{ "configuration": { "url": "http://127.0.0.1:8000", "model_dir": "/root/autodl-tmp/jev9b-deploy/models/JEV-9B", "workload": "correlated", "rows": 128, "cardinality": 8, "correlated_columns": 12, "field_tokens": [ 128 ], "repeat_ratios": [ 0.8, 0.97, 0.983 ], "methods": [ "original", "lexicographic", "random", "cardinality", "solo" ], "repeats": 2, "concurrency": 4, "max_context_tokens": 16384, "seed": 42, "dry_run": false, "out": "validation/correlated-live.json" }, "measurement": "Fresh cache_salt per trial, identical backend and concurrency; alternating method order. Timed scans include normalization, planning, serialization, network and initial cache fills. Kernel/planner warmup is separate.", "cases": [ { "workload": { "family": "equal_ndv_correlated", "cardinality": 8, "correlated_columns": 12, "tokens_per_field": 128, "construction": "A columns are bijective copies of one latent variable; request and B use two other balanced variables. All columns have identical marginal NDV and token length. No identifier column.", "rows": 128, "columns": 14, "synthetic": true, "dataset_sha256": "5a3dc2bd95e1ee9efd88f0d3180397d63462a79078f7d91dc6ae5c7bcbcb4db2", "column_ndv": { "a_00": 8, "customer_request": 8, "b_00": 8, "a_01": 8, "a_02": 8, "a_03": 8, "a_04": 8, "a_05": 8, "a_06": 8, "a_07": 8, "a_08": 8, "a_09": 8, "a_10": 8, "a_11": 8 }, "all_column_ndv_equal": true, "field_token_ranges": { "a_00": [ 128, 128 ], "customer_request": [ 128, 128 ], "b_00": [ 128, 128 ], "a_01": [ 128, 128 ], "a_02": [ 128, 128 ], "a_03": [ 128, 128 ], "a_04": [ 128, 128 ], "a_05": [ 128, 128 ], "a_06": [ 128, 128 ], "a_07": [ 128, 128 ], "a_08": [ 128, 128 ], "a_09": [ 128, 128 ], "a_10": [ 128, 128 ], "a_11": [ 128, 128 ] }, "reusable_value_token_fraction": 1.0, "reuse_definition": "Fraction of cell-value tokens in columns whose whole-field NDV is less than row count; this is not the prefix-cache hit rate.", "prefix_distinct_counts": { "original": [ 8, 56, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 ], "lexicographic": [ 8, 56, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 ], "random": [ 8, 8, 8, 8, 8, 8, 8, 56, 56, 56, 56, 56, 128, 128 ], "cardinality": [ 8, 56, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 ], "solo": [ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 56, 128 ] }, "max_prompt_tokens_across_methods": 1909 }, "summary": [ { "method": "original", "rows": 128, "repeats": 2, "wall_seconds": 29.76774745201692, "planning_seconds": 7.618032395839691e-05, "prompt_tokens": 244352.0, "cached_tokens": 0.0, "cached_fraction": 0.0, "accuracy": 1.0, "rows_per_second": 4.299955856798541, "agreement_with_original": 1.0, "speedup_vs_original": 1.0, "solo_speedup_vs_this_method": 3.1713735202329416 }, { "method": "lexicographic", "rows": 128, "repeats": 2, "wall_seconds": 30.57132820598781, "planning_seconds": 0.0006326981820166111, "prompt_tokens": 244352.0, "cached_tokens": 0.0, "cached_fraction": 0.0, "accuracy": 1.0, "rows_per_second": 4.186929633463863, "agreement_with_original": 1.0, "speedup_vs_original": 0.9737145619399847, "solo_speedup_vs_this_method": 3.2569847922520956 }, { "method": "random", "rows": 128, "repeats": 2, "wall_seconds": 19.473364791367203, "planning_seconds": 0.00020971475169062614, "prompt_tokens": 244352.0, "cached_tokens": 110616.0, "cached_fraction": 0.4526911995809324, "accuracy": 1.0, "rows_per_second": 6.573080788623858, "agreement_with_original": 1.0, "speedup_vs_original": 1.5286391320114001, "solo_speedup_vs_this_method": 2.0746384505151414 }, { "method": "cardinality", "rows": 128, "repeats": 2, "wall_seconds": 30.807700131088495, "planning_seconds": 0.0008569839410483837, "prompt_tokens": 244352.0, "cached_tokens": 0.0, "cached_fraction": 0.0, "accuracy": 1.0, "rows_per_second": 4.154805436801605, "agreement_with_original": 1.0, "speedup_vs_original": 0.966243741835758, "solo_speedup_vs_this_method": 3.282167203699221 }, { "method": "solo", "rows": 128, "repeats": 2, "wall_seconds": 9.386389607563615, "planning_seconds": 0.0008912188932299614, "prompt_tokens": 244352.0, "cached_tokens": 190080.0, "cached_fraction": 0.7778941854374017, "accuracy": 1.0, "rows_per_second": 13.636766142421443, "agreement_with_original": 1.0, "speedup_vs_original": 3.1713735202329416, "solo_speedup_vs_this_method": 1.0 } ], "runs": [ { "method": "original", "rows": 128, "wall_seconds": 28.476774763315916, "rows_per_second": 4.494891049420771, "preparation_seconds": 0.0008212216198444366, "planning_seconds": 4.320219159126282e-05, "inference_seconds": 28.47587933111936, "prompt_tokens": 244352, "cached_tokens": 0, "cached_fraction": 0.0, "latency_p50_seconds": 0.8856474542990327, "latency_p95_seconds": 0.918231365038082, "column_order": [ "a_00", "customer_request", "b_00", "a_01", "a_02", "a_03", "a_04", "a_05", "a_06", "a_07", "a_08", "a_09", "a_10", "a_11" ], "repeat": 0, "cache_salt": "c3376164321a4b20b5d94b2e05cbf4d1", "accuracy": 1.0 }, { "method": "lexicographic", "rows": 128, "wall_seconds": 30.196429044939578, "rows_per_second": 4.238911819987227, "preparation_seconds": 0.0009485054761171341, "planning_seconds": 0.0006207926198840141, "inference_seconds": 30.194795115850866, "prompt_tokens": 244352, "cached_tokens": 0, "cached_fraction": 0.0, "latency_p50_seconds": 0.9393520355224609, "latency_p95_seconds": 0.9534923539496958, "column_order": [ "a_00", "customer_request", "b_00", "a_01", "a_02", "a_03", "a_04", "a_05", "a_06", "a_07", "a_08", "a_09", "a_10", "a_11" ], "repeat": 0, "cache_salt": "a0b1ad0c9faf4907a4893a11c762aadd", "accuracy": 1.0 }, { "method": "random", "rows": 128, "wall_seconds": 19.291211681440473, "rows_per_second": 6.635145687771659, "preparation_seconds": 0.0014984523877501488, "planning_seconds": 0.00027211103588342667, "inference_seconds": 19.289392499253154, "prompt_tokens": 244352, "cached_tokens": 110352, "cached_fraction": 0.45161079099004714, "latency_p50_seconds": 0.5518898940645158, "latency_p95_seconds": 0.9459353646263478, "column_order": [ "a_06", "a_10", "a_05", "a_04", "a_11", "a_09", "a_03", "b_00", "a_07", "a_01", "a_02", "a_00", "customer_request", "a_08" ], "repeat": 0, "cache_salt": "7f2334b9423d495bac31dff3d1c0ad59", "accuracy": 1.0 }, { "method": "cardinality", "rows": 128, "wall_seconds": 30.81115488242358, "rows_per_second": 4.154339572419547, "preparation_seconds": 0.0018439162522554398, "planning_seconds": 0.0009771818295121193, "inference_seconds": 30.808271282352507, "prompt_tokens": 244352, "cached_tokens": 0, "cached_fraction": 0.0, "latency_p50_seconds": 0.9583358974196017, "latency_p95_seconds": 0.9649531967006624, "column_order": [ "a_00", "customer_request", "b_00", "a_01", "a_02", "a_03", "a_04", "a_05", "a_06", "a_07", "a_08", "a_09", "a_10", "a_11" ], "repeat": 0, "cache_salt": "cfa8272983054f7db77533cab105fb17", "accuracy": 1.0 }, { "method": "solo", "rows": 128, "wall_seconds": 9.515077888034284, "rows_per_second": 13.452333391928068, "preparation_seconds": 0.0013711918145418167, "planning_seconds": 0.0008882181718945503, "inference_seconds": 9.512790518812835, "prompt_tokens": 244352, "cached_tokens": 190080, "cached_fraction": 0.7778941854374017, "latency_p50_seconds": 0.2781330840662122, "latency_p95_seconds": 0.4040957856923342, "column_order": [ "a_00", "a_01", "a_02", "a_03", "a_04", "a_05", "a_06", "a_07", "a_08", "a_09", "a_10", "a_11", "customer_request", "b_00" ], "repeat": 0, "cache_salt": "77f701a19395468a9150f37b14bb15bb", "accuracy": 1.0 }, { "method": "solo", "rows": 128, "wall_seconds": 9.257701327092946, "rows_per_second": 13.826326371688411, "preparation_seconds": 0.0010771369561553001, "planning_seconds": 0.0008942196145653725, "inference_seconds": 9.255677881650627, "prompt_tokens": 244352, "cached_tokens": 190080, "cached_fraction": 0.7778941854374017, "latency_p50_seconds": 0.2573451856151223, "latency_p95_seconds": 0.406396303512156, "column_order": [ "a_00", "a_01", "a_02", "a_03", "a_04", "a_05", "a_06", "a_07", "a_08", "a_09", "a_10", "a_11", "customer_request", "b_00" ], "repeat": 1, "cache_salt": "39ca945aefa24c3e98accc494d25d395", "accuracy": 1.0 }, { "method": "cardinality", "rows": 128, "wall_seconds": 30.80424537975341, "rows_per_second": 4.155271405678715, "preparation_seconds": 0.0017188377678394318, "planning_seconds": 0.0007367860525846481, "inference_seconds": 30.80175533797592, "prompt_tokens": 244352, "cached_tokens": 0, "cached_fraction": 0.0, "latency_p50_seconds": 0.9583604247309268, "latency_p95_seconds": 0.9662558281328529, "column_order": [ "a_00", "customer_request", "b_00", "a_01", "a_02", "a_03", "a_04", "a_05", "a_06", "a_07", "a_08", "a_09", "a_10", "a_11" ], "repeat": 1, "cache_salt": "d714172015964b018c8e473d8b3317b7", "accuracy": 1.0 }, { "method": "random", "rows": 128, "wall_seconds": 19.655517901293933, "rows_per_second": 6.512166234580555, "preparation_seconds": 0.0011992454528808594, "planning_seconds": 0.00014731846749782562, "inference_seconds": 19.654139451682568, "prompt_tokens": 244352, "cached_tokens": 110880, "cached_fraction": 0.4537716081718177, "latency_p50_seconds": 0.6073991330340505, "latency_p95_seconds": 1.0103276894893498, "column_order": [ "a_06", "a_10", "a_05", "a_04", "a_11", "a_09", "a_03", "b_00", "a_07", "a_01", "a_02", "a_00", "customer_request", "a_08" ], "repeat": 1, "cache_salt": "729dbaec7f6f4840acc3e69bb7d38fb5", "accuracy": 1.0 }, { "method": "lexicographic", "rows": 128, "wall_seconds": 30.946227367036045, "rows_per_second": 4.136206926998337, "preparation_seconds": 0.0009472277015447617, "planning_seconds": 0.0006446037441492081, "inference_seconds": 30.944572227075696, "prompt_tokens": 244352, "cached_tokens": 0, "cached_fraction": 0.0, "latency_p50_seconds": 0.9643205855973065, "latency_p95_seconds": 0.9676801529247314, "column_order": [ "a_00", "customer_request", "b_00", "a_01", "a_02", "a_03", "a_04", "a_05", "a_06", "a_07", "a_08", "a_09", "a_10", "a_11" ], "repeat": 1, "cache_salt": "ce1a2006878c46aab2f385c432d93f6d", "accuracy": 1.0 }, { "method": "original", "rows": 128, "wall_seconds": 31.058720140717924, "rows_per_second": 4.121225839959588, "preparation_seconds": 0.002361685037612915, "planning_seconds": 0.000109158456325531, "inference_seconds": 31.056185037828982, "prompt_tokens": 244352, "cached_tokens": 0, "cached_fraction": 0.0, "latency_p50_seconds": 0.9671922982670367, "latency_p95_seconds": 0.9698305042926222, "column_order": [ "a_00", "customer_request", "b_00", "a_01", "a_02", "a_03", "a_04", "a_05", "a_06", "a_07", "a_08", "a_09", "a_10", "a_11" ], "repeat": 1, "cache_salt": "79299796a9f147d394f79ec578c71824", "accuracy": 1.0 } ] } ] }