrete bug or change the same code path, so they should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 27, + "component_count": 37, + "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": "94b2c7f3d2b6d97a8fda48c022f431f52c25f39cadeba5bc797688367a75a167", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5524, + "estimated_input_tokens": 1381, + "estimated_eval_tokens": 3018 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5524, + "estimated_input_tokens": 1381, + "estimated_eval_tokens": 3018 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two PRs are not duplicates. PR #46126 is a focused EP + Trainer integration change, while PR #46269 is a broad distributed branch base that touches many unrelated distributed components and infrastructure. They share some filenames because #46269 likely subsumes the same area, but they do not appear to be the same concrete bugfix/change and should not be merged as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46126 is the better canonical PR for this cluster because it is the narrower, clearly scoped change centered on EP/Trainer integration. PR #46269 is too broad and looks like a branch-base/meta integration PR rather than a single discrete fix.", + "best_issue_reason": null, + "best_pr_reason": "PR #46126 is the best representative artifact here: it has a specific title, small targeted diff, and a coherent change set in distributed tensor parallel / trainer integration. PR #46269 spans many files and multiple distributed subsystems, so it is not a good single canonical change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46126", + "right": "pull_request:46269", + "accept": false, + "reason": "Shared distributed filenames are not enough to treat these as the same PR. #46126 is a focused EP + Trainer integration change; #46269 is a large branch-base PR touching many unrelated areas, so they do not look like the same underlying code-path fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 28, + "component_count": 37, + "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": "40439c2bb126b9af18f6a54b74f25f305f195eda8a56b8cca16837fae5d37f2a", + "component_budget": { + "node_count": 10, + "item_count": 10, + "soft_pair_count": 29, + "serialized_chars": 25926, + "estimated_input_tokens": 6482, + "estimated_eval_tokens": 13220 + }, + "budget": { + "node_count": 10, + "item_count": 10, + "soft_pair_count": 29, + "serialized_chars": 25926, + "estimated_input_tokens": 6482, + "estimated_eval_tokens": 13220 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated workflow-maintenance PRs. The only tight subgroup is the pr-ci-security-gate.yml changes, but even those are separate hardening edits rather than the same duplicate fix.", + "confidence": 0.83, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46554 is the most substantive and central workflow-security change in the set, and it anchors the only close mini-cluster around pr-ci-security-gate.yml.", + "best_issue_reason": null, + "best_pr_reason": "PR 46554 best represents the strongest concrete change in this cluster: a real hardening update to the PR CI security gate rather than a minor workflow tweak.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46138", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflows and different fixes: stdin handling for self-comment CI vs PR security-gate hardening." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46558", + "accept": false, + "reason": "Only a broad workflow-safety theme in common; the concrete changes are unrelated." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46563", + "accept": false, + "reason": "One updates comment parsing, the other adds a safety note in the PR security gate wrapper." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46716", + "accept": false, + "reason": "No shared bug/change beyond generic workflow hardening." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46504", + "accept": false, + "reason": "Different files and different shell-safety fixes in separate workflows." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46503", + "accept": false, + "reason": "Unrelated workflows: SSH runner security tweaks versus benchmark trigger logic." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46501", + "accept": false, + "reason": "Different workflow files and different permissions/schedule changes." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46502", + "accept": false, + "reason": "No same underlying bug; SSH runner cleanup versus reviewer assignment behavior." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46554", + "accept": false, + "reason": "Both are workflow hardening, but they fix different workflows and code paths." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46558", + "accept": false, + "reason": "Distinct workflow changes with no concrete duplication." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46503", + "accept": false, + "reason": "Separate workflows with different operational behavior changes." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46502", + "accept": false, + "reason": "Both are chore edits, but they modify different workflow logic entirely." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46504", + "accept": false, + "reason": "No duplicate bug/change; one alters benchmark triggers, the other shell quoting in build image CI." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflow areas and no shared concrete fix." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46503", + "accept": false, + "reason": "Reviewer assignment checkout behavior is unrelated to benchmark dispatch changes." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46504", + "accept": false, + "reason": "Different workflow files and different intent." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46554", + "accept": false, + "reason": "No same underlying code-path problem." + }, + { + "left": "pull_request:46503", + "right": "pull_request:46504", + "accept": false, + "reason": "Benchmark dispatch and build-image message parsing are unrelated changes." + }, + { + "left": "pull_request:46503", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflows and different fixes." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46554", + "accept": false, + "reason": "Similar security posture, but different workflows and different concrete changes." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46558", + "accept": false, + "reason": "No duplicate fix; build-image shell quoting versus PR security-gate permissions." + }, + { + "left": "pull_request:46554", + "right": "pull_request:46558", + "accept": false, + "reason": "Same file, but these are complementary hardening changes rather than the same concrete fix." + }, + { + "left": "pull_request:46554", + "right": "pull_request:46563", + "accept": false, + "reason": "One changes security checks, the other only documents trust assumptions in the wrapper." + }, + { + "left": "pull_request:46554", + "right": "pull_request:46716", + "accept": false, + "reason": "Actual security-gate logic change vs secrets-passing change; related but not the same fix." + }, + { + "left": "pull_request:46558", + "right": "pull_request:46563", + "accept": false, + "reason": "Different edits to the same workflow with different purposes." + }, + { + "left": "pull_request:46558", + "right": "pull_request:46716", + "accept": false, + "reason": "Permissions/approval gating versus secrets inheritance are distinct changes." + }, + { + "left": "pull_request:46563", + "right": "pull_request:46716", + "accept": false, + "reason": "Comment-only safety note is not the same as changing secrets inheritance." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46563", + "accept": false, + "reason": "Different files and unrelated workflow fixes." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46716", + "accept": false, + "reason": "Shell quoting in build CI is unrelated to PR CI secrets handling." + } + ] + }, + "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": 37, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46152-2", + "nodes": [ + "pull_request:46152", + "pull_request:46207" + ], + "soft_pairs": [ + "pull_request:46152|pull_request:46207" + ], + "prepared_review_unit_hash": "561b3c91a7ee420e22ea692f1cdc35ba46ce0cfcbd9447e218469ab759ea657f", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4276, + "estimated_input_tokens": 1069, + "estimated_eval_tokens": 2394 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4276, + "estimated_input_tokens": 1069, + "estimated_eval_tokens": 2394 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs share a broad theme of replacing file-based source inspection with safer source lookup, but they target different code paths and should not be deduplicated.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46152 is the better canonical anchor: it introduces a focused helper in import_utils to avoid reading Transformers\u2019 own source via open(), which is a self-contained utility-level fix.", + "best_issue_reason": null, + "best_pr_reason": "PR 46152 is the clearest representative of the cluster\u2019s theme because it is a direct, reusable source-access change rather than a separate model-internal heuristic update.", + "soft_edge_verdicts": [ + { + "left": "Avoid reading self source code via file open()", + "right": "Use inspect.getsource instead of open() for source-reading in _can_set_*_implementation", + "accept": false, + "reason": "Related pattern, but different concrete code paths: import_utils package resource resolution vs modeling_utils attention-implementation detection. They are not the same underlying bug/change and would not plausibly merge into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 30, + "component_count": 37, + "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-46154-2", + "nodes": [ + "pull_request:46154", + "pull_request:46196" + ], + "soft_pairs": [ + "pull_request:46154|pull_request:46196" + ], + "prepared_review_unit_hash": "402cbd316d8ea27d683ad83cf680a822482ee50d2d56d69a9fc832a2c44f0598", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5700, + "estimated_input_tokens": 1425, + "estimated_eval_tokens": 3106 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5700, + "estimated_input_tokens": 1425, + "estimated_eval_tokens": 3106 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests are about SDPA attention dispatch, but they target different model families and code paths (T5 vs wav2vec2/audio models). They are related by theme, not duplicates, and should not be merged as one PR.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR for duplication purposes: these are separate model-specific changes rather than two copies of the same fix. If a representative must be chosen, neither is clearly canonical over the other.", + "best_issue_reason": null, + "best_pr_reason": "PR #46196 is the broader and more substantial SDPA-dispatch change set, touching many audio models and associated docs/tests, so it is the better representative of the cluster theme.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46154", + "right": "pull_request:46196", + "accept": false, + "reason": "Same high-level theme (SDPA dispatch) but different concrete implementations and model families; not the same underlying bug/change and not plausibly one merged PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 31, + "component_count": 37, + "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-46173-2", + "nodes": [ + "pull_request:46173", + "pull_request:46474" + ], + "soft_pairs": [ + "pull_request:46173|pull_request:46474" + ], + "prepared_review_unit_hash": "8ac88de891156efc041ae6eac2ff2b6f957f37c7ee42f44803b224f91f9b44ec", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3631, + "estimated_input_tokens": 908, + "estimated_eval_tokens": 2072 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3631, + "estimated_input_tokens": 908, + "estimated_eval_tokens": 2072 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests make the same code change in `src/transformers/cache_utils.py` to widen `StaticLayer.get_seq_length` from `int` to `int | torch.Tensor`; they are near-duplicate implementations of the same fix for issue #45987.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46173 is the better canonical representative because it is the earlier, more explicitly titled version of the same fix and already points to the tracking issue #45987.", + "best_issue_reason": null, + "best_pr_reason": "PR #46173 is the strongest standalone representative of the change: it contains the same minimal patch, the same target issue, and a clearer description of the intended fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46173", + "right": "pull_request:46474", + "accept": true, + "reason": "Same underlying bug fix, same file, same diff hunk, and same explicit issue target (#45987); these are functionally duplicate PRs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both PRs touch the same file, make the same type-annotation change for `StaticLayer.get_seq_length`, and both explicitly target issue #45987. The canonical-pr rationale is also conservative and supported by the titles/metadata. No overstatement beyond the visible diff evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 32, + "component_count": 37, + "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": "9faa1aeadd68146d36ec14b61b72665e01c038a1534c8858fa67708060c9bbbe", + "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 boundary bug: the right-side mask used a strict `< right_window_size` check instead of including the edge. They also target the same issue and share the same Moonshine streaming code path, with 46546 extending the fix to Gemma4 and tests.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46546 is the better canonical representative because it addresses the same boundary bug in Moonshine streaming, also applies the same fix to Gemma4, and includes test coverage.", + "best_issue_reason": null, + "best_pr_reason": "PR 46546 is the strongest overall candidate: it covers the shared concrete bug, touches the relevant implementations, and adds tests; 46310 is the narrower subset patch.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46310", + "right": "pull_request:46546", + "accept": true, + "reason": "Same underlying boundary-condition fix in the same sliding-window mask logic, and the smaller PR is a subset of the broader one; they could plausibly be merged into one change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 33, + "component_count": 37, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46376-2", + "nodes": [ + "pull_request:46376", + "pull_request:46417" + ], + "soft_pairs": [ + "pull_request:46376|pull_request:46417" + ], + "prepared_review_unit_hash": "ca6f763e662935a8caf7f447b8503d39170eefb03c6a1b8b47e28e283c07e8ad", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5939, + "estimated_input_tokens": 1485, + "estimated_eval_tokens": 3226 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5939, + "estimated_input_tokens": 1485, + "estimated_eval_tokens": 3226 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two unrelated pull requests: one adds multimodal support to continuous batching, the other adds the HunYuan VL model and its docs/tests. There is no meaningful evidence they fix the same underlying bug or implement the same change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46417 is the better canonical PR candidate because it is a complete, well-scoped model addition with substantial review activity; PR 46376 is a draft WIP for a different subsystem and not a duplicate of 46417.", + "best_issue_reason": null, + "best_pr_reason": "PR 46417 is the strongest representative for the cluster only because it is the more mature and self-contained change; however, the two PRs should not be merged as duplicates since they target different code paths and feature areas.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46376", + "right": "pull_request:46417", + "accept": false, + "reason": "Different goals and file sets: continuous batching/multimodal infrastructure vs. a new HunYuan VL model integration. They do not appear to fix the same concrete problem or be mergeable into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 34, + "component_count": 37, + "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": "b5b6886031257f63c89ec2448c269c459910a8b5f8e1970d72f9fb0e8ef2594e", + "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 behavior that strips real leading spaces, while the other fixes DeepSeek-Coder/DeepSeek-LLM being routed to the wrong tokenizer class. They should not be clustered as duplicates.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: these address different code paths and different bugs, so neither supersedes the other.", + "best_issue_reason": null, + "best_pr_reason": "No global best PR for deduping this cluster; PR #46574 is a focused decode fix and PR #46580 is a separate tokenizer-class routing fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46574", + "right": "pull_request:46580", + "accept": false, + "reason": "Different underlying problems: CodeLlama decoder assignment/leading-space stripping vs DeepSeek checkpoint-to-tokenizer-class override in auto loading. They are not mergeable into one fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 35, + "component_count": 37, + "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": "efb8323e3a772f5d212cf91f06e9d94d7a2910638fa431894a2ea71d15f9e882", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5495, + "estimated_input_tokens": 1374, + "estimated_eval_tokens": 3004 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5495, + "estimated_input_tokens": 1374, + "estimated_eval_tokens": 3004 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs overlap on Mistral tokenizer files, but they are not the same change: one is a refactor that moves `MistralConverter` into the Mistral integration package, while the other adds tekken tokenizer detection/conversion/save utilities. Not a duplicate pair.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "No exact PR duplicate here; if one PR has to represent the cluster, `#46604` is the better canonical PR because it is the broader, feature-adding change, while `#46603` is a supporting refactor/move.", + "best_issue_reason": null, + "best_pr_reason": "`#46604` is the most substantive and central change in the set, with new tokenizer detection/conversion/save behavior. `#46603` is narrower and mainly reorganizes existing Mistral conversion code.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46603", + "right": "pull_request:46604", + "accept": false, + "reason": "Shared Mistral files alone are not enough: `#46603` relocates `MistralConverter`, while `#46604` implements tekken tokenizer support and save utilities. Different underlying code-path changes, so not mergeable as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 36, + "component_count": 37, + "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-46741-4", + "nodes": [ + "pull_request:46741", + "pull_request:46748", + "pull_request:46773", + "pull_request:46781" + ], + "soft_pairs": [ + "pull_request:46741|pull_request:46781", + "pull_request:46741|pull_request:46773", + "pull_request:46741|pull_request:46748", + "pull_request:46748|pull_request:46781", + "pull_request:46748|pull_request:46773" + ], + "prepared_review_unit_hash": "fc138322b12c5095c135a8c72037ba61b5679b6d22485da2f59e86efae1bc528", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 5, + "serialized_chars": 10650, + "estimated_input_tokens": 2663, + "estimated_eval_tokens": 5582 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 5, + "serialized_chars": 10650, + "estimated_input_tokens": 2663, + "estimated_eval_tokens": 5582 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Four distinct PRs about hybrid-model/runtime edge cases: Mamba2 chunked-prefill, cache prefetch on offloaded hybrid layers, padding mask handling for batch size 1, and QuantizedCache layer construction. They share a loose theme but do not fix the same bug or code path.", + "confidence": 0.86, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:46748 is the best representative anchor: it is a focused core cache_utils fix for hybrid models, has a clear runtime bug description, and sits in the shared infrastructure area, while the other PRs are separate model- or cache-specific bugs.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:46748 is the strongest standalone PR for the cluster\u2019s general theme because it touches shared cache infrastructure and addresses a concrete hybrid-model failure mode with tests.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46741", + "right": "pull_request:46781", + "accept": false, + "reason": "Both relate to hybrid/SSM models, but one fixes Mamba2 chunked-prefill/speculative decode in model forward paths while the other fixes QuantizedCache construction in cache_utils; different bug and code path." + }, + { + "left": "pull_request:46741", + "right": "pull_request:46773", + "accept": false, + "reason": "Different failure modes: chunked-prefill/speculative decoding vs padding-mask application for batch-size-1 linear-attention models. Broadly related only by model family." + }, + { + "left": "pull_request:46741", + "right": "pull_request:46748", + "accept": false, + "reason": "One is a Mamba2/Nemotron-H forward-path fix; the other is cache prefetch device handling. Same ecosystem, not the same underlying bug." + }, + { + "left": "pull_request:46748", + "right": "pull_request:46781", + "accept": false, + "reason": "Both modify cache_utils for hybrid models, but one fixes offloaded cache prefetch device mismatch and the other fixes QuantizedCache layer selection/crash. Different concrete code paths." + }, + { + "left": "pull_request:46748", + "right": "pull_request:46773", + "accept": false, + "reason": "Shared hybrid-model context is too broad; cache prefetch mismatch and padding-mask skipping are unrelated bugs." + } + ] + }, + "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": 37, + "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-46753-2", + "nodes": [ + "pull_request:46753", + "pull_request:46778" + ], + "soft_pairs": [ + "pull_request:46753|pull_request:46778" + ], + "prepared_review_unit_hash": "d4befb28cb7d92ac7de5bef007853e0ef853bcea9183629c40bbfb7c71d4676c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4478, + "estimated_input_tokens": 1120, + "estimated_eval_tokens": 2496 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4478, + "estimated_input_tokens": 1120, + "estimated_eval_tokens": 2496 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests implement the same fix in `apply_chat_template` for empty conversations, touching the same code path and tests with only minor wording/test-harness differences. They are duplicate PRs rather than separate changes.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46778 is the better canonical PR: it is the more polished follow-up with clearer error text and the same underlying code-path fix, so it best represents the intended change set.", + "best_issue_reason": null, + "best_pr_reason": "PR 46778 is the strongest representative because it matches PR 46753\u2019s fix while using clearer messaging and slightly more complete test coverage.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46753", + "right": "pull_request:46778", + "accept": true, + "reason": "Same underlying bug fix: both add an empty-conversation guard in `apply_chat_template` and update the same test file for the same issue target. Differences are cosmetic, so they are duplicate PRs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both PRs target the same issue, touch the same files and code path, and differ mainly in the error text plus a test decorator/harness detail. The soft-edge verdict is conservative enough, and these changes look mergeable as a single bugfix PR." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + } + ] +} diff --git a/snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/manifest.json b/snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..50363a031def6d1963cd08491a2d4af573d916b2 --- /dev/null +++ b/snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260620t120016z", + "artifacts": { + "hybrid": "snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-06-20T12:05:23Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260620T120016Z", + "variant": "hybrid" +} diff --git a/snapshots/20260620T120016Z/manifest.json b/snapshots/20260620T120016Z/manifest.json index dfc87a2895fdb40d5196f09ecc038be4df161a30..7561ec638af82f4c256fd3a6ca27e905a7040b94 100644 --- a/snapshots/20260620T120016Z/manifest.json +++ b/snapshots/20260620T120016Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 53 }, "extracted_at": "2026-06-20T12:00:16Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260620t120016z", + "runs": { + "hybrid-model-20260620t120016z": { + "analysis_id": "hybrid-model-20260620t120016z", + "artifacts": { + "hybrid": "snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260620T120016Z/analysis-runs/hybrid-model-20260620t120016z/manifest.json", + "model": null, + "published_at": "2026-06-20T12:05:23Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260620T120016Z", "watermark": {