{ "config_id": "lot_paper_general", "baseline_id": "lot_paper_general", "method_family": "LOT_paper", "variant": "paper_general", "lot_config": "code/cfg/merge/lot_baselines.yaml", "lot_config_sha256": "1eff3fef06e727303a872c951cbbfe25fc187e5299111a9fbfd4fa0e506f5316", "resolved_lot_config": "outputs/lot_configs/lot_paper_general.yaml", "resolved_lot_config_sha256": "a78ddb942302f0565edabaa3c2d968700c04d71f8fe94c3ab996f31e4b1646fa", "base_model": { "alias": "base", "repo": "Qwen/Qwen3-8B-Base", "revision": "49e3418fbbbca6ecbdf9608b4d22e5a407081db4" }, "source_models": [ { "role": "instruction", "alias": "instr_only", "repo": "Qwen/Qwen3-8B", "revision": "b968826d9c46dd6066d109eabc6255188de91218" }, { "role": "reasoning", "alias": "reasoning_only", "repo": "OpenDataArena/Qwen3-8B-ODA-Math-460k", "revision": "8e8758ee23e5d959dd844b3bb5c5344219544795" }, { "role": "uncensored", "alias": "uncensored_only", "repo": "mlabonne/Qwen3-8B-abliterated", "revision": "30c72fa348f37c72d12ecbb259068ddee98aa9ed" } ], "model_config_compatibility": { "strict_fields": [ "model_type", "hidden_size", "intermediate_size", "num_hidden_layers", "num_attention_heads", "num_key_value_heads", "vocab_size", "tie_word_embeddings", "rope_theta" ], "informational_fields": [ "max_position_embeddings" ], "informational_mismatches": [ "instr_only.max_position_embeddings=40960 != base.max_position_embeddings=32768", "uncensored_only.max_position_embeddings=40960 != base.max_position_embeddings=32768" ], "configs": { "base": { "model_type": "qwen3", "hidden_size": 4096, "intermediate_size": 12288, "num_hidden_layers": 36, "num_attention_heads": 32, "num_key_value_heads": 8, "vocab_size": 151936, "tie_word_embeddings": false, "rope_theta": 1000000, "max_position_embeddings": 32768 }, "instr_only": { "model_type": "qwen3", "hidden_size": 4096, "intermediate_size": 12288, "num_hidden_layers": 36, "num_attention_heads": 32, "num_key_value_heads": 8, "vocab_size": 151936, "tie_word_embeddings": false, "rope_theta": 1000000, "max_position_embeddings": 40960 }, "reasoning_only": { "model_type": "qwen3", "hidden_size": 4096, "intermediate_size": 12288, "num_hidden_layers": 36, "num_attention_heads": 32, "num_key_value_heads": 8, "vocab_size": 151936, "tie_word_embeddings": false, "rope_theta": 1000000, "max_position_embeddings": 32768 }, "uncensored_only": { "model_type": "qwen3", "hidden_size": 4096, "intermediate_size": 12288, "num_hidden_layers": 36, "num_attention_heads": 32, "num_key_value_heads": 8, "vocab_size": 151936, "tie_word_embeddings": false, "rope_theta": 1000000, "max_position_embeddings": 40960 } } }, "tokenizer_source": { "alias": "instr_only", "repo": "Qwen/Qwen3-8B", "revision": "b968826d9c46dd6066d109eabc6255188de91218" }, "tokenizer_validation": { "vocab_size": 151669, "pad_token_id": 151643, "eos_token_id": 151645, "chat_template_present": true, "enable_thinking_false_chars": 74, "enable_thinking_true_chars": 55 }, "calibration_config": "code/cfg/calibration/calib_mix.yaml", "calibration_config_sha256": "5056fa7dac3267a5d41183c64de871896b67f80c1ef44133f42c61ca2da7c0a5", "calibration_manifest": "data/calibration/calibration_manifest.json", "calibration_manifest_sha256": "66ad7f4f2b5414120b911fc942948ca9698bbf914a988f3d79505595f526211b", "non_overlap_report": "data/calibration/non_overlap_report.json", "non_overlap_report_sha256": "2f298f021a70209eb9fa7bc85ebebe29179be77dd936b753586b2156a17f6e77", "source_calibrations": { "instruction": { "model_alias": "instr_only", "calibration_id": "general", "calibration_file": "data/calibration/calib_general.jsonl", "calibration_file_sha256": "6540f9aba0cec73a0fcecde5c7b0d26d5ccbc835492964210221fecb18796a31", "module_count": 324, "rows_per_module_min": 9216, "rows_per_module_max": 9216 }, "reasoning": { "model_alias": "reasoning_only", "calibration_id": "general", "calibration_file": "data/calibration/calib_general.jsonl", "calibration_file_sha256": "6540f9aba0cec73a0fcecde5c7b0d26d5ccbc835492964210221fecb18796a31", "module_count": 324, "rows_per_module_min": 9216, "rows_per_module_max": 9216 }, "uncensored": { "model_alias": "uncensored_only", "calibration_id": "general", "calibration_file": "data/calibration/calib_general.jsonl", "calibration_file_sha256": "6540f9aba0cec73a0fcecde5c7b0d26d5ccbc835492964210221fecb18796a31", "module_count": 324, "rows_per_module_min": 9216, "rows_per_module_max": 9216 } }, "feature_settings": { "input_format": "raw_text", "text_field": "text", "max_length": 512, "prompts_per_source": 96, "tokens_per_prompt": 96, "batch_size": 16, "limit_layers": null, "limit_modules": [ "down_proj", "gate_proj", "k_proj", "o_proj", "q_proj", "up_proj", "v_proj" ] }, "solver": { "type": "paper", "output_scale": 1.0, "rms_eps": 1e-06, "div_eps": 1e-10, "pinv_rcond": 0.0001, "fallback_on_nonfinite": true, "max_delta_norm_ratio": 3.0 }, "tensor_settings": { "linear_modules": [ "q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj" ], "norm_modules": { "enabled": true, "strategy": "lot_paper_norm", "strategy_excluded": "mean_delta", "include_suffixes": [ "input_layernorm", "post_attention_layernorm" ] }, "embeddings": { "strategy": "mean_delta" }, "lm_head": { "strategy": "mean_delta" }, "unsupported_float": { "strategy": "mean_delta" }, "non_float": { "strategy": "copy_base" } }, "merge_stats": { "tensor_count": 399, "strategy_counts": { "mean_delta": 78, "lot_paper_norm": 72, "lot_paper": 249 }, "fallback_count": 3, "tensors": [ { "tensor": "model.embed_tokens.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "embedding tensor", "fallback": false, "fallback_reason": null, "delta_norm": 23.47245979309082, "mean_delta_norm": 23.47245979309082, "correction_norm": 0.0, "shape": [ 151936, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.04480249807238579, "mean_delta_norm": 0.044814255088567734, "correction_norm": 0.04480249807238579, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.mlp.down_proj.weight", "strategy": "mean_delta", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": true, "fallback_reason": "LOT paper delta_norm/mean_delta_norm = 3.577 > max_delta_norm_ratio 3.000", "delta_norm": 9.695879936218262, "mean_delta_norm": 9.695879936218262, "correction_norm": 34.68302917480469, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 22.250226974487305, "mean_delta_norm": 9.323307991027832, "correction_norm": 22.250226974487305, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 22.15165901184082, "mean_delta_norm": 9.318660736083984, "correction_norm": 22.15165901184082, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.07153502851724625, "mean_delta_norm": 0.07055182754993439, "correction_norm": 0.07153502851724625, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.05836870148777962, "mean_delta_norm": 0.05836870148777962, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.self_attn.k_proj.weight", "strategy": "mean_delta", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": true, "fallback_reason": "LOT paper delta_norm/mean_delta_norm = 4.876 > max_delta_norm_ratio 3.000", "delta_norm": 2.6395866870880127, "mean_delta_norm": 2.6395866870880127, "correction_norm": 12.870046615600586, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 6.050539970397949, "mean_delta_norm": 5.091900825500488, "correction_norm": 6.050539970397949, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.05779934674501419, "mean_delta_norm": 0.05779934674501419, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.self_attn.q_proj.weight", "strategy": "mean_delta", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": true, "fallback_reason": "LOT paper delta_norm/mean_delta_norm = 5.116 > max_delta_norm_ratio 3.000", "delta_norm": 5.248283863067627, "mean_delta_norm": 5.248283863067627, "correction_norm": 26.84928321838379, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.0.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 5.820149898529053, "mean_delta_norm": 2.4616315364837646, "correction_norm": 5.820149898529053, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.06230851635336876, "mean_delta_norm": 0.06196029111742973, "correction_norm": 0.06230851635336876, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 6.135589599609375, "mean_delta_norm": 7.033468723297119, "correction_norm": 6.135589599609375, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 14.91633415222168, "mean_delta_norm": 5.775701999664307, "correction_norm": 14.91633415222168, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.339476585388184, "mean_delta_norm": 5.62913179397583, "correction_norm": 9.339476585388184, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.07564651966094971, "mean_delta_norm": 0.07497785240411758, "correction_norm": 0.07564651966094971, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.025815490633249283, "mean_delta_norm": 0.025815490633249283, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.167041301727295, "mean_delta_norm": 2.6107921600341797, "correction_norm": 4.167041301727295, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.347004890441895, "mean_delta_norm": 5.162838935852051, "correction_norm": 8.347004890441895, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.024171017110347748, "mean_delta_norm": 0.024171017110347748, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.764363288879395, "mean_delta_norm": 5.409148216247559, "correction_norm": 8.764363288879395, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.1.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.477911949157715, "mean_delta_norm": 2.501331090927124, "correction_norm": 2.477911949157715, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.0659150630235672, "mean_delta_norm": 0.06448334455490112, "correction_norm": 0.0659150630235672, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 11.54808521270752, "mean_delta_norm": 6.383207321166992, "correction_norm": 11.54808521270752, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.57168197631836, "mean_delta_norm": 6.046635150909424, "correction_norm": 10.57168197631836, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.408438682556152, "mean_delta_norm": 5.601812839508057, "correction_norm": 7.408438682556152, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.08489774912595749, "mean_delta_norm": 0.08243715018033981, "correction_norm": 0.08489774912595749, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04085328429937363, "mean_delta_norm": 0.04085328429937363, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.821956157684326, "mean_delta_norm": 2.845982551574707, "correction_norm": 4.821956157684326, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.4689812660217285, "mean_delta_norm": 5.234163284301758, "correction_norm": 7.4689812660217285, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.043157853186130524, "mean_delta_norm": 0.043157853186130524, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.148866653442383, "mean_delta_norm": 5.8178253173828125, "correction_norm": 9.148866653442383, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.2.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.522657632827759, "mean_delta_norm": 2.6065211296081543, "correction_norm": 3.522657632827759, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.06919640302658081, "mean_delta_norm": 0.06723569333553314, "correction_norm": 0.06919640302658081, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 12.864377975463867, "mean_delta_norm": 7.487030982971191, "correction_norm": 12.864377975463867, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.04465389251709, "mean_delta_norm": 7.279012203216553, "correction_norm": 9.04465389251709, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.6931586265563965, "mean_delta_norm": 6.697919845581055, "correction_norm": 4.6931586265563965, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.08671362698078156, "mean_delta_norm": 0.08344621956348419, "correction_norm": 0.08671362698078156, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.021300220862030983, "mean_delta_norm": 0.021300220862030983, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.328273296356201, "mean_delta_norm": 2.850520133972168, "correction_norm": 4.328273296356201, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 6.818342208862305, "mean_delta_norm": 5.273718357086182, "correction_norm": 6.818342208862305, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.023523632436990738, "mean_delta_norm": 0.023523632436990738, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.636321067810059, "mean_delta_norm": 5.902939796447754, "correction_norm": 8.636321067810059, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.3.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.2091987133026123, "mean_delta_norm": 2.617802858352661, "correction_norm": 3.2091987133026123, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.08032214641571045, "mean_delta_norm": 0.07930061966180801, "correction_norm": 0.08032214641571045, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 14.76630687713623, "mean_delta_norm": 9.387441635131836, "correction_norm": 14.76630687713623, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.518941879272461, "mean_delta_norm": 9.702797889709473, "correction_norm": 9.518941879272461, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.736515045166016, "mean_delta_norm": 8.777302742004395, "correction_norm": 4.736515045166016, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.08936761319637299, "mean_delta_norm": 0.08658290654420853, "correction_norm": 0.08936761319637299, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.03788274899125099, "mean_delta_norm": 0.03788274899125099, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.7855682373046875, "mean_delta_norm": 3.0299875736236572, "correction_norm": 4.7855682373046875, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.452361106872559, "mean_delta_norm": 5.468823432922363, "correction_norm": 7.452361106872559, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.038900475949048996, "mean_delta_norm": 0.038900475949048996, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.483064651489258, "mean_delta_norm": 6.187544822692871, "correction_norm": 9.483064651489258, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.4.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.484611749649048, "mean_delta_norm": 2.688159465789795, "correction_norm": 3.484611749649048, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.07428674399852753, "mean_delta_norm": 0.07329247891902924, "correction_norm": 0.07428674399852753, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 22.180747985839844, "mean_delta_norm": 10.341667175292969, "correction_norm": 22.180747985839844, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.224374771118164, "mean_delta_norm": 10.58188247680664, "correction_norm": 9.224374771118164, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 6.099387168884277, "mean_delta_norm": 9.787972450256348, "correction_norm": 6.099387168884277, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.08644147217273712, "mean_delta_norm": 0.08425557613372803, "correction_norm": 0.08644147217273712, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.028221406042575836, "mean_delta_norm": 0.028221406042575836, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.257584095001221, "mean_delta_norm": 3.0953361988067627, "correction_norm": 4.257584095001221, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.56968879699707, "mean_delta_norm": 5.431013584136963, "correction_norm": 8.56968879699707, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.03216938301920891, "mean_delta_norm": 0.03216938301920891, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.708023548126221, "mean_delta_norm": 6.214767932891846, "correction_norm": 7.708023548126221, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.5.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.059596061706543, "mean_delta_norm": 2.800551652908325, "correction_norm": 3.059596061706543, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.0813089907169342, "mean_delta_norm": 0.08001051098108292, "correction_norm": 0.0813089907169342, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.3735175132751465, "mean_delta_norm": 10.950376510620117, "correction_norm": 3.3735175132751465, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.23144817352295, "mean_delta_norm": 11.160828590393066, "correction_norm": 8.23144817352295, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 6.22568941116333, "mean_delta_norm": 10.403380393981934, "correction_norm": 6.22568941116333, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.08602486550807953, "mean_delta_norm": 0.08461999893188477, "correction_norm": 0.08602486550807953, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.05031709745526314, "mean_delta_norm": 0.05031709745526314, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.3363213539123535, "mean_delta_norm": 3.2020859718322754, "correction_norm": 4.3363213539123535, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.395118713378906, "mean_delta_norm": 5.834267616271973, "correction_norm": 10.395118713378906, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04874975606799126, "mean_delta_norm": 0.04874975606799126, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.103529930114746, "mean_delta_norm": 6.5630903244018555, "correction_norm": 8.103529930114746, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.6.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.0858423709869385, "mean_delta_norm": 2.8661069869995117, "correction_norm": 3.0858423709869385, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.0357649326324463, "mean_delta_norm": 3.1656601428985596, "correction_norm": 3.0357649326324463, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 6.0421342849731445, "mean_delta_norm": 6.4129204750061035, "correction_norm": 6.0421342849731445, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.5505003929138184, "mean_delta_norm": 3.084667682647705, "correction_norm": 2.5505003929138184, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.10172207653522491, "mean_delta_norm": 0.10162247717380524, "correction_norm": 0.10172207653522491, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 29.33161163330078, "mean_delta_norm": 12.811537742614746, "correction_norm": 29.33161163330078, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 18.993640899658203, "mean_delta_norm": 14.023926734924316, "correction_norm": 18.993640899658203, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 15.912165641784668, "mean_delta_norm": 12.270183563232422, "correction_norm": 15.912165641784668, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.10502553731203079, "mean_delta_norm": 0.10353638976812363, "correction_norm": 0.10502553731203079, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04411710053682327, "mean_delta_norm": 0.04411710053682327, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.070336818695068, "mean_delta_norm": 3.6965994834899902, "correction_norm": 4.070336818695068, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 11.404263496398926, "mean_delta_norm": 6.716432094573975, "correction_norm": 11.404263496398926, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04809814691543579, "mean_delta_norm": 0.04809814691543579, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.378407001495361, "mean_delta_norm": 7.449376106262207, "correction_norm": 7.378407001495361, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.10.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.0941202640533447, "mean_delta_norm": 3.543776512145996, "correction_norm": 3.0941202640533447, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.10968691110610962, "mean_delta_norm": 0.10866089165210724, "correction_norm": 0.10968691110610962, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 30.931379318237305, "mean_delta_norm": 13.613249778747559, "correction_norm": 30.931379318237305, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.771326065063477, "mean_delta_norm": 14.622610092163086, "correction_norm": 20.771326065063477, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 17.81013298034668, "mean_delta_norm": 12.939362525939941, "correction_norm": 17.81013298034668, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11141763627529144, "mean_delta_norm": 0.10989778488874435, "correction_norm": 0.11141763627529144, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04776467755436897, "mean_delta_norm": 0.04776467755436897, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.304965019226074, "mean_delta_norm": 3.991931438446045, "correction_norm": 4.304965019226074, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 12.762998580932617, "mean_delta_norm": 7.215724468231201, "correction_norm": 12.762998580932617, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.048151321709156036, "mean_delta_norm": 0.048151321709156036, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.94119930267334, "mean_delta_norm": 8.007794380187988, "correction_norm": 7.94119930267334, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.11.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.455146312713623, "mean_delta_norm": 3.804434299468994, "correction_norm": 3.455146312713623, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11099476367235184, "mean_delta_norm": 0.11061286181211472, "correction_norm": 0.11099476367235184, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 30.497100830078125, "mean_delta_norm": 14.123055458068848, "correction_norm": 30.497100830078125, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 22.61113739013672, "mean_delta_norm": 15.212843894958496, "correction_norm": 22.61113739013672, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 19.50568389892578, "mean_delta_norm": 13.551444053649902, "correction_norm": 19.50568389892578, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.10508298128843307, "mean_delta_norm": 0.10403671860694885, "correction_norm": 0.10508298128843307, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.05177808925509453, "mean_delta_norm": 0.05177808925509453, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.6967010498046875, "mean_delta_norm": 4.200942039489746, "correction_norm": 4.6967010498046875, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.20080852508545, "mean_delta_norm": 7.094746112823486, "correction_norm": 10.20080852508545, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.052902545779943466, "mean_delta_norm": 0.052902545779943466, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.56981372833252, "mean_delta_norm": 8.524067878723145, "correction_norm": 9.56981372833252, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.12.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.037657260894775, "mean_delta_norm": 3.800150156021118, "correction_norm": 4.037657260894775, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12880083918571472, "mean_delta_norm": 0.1286391019821167, "correction_norm": 0.12880083918571472, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 33.24708557128906, "mean_delta_norm": 14.960352897644043, "correction_norm": 33.24708557128906, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 23.31303596496582, "mean_delta_norm": 15.825604438781738, "correction_norm": 23.31303596496582, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.736242294311523, "mean_delta_norm": 14.414816856384277, "correction_norm": 20.736242294311523, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11617399007081985, "mean_delta_norm": 0.1149541437625885, "correction_norm": 0.11617399007081985, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04237139970064163, "mean_delta_norm": 0.04237139970064163, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.440920352935791, "mean_delta_norm": 4.348404884338379, "correction_norm": 4.440920352935791, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 11.309403419494629, "mean_delta_norm": 7.551999568939209, "correction_norm": 11.309403419494629, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04107409715652466, "mean_delta_norm": 0.04107409715652466, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.803726196289062, "mean_delta_norm": 8.715754508972168, "correction_norm": 8.803726196289062, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.13.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.094149589538574, "mean_delta_norm": 4.080448627471924, "correction_norm": 4.094149589538574, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11896278709173203, "mean_delta_norm": 0.11855502426624298, "correction_norm": 0.11896278709173203, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 29.010181427001953, "mean_delta_norm": 15.310876846313477, "correction_norm": 29.010181427001953, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 23.731096267700195, "mean_delta_norm": 16.449153900146484, "correction_norm": 23.731096267700195, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.86929702758789, "mean_delta_norm": 14.697593688964844, "correction_norm": 20.86929702758789, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12019939720630646, "mean_delta_norm": 0.1190696656703949, "correction_norm": 0.12019939720630646, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04907158389687538, "mean_delta_norm": 0.04907158389687538, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 5.338240623474121, "mean_delta_norm": 4.6571526527404785, "correction_norm": 5.338240623474121, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 11.750994682312012, "mean_delta_norm": 7.4147114753723145, "correction_norm": 11.750994682312012, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.047031354159116745, "mean_delta_norm": 0.047031354159116745, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.04813289642334, "mean_delta_norm": 9.251814842224121, "correction_norm": 10.04813289642334, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.14.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.1406731605529785, "mean_delta_norm": 4.017894744873047, "correction_norm": 4.1406731605529785, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12072469294071198, "mean_delta_norm": 0.12052858620882034, "correction_norm": 0.12072469294071198, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 34.635704040527344, "mean_delta_norm": 15.514739990234375, "correction_norm": 34.635704040527344, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 22.8719425201416, "mean_delta_norm": 16.408042907714844, "correction_norm": 22.8719425201416, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.674720764160156, "mean_delta_norm": 14.930529594421387, "correction_norm": 20.674720764160156, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12751628458499908, "mean_delta_norm": 0.12684589624404907, "correction_norm": 0.12751628458499908, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04389640688896179, "mean_delta_norm": 0.04389640688896179, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.9027934074401855, "mean_delta_norm": 4.613036632537842, "correction_norm": 4.9027934074401855, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.999247550964355, "mean_delta_norm": 7.226834774017334, "correction_norm": 9.999247550964355, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04478798061609268, "mean_delta_norm": 0.04478798061609268, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.266836166381836, "mean_delta_norm": 9.12621784210205, "correction_norm": 9.266836166381836, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.15.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.990990400314331, "mean_delta_norm": 3.9835057258605957, "correction_norm": 3.990990400314331, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.1322445422410965, "mean_delta_norm": 0.1318608820438385, "correction_norm": 0.1322445422410965, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 1.6544592380523682, "mean_delta_norm": 15.08098316192627, "correction_norm": 1.6544592380523682, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 22.905603408813477, "mean_delta_norm": 16.621980667114258, "correction_norm": 22.905603408813477, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.183238983154297, "mean_delta_norm": 14.839903831481934, "correction_norm": 20.183238983154297, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12420406937599182, "mean_delta_norm": 0.12350689619779587, "correction_norm": 0.12420406937599182, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.08000676333904266, "mean_delta_norm": 0.08000676333904266, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 5.044012069702148, "mean_delta_norm": 4.637606620788574, "correction_norm": 5.044012069702148, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 11.500454902648926, "mean_delta_norm": 7.6167778968811035, "correction_norm": 11.500454902648926, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.08011802285909653, "mean_delta_norm": 0.08011802285909653, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.79908275604248, "mean_delta_norm": 9.409629821777344, "correction_norm": 9.79908275604248, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.16.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.076350688934326, "mean_delta_norm": 3.9488096237182617, "correction_norm": 4.076350688934326, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 23.136343002319336, "mean_delta_norm": 16.605789184570312, "correction_norm": 23.136343002319336, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.0774063766002655, "mean_delta_norm": 0.0774063766002655, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 5.040008068084717, "mean_delta_norm": 4.580779552459717, "correction_norm": 5.040008068084717, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.372756004333496, "mean_delta_norm": 6.919710636138916, "correction_norm": 9.372756004333496, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.06931576132774353, "mean_delta_norm": 0.06931576132774353, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.959024429321289, "mean_delta_norm": 9.292771339416504, "correction_norm": 9.959024429321289, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.079998016357422, "mean_delta_norm": 3.766342878341675, "correction_norm": 4.079998016357422, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.09105321764945984, "mean_delta_norm": 0.09026883542537689, "correction_norm": 0.09105321764945984, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 32.247825622558594, "mean_delta_norm": 11.472127914428711, "correction_norm": 32.247825622558594, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.497215270996094, "mean_delta_norm": 11.708423614501953, "correction_norm": 9.497215270996094, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.188967227935791, "mean_delta_norm": 10.745529174804688, "correction_norm": 7.188967227935791, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.09121686220169067, "mean_delta_norm": 0.08964800089597702, "correction_norm": 0.09121686220169067, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.016531404107809067, "mean_delta_norm": 0.016531404107809067, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.4242525100708, "mean_delta_norm": 6.2448883056640625, "correction_norm": 9.4242525100708, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.7.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.016214074566960335, "mean_delta_norm": 0.016214074566960335, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.09787757694721222, "mean_delta_norm": 0.0973997414112091, "correction_norm": 0.09787757694721222, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 34.25713348388672, "mean_delta_norm": 12.308948516845703, "correction_norm": 34.25713348388672, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 17.885900497436523, "mean_delta_norm": 12.558524131774902, "correction_norm": 17.885900497436523, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 15.952950477600098, "mean_delta_norm": 11.632238388061523, "correction_norm": 15.952950477600098, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.1067078709602356, "mean_delta_norm": 0.10512808710336685, "correction_norm": 0.1067078709602356, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04575887322425842, "mean_delta_norm": 0.04575887322425842, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.8483211994171143, "mean_delta_norm": 3.4495887756347656, "correction_norm": 3.8483211994171143, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.979104995727539, "mean_delta_norm": 6.462774276733398, "correction_norm": 10.979104995727539, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.048497214913368225, "mean_delta_norm": 0.048497214913368225, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 6.990297317504883, "mean_delta_norm": 7.0198073387146, "correction_norm": 6.990297317504883, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.8.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.9454004764556885, "mean_delta_norm": 3.224694013595581, "correction_norm": 2.9454004764556885, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.09047732502222061, "mean_delta_norm": 0.08938659727573395, "correction_norm": 0.09047732502222061, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 30.380931854248047, "mean_delta_norm": 12.598455429077148, "correction_norm": 30.380931854248047, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 18.5018310546875, "mean_delta_norm": 13.477616310119629, "correction_norm": 18.5018310546875, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 15.406779289245605, "mean_delta_norm": 11.948118209838867, "correction_norm": 15.406779289245605, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.10267005115747452, "mean_delta_norm": 0.1010192260146141, "correction_norm": 0.10267005115747452, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04852990061044693, "mean_delta_norm": 0.04852990061044693, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.512322187423706, "mean_delta_norm": 3.577488660812378, "correction_norm": 3.512322187423706, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.511191368103027, "mean_delta_norm": 6.879231929779053, "correction_norm": 10.511191368103027, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.048518531024456024, "mean_delta_norm": 0.048518531024456024, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.012598037719727, "mean_delta_norm": 7.304933547973633, "correction_norm": 7.012598037719727, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.9.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.1966283321380615, "mean_delta_norm": 3.5321555137634277, "correction_norm": 3.1966283321380615, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11087179183959961, "mean_delta_norm": 0.1109766736626625, "correction_norm": 0.11087179183959961, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 27.88714027404785, "mean_delta_norm": 15.114303588867188, "correction_norm": 27.88714027404785, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.83749771118164, "mean_delta_norm": 15.10905933380127, "correction_norm": 20.83749771118164, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.17.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12021013349294662, "mean_delta_norm": 0.11954233050346375, "correction_norm": 0.12021013349294662, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11596867442131042, "mean_delta_norm": 0.11544083058834076, "correction_norm": 0.11596867442131042, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 28.127233505249023, "mean_delta_norm": 15.262646675109863, "correction_norm": 28.127233505249023, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 22.82288932800293, "mean_delta_norm": 16.775936126708984, "correction_norm": 22.82288932800293, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.560373306274414, "mean_delta_norm": 15.250505447387695, "correction_norm": 20.560373306274414, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11829177290201187, "mean_delta_norm": 0.11745259910821915, "correction_norm": 0.11829177290201187, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.06502410024404526, "mean_delta_norm": 0.06502410024404526, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.71043062210083, "mean_delta_norm": 4.693723678588867, "correction_norm": 4.71043062210083, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.270647048950195, "mean_delta_norm": 6.9456682205200195, "correction_norm": 9.270647048950195, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.06794541329145432, "mean_delta_norm": 0.06794541329145432, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.891645431518555, "mean_delta_norm": 9.330567359924316, "correction_norm": 8.891645431518555, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.18.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.732961893081665, "mean_delta_norm": 3.885000705718994, "correction_norm": 3.732961893081665, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12021899223327637, "mean_delta_norm": 0.11974389106035233, "correction_norm": 0.12021899223327637, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 27.191640853881836, "mean_delta_norm": 15.36828899383545, "correction_norm": 27.191640853881836, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 22.373796463012695, "mean_delta_norm": 16.84028434753418, "correction_norm": 22.373796463012695, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.30532455444336, "mean_delta_norm": 15.437763214111328, "correction_norm": 20.30532455444336, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12720710039138794, "mean_delta_norm": 0.12669125199317932, "correction_norm": 0.12720710039138794, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04556182771921158, "mean_delta_norm": 0.04556182771921158, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.71696662902832, "mean_delta_norm": 4.51425313949585, "correction_norm": 4.71696662902832, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.34947395324707, "mean_delta_norm": 7.033217430114746, "correction_norm": 9.34947395324707, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.046861134469509125, "mean_delta_norm": 0.046861134469509125, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.26727294921875, "mean_delta_norm": 9.2641019821167, "correction_norm": 9.26727294921875, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.19.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.694718360900879, "mean_delta_norm": 3.8035154342651367, "correction_norm": 3.694718360900879, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.13535958528518677, "mean_delta_norm": 0.13526692986488342, "correction_norm": 0.13535958528518677, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 28.58302116394043, "mean_delta_norm": 15.316927909851074, "correction_norm": 28.58302116394043, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 22.156475067138672, "mean_delta_norm": 16.949356079101562, "correction_norm": 22.156475067138672, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.078584671020508, "mean_delta_norm": 15.521611213684082, "correction_norm": 20.078584671020508, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.13584522902965546, "mean_delta_norm": 0.13532094657421112, "correction_norm": 0.13584522902965546, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.050597310066223145, "mean_delta_norm": 0.050597310066223145, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.6179962158203125, "mean_delta_norm": 4.756993293762207, "correction_norm": 4.6179962158203125, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.942793846130371, "mean_delta_norm": 7.375676155090332, "correction_norm": 9.942793846130371, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.052674803882837296, "mean_delta_norm": 0.052674803882837296, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.1422119140625, "mean_delta_norm": 9.674200057983398, "correction_norm": 9.1422119140625, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.20.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.7202394008636475, "mean_delta_norm": 3.893605947494507, "correction_norm": 3.7202394008636475, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.1236923485994339, "mean_delta_norm": 0.12315146625041962, "correction_norm": 0.1236923485994339, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 28.92435646057129, "mean_delta_norm": 15.187925338745117, "correction_norm": 28.92435646057129, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 21.66499900817871, "mean_delta_norm": 16.750429153442383, "correction_norm": 21.66499900817871, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 19.518970489501953, "mean_delta_norm": 15.304078102111816, "correction_norm": 19.518970489501953, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.13516342639923096, "mean_delta_norm": 0.1347334086894989, "correction_norm": 0.13516342639923096, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.046919483691453934, "mean_delta_norm": 0.046919483691453934, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.504430294036865, "mean_delta_norm": 4.711413860321045, "correction_norm": 4.504430294036865, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.749715805053711, "mean_delta_norm": 7.063665390014648, "correction_norm": 8.749715805053711, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.042686279863119125, "mean_delta_norm": 0.042686279863119125, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.550975799560547, "mean_delta_norm": 9.418481826782227, "correction_norm": 8.550975799560547, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.21.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.4210660457611084, "mean_delta_norm": 3.7870254516601562, "correction_norm": 3.4210660457611084, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12061836570501328, "mean_delta_norm": 0.11957519501447678, "correction_norm": 0.12061836570501328, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 23.548389434814453, "mean_delta_norm": 14.66065502166748, "correction_norm": 23.548389434814453, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 21.42475700378418, "mean_delta_norm": 16.601533889770508, "correction_norm": 21.42475700378418, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 19.222814559936523, "mean_delta_norm": 15.201481819152832, "correction_norm": 19.222814559936523, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.13749533891677856, "mean_delta_norm": 0.13704165816307068, "correction_norm": 0.13749533891677856, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.05399591475725174, "mean_delta_norm": 0.05399591475725174, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.495304107666016, "mean_delta_norm": 4.567840576171875, "correction_norm": 4.495304107666016, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.451478958129883, "mean_delta_norm": 6.839122772216797, "correction_norm": 9.451478958129883, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.050184216350317, "mean_delta_norm": 0.050184216350317, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.889542579650879, "mean_delta_norm": 9.391939163208008, "correction_norm": 8.889542579650879, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.22.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.519056558609009, "mean_delta_norm": 3.880906105041504, "correction_norm": 3.519056558609009, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12892718613147736, "mean_delta_norm": 0.12792803347110748, "correction_norm": 0.12892718613147736, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 24.228857040405273, "mean_delta_norm": 14.436335563659668, "correction_norm": 24.228857040405273, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 21.531831741333008, "mean_delta_norm": 16.497461318969727, "correction_norm": 21.531831741333008, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 19.063745498657227, "mean_delta_norm": 15.0331392288208, "correction_norm": 19.063745498657227, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12610352039337158, "mean_delta_norm": 0.12584511935710907, "correction_norm": 0.12610352039337158, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.05534135922789574, "mean_delta_norm": 0.05534135922789574, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.6349897384643555, "mean_delta_norm": 4.6994757652282715, "correction_norm": 4.6349897384643555, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 11.29124927520752, "mean_delta_norm": 7.337616443634033, "correction_norm": 11.29124927520752, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.052082087844610214, "mean_delta_norm": 0.052082087844610214, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.957193374633789, "mean_delta_norm": 9.46519947052002, "correction_norm": 8.957193374633789, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.23.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.326539993286133, "mean_delta_norm": 3.7741546630859375, "correction_norm": 3.326539993286133, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12197718024253845, "mean_delta_norm": 0.12222430109977722, "correction_norm": 0.12197718024253845, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 23.208498001098633, "mean_delta_norm": 14.029741287231445, "correction_norm": 23.208498001098633, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 21.13907241821289, "mean_delta_norm": 16.264314651489258, "correction_norm": 21.13907241821289, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 18.51018524169922, "mean_delta_norm": 14.704804420471191, "correction_norm": 18.51018524169922, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12107730656862259, "mean_delta_norm": 0.12142354249954224, "correction_norm": 0.12107730656862259, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04876039922237396, "mean_delta_norm": 0.04876039922237396, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.9818708896636963, "mean_delta_norm": 4.440162658691406, "correction_norm": 3.9818708896636963, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.056170463562012, "mean_delta_norm": 7.4803056716918945, "correction_norm": 10.056170463562012, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04909783974289894, "mean_delta_norm": 0.04909783974289894, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.45444393157959, "mean_delta_norm": 9.37988567352295, "correction_norm": 8.45444393157959, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.24.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.082794666290283, "mean_delta_norm": 3.707090377807617, "correction_norm": 3.082794666290283, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.1365073025226593, "mean_delta_norm": 0.1349526345729828, "correction_norm": 0.1365073025226593, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 24.728012084960938, "mean_delta_norm": 13.916533470153809, "correction_norm": 24.728012084960938, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 21.000307083129883, "mean_delta_norm": 16.035295486450195, "correction_norm": 21.000307083129883, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 18.213510513305664, "mean_delta_norm": 14.42565631866455, "correction_norm": 18.213510513305664, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12137258052825928, "mean_delta_norm": 0.12151110172271729, "correction_norm": 0.12137258052825928, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.03914468735456467, "mean_delta_norm": 0.03914468735456467, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.9729926586151123, "mean_delta_norm": 4.409243583679199, "correction_norm": 3.9729926586151123, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 11.1517972946167, "mean_delta_norm": 7.3590006828308105, "correction_norm": 11.1517972946167, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04176881164312363, "mean_delta_norm": 0.04176881164312363, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.694042682647705, "mean_delta_norm": 8.790257453918457, "correction_norm": 7.694042682647705, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.25.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.05841064453125, "mean_delta_norm": 3.7716119289398193, "correction_norm": 3.05841064453125, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.10898488759994507, "mean_delta_norm": 0.10907620191574097, "correction_norm": 0.10898488759994507, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 23.735458374023438, "mean_delta_norm": 13.731210708618164, "correction_norm": 23.735458374023438, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.76694107055664, "mean_delta_norm": 15.880166053771973, "correction_norm": 20.76694107055664, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 17.694774627685547, "mean_delta_norm": 14.192238807678223, "correction_norm": 17.694774627685547, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.1215929239988327, "mean_delta_norm": 0.12229367345571518, "correction_norm": 0.1215929239988327, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.06122780218720436, "mean_delta_norm": 0.06122780218720436, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.258833885192871, "mean_delta_norm": 4.538130283355713, "correction_norm": 4.258833885192871, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 12.118030548095703, "mean_delta_norm": 8.056065559387207, "correction_norm": 12.118030548095703, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.06345506012439728, "mean_delta_norm": 0.06345506012439728, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.705382347106934, "mean_delta_norm": 9.396977424621582, "correction_norm": 8.705382347106934, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.26.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.274338483810425, "mean_delta_norm": 3.7956011295318604, "correction_norm": 3.274338483810425, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.67706298828125, "mean_delta_norm": 15.660127639770508, "correction_norm": 20.67706298828125, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 17.642215728759766, "mean_delta_norm": 14.018940925598145, "correction_norm": 17.642215728759766, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04986026510596275, "mean_delta_norm": 0.04986026510596275, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.9326603412628174, "mean_delta_norm": 4.333900451660156, "correction_norm": 3.9326603412628174, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.73476505279541, "mean_delta_norm": 7.547715663909912, "correction_norm": 10.73476505279541, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.053514182567596436, "mean_delta_norm": 0.053514182567596436, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.127814292907715, "mean_delta_norm": 9.270837783813477, "correction_norm": 8.127814292907715, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.949017286300659, "mean_delta_norm": 3.635890007019043, "correction_norm": 2.949017286300659, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11429273337125778, "mean_delta_norm": 0.11116810888051987, "correction_norm": 0.11429273337125778, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 23.659957885742188, "mean_delta_norm": 13.65848159790039, "correction_norm": 23.659957885742188, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.27.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12154988199472427, "mean_delta_norm": 0.12226466089487076, "correction_norm": 0.12154988199472427, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11998438090085983, "mean_delta_norm": 0.11981289833784103, "correction_norm": 0.11998438090085983, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 22.56531524658203, "mean_delta_norm": 13.663535118103027, "correction_norm": 22.56531524658203, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.466445922851562, "mean_delta_norm": 15.47217082977295, "correction_norm": 20.466445922851562, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 17.74509620666504, "mean_delta_norm": 14.008172035217285, "correction_norm": 17.74509620666504, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11675440520048141, "mean_delta_norm": 0.11716267466545105, "correction_norm": 0.11675440520048141, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.04718874394893646, "mean_delta_norm": 0.04718874394893646, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.7980847358703613, "mean_delta_norm": 4.462097644805908, "correction_norm": 3.7980847358703613, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.26573371887207, "mean_delta_norm": 7.4708380699157715, "correction_norm": 10.26573371887207, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.046761270612478256, "mean_delta_norm": 0.046761270612478256, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.170965194702148, "mean_delta_norm": 8.991769790649414, "correction_norm": 7.170965194702148, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.28.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.8442962169647217, "mean_delta_norm": 3.931070566177368, "correction_norm": 2.8442962169647217, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.10938101261854172, "mean_delta_norm": 0.11054296791553497, "correction_norm": 0.10938101261854172, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.79505729675293, "mean_delta_norm": 13.421605110168457, "correction_norm": 20.79505729675293, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.25730323791504, "mean_delta_norm": 15.188630104064941, "correction_norm": 20.25730323791504, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 17.499067306518555, "mean_delta_norm": 13.745729446411133, "correction_norm": 17.499067306518555, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12140744924545288, "mean_delta_norm": 0.12243523448705673, "correction_norm": 0.12140744924545288, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.06185254827141762, "mean_delta_norm": 0.06185254827141762, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.841519594192505, "mean_delta_norm": 4.275417804718018, "correction_norm": 3.841519594192505, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 13.38206958770752, "mean_delta_norm": 7.9908270835876465, "correction_norm": 13.38206958770752, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.06631797552108765, "mean_delta_norm": 0.06631797552108765, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.992615222930908, "mean_delta_norm": 9.123266220092773, "correction_norm": 7.992615222930908, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.29.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.947432518005371, "mean_delta_norm": 3.667443037033081, "correction_norm": 2.947432518005371, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.10039788484573364, "mean_delta_norm": 0.09741289168596268, "correction_norm": 0.10039788484573364, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 19.449705123901367, "mean_delta_norm": 13.478584289550781, "correction_norm": 19.449705123901367, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.365243911743164, "mean_delta_norm": 15.047507286071777, "correction_norm": 20.365243911743164, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 17.991880416870117, "mean_delta_norm": 13.810097694396973, "correction_norm": 17.991880416870117, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.11988095939159393, "mean_delta_norm": 0.12113708257675171, "correction_norm": 0.11988095939159393, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.03671497479081154, "mean_delta_norm": 0.03671497479081154, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.935166120529175, "mean_delta_norm": 4.202287197113037, "correction_norm": 3.935166120529175, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 9.533612251281738, "mean_delta_norm": 7.537069320678711, "correction_norm": 9.533612251281738, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.03843076527118683, "mean_delta_norm": 0.03843076527118683, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.847126007080078, "mean_delta_norm": 8.600102424621582, "correction_norm": 7.847126007080078, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.30.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.968794107437134, "mean_delta_norm": 3.6202914714813232, "correction_norm": 2.968794107437134, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.101173534989357, "mean_delta_norm": 0.09892406314611435, "correction_norm": 0.101173534989357, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 13.934383392333984, "mean_delta_norm": 13.566493034362793, "correction_norm": 13.934383392333984, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.27193832397461, "mean_delta_norm": 14.904797554016113, "correction_norm": 20.27193832397461, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 18.163496017456055, "mean_delta_norm": 13.79093074798584, "correction_norm": 18.163496017456055, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12135640531778336, "mean_delta_norm": 0.12323465198278427, "correction_norm": 0.12135640531778336, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.032082486897706985, "mean_delta_norm": 0.032082486897706985, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.4536149501800537, "mean_delta_norm": 4.057459831237793, "correction_norm": 3.4536149501800537, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 12.007843017578125, "mean_delta_norm": 7.531981945037842, "correction_norm": 12.007843017578125, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.03347129747271538, "mean_delta_norm": 0.03347129747271538, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.040407657623291, "mean_delta_norm": 8.36264705657959, "correction_norm": 7.040407657623291, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.31.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.8285562992095947, "mean_delta_norm": 3.679370880126953, "correction_norm": 2.8285562992095947, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.03861597552895546, "mean_delta_norm": 0.037217725068330765, "correction_norm": 0.03861597552895546, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 23.885814666748047, "mean_delta_norm": 13.502298355102539, "correction_norm": 23.885814666748047, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.473012924194336, "mean_delta_norm": 15.006918907165527, "correction_norm": 20.473012924194336, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 18.287317276000977, "mean_delta_norm": 13.82260799407959, "correction_norm": 18.287317276000977, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12745963037014008, "mean_delta_norm": 0.12973551452159882, "correction_norm": 0.12745963037014008, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.02311064675450325, "mean_delta_norm": 0.02311064675450325, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.234774351119995, "mean_delta_norm": 3.920889139175415, "correction_norm": 3.234774351119995, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.086566925048828, "mean_delta_norm": 7.42902135848999, "correction_norm": 10.086566925048828, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.02367759495973587, "mean_delta_norm": 0.02367759495973587, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 6.612943649291992, "mean_delta_norm": 8.33387279510498, "correction_norm": 6.612943649291992, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.32.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.618835210800171, "mean_delta_norm": 3.638791561126709, "correction_norm": 2.618835210800171, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.02964836359024048, "mean_delta_norm": 0.029919598251581192, "correction_norm": 0.02964836359024048, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 12.378140449523926, "mean_delta_norm": 13.085491180419922, "correction_norm": 12.378140449523926, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 20.570987701416016, "mean_delta_norm": 15.046502113342285, "correction_norm": 20.570987701416016, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 18.101335525512695, "mean_delta_norm": 13.668618202209473, "correction_norm": 18.101335525512695, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.12484102696180344, "mean_delta_norm": 0.12780535221099854, "correction_norm": 0.12484102696180344, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.05595448985695839, "mean_delta_norm": 0.05595448985695839, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.05159068107605, "mean_delta_norm": 3.939868211746216, "correction_norm": 3.05159068107605, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 11.80810260772705, "mean_delta_norm": 7.380847930908203, "correction_norm": 11.80810260772705, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.0649559423327446, "mean_delta_norm": 0.0649559423327446, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 6.47959041595459, "mean_delta_norm": 8.542743682861328, "correction_norm": 6.47959041595459, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.33.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.456451416015625, "mean_delta_norm": 3.5535032749176025, "correction_norm": 2.456451416015625, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.0, "mean_delta_norm": 0.0, "correction_norm": 0.0, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.723296165466309, "mean_delta_norm": 11.994125366210938, "correction_norm": 8.723296165466309, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 17.74866485595703, "mean_delta_norm": 14.661887168884277, "correction_norm": 17.74866485595703, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 15.221205711364746, "mean_delta_norm": 12.94886589050293, "correction_norm": 15.221205711364746, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.08243239670991898, "mean_delta_norm": 0.08492795377969742, "correction_norm": 0.08243239670991898, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.032056502997875214, "mean_delta_norm": 0.032056502997875214, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 3.4883170127868652, "mean_delta_norm": 3.8148138523101807, "correction_norm": 3.4883170127868652, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.054081916809082, "mean_delta_norm": 6.924426078796387, "correction_norm": 7.054081916809082, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.03160680830478668, "mean_delta_norm": 0.03160680830478668, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.035632133483887, "mean_delta_norm": 8.335152626037598, "correction_norm": 7.035632133483887, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.34.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.379072904586792, "mean_delta_norm": 3.2492215633392334, "correction_norm": 2.379072904586792, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.input_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.016816269606351852, "mean_delta_norm": 0.01403897162526846, "correction_norm": 0.016816269606351852, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.mlp.down_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 1.583936333656311, "mean_delta_norm": 10.264862060546875, "correction_norm": 1.583936333656311, "shape": [ 4096, 12288 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.mlp.gate_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 12.646941184997559, "mean_delta_norm": 13.405177116394043, "correction_norm": 12.646941184997559, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.mlp.up_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 10.083135604858398, "mean_delta_norm": 11.45141315460205, "correction_norm": 10.083135604858398, "shape": [ 12288, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.post_attention_layernorm.weight", "strategy": "lot_paper_norm", "requested_strategy": "lot_paper_norm", "strategy_reason": "norm LOT requested", "fallback": false, "fallback_reason": null, "delta_norm": 0.038605183362960815, "mean_delta_norm": 0.03934391960501671, "correction_norm": 0.038605183362960815, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.self_attn.k_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.05124904587864876, "mean_delta_norm": 0.05124904587864876, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.self_attn.k_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 4.655656337738037, "mean_delta_norm": 3.922163724899292, "correction_norm": 4.655656337738037, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.self_attn.o_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 7.69063138961792, "mean_delta_norm": 6.284510612487793, "correction_norm": 7.69063138961792, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.self_attn.q_norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.05115547031164169, "mean_delta_norm": 0.05115547031164169, "correction_norm": 0.0, "shape": [ 128 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.self_attn.q_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 8.880091667175293, "mean_delta_norm": 7.9699177742004395, "correction_norm": 8.880091667175293, "shape": [ 4096, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.layers.35.self_attn.v_proj.weight", "strategy": "lot_paper", "requested_strategy": "lot_paper", "strategy_reason": null, "fallback": false, "fallback_reason": null, "delta_norm": 2.9842488765716553, "mean_delta_norm": 3.1581757068634033, "correction_norm": 2.9842488765716553, "shape": [ 1024, 4096 ], "dtype": "bfloat16" }, { "tensor": "model.norm.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "norm suffix excluded from LOT", "fallback": false, "fallback_reason": null, "delta_norm": 0.3012599050998688, "mean_delta_norm": 0.3012599050998688, "correction_norm": 0.0, "shape": [ 4096 ], "dtype": "bfloat16" }, { "tensor": "lm_head.weight", "strategy": "mean_delta", "requested_strategy": "mean_delta", "strategy_reason": "lm_head tensor", "fallback": false, "fallback_reason": null, "delta_norm": 37.18855667114258, "mean_delta_norm": 37.18855667114258, "correction_norm": 0.0, "shape": [ 151936, 4096 ], "dtype": "bfloat16" } ] }, "smoke_validation": { "chat_template_chars": { "false": 76, "true": 57 }, "forward_logits_finite": true }, "created_at_utc": "2026-05-26T13:12:53.682891+00:00", "git_commit": null, "run": { "exit_code": 0 } }