CoderBak/editlens_roberta_modelkit
0267
1{2 "variant": "fp32",3 "model": "onnx/model.onnx",4 "model_sha256": "ddd1173f2ef517ad499965e5029fae8099a8054a2bc76d8134e5889cc4ed1b3e",5 "reference_npz_sha256": "401ca4e9bf9b65c3f91552b277f254254fa1469cf870ac81424985df3d92286c",6 "fixtures_sha256": "b4cfec82a9c2839e3231a51a4fe991c5ce83ae516aa61e8fe5f93c235953651b",7 "provider": "CPUExecutionProvider",8 "onnxruntime": "1.30.0",9 "platform": {10 "os": "Darwin",11 "version": "27.2",12 "machine": "arm64"13 },14 "threads": 4,15 "load_seconds": 1.27055700001074,16 "cases": [17 {18 "id": "empty",19 "shape": [20 1,21 222 ],23 "max_absolute_logit_difference": 3.337860107421875e-06,24 "max_absolute_probability_difference": 5.654622786677166e-07,25 "argmax_agreements": 1,26 "samples": 1,27 "single_run_seconds": 0.03646879098960198528 },29 {30 "id": "minimal",31 "shape": [32 1,33 434 ],35 "max_absolute_logit_difference": 5.7220458984375e-06,36 "max_absolute_probability_difference": 5.157515448050987e-07,37 "argmax_agreements": 1,38 "samples": 1,39 "single_run_seconds": 0.0199893749959301240 },41 {42 "id": "short",43 "shape": [44 1,45 3346 ],47 "max_absolute_logit_difference": 2.1457672119140625e-06,48 "max_absolute_probability_difference": 1.651584634609904e-07,49 "argmax_agreements": 1,50 "samples": 1,51 "single_run_seconds": 0.0419480830023530952 },53 {54 "id": "mixed_padding",55 "shape": [56 4,57 4258 ],59 "max_absolute_logit_difference": 1.3828277587890625e-05,60 "max_absolute_probability_difference": 2.085083897984674e-06,61 "argmax_agreements": 4,62 "samples": 4,63 "single_run_seconds": 0.125619375001406364 },65 {66 "id": "technical",67 "shape": [68 4,69 7570 ],71 "max_absolute_logit_difference": 1.7523765563964844e-05,72 "max_absolute_probability_difference": 4.016848888621638e-06,73 "argmax_agreements": 4,74 "samples": 4,75 "single_run_seconds": 0.2091721250035334476 },77 {78 "id": "formatting",79 "shape": [80 4,81 9382 ],83 "max_absolute_logit_difference": 5.841255187988281e-06,84 "max_absolute_probability_difference": 1.8292948700815437e-06,85 "argmax_agreements": 4,86 "samples": 4,87 "single_run_seconds": 0.249990291020367388 },89 {90 "id": "repetition",91 "shape": [92 2,93 51294 ],95 "max_absolute_logit_difference": 8.344650268554688e-06,96 "max_absolute_probability_difference": 9.791367671097007e-07,97 "argmax_agreements": 2,98 "samples": 2,99 "single_run_seconds": 0.780980624986114100 },101 {102 "id": "long_512",103 "shape": [104 1,105 512106 ],107 "max_absolute_logit_difference": 3.933906555175781e-06,108 "max_absolute_probability_difference": 1.4080306614400184e-07,109 "argmax_agreements": 1,110 "samples": 1,111 "single_run_seconds": 0.3768378750246484112 },113 {114 "id": "long_mixed_batch",115 "shape": [116 2,117 512118 ],119 "max_absolute_logit_difference": 2.384185791015625e-06,120 "max_absolute_probability_difference": 5.780850338421573e-07,121 "argmax_agreements": 2,122 "samples": 2,123 "single_run_seconds": 0.7969407500058878124 },125 {126 "id": "whitespace",127 "shape": [128 2,129 17130 ],131 "max_absolute_logit_difference": 2.0265579223632812e-05,132 "max_absolute_probability_difference": 1.538472105155897e-06,133 "argmax_agreements": 2,134 "samples": 2,135 "single_run_seconds": 0.03817816698574461136 },137 {138 "id": "sequence_128",139 "shape": [140 1,141 129142 ],143 "max_absolute_logit_difference": 2.2649765014648438e-06,144 "max_absolute_probability_difference": 5.777336291590274e-07,145 "argmax_agreements": 1,146 "samples": 1,147 "single_run_seconds": 0.09674633399117738148 },149 {150 "id": "sequence_256",151 "shape": [152 1,153 257154 ],155 "max_absolute_logit_difference": 2.6226043701171875e-06,156 "max_absolute_probability_difference": 8.311704222663252e-07,157 "argmax_agreements": 1,158 "samples": 1,159 "single_run_seconds": 0.18392837498686276160 }161 ],162 "max_absolute_probability_difference": 4.016848888621638e-06,163 "argmax_disagreements": 0,164 "samples": 24,165 "acceptance_probability_tolerance": 0.0001,166 "passed": true,167 "limitations": "Synthetic unlabeled conversion fixtures. Timing is one run per shape, not a comparative performance benchmark. No Windows, Linux, CUDA, DirectML, WinML or CoreML validation is implied."168}169 