| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 22.900763358778626, | |
| "eval_steps": 100, | |
| "global_step": 3000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 4.998727735368957e-05, | |
| "loss": 13.0747, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.998727735368957e-05, | |
| "loss": 10.7662, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.9961832061068705e-05, | |
| "loss": 7.722, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.993638676844784e-05, | |
| "loss": 3.4523, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.9910941475826974e-05, | |
| "loss": 2.5227, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.988549618320611e-05, | |
| "loss": 0.8054, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.9860050890585244e-05, | |
| "loss": 0.2351, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.983460559796438e-05, | |
| "loss": 0.2044, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.9809160305343514e-05, | |
| "loss": 0.5828, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.978371501272265e-05, | |
| "loss": 0.6078, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.975826972010178e-05, | |
| "loss": 0.8824, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.973282442748092e-05, | |
| "loss": 0.3526, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.970737913486005e-05, | |
| "loss": 0.1536, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.968193384223919e-05, | |
| "loss": 0.3102, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.965648854961832e-05, | |
| "loss": 0.8505, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.963104325699746e-05, | |
| "loss": 0.2302, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.960559796437659e-05, | |
| "loss": 0.1172, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.958015267175573e-05, | |
| "loss": 0.4224, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.955470737913486e-05, | |
| "loss": 0.2523, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 4.9529262086514e-05, | |
| "loss": 0.4299, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.950381679389313e-05, | |
| "loss": 0.5534, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.9478371501272266e-05, | |
| "loss": 0.5437, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 4.94529262086514e-05, | |
| "loss": 0.7888, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.9427480916030536e-05, | |
| "loss": 0.3401, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.940203562340967e-05, | |
| "loss": 0.442, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.9376590330788806e-05, | |
| "loss": 0.5163, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.935114503816794e-05, | |
| "loss": 0.5436, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.9325699745547075e-05, | |
| "loss": 0.5621, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.930025445292621e-05, | |
| "loss": 0.2313, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.9274809160305345e-05, | |
| "loss": 0.1979, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.924936386768448e-05, | |
| "loss": 0.3287, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.9223918575063615e-05, | |
| "loss": 0.0833, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.9198473282442756e-05, | |
| "loss": 0.7452, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.9173027989821884e-05, | |
| "loss": 0.3994, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.914758269720102e-05, | |
| "loss": 0.413, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.9122137404580154e-05, | |
| "loss": 0.4078, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.909669211195929e-05, | |
| "loss": 0.1743, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.9071246819338424e-05, | |
| "loss": 0.3877, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 4.904580152671756e-05, | |
| "loss": 0.4874, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 4.902035623409669e-05, | |
| "loss": 0.1321, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 4.899491094147583e-05, | |
| "loss": 0.2902, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 4.896946564885496e-05, | |
| "loss": 0.0824, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 4.89440203562341e-05, | |
| "loss": 0.086, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 4.891857506361323e-05, | |
| "loss": 0.2476, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.889312977099237e-05, | |
| "loss": 0.4408, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.886768447837151e-05, | |
| "loss": 0.1607, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.884223918575064e-05, | |
| "loss": 0.5495, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.881679389312977e-05, | |
| "loss": 0.4271, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 4.879134860050891e-05, | |
| "loss": 0.1442, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 4.876590330788804e-05, | |
| "loss": 0.2471, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "eval_cer": 0.06833333333333333, | |
| "eval_loss": 0.19153563678264618, | |
| "eval_runtime": 9.1932, | |
| "eval_samples_per_second": 28.391, | |
| "eval_steps_per_second": 3.59, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.8740458015267176e-05, | |
| "loss": 0.2194, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.871501272264631e-05, | |
| "loss": 0.1366, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.8689567430025446e-05, | |
| "loss": 0.0784, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 4.866412213740458e-05, | |
| "loss": 0.5527, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 4.8638676844783716e-05, | |
| "loss": 0.2214, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 4.861323155216285e-05, | |
| "loss": 0.3578, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.858778625954199e-05, | |
| "loss": 0.1092, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 4.856234096692112e-05, | |
| "loss": 0.2762, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.853689567430026e-05, | |
| "loss": 0.1063, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.851145038167939e-05, | |
| "loss": 0.3029, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.848600508905853e-05, | |
| "loss": 0.482, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 4.846055979643766e-05, | |
| "loss": 0.5909, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.8435114503816794e-05, | |
| "loss": 0.6219, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.840966921119593e-05, | |
| "loss": 0.6025, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.8384223918575064e-05, | |
| "loss": 0.5285, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 4.83587786259542e-05, | |
| "loss": 0.1767, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 4.8333333333333334e-05, | |
| "loss": 0.5803, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 4.830788804071247e-05, | |
| "loss": 0.1041, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 4.82824427480916e-05, | |
| "loss": 0.3941, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 4.8256997455470745e-05, | |
| "loss": 0.2762, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 4.823155216284987e-05, | |
| "loss": 0.1117, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 4.8206106870229015e-05, | |
| "loss": 0.13, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 4.818066157760814e-05, | |
| "loss": 0.5239, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 4.8155216284987284e-05, | |
| "loss": 0.2415, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 4.812977099236641e-05, | |
| "loss": 0.1007, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 4.8104325699745554e-05, | |
| "loss": 0.1092, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 4.807888040712468e-05, | |
| "loss": 0.1845, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 4.805343511450382e-05, | |
| "loss": 0.0917, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 4.802798982188295e-05, | |
| "loss": 0.1885, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 4.8002544529262086e-05, | |
| "loss": 0.2213, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 4.797709923664123e-05, | |
| "loss": 0.1772, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 4.7951653944020356e-05, | |
| "loss": 0.1693, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 4.79262086513995e-05, | |
| "loss": 0.1783, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 4.7900763358778626e-05, | |
| "loss": 0.7147, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 4.787531806615777e-05, | |
| "loss": 0.0766, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 4.7849872773536895e-05, | |
| "loss": 0.2119, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 4.782442748091604e-05, | |
| "loss": 0.0908, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 4.7798982188295165e-05, | |
| "loss": 0.4874, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 4.7773536895674307e-05, | |
| "loss": 0.0256, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 4.7748091603053435e-05, | |
| "loss": 0.2564, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 4.772264631043257e-05, | |
| "loss": 0.1598, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 4.7697201017811704e-05, | |
| "loss": 0.163, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 4.767175572519084e-05, | |
| "loss": 0.1982, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 4.764631043256998e-05, | |
| "loss": 0.0347, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 4.762086513994911e-05, | |
| "loss": 0.3642, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 4.759541984732825e-05, | |
| "loss": 0.2434, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 4.756997455470738e-05, | |
| "loss": 0.7607, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 4.754452926208652e-05, | |
| "loss": 0.245, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 4.751908396946565e-05, | |
| "loss": 0.1778, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 4.749363867684479e-05, | |
| "loss": 0.1461, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "eval_cer": 0.09, | |
| "eval_loss": 0.2961411774158478, | |
| "eval_runtime": 9.3409, | |
| "eval_samples_per_second": 27.942, | |
| "eval_steps_per_second": 3.533, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 4.746819338422392e-05, | |
| "loss": 0.2253, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 4.744274809160306e-05, | |
| "loss": 0.5202, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 4.741730279898219e-05, | |
| "loss": 0.6376, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 4.739185750636133e-05, | |
| "loss": 0.0297, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 4.736641221374046e-05, | |
| "loss": 0.158, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 4.734096692111959e-05, | |
| "loss": 0.5359, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 4.7315521628498734e-05, | |
| "loss": 0.1949, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 4.729007633587786e-05, | |
| "loss": 0.0096, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 4.7264631043257e-05, | |
| "loss": 0.1996, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 4.723918575063613e-05, | |
| "loss": 0.2097, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 4.721374045801527e-05, | |
| "loss": 0.2805, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 4.71882951653944e-05, | |
| "loss": 0.5995, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 4.716284987277354e-05, | |
| "loss": 0.0177, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 4.713740458015267e-05, | |
| "loss": 0.3505, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 4.711195928753181e-05, | |
| "loss": 0.4022, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 4.708651399491094e-05, | |
| "loss": 0.2403, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 4.706106870229008e-05, | |
| "loss": 0.0289, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 4.7035623409669217e-05, | |
| "loss": 0.1095, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 4.7010178117048345e-05, | |
| "loss": 0.5152, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 4.6984732824427486e-05, | |
| "loss": 0.0785, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 4.6959287531806614e-05, | |
| "loss": 0.025, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 4.6933842239185756e-05, | |
| "loss": 0.0908, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 4.6908396946564884e-05, | |
| "loss": 0.236, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 4.6882951653944026e-05, | |
| "loss": 0.2633, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 4.6857506361323154e-05, | |
| "loss": 0.0273, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 4.6832061068702295e-05, | |
| "loss": 0.0652, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 4.680661577608142e-05, | |
| "loss": 0.2345, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 4.6781170483460565e-05, | |
| "loss": 0.1733, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 4.675572519083969e-05, | |
| "loss": 0.1278, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 4.6730279898218835e-05, | |
| "loss": 0.0199, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 4.670483460559797e-05, | |
| "loss": 0.0073, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 4.6679389312977104e-05, | |
| "loss": 0.0656, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 4.665394402035624e-05, | |
| "loss": 0.0056, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 4.662849872773537e-05, | |
| "loss": 0.2293, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 4.660305343511451e-05, | |
| "loss": 0.005, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 4.657760814249364e-05, | |
| "loss": 0.2217, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 4.655216284987278e-05, | |
| "loss": 0.1531, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 4.6526717557251906e-05, | |
| "loss": 0.0001, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 4.650127226463105e-05, | |
| "loss": 0.0067, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 4.6475826972010176e-05, | |
| "loss": 0.1709, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 4.645038167938932e-05, | |
| "loss": 0.1888, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 4.642493638676845e-05, | |
| "loss": 0.0052, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 4.639949109414759e-05, | |
| "loss": 0.321, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 4.637404580152672e-05, | |
| "loss": 0.0088, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 4.634860050890586e-05, | |
| "loss": 0.024, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 4.632315521628499e-05, | |
| "loss": 0.3312, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 4.629770992366412e-05, | |
| "loss": 0.5438, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 4.6284987277353694e-05, | |
| "loss": 1.0535, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 4.625954198473283e-05, | |
| "loss": 0.8874, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 4.6234096692111964e-05, | |
| "loss": 0.1959, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "eval_cer": 0.06666666666666667, | |
| "eval_loss": 0.21715956926345825, | |
| "eval_runtime": 9.1608, | |
| "eval_samples_per_second": 28.491, | |
| "eval_steps_per_second": 3.602, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 4.62086513994911e-05, | |
| "loss": 0.029, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 4.618320610687023e-05, | |
| "loss": 0.592, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 4.615776081424937e-05, | |
| "loss": 0.1281, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 4.61323155216285e-05, | |
| "loss": 0.1893, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 4.610687022900764e-05, | |
| "loss": 0.0201, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 4.6081424936386766e-05, | |
| "loss": 0.04, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 4.605597964376591e-05, | |
| "loss": 0.0411, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 4.6030534351145036e-05, | |
| "loss": 1.5439, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 4.600508905852418e-05, | |
| "loss": 0.0094, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 4.5979643765903305e-05, | |
| "loss": 0.006, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 4.595419847328245e-05, | |
| "loss": 5.4483, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 4.592875318066158e-05, | |
| "loss": 0.2359, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 4.5903307888040716e-05, | |
| "loss": 0.0957, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 4.587786259541985e-05, | |
| "loss": 3.4057, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 4.5852417302798986e-05, | |
| "loss": 4.2062, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 4.582697201017812e-05, | |
| "loss": 2.477, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 4.5801526717557256e-05, | |
| "loss": 1.9953, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 4.577608142493639e-05, | |
| "loss": 4.1759, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 4.575063613231552e-05, | |
| "loss": 4.7143, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 4.572519083969466e-05, | |
| "loss": 1.8038, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 4.569974554707379e-05, | |
| "loss": 1.516, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 4.567430025445293e-05, | |
| "loss": 3.2261, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 4.5648854961832065e-05, | |
| "loss": 2.0044, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 4.56234096692112e-05, | |
| "loss": 1.2701, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 4.5597964376590334e-05, | |
| "loss": 1.3727, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 4.557251908396947e-05, | |
| "loss": 1.2656, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 4.5547073791348604e-05, | |
| "loss": 1.1259, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.552162849872774e-05, | |
| "loss": 1.3548, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.5496183206106874e-05, | |
| "loss": 1.2812, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.547073791348601e-05, | |
| "loss": 1.0272, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.544529262086514e-05, | |
| "loss": 1.0934, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.541984732824428e-05, | |
| "loss": 1.2292, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.539440203562341e-05, | |
| "loss": 1.1085, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.536895674300254e-05, | |
| "loss": 1.0788, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.534351145038168e-05, | |
| "loss": 0.9239, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.531806615776082e-05, | |
| "loss": 1.0381, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.529262086513995e-05, | |
| "loss": 1.3701, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 4.526717557251909e-05, | |
| "loss": 0.8537, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 4.524173027989822e-05, | |
| "loss": 0.9638, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 4.521628498727736e-05, | |
| "loss": 0.8149, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 4.519083969465649e-05, | |
| "loss": 1.1826, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 4.5165394402035626e-05, | |
| "loss": 1.1008, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 4.513994910941476e-05, | |
| "loss": 0.9321, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 4.5114503816793896e-05, | |
| "loss": 0.8994, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 4.508905852417303e-05, | |
| "loss": 0.9649, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 4.5063613231552166e-05, | |
| "loss": 1.1236, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 4.5038167938931294e-05, | |
| "loss": 0.915, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 4.5012722646310435e-05, | |
| "loss": 1.2086, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 4.498727735368957e-05, | |
| "loss": 1.1753, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 4.4961832061068705e-05, | |
| "loss": 1.4127, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "eval_cer": 0.105, | |
| "eval_loss": 1.2658765316009521, | |
| "eval_runtime": 9.0281, | |
| "eval_samples_per_second": 28.91, | |
| "eval_steps_per_second": 3.655, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 4.493638676844784e-05, | |
| "loss": 1.1454, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 4.4910941475826975e-05, | |
| "loss": 1.0668, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 4.488549618320611e-05, | |
| "loss": 1.1689, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 4.4860050890585244e-05, | |
| "loss": 0.8271, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 4.483460559796438e-05, | |
| "loss": 3.0489, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 4.4809160305343514e-05, | |
| "loss": 4.085, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 4.478371501272265e-05, | |
| "loss": 2.5191, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 4.4758269720101784e-05, | |
| "loss": 2.7557, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 4.473282442748092e-05, | |
| "loss": 2.4464, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 4.470737913486005e-05, | |
| "loss": 2.3328, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 4.468193384223919e-05, | |
| "loss": 2.1459, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 4.465648854961832e-05, | |
| "loss": 2.0175, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 4.463104325699746e-05, | |
| "loss": 1.8922, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 4.460559796437659e-05, | |
| "loss": 1.5296, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 4.458015267175573e-05, | |
| "loss": 1.6281, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 4.455470737913486e-05, | |
| "loss": 1.8866, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 4.4529262086514e-05, | |
| "loss": 1.7002, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 4.450381679389313e-05, | |
| "loss": 1.6532, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 4.447837150127227e-05, | |
| "loss": 1.5812, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 4.44529262086514e-05, | |
| "loss": 1.5947, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 4.4427480916030536e-05, | |
| "loss": 1.6216, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 4.440203562340967e-05, | |
| "loss": 1.7792, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 4.4376590330788806e-05, | |
| "loss": 1.6228, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 4.435114503816794e-05, | |
| "loss": 1.6892, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 4.4325699745547076e-05, | |
| "loss": 1.6365, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 4.430025445292621e-05, | |
| "loss": 1.6137, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 4.4274809160305345e-05, | |
| "loss": 1.5058, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 4.424936386768448e-05, | |
| "loss": 1.5183, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 4.4223918575063615e-05, | |
| "loss": 1.7357, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 4.419847328244275e-05, | |
| "loss": 1.4155, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 4.4173027989821885e-05, | |
| "loss": 1.4845, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 4.414758269720102e-05, | |
| "loss": 1.5083, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 4.4122137404580154e-05, | |
| "loss": 1.5875, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 4.409669211195929e-05, | |
| "loss": 1.6002, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 4.4071246819338424e-05, | |
| "loss": 1.3794, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 4.404580152671756e-05, | |
| "loss": 1.519, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.4020356234096694e-05, | |
| "loss": 1.6349, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.399491094147583e-05, | |
| "loss": 1.5777, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.396946564885496e-05, | |
| "loss": 1.79, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 4.39440203562341e-05, | |
| "loss": 1.7578, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 4.391857506361323e-05, | |
| "loss": 1.7805, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 4.389312977099237e-05, | |
| "loss": 1.6965, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 4.38676844783715e-05, | |
| "loss": 1.4909, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 4.384223918575064e-05, | |
| "loss": 1.4643, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 4.381679389312977e-05, | |
| "loss": 1.4647, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 4.379134860050891e-05, | |
| "loss": 1.4439, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 4.376590330788805e-05, | |
| "loss": 1.4828, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 4.374045801526718e-05, | |
| "loss": 1.418, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 4.371501272264631e-05, | |
| "loss": 1.4172, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 4.3689567430025446e-05, | |
| "loss": 1.6268, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "eval_cer": 1.4616666666666667, | |
| "eval_loss": 1.5408730506896973, | |
| "eval_runtime": 12.3957, | |
| "eval_samples_per_second": 21.056, | |
| "eval_steps_per_second": 2.662, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 4.366412213740458e-05, | |
| "loss": 1.4824, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 4.3638676844783716e-05, | |
| "loss": 1.5388, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 4.361323155216285e-05, | |
| "loss": 1.4597, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 4.3587786259541986e-05, | |
| "loss": 1.5521, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 4.356234096692112e-05, | |
| "loss": 1.6686, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 4.3536895674300255e-05, | |
| "loss": 1.4682, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 4.351145038167939e-05, | |
| "loss": 1.9896, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 4.3486005089058525e-05, | |
| "loss": 1.9093, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 4.346055979643766e-05, | |
| "loss": 1.4429, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 4.34351145038168e-05, | |
| "loss": 1.3953, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 4.340966921119593e-05, | |
| "loss": 1.598, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 4.3384223918575064e-05, | |
| "loss": 1.8409, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 4.33587786259542e-05, | |
| "loss": 1.6732, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 4.3333333333333334e-05, | |
| "loss": 1.7087, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 4.330788804071247e-05, | |
| "loss": 1.4471, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 4.3282442748091604e-05, | |
| "loss": 1.505, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 4.325699745547074e-05, | |
| "loss": 1.5021, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 4.323155216284987e-05, | |
| "loss": 1.5206, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 4.320610687022901e-05, | |
| "loss": 1.6, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 4.318066157760814e-05, | |
| "loss": 1.5293, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 4.3155216284987285e-05, | |
| "loss": 1.4451, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 4.312977099236641e-05, | |
| "loss": 1.4265, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 4.3104325699745554e-05, | |
| "loss": 1.3854, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 4.307888040712468e-05, | |
| "loss": 1.373, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 4.3053435114503824e-05, | |
| "loss": 1.4177, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 4.21, | |
| "learning_rate": 4.302798982188295e-05, | |
| "loss": 1.4468, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 4.300254452926209e-05, | |
| "loss": 1.5429, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 4.297709923664122e-05, | |
| "loss": 1.308, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 4.26, | |
| "learning_rate": 4.2951653944020356e-05, | |
| "loss": 1.4762, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 4.292620865139949e-05, | |
| "loss": 1.3894, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 4.2900763358778626e-05, | |
| "loss": 1.4224, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 4.287531806615776e-05, | |
| "loss": 1.2713, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 4.2849872773536896e-05, | |
| "loss": 1.4202, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 4.34, | |
| "learning_rate": 4.282442748091604e-05, | |
| "loss": 1.3968, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 4.2798982188295165e-05, | |
| "loss": 1.283, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 4.277353689567431e-05, | |
| "loss": 1.409, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 4.2748091603053435e-05, | |
| "loss": 1.4048, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 4.272264631043258e-05, | |
| "loss": 1.5048, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 4.2697201017811705e-05, | |
| "loss": 1.2965, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 4.267175572519084e-05, | |
| "loss": 1.4138, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 4.2646310432569974e-05, | |
| "loss": 1.2395, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 4.262086513994911e-05, | |
| "loss": 1.4204, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 4.2595419847328244e-05, | |
| "loss": 1.4236, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 4.49, | |
| "learning_rate": 4.256997455470738e-05, | |
| "loss": 1.3041, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 4.254452926208652e-05, | |
| "loss": 1.4454, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 4.52, | |
| "learning_rate": 4.251908396946565e-05, | |
| "loss": 1.4345, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 4.249363867684479e-05, | |
| "loss": 1.3708, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 4.246819338422392e-05, | |
| "loss": 1.4303, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 4.244274809160306e-05, | |
| "loss": 1.4196, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 4.58, | |
| "learning_rate": 4.241730279898219e-05, | |
| "loss": 1.3672, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 4.58, | |
| "eval_cer": 1.4533333333333334, | |
| "eval_loss": 1.5275843143463135, | |
| "eval_runtime": 11.8906, | |
| "eval_samples_per_second": 21.95, | |
| "eval_steps_per_second": 2.775, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 4.239185750636133e-05, | |
| "loss": 1.4312, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 4.61, | |
| "learning_rate": 4.236641221374046e-05, | |
| "loss": 1.4523, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 4.23409669211196e-05, | |
| "loss": 1.555, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 4.231552162849873e-05, | |
| "loss": 1.2982, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 4.229007633587786e-05, | |
| "loss": 1.2546, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 4.2264631043257e-05, | |
| "loss": 1.4486, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 4.223918575063613e-05, | |
| "loss": 1.5046, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 4.221374045801527e-05, | |
| "loss": 1.4805, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 4.21882951653944e-05, | |
| "loss": 1.4282, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 4.216284987277354e-05, | |
| "loss": 1.4071, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 4.213740458015267e-05, | |
| "loss": 1.4593, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "learning_rate": 4.211195928753181e-05, | |
| "loss": 1.5556, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 4.78, | |
| "learning_rate": 4.208651399491094e-05, | |
| "loss": 1.3197, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 4.206106870229008e-05, | |
| "loss": 1.1628, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 4.81, | |
| "learning_rate": 4.203562340966921e-05, | |
| "loss": 1.5496, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 4.201017811704835e-05, | |
| "loss": 1.4464, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 4.198473282442748e-05, | |
| "loss": 1.2623, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 4.1959287531806615e-05, | |
| "loss": 1.4, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 4.193384223918575e-05, | |
| "loss": 1.3183, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 4.89, | |
| "learning_rate": 4.1908396946564884e-05, | |
| "loss": 1.4607, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 4.1882951653944026e-05, | |
| "loss": 1.4597, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 4.1857506361323154e-05, | |
| "loss": 1.3958, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 4.1832061068702296e-05, | |
| "loss": 1.4509, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 4.1806615776081424e-05, | |
| "loss": 1.4918, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 4.96, | |
| "learning_rate": 4.1781170483460565e-05, | |
| "loss": 1.3633, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 4.175572519083969e-05, | |
| "loss": 1.4143, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 4.99, | |
| "learning_rate": 4.1730279898218835e-05, | |
| "loss": 1.3187, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 4.170483460559796e-05, | |
| "loss": 1.4254, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 4.1679389312977105e-05, | |
| "loss": 1.4741, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 4.165394402035623e-05, | |
| "loss": 1.2889, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 4.1628498727735374e-05, | |
| "loss": 1.4764, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 4.160305343511451e-05, | |
| "loss": 1.513, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 4.157760814249364e-05, | |
| "loss": 1.2969, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 4.155216284987278e-05, | |
| "loss": 1.4052, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 4.152671755725191e-05, | |
| "loss": 1.4238, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "learning_rate": 4.150127226463105e-05, | |
| "loss": 1.5067, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "learning_rate": 4.1475826972010176e-05, | |
| "loss": 1.3132, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 4.145038167938932e-05, | |
| "loss": 1.4175, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 4.1424936386768446e-05, | |
| "loss": 1.1908, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 5.19, | |
| "learning_rate": 4.139949109414759e-05, | |
| "loss": 1.4047, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 4.1374045801526716e-05, | |
| "loss": 1.2117, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 4.134860050890586e-05, | |
| "loss": 1.3999, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 4.1323155216284985e-05, | |
| "loss": 1.3811, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 4.129770992366413e-05, | |
| "loss": 1.3761, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 5.27, | |
| "learning_rate": 4.127226463104326e-05, | |
| "loss": 1.269, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "learning_rate": 4.124681933842239e-05, | |
| "loss": 1.2983, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "learning_rate": 4.122137404580153e-05, | |
| "loss": 1.2791, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 5.31, | |
| "learning_rate": 4.119592875318066e-05, | |
| "loss": 1.4508, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 4.11704834605598e-05, | |
| "loss": 1.2673, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 5.34, | |
| "learning_rate": 4.114503816793893e-05, | |
| "loss": 1.3251, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 5.34, | |
| "eval_cer": 2.08, | |
| "eval_loss": 1.4844698905944824, | |
| "eval_runtime": 13.3113, | |
| "eval_samples_per_second": 19.607, | |
| "eval_steps_per_second": 2.479, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 5.36, | |
| "learning_rate": 4.111959287531807e-05, | |
| "loss": 1.2549, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 4.10941475826972e-05, | |
| "loss": 1.4543, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 5.39, | |
| "learning_rate": 4.106870229007634e-05, | |
| "loss": 1.4231, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 4.104325699745547e-05, | |
| "loss": 1.5034, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 5.42, | |
| "learning_rate": 4.101781170483461e-05, | |
| "loss": 1.405, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 4.0992366412213745e-05, | |
| "loss": 1.3133, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 4.096692111959288e-05, | |
| "loss": 1.3101, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 4.0941475826972015e-05, | |
| "loss": 1.5093, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 5.48, | |
| "learning_rate": 4.091603053435115e-05, | |
| "loss": 1.3983, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 4.0890585241730284e-05, | |
| "loss": 1.3322, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 5.51, | |
| "learning_rate": 4.086513994910941e-05, | |
| "loss": 1.1801, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 4.0839694656488554e-05, | |
| "loss": 1.2332, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 4.081424936386768e-05, | |
| "loss": 1.1875, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 5.56, | |
| "learning_rate": 4.0788804071246824e-05, | |
| "loss": 1.3077, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 4.076335877862595e-05, | |
| "loss": 1.2804, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 5.59, | |
| "learning_rate": 4.073791348600509e-05, | |
| "loss": 1.1942, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 4.071246819338422e-05, | |
| "loss": 1.3138, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 5.62, | |
| "learning_rate": 4.068702290076336e-05, | |
| "loss": 1.2047, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 4.06615776081425e-05, | |
| "loss": 1.1966, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 4.063613231552163e-05, | |
| "loss": 1.3232, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 4.061068702290077e-05, | |
| "loss": 1.2763, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 5.68, | |
| "learning_rate": 4.05852417302799e-05, | |
| "loss": 1.1923, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 4.055979643765904e-05, | |
| "loss": 1.2184, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "learning_rate": 4.053435114503817e-05, | |
| "loss": 1.3541, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 4.050890585241731e-05, | |
| "loss": 1.1989, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 5.74, | |
| "learning_rate": 4.0483460559796435e-05, | |
| "loss": 1.2263, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 4.0458015267175576e-05, | |
| "loss": 1.2974, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 5.77, | |
| "learning_rate": 4.0432569974554704e-05, | |
| "loss": 1.2148, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 5.79, | |
| "learning_rate": 4.0407124681933846e-05, | |
| "loss": 1.3899, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 4.0381679389312974e-05, | |
| "loss": 1.2513, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 4.0356234096692116e-05, | |
| "loss": 1.163, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 5.83, | |
| "learning_rate": 4.033078880407125e-05, | |
| "loss": 1.3232, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 4.0305343511450385e-05, | |
| "loss": 1.3787, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 5.86, | |
| "learning_rate": 4.027989821882952e-05, | |
| "loss": 1.2339, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 4.0254452926208655e-05, | |
| "loss": 1.3441, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 5.89, | |
| "learning_rate": 4.022900763358779e-05, | |
| "loss": 1.1968, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 4.0203562340966925e-05, | |
| "loss": 1.2653, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 5.92, | |
| "learning_rate": 4.017811704834606e-05, | |
| "loss": 1.1534, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 5.94, | |
| "learning_rate": 4.015267175572519e-05, | |
| "loss": 1.367, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 4.012722646310433e-05, | |
| "loss": 1.3714, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 4.010178117048346e-05, | |
| "loss": 1.3147, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "learning_rate": 4.00763358778626e-05, | |
| "loss": 1.3468, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 4.0050890585241734e-05, | |
| "loss": 1.1087, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 4.002544529262087e-05, | |
| "loss": 1.4632, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 4e-05, | |
| "loss": 1.2951, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 3.997455470737914e-05, | |
| "loss": 1.1655, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 3.994910941475827e-05, | |
| "loss": 1.3331, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 3.992366412213741e-05, | |
| "loss": 1.1769, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 3.989821882951654e-05, | |
| "loss": 1.1607, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 3.987277353689568e-05, | |
| "loss": 1.2477, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "eval_cer": 1.2383333333333333, | |
| "eval_loss": 1.4819015264511108, | |
| "eval_runtime": 12.5927, | |
| "eval_samples_per_second": 20.726, | |
| "eval_steps_per_second": 2.621, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 3.984732824427481e-05, | |
| "loss": 1.3543, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 3.982188295165395e-05, | |
| "loss": 1.32, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 3.979643765903308e-05, | |
| "loss": 1.1779, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "learning_rate": 3.977099236641221e-05, | |
| "loss": 1.2045, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "learning_rate": 3.974554707379135e-05, | |
| "loss": 1.1947, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 3.9720101781170486e-05, | |
| "loss": 1.0608, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 6.21, | |
| "learning_rate": 3.969465648854962e-05, | |
| "loss": 1.3838, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 6.23, | |
| "learning_rate": 3.9669211195928756e-05, | |
| "loss": 1.2513, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 3.964376590330789e-05, | |
| "loss": 1.2486, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 6.26, | |
| "learning_rate": 3.9618320610687026e-05, | |
| "loss": 1.0845, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 6.27, | |
| "learning_rate": 3.959287531806616e-05, | |
| "loss": 0.9877, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 3.9567430025445295e-05, | |
| "loss": 1.2315, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 6.31, | |
| "learning_rate": 3.954198473282443e-05, | |
| "loss": 1.2069, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 6.32, | |
| "learning_rate": 3.9516539440203565e-05, | |
| "loss": 1.1374, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 3.94910941475827e-05, | |
| "loss": 1.2448, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 6.35, | |
| "learning_rate": 3.9465648854961835e-05, | |
| "loss": 1.1527, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 6.37, | |
| "learning_rate": 3.944020356234097e-05, | |
| "loss": 1.3363, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 3.9414758269720104e-05, | |
| "loss": 1.1996, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 3.938931297709924e-05, | |
| "loss": 0.9644, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 3.9363867684478374e-05, | |
| "loss": 1.3315, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 6.43, | |
| "learning_rate": 3.933842239185751e-05, | |
| "loss": 1.1744, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 6.44, | |
| "learning_rate": 3.9312977099236644e-05, | |
| "loss": 1.3106, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 3.928753180661578e-05, | |
| "loss": 1.1695, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 3.926208651399491e-05, | |
| "loss": 0.986, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 6.49, | |
| "learning_rate": 3.923664122137405e-05, | |
| "loss": 1.22, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "learning_rate": 3.921119592875318e-05, | |
| "loss": 1.2478, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 6.52, | |
| "learning_rate": 3.918575063613232e-05, | |
| "loss": 1.2913, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 3.916030534351145e-05, | |
| "loss": 1.078, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 3.913486005089059e-05, | |
| "loss": 1.155, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 6.56, | |
| "learning_rate": 3.910941475826972e-05, | |
| "loss": 1.2021, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 3.908396946564886e-05, | |
| "loss": 1.2817, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 3.905852417302799e-05, | |
| "loss": 1.5194, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 6.61, | |
| "learning_rate": 3.903307888040713e-05, | |
| "loss": 1.1239, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 6.63, | |
| "learning_rate": 3.900763358778626e-05, | |
| "loss": 1.1291, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 6.64, | |
| "learning_rate": 3.8982188295165396e-05, | |
| "loss": 1.1058, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 3.895674300254453e-05, | |
| "loss": 1.338, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 3.8931297709923666e-05, | |
| "loss": 1.4552, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 6.69, | |
| "learning_rate": 3.89058524173028e-05, | |
| "loss": 1.1117, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 3.8880407124681936e-05, | |
| "loss": 1.329, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 3.885496183206107e-05, | |
| "loss": 1.0848, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 3.8829516539440205e-05, | |
| "loss": 1.141, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 3.880407124681934e-05, | |
| "loss": 1.4024, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 6.76, | |
| "learning_rate": 3.8778625954198475e-05, | |
| "loss": 1.3653, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 3.875318066157761e-05, | |
| "loss": 1.1886, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 6.79, | |
| "learning_rate": 3.8727735368956745e-05, | |
| "loss": 1.103, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 3.870229007633588e-05, | |
| "loss": 1.0959, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 3.8676844783715014e-05, | |
| "loss": 1.2043, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 3.865139949109415e-05, | |
| "loss": 1.2143, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 3.8625954198473284e-05, | |
| "loss": 1.1641, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 6.87, | |
| "learning_rate": 3.860050890585242e-05, | |
| "loss": 1.1487, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 6.87, | |
| "eval_cer": 1.6783333333333332, | |
| "eval_loss": 1.4798498153686523, | |
| "eval_runtime": 13.2397, | |
| "eval_samples_per_second": 19.713, | |
| "eval_steps_per_second": 2.493, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 6.89, | |
| "learning_rate": 3.8575063613231554e-05, | |
| "loss": 1.5181, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 3.854961832061069e-05, | |
| "loss": 1.4416, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 6.92, | |
| "learning_rate": 3.852417302798982e-05, | |
| "loss": 1.183, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "learning_rate": 3.849872773536896e-05, | |
| "loss": 0.9592, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 3.847328244274809e-05, | |
| "loss": 1.1535, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "learning_rate": 3.844783715012723e-05, | |
| "loss": 1.1211, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 3.842239185750636e-05, | |
| "loss": 0.8703, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "learning_rate": 3.83969465648855e-05, | |
| "loss": 1.1494, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 3.837150127226463e-05, | |
| "loss": 1.0239, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 3.834605597964377e-05, | |
| "loss": 1.2222, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 3.83206106870229e-05, | |
| "loss": 1.2261, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 3.829516539440204e-05, | |
| "loss": 1.2203, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 3.826972010178117e-05, | |
| "loss": 1.1816, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 3.8244274809160306e-05, | |
| "loss": 1.1933, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 3.821882951653944e-05, | |
| "loss": 1.0713, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 3.8193384223918576e-05, | |
| "loss": 1.2109, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 7.13, | |
| "learning_rate": 3.816793893129771e-05, | |
| "loss": 1.1673, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 3.8142493638676846e-05, | |
| "loss": 1.167, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 7.16, | |
| "learning_rate": 3.811704834605598e-05, | |
| "loss": 1.2627, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 3.8091603053435115e-05, | |
| "loss": 1.1286, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 7.19, | |
| "learning_rate": 3.806615776081425e-05, | |
| "loss": 1.0552, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 7.21, | |
| "learning_rate": 3.8040712468193385e-05, | |
| "loss": 1.1211, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 7.22, | |
| "learning_rate": 3.801526717557252e-05, | |
| "loss": 1.2985, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 7.24, | |
| "learning_rate": 3.7989821882951655e-05, | |
| "loss": 1.2846, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 3.796437659033079e-05, | |
| "loss": 1.0098, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 7.27, | |
| "learning_rate": 3.7938931297709924e-05, | |
| "loss": 1.0775, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 3.791348600508906e-05, | |
| "loss": 1.1962, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 7.3, | |
| "learning_rate": 3.78880407124682e-05, | |
| "loss": 1.0683, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 7.31, | |
| "learning_rate": 3.786259541984733e-05, | |
| "loss": 1.0593, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 3.783715012722647e-05, | |
| "loss": 1.1983, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 7.34, | |
| "learning_rate": 3.78117048346056e-05, | |
| "loss": 1.0894, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 7.36, | |
| "learning_rate": 3.778625954198473e-05, | |
| "loss": 1.141, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 7.37, | |
| "learning_rate": 3.776081424936387e-05, | |
| "loss": 1.1035, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 7.39, | |
| "learning_rate": 3.7735368956743e-05, | |
| "loss": 1.2687, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 3.770992366412214e-05, | |
| "loss": 1.1224, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 7.42, | |
| "learning_rate": 3.768447837150127e-05, | |
| "loss": 1.0745, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 3.765903307888041e-05, | |
| "loss": 0.9383, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 7.45, | |
| "learning_rate": 3.763358778625954e-05, | |
| "loss": 1.1412, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 7.47, | |
| "learning_rate": 3.760814249363868e-05, | |
| "loss": 0.9986, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 7.48, | |
| "learning_rate": 3.758269720101781e-05, | |
| "loss": 1.1864, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 7.5, | |
| "learning_rate": 3.7557251908396954e-05, | |
| "loss": 1.1357, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 7.51, | |
| "learning_rate": 3.753180661577608e-05, | |
| "loss": 0.9112, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 7.53, | |
| "learning_rate": 3.750636132315522e-05, | |
| "loss": 1.0888, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 7.54, | |
| "learning_rate": 3.748091603053435e-05, | |
| "loss": 1.246, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 7.56, | |
| "learning_rate": 3.745547073791349e-05, | |
| "loss": 1.0594, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 3.743002544529262e-05, | |
| "loss": 0.9861, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 3.7404580152671756e-05, | |
| "loss": 1.3891, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 7.6, | |
| "learning_rate": 3.737913486005089e-05, | |
| "loss": 1.1963, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "learning_rate": 3.7353689567430025e-05, | |
| "loss": 1.0723, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "learning_rate": 3.732824427480916e-05, | |
| "loss": 0.879, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "eval_cer": 1.0783333333333334, | |
| "eval_loss": 1.5497722625732422, | |
| "eval_runtime": 12.6235, | |
| "eval_samples_per_second": 20.676, | |
| "eval_steps_per_second": 2.614, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 3.7302798982188295e-05, | |
| "loss": 1.1869, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 7.66, | |
| "learning_rate": 3.727735368956743e-05, | |
| "loss": 0.912, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 7.68, | |
| "learning_rate": 3.7251908396946565e-05, | |
| "loss": 1.1385, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 7.69, | |
| "learning_rate": 3.7226463104325706e-05, | |
| "loss": 1.0039, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 3.7201017811704834e-05, | |
| "loss": 0.9042, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 7.73, | |
| "learning_rate": 3.7175572519083976e-05, | |
| "loss": 1.108, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 7.74, | |
| "learning_rate": 3.7150127226463104e-05, | |
| "loss": 1.1434, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 7.76, | |
| "learning_rate": 3.7124681933842246e-05, | |
| "loss": 1.0162, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 7.77, | |
| "learning_rate": 3.7099236641221374e-05, | |
| "loss": 0.9161, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 7.79, | |
| "learning_rate": 3.707379134860051e-05, | |
| "loss": 1.1173, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "learning_rate": 3.704834605597964e-05, | |
| "loss": 1.0893, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 7.82, | |
| "learning_rate": 3.702290076335878e-05, | |
| "loss": 0.9602, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 7.83, | |
| "learning_rate": 3.699745547073791e-05, | |
| "loss": 0.9314, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 7.85, | |
| "learning_rate": 3.697201017811705e-05, | |
| "loss": 0.9972, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 7.86, | |
| "learning_rate": 3.694656488549619e-05, | |
| "loss": 1.0091, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 7.88, | |
| "learning_rate": 3.692111959287532e-05, | |
| "loss": 1.1803, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 7.89, | |
| "learning_rate": 3.689567430025446e-05, | |
| "loss": 1.1259, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "learning_rate": 3.687022900763359e-05, | |
| "loss": 1.1641, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 7.92, | |
| "learning_rate": 3.684478371501273e-05, | |
| "loss": 0.9592, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 7.94, | |
| "learning_rate": 3.681933842239186e-05, | |
| "loss": 1.3188, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 7.95, | |
| "learning_rate": 3.6793893129771e-05, | |
| "loss": 1.1046, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 7.97, | |
| "learning_rate": 3.6768447837150126e-05, | |
| "loss": 1.0569, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "learning_rate": 3.674300254452927e-05, | |
| "loss": 1.0293, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 3.6717557251908396e-05, | |
| "loss": 0.9867, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 3.669211195928753e-05, | |
| "loss": 1.0242, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 3.6666666666666666e-05, | |
| "loss": 1.0203, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 3.66412213740458e-05, | |
| "loss": 1.2455, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 3.661577608142494e-05, | |
| "loss": 1.1691, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 3.659033078880407e-05, | |
| "loss": 1.0646, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 3.656488549618321e-05, | |
| "loss": 1.2154, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 8.11, | |
| "learning_rate": 3.653944020356234e-05, | |
| "loss": 0.9653, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 3.651399491094148e-05, | |
| "loss": 1.1375, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 8.14, | |
| "learning_rate": 3.648854961832061e-05, | |
| "loss": 1.0571, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 3.646310432569975e-05, | |
| "loss": 0.9675, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 8.17, | |
| "learning_rate": 3.643765903307888e-05, | |
| "loss": 0.9511, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 8.18, | |
| "learning_rate": 3.641221374045802e-05, | |
| "loss": 1.2214, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 8.2, | |
| "learning_rate": 3.638676844783715e-05, | |
| "loss": 1.0093, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 8.21, | |
| "learning_rate": 3.6361323155216284e-05, | |
| "loss": 0.9248, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 8.23, | |
| "learning_rate": 3.633587786259542e-05, | |
| "loss": 0.9676, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 8.24, | |
| "learning_rate": 3.631043256997455e-05, | |
| "loss": 0.9328, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 8.26, | |
| "learning_rate": 3.6284987277353695e-05, | |
| "loss": 0.9463, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 3.625954198473282e-05, | |
| "loss": 1.0143, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 8.29, | |
| "learning_rate": 3.6234096692111965e-05, | |
| "loss": 0.8562, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 8.31, | |
| "learning_rate": 3.620865139949109e-05, | |
| "loss": 0.9518, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 8.32, | |
| "learning_rate": 3.6183206106870234e-05, | |
| "loss": 1.0586, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 8.34, | |
| "learning_rate": 3.615776081424936e-05, | |
| "loss": 0.8875, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 8.35, | |
| "learning_rate": 3.6132315521628504e-05, | |
| "loss": 0.8863, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 8.37, | |
| "learning_rate": 3.610687022900763e-05, | |
| "loss": 0.882, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 8.38, | |
| "learning_rate": 3.6081424936386774e-05, | |
| "loss": 0.8965, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "learning_rate": 3.60559796437659e-05, | |
| "loss": 0.9638, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "eval_cer": 2.155, | |
| "eval_loss": 1.631682276725769, | |
| "eval_runtime": 13.7528, | |
| "eval_samples_per_second": 18.978, | |
| "eval_steps_per_second": 2.4, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 8.41, | |
| "learning_rate": 3.603053435114504e-05, | |
| "loss": 1.1097, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 8.43, | |
| "learning_rate": 3.600508905852418e-05, | |
| "loss": 0.9083, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 8.44, | |
| "learning_rate": 3.5979643765903306e-05, | |
| "loss": 0.96, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 3.595419847328245e-05, | |
| "loss": 0.9702, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 8.47, | |
| "learning_rate": 3.5928753180661576e-05, | |
| "loss": 1.0194, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 8.49, | |
| "learning_rate": 3.590330788804072e-05, | |
| "loss": 0.9183, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 8.5, | |
| "learning_rate": 3.5877862595419845e-05, | |
| "loss": 0.9661, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "learning_rate": 3.585241730279899e-05, | |
| "loss": 1.1686, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 8.53, | |
| "learning_rate": 3.5826972010178115e-05, | |
| "loss": 0.7814, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 8.55, | |
| "learning_rate": 3.580152671755726e-05, | |
| "loss": 0.8574, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 8.56, | |
| "learning_rate": 3.5776081424936385e-05, | |
| "loss": 0.8797, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "learning_rate": 3.5750636132315526e-05, | |
| "loss": 0.9171, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 8.6, | |
| "learning_rate": 3.5725190839694654e-05, | |
| "loss": 0.9354, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 8.61, | |
| "learning_rate": 3.5699745547073796e-05, | |
| "loss": 1.018, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 8.63, | |
| "learning_rate": 3.567430025445293e-05, | |
| "loss": 0.9923, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 3.564885496183206e-05, | |
| "loss": 0.9084, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 8.66, | |
| "learning_rate": 3.56234096692112e-05, | |
| "loss": 1.002, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 8.67, | |
| "learning_rate": 3.559796437659033e-05, | |
| "loss": 0.8637, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 8.69, | |
| "learning_rate": 3.557251908396947e-05, | |
| "loss": 0.8584, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 8.7, | |
| "learning_rate": 3.55470737913486e-05, | |
| "loss": 0.819, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 8.72, | |
| "learning_rate": 3.552162849872774e-05, | |
| "loss": 0.9845, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 8.73, | |
| "learning_rate": 3.549618320610687e-05, | |
| "loss": 0.9483, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 8.75, | |
| "learning_rate": 3.547073791348601e-05, | |
| "loss": 0.7486, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 8.76, | |
| "learning_rate": 3.544529262086514e-05, | |
| "loss": 0.863, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "learning_rate": 3.541984732824428e-05, | |
| "loss": 0.9741, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 8.79, | |
| "learning_rate": 3.5394402035623414e-05, | |
| "loss": 1.0203, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 8.81, | |
| "learning_rate": 3.536895674300255e-05, | |
| "loss": 1.056, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 3.5343511450381684e-05, | |
| "loss": 0.9974, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 8.84, | |
| "learning_rate": 3.531806615776082e-05, | |
| "loss": 0.8415, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 8.85, | |
| "learning_rate": 3.529262086513995e-05, | |
| "loss": 1.1325, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 8.87, | |
| "learning_rate": 3.526717557251908e-05, | |
| "loss": 0.7591, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 8.89, | |
| "learning_rate": 3.524173027989822e-05, | |
| "loss": 1.0119, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 8.9, | |
| "learning_rate": 3.521628498727735e-05, | |
| "loss": 0.8473, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 8.92, | |
| "learning_rate": 3.519083969465649e-05, | |
| "loss": 0.8779, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 8.93, | |
| "learning_rate": 3.516539440203562e-05, | |
| "loss": 0.7857, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 8.95, | |
| "learning_rate": 3.513994910941476e-05, | |
| "loss": 0.9987, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "learning_rate": 3.511450381679389e-05, | |
| "loss": 1.1549, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 8.98, | |
| "learning_rate": 3.508905852417303e-05, | |
| "loss": 0.8328, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 8.99, | |
| "learning_rate": 3.506361323155217e-05, | |
| "loss": 0.8897, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 3.50381679389313e-05, | |
| "loss": 0.8221, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 3.5012722646310436e-05, | |
| "loss": 0.8064, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 3.498727735368957e-05, | |
| "loss": 0.6959, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 9.05, | |
| "learning_rate": 3.4961832061068706e-05, | |
| "loss": 0.8445, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 9.07, | |
| "learning_rate": 3.493638676844784e-05, | |
| "loss": 0.8955, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 3.4910941475826976e-05, | |
| "loss": 0.5805, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 9.1, | |
| "learning_rate": 3.4885496183206104e-05, | |
| "loss": 0.8472, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 9.11, | |
| "learning_rate": 3.4860050890585245e-05, | |
| "loss": 0.88, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 9.13, | |
| "learning_rate": 3.483460559796437e-05, | |
| "loss": 0.8396, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 9.15, | |
| "learning_rate": 3.4809160305343515e-05, | |
| "loss": 0.8501, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 9.16, | |
| "learning_rate": 3.478371501272264e-05, | |
| "loss": 0.8663, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 9.16, | |
| "eval_cer": 0.785, | |
| "eval_loss": 1.9156314134597778, | |
| "eval_runtime": 12.5379, | |
| "eval_samples_per_second": 20.817, | |
| "eval_steps_per_second": 2.632, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 9.18, | |
| "learning_rate": 3.4758269720101785e-05, | |
| "loss": 0.8473, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 9.19, | |
| "learning_rate": 3.473282442748092e-05, | |
| "loss": 1.0161, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 9.21, | |
| "learning_rate": 3.4707379134860054e-05, | |
| "loss": 0.7519, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 9.22, | |
| "learning_rate": 3.468193384223919e-05, | |
| "loss": 0.7459, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 9.24, | |
| "learning_rate": 3.4656488549618324e-05, | |
| "loss": 0.7898, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 9.25, | |
| "learning_rate": 3.463104325699746e-05, | |
| "loss": 0.6242, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 3.4605597964376594e-05, | |
| "loss": 0.8839, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 9.28, | |
| "learning_rate": 3.458015267175573e-05, | |
| "loss": 0.9561, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 9.3, | |
| "learning_rate": 3.4554707379134856e-05, | |
| "loss": 1.0197, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 9.31, | |
| "learning_rate": 3.4529262086514e-05, | |
| "loss": 0.996, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 3.4503816793893126e-05, | |
| "loss": 0.7677, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 9.34, | |
| "learning_rate": 3.447837150127227e-05, | |
| "loss": 0.7711, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 9.36, | |
| "learning_rate": 3.44529262086514e-05, | |
| "loss": 0.9012, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 9.37, | |
| "learning_rate": 3.442748091603054e-05, | |
| "loss": 0.8905, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 9.39, | |
| "learning_rate": 3.440203562340967e-05, | |
| "loss": 0.8825, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 9.4, | |
| "learning_rate": 3.437659033078881e-05, | |
| "loss": 0.6976, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 9.42, | |
| "learning_rate": 3.435114503816794e-05, | |
| "loss": 0.7713, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 9.44, | |
| "learning_rate": 3.432569974554708e-05, | |
| "loss": 0.7792, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 9.45, | |
| "learning_rate": 3.430025445292621e-05, | |
| "loss": 0.7397, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 9.47, | |
| "learning_rate": 3.4274809160305346e-05, | |
| "loss": 0.7844, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 9.48, | |
| "learning_rate": 3.424936386768448e-05, | |
| "loss": 0.8553, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 9.5, | |
| "learning_rate": 3.4223918575063616e-05, | |
| "loss": 0.8848, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 9.51, | |
| "learning_rate": 3.419847328244275e-05, | |
| "loss": 0.698, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 9.53, | |
| "learning_rate": 3.417302798982188e-05, | |
| "loss": 0.7525, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 9.54, | |
| "learning_rate": 3.414758269720102e-05, | |
| "loss": 0.6092, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 9.56, | |
| "learning_rate": 3.4122137404580155e-05, | |
| "loss": 0.9465, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 9.57, | |
| "learning_rate": 3.409669211195929e-05, | |
| "loss": 0.7059, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 9.59, | |
| "learning_rate": 3.4071246819338425e-05, | |
| "loss": 0.8609, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 9.6, | |
| "learning_rate": 3.404580152671756e-05, | |
| "loss": 0.579, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 9.62, | |
| "learning_rate": 3.4020356234096695e-05, | |
| "loss": 0.8543, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 9.63, | |
| "learning_rate": 3.399491094147583e-05, | |
| "loss": 1.0405, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 9.65, | |
| "learning_rate": 3.3969465648854964e-05, | |
| "loss": 1.0546, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 9.66, | |
| "learning_rate": 3.39440203562341e-05, | |
| "loss": 0.8783, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 9.68, | |
| "learning_rate": 3.3918575063613234e-05, | |
| "loss": 0.7796, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 9.69, | |
| "learning_rate": 3.389312977099237e-05, | |
| "loss": 0.6612, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 9.71, | |
| "learning_rate": 3.3867684478371504e-05, | |
| "loss": 0.7545, | |
| "step": 1272 | |
| }, | |
| { | |
| "epoch": 9.73, | |
| "learning_rate": 3.384223918575064e-05, | |
| "loss": 0.8692, | |
| "step": 1274 | |
| }, | |
| { | |
| "epoch": 9.74, | |
| "learning_rate": 3.381679389312977e-05, | |
| "loss": 0.7158, | |
| "step": 1276 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 3.379134860050891e-05, | |
| "loss": 0.8735, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 9.77, | |
| "learning_rate": 3.376590330788804e-05, | |
| "loss": 0.7916, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 9.79, | |
| "learning_rate": 3.374045801526718e-05, | |
| "loss": 0.802, | |
| "step": 1282 | |
| }, | |
| { | |
| "epoch": 9.8, | |
| "learning_rate": 3.371501272264631e-05, | |
| "loss": 0.8425, | |
| "step": 1284 | |
| }, | |
| { | |
| "epoch": 9.82, | |
| "learning_rate": 3.368956743002545e-05, | |
| "loss": 0.7276, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 9.83, | |
| "learning_rate": 3.366412213740458e-05, | |
| "loss": 0.891, | |
| "step": 1288 | |
| }, | |
| { | |
| "epoch": 9.85, | |
| "learning_rate": 3.363867684478372e-05, | |
| "loss": 0.7913, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 9.86, | |
| "learning_rate": 3.361323155216285e-05, | |
| "loss": 0.7408, | |
| "step": 1292 | |
| }, | |
| { | |
| "epoch": 9.88, | |
| "learning_rate": 3.358778625954199e-05, | |
| "loss": 0.8176, | |
| "step": 1294 | |
| }, | |
| { | |
| "epoch": 9.89, | |
| "learning_rate": 3.356234096692112e-05, | |
| "loss": 0.7871, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 9.91, | |
| "learning_rate": 3.3536895674300256e-05, | |
| "loss": 0.7954, | |
| "step": 1298 | |
| }, | |
| { | |
| "epoch": 9.92, | |
| "learning_rate": 3.351145038167939e-05, | |
| "loss": 0.9755, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 9.92, | |
| "eval_cer": 0.485, | |
| "eval_loss": 2.777585506439209, | |
| "eval_runtime": 12.8482, | |
| "eval_samples_per_second": 20.314, | |
| "eval_steps_per_second": 2.568, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 9.94, | |
| "learning_rate": 3.3486005089058526e-05, | |
| "loss": 0.8733, | |
| "step": 1302 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "learning_rate": 3.346055979643766e-05, | |
| "loss": 0.9032, | |
| "step": 1304 | |
| }, | |
| { | |
| "epoch": 9.97, | |
| "learning_rate": 3.3435114503816796e-05, | |
| "loss": 0.6854, | |
| "step": 1306 | |
| }, | |
| { | |
| "epoch": 9.98, | |
| "learning_rate": 3.340966921119593e-05, | |
| "loss": 0.9199, | |
| "step": 1308 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 3.3384223918575065e-05, | |
| "loss": 0.8205, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 3.33587786259542e-05, | |
| "loss": 0.9542, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 3.3333333333333335e-05, | |
| "loss": 0.6914, | |
| "step": 1314 | |
| }, | |
| { | |
| "epoch": 10.05, | |
| "learning_rate": 3.330788804071247e-05, | |
| "loss": 0.7872, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 10.06, | |
| "learning_rate": 3.3282442748091605e-05, | |
| "loss": 0.744, | |
| "step": 1318 | |
| }, | |
| { | |
| "epoch": 10.08, | |
| "learning_rate": 3.325699745547074e-05, | |
| "loss": 0.8267, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 10.09, | |
| "learning_rate": 3.3231552162849874e-05, | |
| "loss": 0.9622, | |
| "step": 1322 | |
| }, | |
| { | |
| "epoch": 10.11, | |
| "learning_rate": 3.320610687022901e-05, | |
| "loss": 0.9869, | |
| "step": 1324 | |
| }, | |
| { | |
| "epoch": 10.12, | |
| "learning_rate": 3.3180661577608144e-05, | |
| "loss": 0.883, | |
| "step": 1326 | |
| }, | |
| { | |
| "epoch": 10.14, | |
| "learning_rate": 3.315521628498728e-05, | |
| "loss": 0.7592, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 10.15, | |
| "learning_rate": 3.3129770992366414e-05, | |
| "loss": 0.8645, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 10.17, | |
| "learning_rate": 3.310432569974555e-05, | |
| "loss": 0.7664, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 10.18, | |
| "learning_rate": 3.307888040712468e-05, | |
| "loss": 0.8177, | |
| "step": 1334 | |
| }, | |
| { | |
| "epoch": 10.2, | |
| "learning_rate": 3.305343511450382e-05, | |
| "loss": 0.9408, | |
| "step": 1336 | |
| }, | |
| { | |
| "epoch": 10.21, | |
| "learning_rate": 3.302798982188295e-05, | |
| "loss": 0.8006, | |
| "step": 1338 | |
| }, | |
| { | |
| "epoch": 10.23, | |
| "learning_rate": 3.300254452926209e-05, | |
| "loss": 0.8512, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 10.24, | |
| "learning_rate": 3.297709923664122e-05, | |
| "loss": 0.6951, | |
| "step": 1342 | |
| }, | |
| { | |
| "epoch": 10.26, | |
| "learning_rate": 3.295165394402036e-05, | |
| "loss": 0.8256, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 10.27, | |
| "learning_rate": 3.292620865139949e-05, | |
| "loss": 0.879, | |
| "step": 1346 | |
| }, | |
| { | |
| "epoch": 10.29, | |
| "learning_rate": 3.290076335877863e-05, | |
| "loss": 0.8452, | |
| "step": 1348 | |
| }, | |
| { | |
| "epoch": 10.31, | |
| "learning_rate": 3.287531806615776e-05, | |
| "loss": 0.7518, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 10.32, | |
| "learning_rate": 3.28498727735369e-05, | |
| "loss": 0.7397, | |
| "step": 1352 | |
| }, | |
| { | |
| "epoch": 10.34, | |
| "learning_rate": 3.282442748091603e-05, | |
| "loss": 0.7552, | |
| "step": 1354 | |
| }, | |
| { | |
| "epoch": 10.35, | |
| "learning_rate": 3.2798982188295166e-05, | |
| "loss": 1.0041, | |
| "step": 1356 | |
| }, | |
| { | |
| "epoch": 10.37, | |
| "learning_rate": 3.27735368956743e-05, | |
| "loss": 1.0255, | |
| "step": 1358 | |
| }, | |
| { | |
| "epoch": 10.38, | |
| "learning_rate": 3.2748091603053436e-05, | |
| "loss": 0.8253, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 10.4, | |
| "learning_rate": 3.272264631043257e-05, | |
| "loss": 0.8581, | |
| "step": 1362 | |
| }, | |
| { | |
| "epoch": 10.41, | |
| "learning_rate": 3.2697201017811706e-05, | |
| "loss": 0.8025, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 10.43, | |
| "learning_rate": 3.267175572519084e-05, | |
| "loss": 0.5604, | |
| "step": 1366 | |
| }, | |
| { | |
| "epoch": 10.44, | |
| "learning_rate": 3.2646310432569975e-05, | |
| "loss": 0.7864, | |
| "step": 1368 | |
| }, | |
| { | |
| "epoch": 10.46, | |
| "learning_rate": 3.262086513994911e-05, | |
| "loss": 0.7906, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 10.47, | |
| "learning_rate": 3.2595419847328245e-05, | |
| "loss": 0.7996, | |
| "step": 1372 | |
| }, | |
| { | |
| "epoch": 10.49, | |
| "learning_rate": 3.256997455470738e-05, | |
| "loss": 0.66, | |
| "step": 1374 | |
| }, | |
| { | |
| "epoch": 10.5, | |
| "learning_rate": 3.2544529262086515e-05, | |
| "loss": 0.6681, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 10.52, | |
| "learning_rate": 3.251908396946565e-05, | |
| "loss": 0.6779, | |
| "step": 1378 | |
| }, | |
| { | |
| "epoch": 10.53, | |
| "learning_rate": 3.2493638676844784e-05, | |
| "loss": 0.7656, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 10.55, | |
| "learning_rate": 3.246819338422392e-05, | |
| "loss": 0.8681, | |
| "step": 1382 | |
| }, | |
| { | |
| "epoch": 10.56, | |
| "learning_rate": 3.2442748091603054e-05, | |
| "loss": 0.7303, | |
| "step": 1384 | |
| }, | |
| { | |
| "epoch": 10.58, | |
| "learning_rate": 3.241730279898219e-05, | |
| "loss": 0.8804, | |
| "step": 1386 | |
| }, | |
| { | |
| "epoch": 10.6, | |
| "learning_rate": 3.2391857506361324e-05, | |
| "loss": 0.8637, | |
| "step": 1388 | |
| }, | |
| { | |
| "epoch": 10.61, | |
| "learning_rate": 3.236641221374046e-05, | |
| "loss": 0.9304, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 10.63, | |
| "learning_rate": 3.234096692111959e-05, | |
| "loss": 0.7039, | |
| "step": 1392 | |
| }, | |
| { | |
| "epoch": 10.64, | |
| "learning_rate": 3.231552162849873e-05, | |
| "loss": 0.7348, | |
| "step": 1394 | |
| }, | |
| { | |
| "epoch": 10.66, | |
| "learning_rate": 3.229007633587787e-05, | |
| "loss": 0.7779, | |
| "step": 1396 | |
| }, | |
| { | |
| "epoch": 10.67, | |
| "learning_rate": 3.2264631043257e-05, | |
| "loss": 0.6528, | |
| "step": 1398 | |
| }, | |
| { | |
| "epoch": 10.69, | |
| "learning_rate": 3.223918575063614e-05, | |
| "loss": 0.6447, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 10.69, | |
| "eval_cer": 0.8266666666666667, | |
| "eval_loss": 2.4866857528686523, | |
| "eval_runtime": 13.2056, | |
| "eval_samples_per_second": 19.764, | |
| "eval_steps_per_second": 2.499, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 10.7, | |
| "learning_rate": 3.221374045801527e-05, | |
| "loss": 0.771, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 10.72, | |
| "learning_rate": 3.21882951653944e-05, | |
| "loss": 0.7643, | |
| "step": 1404 | |
| }, | |
| { | |
| "epoch": 10.73, | |
| "learning_rate": 3.216284987277354e-05, | |
| "loss": 0.6259, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 10.75, | |
| "learning_rate": 3.213740458015267e-05, | |
| "loss": 0.9097, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 10.76, | |
| "learning_rate": 3.211195928753181e-05, | |
| "loss": 0.7458, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 10.78, | |
| "learning_rate": 3.208651399491094e-05, | |
| "loss": 0.6047, | |
| "step": 1412 | |
| }, | |
| { | |
| "epoch": 10.79, | |
| "learning_rate": 3.2061068702290076e-05, | |
| "loss": 0.7833, | |
| "step": 1414 | |
| }, | |
| { | |
| "epoch": 10.81, | |
| "learning_rate": 3.203562340966921e-05, | |
| "loss": 0.8714, | |
| "step": 1416 | |
| }, | |
| { | |
| "epoch": 10.82, | |
| "learning_rate": 3.2010178117048346e-05, | |
| "loss": 0.8053, | |
| "step": 1418 | |
| }, | |
| { | |
| "epoch": 10.84, | |
| "learning_rate": 3.198473282442748e-05, | |
| "loss": 0.7495, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 10.85, | |
| "learning_rate": 3.195928753180662e-05, | |
| "loss": 0.7018, | |
| "step": 1422 | |
| }, | |
| { | |
| "epoch": 10.87, | |
| "learning_rate": 3.193384223918575e-05, | |
| "loss": 0.5693, | |
| "step": 1424 | |
| }, | |
| { | |
| "epoch": 10.89, | |
| "learning_rate": 3.190839694656489e-05, | |
| "loss": 0.8921, | |
| "step": 1426 | |
| }, | |
| { | |
| "epoch": 10.9, | |
| "learning_rate": 3.188295165394402e-05, | |
| "loss": 0.6977, | |
| "step": 1428 | |
| }, | |
| { | |
| "epoch": 10.92, | |
| "learning_rate": 3.185750636132316e-05, | |
| "loss": 0.7847, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 10.93, | |
| "learning_rate": 3.183206106870229e-05, | |
| "loss": 0.7979, | |
| "step": 1432 | |
| }, | |
| { | |
| "epoch": 10.95, | |
| "learning_rate": 3.1806615776081425e-05, | |
| "loss": 0.7237, | |
| "step": 1434 | |
| }, | |
| { | |
| "epoch": 10.96, | |
| "learning_rate": 3.178117048346056e-05, | |
| "loss": 0.6758, | |
| "step": 1436 | |
| }, | |
| { | |
| "epoch": 10.98, | |
| "learning_rate": 3.1755725190839694e-05, | |
| "loss": 0.6873, | |
| "step": 1438 | |
| }, | |
| { | |
| "epoch": 10.99, | |
| "learning_rate": 3.173027989821883e-05, | |
| "loss": 0.9999, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 3.1704834605597964e-05, | |
| "loss": 1.0997, | |
| "step": 1442 | |
| }, | |
| { | |
| "epoch": 11.02, | |
| "learning_rate": 3.16793893129771e-05, | |
| "loss": 0.8083, | |
| "step": 1444 | |
| }, | |
| { | |
| "epoch": 11.04, | |
| "learning_rate": 3.1653944020356234e-05, | |
| "loss": 0.6659, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 11.05, | |
| "learning_rate": 3.1628498727735375e-05, | |
| "loss": 0.7703, | |
| "step": 1448 | |
| }, | |
| { | |
| "epoch": 11.07, | |
| "learning_rate": 3.16030534351145e-05, | |
| "loss": 0.7253, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 11.08, | |
| "learning_rate": 3.1577608142493645e-05, | |
| "loss": 0.7858, | |
| "step": 1452 | |
| }, | |
| { | |
| "epoch": 11.1, | |
| "learning_rate": 3.155216284987277e-05, | |
| "loss": 0.8834, | |
| "step": 1454 | |
| }, | |
| { | |
| "epoch": 11.11, | |
| "learning_rate": 3.1526717557251914e-05, | |
| "loss": 0.7663, | |
| "step": 1456 | |
| }, | |
| { | |
| "epoch": 11.13, | |
| "learning_rate": 3.150127226463104e-05, | |
| "loss": 0.8507, | |
| "step": 1458 | |
| }, | |
| { | |
| "epoch": 11.15, | |
| "learning_rate": 3.147582697201018e-05, | |
| "loss": 0.7357, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 11.16, | |
| "learning_rate": 3.145038167938931e-05, | |
| "loss": 0.68, | |
| "step": 1462 | |
| }, | |
| { | |
| "epoch": 11.18, | |
| "learning_rate": 3.142493638676845e-05, | |
| "loss": 0.6483, | |
| "step": 1464 | |
| }, | |
| { | |
| "epoch": 11.19, | |
| "learning_rate": 3.139949109414758e-05, | |
| "loss": 0.7656, | |
| "step": 1466 | |
| }, | |
| { | |
| "epoch": 11.21, | |
| "learning_rate": 3.137404580152672e-05, | |
| "loss": 0.5926, | |
| "step": 1468 | |
| }, | |
| { | |
| "epoch": 11.22, | |
| "learning_rate": 3.134860050890586e-05, | |
| "loss": 0.5725, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 11.24, | |
| "learning_rate": 3.1323155216284986e-05, | |
| "loss": 0.9918, | |
| "step": 1472 | |
| }, | |
| { | |
| "epoch": 11.25, | |
| "learning_rate": 3.129770992366413e-05, | |
| "loss": 0.875, | |
| "step": 1474 | |
| }, | |
| { | |
| "epoch": 11.27, | |
| "learning_rate": 3.1272264631043256e-05, | |
| "loss": 0.6928, | |
| "step": 1476 | |
| }, | |
| { | |
| "epoch": 11.28, | |
| "learning_rate": 3.12468193384224e-05, | |
| "loss": 0.7292, | |
| "step": 1478 | |
| }, | |
| { | |
| "epoch": 11.3, | |
| "learning_rate": 3.1221374045801526e-05, | |
| "loss": 0.7387, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 11.31, | |
| "learning_rate": 3.119592875318067e-05, | |
| "loss": 0.67, | |
| "step": 1482 | |
| }, | |
| { | |
| "epoch": 11.33, | |
| "learning_rate": 3.1170483460559795e-05, | |
| "loss": 0.6382, | |
| "step": 1484 | |
| }, | |
| { | |
| "epoch": 11.34, | |
| "learning_rate": 3.114503816793894e-05, | |
| "loss": 0.7604, | |
| "step": 1486 | |
| }, | |
| { | |
| "epoch": 11.36, | |
| "learning_rate": 3.1119592875318065e-05, | |
| "loss": 0.5445, | |
| "step": 1488 | |
| }, | |
| { | |
| "epoch": 11.37, | |
| "learning_rate": 3.10941475826972e-05, | |
| "loss": 0.7119, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 11.39, | |
| "learning_rate": 3.1068702290076335e-05, | |
| "loss": 0.7391, | |
| "step": 1492 | |
| }, | |
| { | |
| "epoch": 11.4, | |
| "learning_rate": 3.104325699745547e-05, | |
| "loss": 0.8999, | |
| "step": 1494 | |
| }, | |
| { | |
| "epoch": 11.42, | |
| "learning_rate": 3.101781170483461e-05, | |
| "loss": 0.8159, | |
| "step": 1496 | |
| }, | |
| { | |
| "epoch": 11.44, | |
| "learning_rate": 3.099236641221374e-05, | |
| "loss": 0.6531, | |
| "step": 1498 | |
| }, | |
| { | |
| "epoch": 11.45, | |
| "learning_rate": 3.096692111959288e-05, | |
| "loss": 0.7862, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 11.45, | |
| "eval_cer": 0.61, | |
| "eval_loss": 3.6975245475769043, | |
| "eval_runtime": 13.3376, | |
| "eval_samples_per_second": 19.569, | |
| "eval_steps_per_second": 2.474, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 11.47, | |
| "learning_rate": 3.094147582697201e-05, | |
| "loss": 0.7377, | |
| "step": 1502 | |
| }, | |
| { | |
| "epoch": 11.48, | |
| "learning_rate": 3.091603053435115e-05, | |
| "loss": 1.1392, | |
| "step": 1504 | |
| }, | |
| { | |
| "epoch": 11.5, | |
| "learning_rate": 3.089058524173028e-05, | |
| "loss": 0.8604, | |
| "step": 1506 | |
| }, | |
| { | |
| "epoch": 11.51, | |
| "learning_rate": 3.086513994910942e-05, | |
| "loss": 0.7484, | |
| "step": 1508 | |
| }, | |
| { | |
| "epoch": 11.53, | |
| "learning_rate": 3.083969465648855e-05, | |
| "loss": 0.736, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 11.54, | |
| "learning_rate": 3.081424936386769e-05, | |
| "loss": 0.7803, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 11.56, | |
| "learning_rate": 3.078880407124682e-05, | |
| "loss": 0.7426, | |
| "step": 1514 | |
| }, | |
| { | |
| "epoch": 11.57, | |
| "learning_rate": 3.076335877862595e-05, | |
| "loss": 0.6179, | |
| "step": 1516 | |
| }, | |
| { | |
| "epoch": 11.59, | |
| "learning_rate": 3.0737913486005094e-05, | |
| "loss": 0.789, | |
| "step": 1518 | |
| }, | |
| { | |
| "epoch": 11.6, | |
| "learning_rate": 3.071246819338422e-05, | |
| "loss": 0.7082, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 11.62, | |
| "learning_rate": 3.0687022900763364e-05, | |
| "loss": 0.6718, | |
| "step": 1522 | |
| }, | |
| { | |
| "epoch": 11.63, | |
| "learning_rate": 3.066157760814249e-05, | |
| "loss": 0.7069, | |
| "step": 1524 | |
| }, | |
| { | |
| "epoch": 11.65, | |
| "learning_rate": 3.0636132315521633e-05, | |
| "loss": 0.7823, | |
| "step": 1526 | |
| }, | |
| { | |
| "epoch": 11.66, | |
| "learning_rate": 3.061068702290076e-05, | |
| "loss": 0.5196, | |
| "step": 1528 | |
| }, | |
| { | |
| "epoch": 11.68, | |
| "learning_rate": 3.05852417302799e-05, | |
| "loss": 0.6285, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 11.69, | |
| "learning_rate": 3.055979643765903e-05, | |
| "loss": 0.6355, | |
| "step": 1532 | |
| }, | |
| { | |
| "epoch": 11.71, | |
| "learning_rate": 3.053435114503817e-05, | |
| "loss": 0.7717, | |
| "step": 1534 | |
| }, | |
| { | |
| "epoch": 11.73, | |
| "learning_rate": 3.05089058524173e-05, | |
| "loss": 0.7217, | |
| "step": 1536 | |
| }, | |
| { | |
| "epoch": 11.74, | |
| "learning_rate": 3.048346055979644e-05, | |
| "loss": 0.5821, | |
| "step": 1538 | |
| }, | |
| { | |
| "epoch": 11.76, | |
| "learning_rate": 3.045801526717557e-05, | |
| "loss": 0.5953, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 11.77, | |
| "learning_rate": 3.043256997455471e-05, | |
| "loss": 0.7736, | |
| "step": 1542 | |
| }, | |
| { | |
| "epoch": 11.79, | |
| "learning_rate": 3.0407124681933847e-05, | |
| "loss": 0.5606, | |
| "step": 1544 | |
| }, | |
| { | |
| "epoch": 11.8, | |
| "learning_rate": 3.038167938931298e-05, | |
| "loss": 0.8974, | |
| "step": 1546 | |
| }, | |
| { | |
| "epoch": 11.82, | |
| "learning_rate": 3.0356234096692117e-05, | |
| "loss": 0.8495, | |
| "step": 1548 | |
| }, | |
| { | |
| "epoch": 11.83, | |
| "learning_rate": 3.0330788804071248e-05, | |
| "loss": 0.8835, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 11.85, | |
| "learning_rate": 3.0305343511450386e-05, | |
| "loss": 0.5954, | |
| "step": 1552 | |
| }, | |
| { | |
| "epoch": 11.86, | |
| "learning_rate": 3.0279898218829518e-05, | |
| "loss": 0.7219, | |
| "step": 1554 | |
| }, | |
| { | |
| "epoch": 11.88, | |
| "learning_rate": 3.0254452926208652e-05, | |
| "loss": 0.8032, | |
| "step": 1556 | |
| }, | |
| { | |
| "epoch": 11.89, | |
| "learning_rate": 3.0229007633587787e-05, | |
| "loss": 0.6683, | |
| "step": 1558 | |
| }, | |
| { | |
| "epoch": 11.91, | |
| "learning_rate": 3.0203562340966922e-05, | |
| "loss": 0.6912, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 11.92, | |
| "learning_rate": 3.0178117048346054e-05, | |
| "loss": 0.6155, | |
| "step": 1562 | |
| }, | |
| { | |
| "epoch": 11.94, | |
| "learning_rate": 3.0152671755725192e-05, | |
| "loss": 0.7019, | |
| "step": 1564 | |
| }, | |
| { | |
| "epoch": 11.95, | |
| "learning_rate": 3.0127226463104323e-05, | |
| "loss": 0.6955, | |
| "step": 1566 | |
| }, | |
| { | |
| "epoch": 11.97, | |
| "learning_rate": 3.010178117048346e-05, | |
| "loss": 0.598, | |
| "step": 1568 | |
| }, | |
| { | |
| "epoch": 11.98, | |
| "learning_rate": 3.00763358778626e-05, | |
| "loss": 0.6676, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 3.005089058524173e-05, | |
| "loss": 0.4945, | |
| "step": 1572 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 3.002544529262087e-05, | |
| "loss": 0.6349, | |
| "step": 1574 | |
| }, | |
| { | |
| "epoch": 12.03, | |
| "learning_rate": 3e-05, | |
| "loss": 0.5586, | |
| "step": 1576 | |
| }, | |
| { | |
| "epoch": 12.05, | |
| "learning_rate": 2.997455470737914e-05, | |
| "loss": 0.6804, | |
| "step": 1578 | |
| }, | |
| { | |
| "epoch": 12.06, | |
| "learning_rate": 2.994910941475827e-05, | |
| "loss": 0.5436, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 12.08, | |
| "learning_rate": 2.992366412213741e-05, | |
| "loss": 1.0288, | |
| "step": 1582 | |
| }, | |
| { | |
| "epoch": 12.09, | |
| "learning_rate": 2.989821882951654e-05, | |
| "loss": 0.6567, | |
| "step": 1584 | |
| }, | |
| { | |
| "epoch": 12.11, | |
| "learning_rate": 2.9872773536895675e-05, | |
| "loss": 0.8401, | |
| "step": 1586 | |
| }, | |
| { | |
| "epoch": 12.12, | |
| "learning_rate": 2.984732824427481e-05, | |
| "loss": 0.8642, | |
| "step": 1588 | |
| }, | |
| { | |
| "epoch": 12.14, | |
| "learning_rate": 2.9821882951653945e-05, | |
| "loss": 0.652, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 12.15, | |
| "learning_rate": 2.9796437659033083e-05, | |
| "loss": 0.8162, | |
| "step": 1592 | |
| }, | |
| { | |
| "epoch": 12.17, | |
| "learning_rate": 2.9770992366412214e-05, | |
| "loss": 0.8508, | |
| "step": 1594 | |
| }, | |
| { | |
| "epoch": 12.18, | |
| "learning_rate": 2.9745547073791352e-05, | |
| "loss": 0.5377, | |
| "step": 1596 | |
| }, | |
| { | |
| "epoch": 12.2, | |
| "learning_rate": 2.9720101781170484e-05, | |
| "loss": 1.0559, | |
| "step": 1598 | |
| }, | |
| { | |
| "epoch": 12.21, | |
| "learning_rate": 2.9694656488549622e-05, | |
| "loss": 0.7416, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 12.21, | |
| "eval_cer": 0.155, | |
| "eval_loss": 4.830371379852295, | |
| "eval_runtime": 11.8584, | |
| "eval_samples_per_second": 22.01, | |
| "eval_steps_per_second": 2.783, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 12.23, | |
| "learning_rate": 2.9669211195928754e-05, | |
| "loss": 0.7896, | |
| "step": 1602 | |
| }, | |
| { | |
| "epoch": 12.24, | |
| "learning_rate": 2.9643765903307892e-05, | |
| "loss": 0.8071, | |
| "step": 1604 | |
| }, | |
| { | |
| "epoch": 12.26, | |
| "learning_rate": 2.9618320610687023e-05, | |
| "loss": 0.6148, | |
| "step": 1606 | |
| }, | |
| { | |
| "epoch": 12.27, | |
| "learning_rate": 2.959287531806616e-05, | |
| "loss": 0.7511, | |
| "step": 1608 | |
| }, | |
| { | |
| "epoch": 12.29, | |
| "learning_rate": 2.9567430025445293e-05, | |
| "loss": 0.7985, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 12.31, | |
| "learning_rate": 2.9541984732824428e-05, | |
| "loss": 0.6049, | |
| "step": 1612 | |
| }, | |
| { | |
| "epoch": 12.32, | |
| "learning_rate": 2.9516539440203562e-05, | |
| "loss": 0.6502, | |
| "step": 1614 | |
| }, | |
| { | |
| "epoch": 12.34, | |
| "learning_rate": 2.9491094147582697e-05, | |
| "loss": 0.7028, | |
| "step": 1616 | |
| }, | |
| { | |
| "epoch": 12.35, | |
| "learning_rate": 2.9465648854961836e-05, | |
| "loss": 0.8401, | |
| "step": 1618 | |
| }, | |
| { | |
| "epoch": 12.37, | |
| "learning_rate": 2.9440203562340967e-05, | |
| "loss": 0.7012, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 12.38, | |
| "learning_rate": 2.9414758269720105e-05, | |
| "loss": 0.6168, | |
| "step": 1622 | |
| }, | |
| { | |
| "epoch": 12.4, | |
| "learning_rate": 2.9389312977099237e-05, | |
| "loss": 0.8844, | |
| "step": 1624 | |
| }, | |
| { | |
| "epoch": 12.41, | |
| "learning_rate": 2.9363867684478375e-05, | |
| "loss": 0.581, | |
| "step": 1626 | |
| }, | |
| { | |
| "epoch": 12.43, | |
| "learning_rate": 2.9338422391857506e-05, | |
| "loss": 0.628, | |
| "step": 1628 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "learning_rate": 2.9312977099236644e-05, | |
| "loss": 0.8216, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 12.46, | |
| "learning_rate": 2.9287531806615776e-05, | |
| "loss": 0.5529, | |
| "step": 1632 | |
| }, | |
| { | |
| "epoch": 12.47, | |
| "learning_rate": 2.9262086513994914e-05, | |
| "loss": 0.8191, | |
| "step": 1634 | |
| }, | |
| { | |
| "epoch": 12.49, | |
| "learning_rate": 2.9236641221374046e-05, | |
| "loss": 0.5529, | |
| "step": 1636 | |
| }, | |
| { | |
| "epoch": 12.5, | |
| "learning_rate": 2.9211195928753184e-05, | |
| "loss": 0.7198, | |
| "step": 1638 | |
| }, | |
| { | |
| "epoch": 12.52, | |
| "learning_rate": 2.918575063613232e-05, | |
| "loss": 0.7897, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 12.53, | |
| "learning_rate": 2.916030534351145e-05, | |
| "loss": 0.6908, | |
| "step": 1642 | |
| }, | |
| { | |
| "epoch": 12.55, | |
| "learning_rate": 2.9134860050890588e-05, | |
| "loss": 0.7455, | |
| "step": 1644 | |
| }, | |
| { | |
| "epoch": 12.56, | |
| "learning_rate": 2.910941475826972e-05, | |
| "loss": 0.8934, | |
| "step": 1646 | |
| }, | |
| { | |
| "epoch": 12.58, | |
| "learning_rate": 2.9083969465648858e-05, | |
| "loss": 0.9264, | |
| "step": 1648 | |
| }, | |
| { | |
| "epoch": 12.6, | |
| "learning_rate": 2.905852417302799e-05, | |
| "loss": 0.6212, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 12.61, | |
| "learning_rate": 2.9033078880407128e-05, | |
| "loss": 0.758, | |
| "step": 1652 | |
| }, | |
| { | |
| "epoch": 12.63, | |
| "learning_rate": 2.900763358778626e-05, | |
| "loss": 0.5884, | |
| "step": 1654 | |
| }, | |
| { | |
| "epoch": 12.64, | |
| "learning_rate": 2.8982188295165397e-05, | |
| "loss": 0.8146, | |
| "step": 1656 | |
| }, | |
| { | |
| "epoch": 12.66, | |
| "learning_rate": 2.895674300254453e-05, | |
| "loss": 0.7774, | |
| "step": 1658 | |
| }, | |
| { | |
| "epoch": 12.67, | |
| "learning_rate": 2.8931297709923667e-05, | |
| "loss": 0.6874, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 12.69, | |
| "learning_rate": 2.89058524173028e-05, | |
| "loss": 0.6767, | |
| "step": 1662 | |
| }, | |
| { | |
| "epoch": 12.7, | |
| "learning_rate": 2.8880407124681937e-05, | |
| "loss": 0.7755, | |
| "step": 1664 | |
| }, | |
| { | |
| "epoch": 12.72, | |
| "learning_rate": 2.885496183206107e-05, | |
| "loss": 0.6158, | |
| "step": 1666 | |
| }, | |
| { | |
| "epoch": 12.73, | |
| "learning_rate": 2.8829516539440203e-05, | |
| "loss": 0.6351, | |
| "step": 1668 | |
| }, | |
| { | |
| "epoch": 12.75, | |
| "learning_rate": 2.880407124681934e-05, | |
| "loss": 0.8882, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 12.76, | |
| "learning_rate": 2.8778625954198472e-05, | |
| "loss": 0.5635, | |
| "step": 1672 | |
| }, | |
| { | |
| "epoch": 12.78, | |
| "learning_rate": 2.875318066157761e-05, | |
| "loss": 0.8622, | |
| "step": 1674 | |
| }, | |
| { | |
| "epoch": 12.79, | |
| "learning_rate": 2.8727735368956742e-05, | |
| "loss": 0.7301, | |
| "step": 1676 | |
| }, | |
| { | |
| "epoch": 12.81, | |
| "learning_rate": 2.870229007633588e-05, | |
| "loss": 0.7363, | |
| "step": 1678 | |
| }, | |
| { | |
| "epoch": 12.82, | |
| "learning_rate": 2.8676844783715012e-05, | |
| "loss": 0.7507, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 12.84, | |
| "learning_rate": 2.865139949109415e-05, | |
| "loss": 0.6413, | |
| "step": 1682 | |
| }, | |
| { | |
| "epoch": 12.85, | |
| "learning_rate": 2.862595419847328e-05, | |
| "loss": 0.7979, | |
| "step": 1684 | |
| }, | |
| { | |
| "epoch": 12.87, | |
| "learning_rate": 2.860050890585242e-05, | |
| "loss": 0.6187, | |
| "step": 1686 | |
| }, | |
| { | |
| "epoch": 12.89, | |
| "learning_rate": 2.857506361323155e-05, | |
| "loss": 0.5141, | |
| "step": 1688 | |
| }, | |
| { | |
| "epoch": 12.9, | |
| "learning_rate": 2.854961832061069e-05, | |
| "loss": 0.6732, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 12.92, | |
| "learning_rate": 2.8524173027989824e-05, | |
| "loss": 0.6868, | |
| "step": 1692 | |
| }, | |
| { | |
| "epoch": 12.93, | |
| "learning_rate": 2.849872773536896e-05, | |
| "loss": 0.643, | |
| "step": 1694 | |
| }, | |
| { | |
| "epoch": 12.95, | |
| "learning_rate": 2.8473282442748094e-05, | |
| "loss": 0.6161, | |
| "step": 1696 | |
| }, | |
| { | |
| "epoch": 12.96, | |
| "learning_rate": 2.8447837150127225e-05, | |
| "loss": 0.6776, | |
| "step": 1698 | |
| }, | |
| { | |
| "epoch": 12.98, | |
| "learning_rate": 2.8422391857506363e-05, | |
| "loss": 0.5768, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 12.98, | |
| "eval_cer": 0.985, | |
| "eval_loss": 2.614406108856201, | |
| "eval_runtime": 12.9478, | |
| "eval_samples_per_second": 20.158, | |
| "eval_steps_per_second": 2.549, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 12.99, | |
| "learning_rate": 2.8396946564885495e-05, | |
| "loss": 0.7055, | |
| "step": 1702 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 2.8371501272264633e-05, | |
| "loss": 0.6081, | |
| "step": 1704 | |
| }, | |
| { | |
| "epoch": 13.02, | |
| "learning_rate": 2.8346055979643765e-05, | |
| "loss": 0.6232, | |
| "step": 1706 | |
| }, | |
| { | |
| "epoch": 13.04, | |
| "learning_rate": 2.8320610687022903e-05, | |
| "loss": 1.0069, | |
| "step": 1708 | |
| }, | |
| { | |
| "epoch": 13.05, | |
| "learning_rate": 2.8295165394402034e-05, | |
| "loss": 0.8088, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 13.07, | |
| "learning_rate": 2.8269720101781172e-05, | |
| "loss": 0.5509, | |
| "step": 1712 | |
| }, | |
| { | |
| "epoch": 13.08, | |
| "learning_rate": 2.824427480916031e-05, | |
| "loss": 0.6812, | |
| "step": 1714 | |
| }, | |
| { | |
| "epoch": 13.1, | |
| "learning_rate": 2.8218829516539442e-05, | |
| "loss": 0.6886, | |
| "step": 1716 | |
| }, | |
| { | |
| "epoch": 13.11, | |
| "learning_rate": 2.8193384223918577e-05, | |
| "loss": 0.6288, | |
| "step": 1718 | |
| }, | |
| { | |
| "epoch": 13.13, | |
| "learning_rate": 2.8167938931297712e-05, | |
| "loss": 0.5272, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 13.15, | |
| "learning_rate": 2.8142493638676847e-05, | |
| "loss": 1.008, | |
| "step": 1722 | |
| }, | |
| { | |
| "epoch": 13.16, | |
| "learning_rate": 2.8117048346055978e-05, | |
| "loss": 0.8586, | |
| "step": 1724 | |
| }, | |
| { | |
| "epoch": 13.18, | |
| "learning_rate": 2.8091603053435116e-05, | |
| "loss": 0.5503, | |
| "step": 1726 | |
| }, | |
| { | |
| "epoch": 13.19, | |
| "learning_rate": 2.8066157760814248e-05, | |
| "loss": 0.7417, | |
| "step": 1728 | |
| }, | |
| { | |
| "epoch": 13.21, | |
| "learning_rate": 2.8040712468193386e-05, | |
| "loss": 0.6712, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 13.22, | |
| "learning_rate": 2.8015267175572517e-05, | |
| "loss": 0.6285, | |
| "step": 1732 | |
| }, | |
| { | |
| "epoch": 13.24, | |
| "learning_rate": 2.7989821882951656e-05, | |
| "loss": 0.6874, | |
| "step": 1734 | |
| }, | |
| { | |
| "epoch": 13.25, | |
| "learning_rate": 2.7964376590330787e-05, | |
| "loss": 0.6408, | |
| "step": 1736 | |
| }, | |
| { | |
| "epoch": 13.27, | |
| "learning_rate": 2.7938931297709925e-05, | |
| "loss": 0.6697, | |
| "step": 1738 | |
| }, | |
| { | |
| "epoch": 13.28, | |
| "learning_rate": 2.7913486005089063e-05, | |
| "loss": 0.5801, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 13.3, | |
| "learning_rate": 2.7888040712468195e-05, | |
| "loss": 0.6358, | |
| "step": 1742 | |
| }, | |
| { | |
| "epoch": 13.31, | |
| "learning_rate": 2.7862595419847333e-05, | |
| "loss": 0.745, | |
| "step": 1744 | |
| }, | |
| { | |
| "epoch": 13.33, | |
| "learning_rate": 2.7837150127226465e-05, | |
| "loss": 0.7081, | |
| "step": 1746 | |
| }, | |
| { | |
| "epoch": 13.34, | |
| "learning_rate": 2.78117048346056e-05, | |
| "loss": 0.659, | |
| "step": 1748 | |
| }, | |
| { | |
| "epoch": 13.36, | |
| "learning_rate": 2.7786259541984734e-05, | |
| "loss": 0.5541, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 13.37, | |
| "learning_rate": 2.776081424936387e-05, | |
| "loss": 0.7937, | |
| "step": 1752 | |
| }, | |
| { | |
| "epoch": 13.39, | |
| "learning_rate": 2.7735368956743e-05, | |
| "loss": 0.9197, | |
| "step": 1754 | |
| }, | |
| { | |
| "epoch": 13.4, | |
| "learning_rate": 2.770992366412214e-05, | |
| "loss": 0.6734, | |
| "step": 1756 | |
| }, | |
| { | |
| "epoch": 13.42, | |
| "learning_rate": 2.768447837150127e-05, | |
| "loss": 0.6809, | |
| "step": 1758 | |
| }, | |
| { | |
| "epoch": 13.44, | |
| "learning_rate": 2.7659033078880408e-05, | |
| "loss": 0.712, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 13.45, | |
| "learning_rate": 2.7633587786259547e-05, | |
| "loss": 0.6177, | |
| "step": 1762 | |
| }, | |
| { | |
| "epoch": 13.47, | |
| "learning_rate": 2.7608142493638678e-05, | |
| "loss": 0.6693, | |
| "step": 1764 | |
| }, | |
| { | |
| "epoch": 13.48, | |
| "learning_rate": 2.7582697201017816e-05, | |
| "loss": 0.6131, | |
| "step": 1766 | |
| }, | |
| { | |
| "epoch": 13.5, | |
| "learning_rate": 2.7557251908396948e-05, | |
| "loss": 0.61, | |
| "step": 1768 | |
| }, | |
| { | |
| "epoch": 13.51, | |
| "learning_rate": 2.7531806615776086e-05, | |
| "loss": 0.5576, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 13.53, | |
| "learning_rate": 2.7506361323155217e-05, | |
| "loss": 0.7008, | |
| "step": 1772 | |
| }, | |
| { | |
| "epoch": 13.54, | |
| "learning_rate": 2.7480916030534355e-05, | |
| "loss": 0.6033, | |
| "step": 1774 | |
| }, | |
| { | |
| "epoch": 13.56, | |
| "learning_rate": 2.7455470737913487e-05, | |
| "loss": 0.4839, | |
| "step": 1776 | |
| }, | |
| { | |
| "epoch": 13.57, | |
| "learning_rate": 2.7430025445292622e-05, | |
| "loss": 0.6713, | |
| "step": 1778 | |
| }, | |
| { | |
| "epoch": 13.59, | |
| "learning_rate": 2.7404580152671757e-05, | |
| "loss": 0.6208, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 13.6, | |
| "learning_rate": 2.737913486005089e-05, | |
| "loss": 0.6956, | |
| "step": 1782 | |
| }, | |
| { | |
| "epoch": 13.62, | |
| "learning_rate": 2.7353689567430023e-05, | |
| "loss": 0.6163, | |
| "step": 1784 | |
| }, | |
| { | |
| "epoch": 13.63, | |
| "learning_rate": 2.732824427480916e-05, | |
| "loss": 0.682, | |
| "step": 1786 | |
| }, | |
| { | |
| "epoch": 13.65, | |
| "learning_rate": 2.73027989821883e-05, | |
| "loss": 0.6238, | |
| "step": 1788 | |
| }, | |
| { | |
| "epoch": 13.66, | |
| "learning_rate": 2.727735368956743e-05, | |
| "loss": 0.8029, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 13.68, | |
| "learning_rate": 2.725190839694657e-05, | |
| "loss": 0.6024, | |
| "step": 1792 | |
| }, | |
| { | |
| "epoch": 13.69, | |
| "learning_rate": 2.72264631043257e-05, | |
| "loss": 0.6567, | |
| "step": 1794 | |
| }, | |
| { | |
| "epoch": 13.71, | |
| "learning_rate": 2.720101781170484e-05, | |
| "loss": 0.6399, | |
| "step": 1796 | |
| }, | |
| { | |
| "epoch": 13.73, | |
| "learning_rate": 2.717557251908397e-05, | |
| "loss": 0.7036, | |
| "step": 1798 | |
| }, | |
| { | |
| "epoch": 13.74, | |
| "learning_rate": 2.7150127226463108e-05, | |
| "loss": 0.6198, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 13.74, | |
| "eval_cer": 0.9283333333333333, | |
| "eval_loss": 1.9728145599365234, | |
| "eval_runtime": 13.0399, | |
| "eval_samples_per_second": 20.016, | |
| "eval_steps_per_second": 2.531, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 13.76, | |
| "learning_rate": 2.712468193384224e-05, | |
| "loss": 0.7954, | |
| "step": 1802 | |
| }, | |
| { | |
| "epoch": 13.77, | |
| "learning_rate": 2.7099236641221375e-05, | |
| "loss": 0.6562, | |
| "step": 1804 | |
| }, | |
| { | |
| "epoch": 13.79, | |
| "learning_rate": 2.707379134860051e-05, | |
| "loss": 0.6525, | |
| "step": 1806 | |
| }, | |
| { | |
| "epoch": 13.8, | |
| "learning_rate": 2.7048346055979644e-05, | |
| "loss": 0.6105, | |
| "step": 1808 | |
| }, | |
| { | |
| "epoch": 13.82, | |
| "learning_rate": 2.7022900763358776e-05, | |
| "loss": 0.6545, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 13.83, | |
| "learning_rate": 2.6997455470737914e-05, | |
| "loss": 0.593, | |
| "step": 1812 | |
| }, | |
| { | |
| "epoch": 13.85, | |
| "learning_rate": 2.6972010178117052e-05, | |
| "loss": 0.6329, | |
| "step": 1814 | |
| }, | |
| { | |
| "epoch": 13.86, | |
| "learning_rate": 2.6946564885496183e-05, | |
| "loss": 0.7738, | |
| "step": 1816 | |
| }, | |
| { | |
| "epoch": 13.88, | |
| "learning_rate": 2.6921119592875322e-05, | |
| "loss": 0.6553, | |
| "step": 1818 | |
| }, | |
| { | |
| "epoch": 13.89, | |
| "learning_rate": 2.6895674300254453e-05, | |
| "loss": 0.7885, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 13.91, | |
| "learning_rate": 2.687022900763359e-05, | |
| "loss": 0.7487, | |
| "step": 1822 | |
| }, | |
| { | |
| "epoch": 13.92, | |
| "learning_rate": 2.6844783715012723e-05, | |
| "loss": 0.538, | |
| "step": 1824 | |
| }, | |
| { | |
| "epoch": 13.94, | |
| "learning_rate": 2.681933842239186e-05, | |
| "loss": 0.641, | |
| "step": 1826 | |
| }, | |
| { | |
| "epoch": 13.95, | |
| "learning_rate": 2.6793893129770992e-05, | |
| "loss": 1.0136, | |
| "step": 1828 | |
| }, | |
| { | |
| "epoch": 13.97, | |
| "learning_rate": 2.676844783715013e-05, | |
| "loss": 0.6821, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 13.98, | |
| "learning_rate": 2.6743002544529262e-05, | |
| "loss": 0.5798, | |
| "step": 1832 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "learning_rate": 2.6717557251908397e-05, | |
| "loss": 0.6055, | |
| "step": 1834 | |
| }, | |
| { | |
| "epoch": 14.02, | |
| "learning_rate": 2.6692111959287535e-05, | |
| "loss": 0.6657, | |
| "step": 1836 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "learning_rate": 2.6666666666666667e-05, | |
| "loss": 0.5488, | |
| "step": 1838 | |
| }, | |
| { | |
| "epoch": 14.05, | |
| "learning_rate": 2.6641221374045805e-05, | |
| "loss": 0.6495, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 14.06, | |
| "learning_rate": 2.6615776081424936e-05, | |
| "loss": 0.7168, | |
| "step": 1842 | |
| }, | |
| { | |
| "epoch": 14.08, | |
| "learning_rate": 2.6590330788804074e-05, | |
| "loss": 0.6762, | |
| "step": 1844 | |
| }, | |
| { | |
| "epoch": 14.09, | |
| "learning_rate": 2.6564885496183206e-05, | |
| "loss": 0.6912, | |
| "step": 1846 | |
| }, | |
| { | |
| "epoch": 14.11, | |
| "learning_rate": 2.6539440203562344e-05, | |
| "loss": 0.6808, | |
| "step": 1848 | |
| }, | |
| { | |
| "epoch": 14.12, | |
| "learning_rate": 2.6513994910941476e-05, | |
| "loss": 0.714, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 14.14, | |
| "learning_rate": 2.6488549618320614e-05, | |
| "loss": 0.6953, | |
| "step": 1852 | |
| }, | |
| { | |
| "epoch": 14.15, | |
| "learning_rate": 2.6463104325699745e-05, | |
| "loss": 0.7978, | |
| "step": 1854 | |
| }, | |
| { | |
| "epoch": 14.17, | |
| "learning_rate": 2.6437659033078883e-05, | |
| "loss": 0.7208, | |
| "step": 1856 | |
| }, | |
| { | |
| "epoch": 14.18, | |
| "learning_rate": 2.6412213740458015e-05, | |
| "loss": 0.7389, | |
| "step": 1858 | |
| }, | |
| { | |
| "epoch": 14.2, | |
| "learning_rate": 2.638676844783715e-05, | |
| "loss": 0.6913, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 14.21, | |
| "learning_rate": 2.6361323155216288e-05, | |
| "loss": 0.589, | |
| "step": 1862 | |
| }, | |
| { | |
| "epoch": 14.23, | |
| "learning_rate": 2.633587786259542e-05, | |
| "loss": 0.596, | |
| "step": 1864 | |
| }, | |
| { | |
| "epoch": 14.24, | |
| "learning_rate": 2.6310432569974558e-05, | |
| "loss": 0.6922, | |
| "step": 1866 | |
| }, | |
| { | |
| "epoch": 14.26, | |
| "learning_rate": 2.628498727735369e-05, | |
| "loss": 0.5771, | |
| "step": 1868 | |
| }, | |
| { | |
| "epoch": 14.27, | |
| "learning_rate": 2.6259541984732827e-05, | |
| "loss": 0.4962, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 14.29, | |
| "learning_rate": 2.623409669211196e-05, | |
| "loss": 0.7108, | |
| "step": 1872 | |
| }, | |
| { | |
| "epoch": 14.31, | |
| "learning_rate": 2.6208651399491097e-05, | |
| "loss": 0.6828, | |
| "step": 1874 | |
| }, | |
| { | |
| "epoch": 14.32, | |
| "learning_rate": 2.618320610687023e-05, | |
| "loss": 0.4449, | |
| "step": 1876 | |
| }, | |
| { | |
| "epoch": 14.34, | |
| "learning_rate": 2.6157760814249367e-05, | |
| "loss": 0.7032, | |
| "step": 1878 | |
| }, | |
| { | |
| "epoch": 14.35, | |
| "learning_rate": 2.6132315521628498e-05, | |
| "loss": 0.6961, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 14.37, | |
| "learning_rate": 2.6106870229007636e-05, | |
| "loss": 0.6392, | |
| "step": 1882 | |
| }, | |
| { | |
| "epoch": 14.38, | |
| "learning_rate": 2.608142493638677e-05, | |
| "loss": 0.662, | |
| "step": 1884 | |
| }, | |
| { | |
| "epoch": 14.4, | |
| "learning_rate": 2.6055979643765906e-05, | |
| "loss": 0.8816, | |
| "step": 1886 | |
| }, | |
| { | |
| "epoch": 14.41, | |
| "learning_rate": 2.603053435114504e-05, | |
| "loss": 0.6262, | |
| "step": 1888 | |
| }, | |
| { | |
| "epoch": 14.43, | |
| "learning_rate": 2.6005089058524172e-05, | |
| "loss": 0.5687, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 14.44, | |
| "learning_rate": 2.597964376590331e-05, | |
| "loss": 0.7207, | |
| "step": 1892 | |
| }, | |
| { | |
| "epoch": 14.46, | |
| "learning_rate": 2.5954198473282442e-05, | |
| "loss": 0.5061, | |
| "step": 1894 | |
| }, | |
| { | |
| "epoch": 14.47, | |
| "learning_rate": 2.592875318066158e-05, | |
| "loss": 0.5091, | |
| "step": 1896 | |
| }, | |
| { | |
| "epoch": 14.49, | |
| "learning_rate": 2.590330788804071e-05, | |
| "loss": 0.5773, | |
| "step": 1898 | |
| }, | |
| { | |
| "epoch": 14.5, | |
| "learning_rate": 2.587786259541985e-05, | |
| "loss": 0.5782, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 14.5, | |
| "eval_cer": 0.235, | |
| "eval_loss": 4.354420185089111, | |
| "eval_runtime": 12.4381, | |
| "eval_samples_per_second": 20.984, | |
| "eval_steps_per_second": 2.653, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 14.52, | |
| "learning_rate": 2.585241730279898e-05, | |
| "loss": 0.4823, | |
| "step": 1902 | |
| }, | |
| { | |
| "epoch": 14.53, | |
| "learning_rate": 2.582697201017812e-05, | |
| "loss": 0.5894, | |
| "step": 1904 | |
| }, | |
| { | |
| "epoch": 14.55, | |
| "learning_rate": 2.580152671755725e-05, | |
| "loss": 0.4522, | |
| "step": 1906 | |
| }, | |
| { | |
| "epoch": 14.56, | |
| "learning_rate": 2.577608142493639e-05, | |
| "loss": 0.5878, | |
| "step": 1908 | |
| }, | |
| { | |
| "epoch": 14.58, | |
| "learning_rate": 2.5750636132315524e-05, | |
| "loss": 0.7595, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 14.6, | |
| "learning_rate": 2.572519083969466e-05, | |
| "loss": 0.599, | |
| "step": 1912 | |
| }, | |
| { | |
| "epoch": 14.61, | |
| "learning_rate": 2.5699745547073793e-05, | |
| "loss": 0.6153, | |
| "step": 1914 | |
| }, | |
| { | |
| "epoch": 14.63, | |
| "learning_rate": 2.5674300254452925e-05, | |
| "loss": 0.59, | |
| "step": 1916 | |
| }, | |
| { | |
| "epoch": 14.64, | |
| "learning_rate": 2.5648854961832063e-05, | |
| "loss": 0.6025, | |
| "step": 1918 | |
| }, | |
| { | |
| "epoch": 14.66, | |
| "learning_rate": 2.5623409669211195e-05, | |
| "loss": 0.6712, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 14.67, | |
| "learning_rate": 2.5597964376590333e-05, | |
| "loss": 0.6139, | |
| "step": 1922 | |
| }, | |
| { | |
| "epoch": 14.69, | |
| "learning_rate": 2.5572519083969464e-05, | |
| "loss": 0.6953, | |
| "step": 1924 | |
| }, | |
| { | |
| "epoch": 14.7, | |
| "learning_rate": 2.5547073791348602e-05, | |
| "loss": 0.6114, | |
| "step": 1926 | |
| }, | |
| { | |
| "epoch": 14.72, | |
| "learning_rate": 2.5521628498727734e-05, | |
| "loss": 0.6656, | |
| "step": 1928 | |
| }, | |
| { | |
| "epoch": 14.73, | |
| "learning_rate": 2.5496183206106872e-05, | |
| "loss": 0.6653, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 14.75, | |
| "learning_rate": 2.5470737913486003e-05, | |
| "loss": 0.7021, | |
| "step": 1932 | |
| }, | |
| { | |
| "epoch": 14.76, | |
| "learning_rate": 2.5445292620865142e-05, | |
| "loss": 0.4988, | |
| "step": 1934 | |
| }, | |
| { | |
| "epoch": 14.78, | |
| "learning_rate": 2.541984732824428e-05, | |
| "loss": 0.6053, | |
| "step": 1936 | |
| }, | |
| { | |
| "epoch": 14.79, | |
| "learning_rate": 2.539440203562341e-05, | |
| "loss": 0.537, | |
| "step": 1938 | |
| }, | |
| { | |
| "epoch": 14.81, | |
| "learning_rate": 2.5368956743002546e-05, | |
| "loss": 0.8234, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 14.82, | |
| "learning_rate": 2.534351145038168e-05, | |
| "loss": 0.6854, | |
| "step": 1942 | |
| }, | |
| { | |
| "epoch": 14.84, | |
| "learning_rate": 2.5318066157760816e-05, | |
| "loss": 0.8084, | |
| "step": 1944 | |
| }, | |
| { | |
| "epoch": 14.85, | |
| "learning_rate": 2.5292620865139947e-05, | |
| "loss": 0.7823, | |
| "step": 1946 | |
| }, | |
| { | |
| "epoch": 14.87, | |
| "learning_rate": 2.5267175572519086e-05, | |
| "loss": 0.8129, | |
| "step": 1948 | |
| }, | |
| { | |
| "epoch": 14.89, | |
| "learning_rate": 2.5241730279898217e-05, | |
| "loss": 0.6747, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 14.9, | |
| "learning_rate": 2.5216284987277355e-05, | |
| "loss": 0.5221, | |
| "step": 1952 | |
| }, | |
| { | |
| "epoch": 14.92, | |
| "learning_rate": 2.5190839694656487e-05, | |
| "loss": 0.6249, | |
| "step": 1954 | |
| }, | |
| { | |
| "epoch": 14.93, | |
| "learning_rate": 2.5165394402035625e-05, | |
| "loss": 0.5185, | |
| "step": 1956 | |
| }, | |
| { | |
| "epoch": 14.95, | |
| "learning_rate": 2.5139949109414763e-05, | |
| "loss": 0.5373, | |
| "step": 1958 | |
| }, | |
| { | |
| "epoch": 14.96, | |
| "learning_rate": 2.5114503816793894e-05, | |
| "loss": 0.7259, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 14.98, | |
| "learning_rate": 2.5089058524173033e-05, | |
| "loss": 0.4947, | |
| "step": 1962 | |
| }, | |
| { | |
| "epoch": 14.99, | |
| "learning_rate": 2.5063613231552164e-05, | |
| "loss": 0.5521, | |
| "step": 1964 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 2.50381679389313e-05, | |
| "loss": 0.4607, | |
| "step": 1966 | |
| }, | |
| { | |
| "epoch": 15.02, | |
| "learning_rate": 2.5012722646310434e-05, | |
| "loss": 0.5285, | |
| "step": 1968 | |
| }, | |
| { | |
| "epoch": 15.04, | |
| "learning_rate": 2.498727735368957e-05, | |
| "loss": 0.8805, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 15.05, | |
| "learning_rate": 2.4961832061068703e-05, | |
| "loss": 0.5849, | |
| "step": 1972 | |
| }, | |
| { | |
| "epoch": 15.07, | |
| "learning_rate": 2.4936386768447838e-05, | |
| "loss": 0.6189, | |
| "step": 1974 | |
| }, | |
| { | |
| "epoch": 15.08, | |
| "learning_rate": 2.4910941475826973e-05, | |
| "loss": 0.6684, | |
| "step": 1976 | |
| }, | |
| { | |
| "epoch": 15.1, | |
| "learning_rate": 2.4885496183206108e-05, | |
| "loss": 0.7003, | |
| "step": 1978 | |
| }, | |
| { | |
| "epoch": 15.11, | |
| "learning_rate": 2.4860050890585243e-05, | |
| "loss": 0.6328, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 15.13, | |
| "learning_rate": 2.4834605597964378e-05, | |
| "loss": 0.6727, | |
| "step": 1982 | |
| }, | |
| { | |
| "epoch": 15.15, | |
| "learning_rate": 2.4809160305343512e-05, | |
| "loss": 0.7608, | |
| "step": 1984 | |
| }, | |
| { | |
| "epoch": 15.16, | |
| "learning_rate": 2.4783715012722647e-05, | |
| "loss": 0.5894, | |
| "step": 1986 | |
| }, | |
| { | |
| "epoch": 15.18, | |
| "learning_rate": 2.4758269720101782e-05, | |
| "loss": 0.771, | |
| "step": 1988 | |
| }, | |
| { | |
| "epoch": 15.19, | |
| "learning_rate": 2.4732824427480917e-05, | |
| "loss": 0.6449, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 15.21, | |
| "learning_rate": 2.4707379134860052e-05, | |
| "loss": 0.6029, | |
| "step": 1992 | |
| }, | |
| { | |
| "epoch": 15.22, | |
| "learning_rate": 2.468193384223919e-05, | |
| "loss": 0.6521, | |
| "step": 1994 | |
| }, | |
| { | |
| "epoch": 15.24, | |
| "learning_rate": 2.465648854961832e-05, | |
| "loss": 0.5042, | |
| "step": 1996 | |
| }, | |
| { | |
| "epoch": 15.25, | |
| "learning_rate": 2.4631043256997456e-05, | |
| "loss": 0.6181, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 15.27, | |
| "learning_rate": 2.460559796437659e-05, | |
| "loss": 0.6579, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 15.27, | |
| "eval_cer": 0.38666666666666666, | |
| "eval_loss": 3.3335180282592773, | |
| "eval_runtime": 12.8303, | |
| "eval_samples_per_second": 20.342, | |
| "eval_steps_per_second": 2.572, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 15.28, | |
| "learning_rate": 2.4580152671755726e-05, | |
| "loss": 0.66, | |
| "step": 2002 | |
| }, | |
| { | |
| "epoch": 15.3, | |
| "learning_rate": 2.455470737913486e-05, | |
| "loss": 0.6787, | |
| "step": 2004 | |
| }, | |
| { | |
| "epoch": 15.31, | |
| "learning_rate": 2.4529262086513996e-05, | |
| "loss": 0.7532, | |
| "step": 2006 | |
| }, | |
| { | |
| "epoch": 15.33, | |
| "learning_rate": 2.450381679389313e-05, | |
| "loss": 0.6325, | |
| "step": 2008 | |
| }, | |
| { | |
| "epoch": 15.34, | |
| "learning_rate": 2.4478371501272265e-05, | |
| "loss": 0.5222, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 15.36, | |
| "learning_rate": 2.44529262086514e-05, | |
| "loss": 0.5865, | |
| "step": 2012 | |
| }, | |
| { | |
| "epoch": 15.37, | |
| "learning_rate": 2.4427480916030535e-05, | |
| "loss": 0.6388, | |
| "step": 2014 | |
| }, | |
| { | |
| "epoch": 15.39, | |
| "learning_rate": 2.440203562340967e-05, | |
| "loss": 0.455, | |
| "step": 2016 | |
| }, | |
| { | |
| "epoch": 15.4, | |
| "learning_rate": 2.4376590330788808e-05, | |
| "loss": 0.7409, | |
| "step": 2018 | |
| }, | |
| { | |
| "epoch": 15.42, | |
| "learning_rate": 2.4351145038167943e-05, | |
| "loss": 0.5731, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 15.44, | |
| "learning_rate": 2.4325699745547078e-05, | |
| "loss": 0.5482, | |
| "step": 2022 | |
| }, | |
| { | |
| "epoch": 15.45, | |
| "learning_rate": 2.430025445292621e-05, | |
| "loss": 0.5084, | |
| "step": 2024 | |
| }, | |
| { | |
| "epoch": 15.47, | |
| "learning_rate": 2.4274809160305344e-05, | |
| "loss": 0.4516, | |
| "step": 2026 | |
| }, | |
| { | |
| "epoch": 15.48, | |
| "learning_rate": 2.424936386768448e-05, | |
| "loss": 0.5458, | |
| "step": 2028 | |
| }, | |
| { | |
| "epoch": 15.5, | |
| "learning_rate": 2.4223918575063613e-05, | |
| "loss": 0.5029, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 15.51, | |
| "learning_rate": 2.4198473282442748e-05, | |
| "loss": 0.4957, | |
| "step": 2032 | |
| }, | |
| { | |
| "epoch": 15.53, | |
| "learning_rate": 2.4173027989821883e-05, | |
| "loss": 0.5328, | |
| "step": 2034 | |
| }, | |
| { | |
| "epoch": 15.54, | |
| "learning_rate": 2.4147582697201018e-05, | |
| "loss": 0.4934, | |
| "step": 2036 | |
| }, | |
| { | |
| "epoch": 15.56, | |
| "learning_rate": 2.4122137404580153e-05, | |
| "loss": 0.6305, | |
| "step": 2038 | |
| }, | |
| { | |
| "epoch": 15.57, | |
| "learning_rate": 2.4096692111959288e-05, | |
| "loss": 0.6244, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 15.59, | |
| "learning_rate": 2.4071246819338422e-05, | |
| "loss": 0.7726, | |
| "step": 2042 | |
| }, | |
| { | |
| "epoch": 15.6, | |
| "learning_rate": 2.404580152671756e-05, | |
| "loss": 0.6559, | |
| "step": 2044 | |
| }, | |
| { | |
| "epoch": 15.62, | |
| "learning_rate": 2.4020356234096695e-05, | |
| "loss": 0.5439, | |
| "step": 2046 | |
| }, | |
| { | |
| "epoch": 15.63, | |
| "learning_rate": 2.399491094147583e-05, | |
| "loss": 0.6267, | |
| "step": 2048 | |
| }, | |
| { | |
| "epoch": 15.65, | |
| "learning_rate": 2.3969465648854965e-05, | |
| "loss": 0.7257, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 15.66, | |
| "learning_rate": 2.3944020356234097e-05, | |
| "loss": 0.6406, | |
| "step": 2052 | |
| }, | |
| { | |
| "epoch": 15.68, | |
| "learning_rate": 2.391857506361323e-05, | |
| "loss": 0.6335, | |
| "step": 2054 | |
| }, | |
| { | |
| "epoch": 15.69, | |
| "learning_rate": 2.3893129770992366e-05, | |
| "loss": 0.6477, | |
| "step": 2056 | |
| }, | |
| { | |
| "epoch": 15.71, | |
| "learning_rate": 2.38676844783715e-05, | |
| "loss": 0.4761, | |
| "step": 2058 | |
| }, | |
| { | |
| "epoch": 15.73, | |
| "learning_rate": 2.3842239185750636e-05, | |
| "loss": 0.6198, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 15.74, | |
| "learning_rate": 2.381679389312977e-05, | |
| "loss": 0.6266, | |
| "step": 2062 | |
| }, | |
| { | |
| "epoch": 15.76, | |
| "learning_rate": 2.3791348600508906e-05, | |
| "loss": 0.7595, | |
| "step": 2064 | |
| }, | |
| { | |
| "epoch": 15.77, | |
| "learning_rate": 2.376590330788804e-05, | |
| "loss": 0.703, | |
| "step": 2066 | |
| }, | |
| { | |
| "epoch": 15.79, | |
| "learning_rate": 2.374045801526718e-05, | |
| "loss": 0.5967, | |
| "step": 2068 | |
| }, | |
| { | |
| "epoch": 15.8, | |
| "learning_rate": 2.3715012722646313e-05, | |
| "loss": 0.6107, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 15.82, | |
| "learning_rate": 2.3689567430025448e-05, | |
| "loss": 0.4899, | |
| "step": 2072 | |
| }, | |
| { | |
| "epoch": 15.83, | |
| "learning_rate": 2.3664122137404583e-05, | |
| "loss": 0.6638, | |
| "step": 2074 | |
| }, | |
| { | |
| "epoch": 15.85, | |
| "learning_rate": 2.3638676844783718e-05, | |
| "loss": 0.5411, | |
| "step": 2076 | |
| }, | |
| { | |
| "epoch": 15.86, | |
| "learning_rate": 2.3613231552162853e-05, | |
| "loss": 0.5502, | |
| "step": 2078 | |
| }, | |
| { | |
| "epoch": 15.88, | |
| "learning_rate": 2.3587786259541984e-05, | |
| "loss": 0.6979, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 15.89, | |
| "learning_rate": 2.356234096692112e-05, | |
| "loss": 0.5588, | |
| "step": 2082 | |
| }, | |
| { | |
| "epoch": 15.91, | |
| "learning_rate": 2.3536895674300254e-05, | |
| "loss": 0.6608, | |
| "step": 2084 | |
| }, | |
| { | |
| "epoch": 15.92, | |
| "learning_rate": 2.351145038167939e-05, | |
| "loss": 0.6453, | |
| "step": 2086 | |
| }, | |
| { | |
| "epoch": 15.94, | |
| "learning_rate": 2.3486005089058523e-05, | |
| "loss": 0.5868, | |
| "step": 2088 | |
| }, | |
| { | |
| "epoch": 15.95, | |
| "learning_rate": 2.3460559796437658e-05, | |
| "loss": 0.5969, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 15.97, | |
| "learning_rate": 2.3435114503816796e-05, | |
| "loss": 0.5101, | |
| "step": 2092 | |
| }, | |
| { | |
| "epoch": 15.98, | |
| "learning_rate": 2.340966921119593e-05, | |
| "loss": 0.5134, | |
| "step": 2094 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "learning_rate": 2.3384223918575066e-05, | |
| "loss": 0.6774, | |
| "step": 2096 | |
| }, | |
| { | |
| "epoch": 16.02, | |
| "learning_rate": 2.33587786259542e-05, | |
| "loss": 0.5242, | |
| "step": 2098 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "learning_rate": 2.3333333333333336e-05, | |
| "loss": 0.5213, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "eval_cer": 0.23833333333333334, | |
| "eval_loss": 4.873970031738281, | |
| "eval_runtime": 11.9775, | |
| "eval_samples_per_second": 21.791, | |
| "eval_steps_per_second": 2.755, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 16.05, | |
| "learning_rate": 2.330788804071247e-05, | |
| "loss": 0.5324, | |
| "step": 2102 | |
| }, | |
| { | |
| "epoch": 16.06, | |
| "learning_rate": 2.3282442748091605e-05, | |
| "loss": 0.5842, | |
| "step": 2104 | |
| }, | |
| { | |
| "epoch": 16.08, | |
| "learning_rate": 2.325699745547074e-05, | |
| "loss": 0.6585, | |
| "step": 2106 | |
| }, | |
| { | |
| "epoch": 16.09, | |
| "learning_rate": 2.3231552162849872e-05, | |
| "loss": 0.6619, | |
| "step": 2108 | |
| }, | |
| { | |
| "epoch": 16.11, | |
| "learning_rate": 2.3206106870229007e-05, | |
| "loss": 0.6788, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 16.12, | |
| "learning_rate": 2.318066157760814e-05, | |
| "loss": 0.4784, | |
| "step": 2112 | |
| }, | |
| { | |
| "epoch": 16.14, | |
| "learning_rate": 2.3155216284987276e-05, | |
| "loss": 0.6589, | |
| "step": 2114 | |
| }, | |
| { | |
| "epoch": 16.15, | |
| "learning_rate": 2.3129770992366414e-05, | |
| "loss": 0.8936, | |
| "step": 2116 | |
| }, | |
| { | |
| "epoch": 16.17, | |
| "learning_rate": 2.310432569974555e-05, | |
| "loss": 0.5156, | |
| "step": 2118 | |
| }, | |
| { | |
| "epoch": 16.18, | |
| "learning_rate": 2.3078880407124684e-05, | |
| "loss": 0.7076, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 16.2, | |
| "learning_rate": 2.305343511450382e-05, | |
| "loss": 0.7453, | |
| "step": 2122 | |
| }, | |
| { | |
| "epoch": 16.21, | |
| "learning_rate": 2.3027989821882954e-05, | |
| "loss": 0.4477, | |
| "step": 2124 | |
| }, | |
| { | |
| "epoch": 16.23, | |
| "learning_rate": 2.300254452926209e-05, | |
| "loss": 0.4964, | |
| "step": 2126 | |
| }, | |
| { | |
| "epoch": 16.24, | |
| "learning_rate": 2.2977099236641223e-05, | |
| "loss": 0.7308, | |
| "step": 2128 | |
| }, | |
| { | |
| "epoch": 16.26, | |
| "learning_rate": 2.2951653944020358e-05, | |
| "loss": 0.567, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 16.27, | |
| "learning_rate": 2.2926208651399493e-05, | |
| "loss": 0.46, | |
| "step": 2132 | |
| }, | |
| { | |
| "epoch": 16.29, | |
| "learning_rate": 2.2900763358778628e-05, | |
| "loss": 0.6041, | |
| "step": 2134 | |
| }, | |
| { | |
| "epoch": 16.31, | |
| "learning_rate": 2.287531806615776e-05, | |
| "loss": 0.5844, | |
| "step": 2136 | |
| }, | |
| { | |
| "epoch": 16.32, | |
| "learning_rate": 2.2849872773536894e-05, | |
| "loss": 0.6429, | |
| "step": 2138 | |
| }, | |
| { | |
| "epoch": 16.34, | |
| "learning_rate": 2.2824427480916032e-05, | |
| "loss": 0.6305, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 16.35, | |
| "learning_rate": 2.2798982188295167e-05, | |
| "loss": 0.6734, | |
| "step": 2142 | |
| }, | |
| { | |
| "epoch": 16.37, | |
| "learning_rate": 2.2773536895674302e-05, | |
| "loss": 0.5662, | |
| "step": 2144 | |
| }, | |
| { | |
| "epoch": 16.38, | |
| "learning_rate": 2.2748091603053437e-05, | |
| "loss": 0.525, | |
| "step": 2146 | |
| }, | |
| { | |
| "epoch": 16.4, | |
| "learning_rate": 2.272264631043257e-05, | |
| "loss": 0.4687, | |
| "step": 2148 | |
| }, | |
| { | |
| "epoch": 16.41, | |
| "learning_rate": 2.2697201017811707e-05, | |
| "loss": 0.6332, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 16.43, | |
| "learning_rate": 2.267175572519084e-05, | |
| "loss": 0.6017, | |
| "step": 2152 | |
| }, | |
| { | |
| "epoch": 16.44, | |
| "learning_rate": 2.2646310432569976e-05, | |
| "loss": 0.5675, | |
| "step": 2154 | |
| }, | |
| { | |
| "epoch": 16.46, | |
| "learning_rate": 2.262086513994911e-05, | |
| "loss": 0.5413, | |
| "step": 2156 | |
| }, | |
| { | |
| "epoch": 16.47, | |
| "learning_rate": 2.2595419847328246e-05, | |
| "loss": 0.5284, | |
| "step": 2158 | |
| }, | |
| { | |
| "epoch": 16.49, | |
| "learning_rate": 2.256997455470738e-05, | |
| "loss": 0.6012, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 16.5, | |
| "learning_rate": 2.2544529262086515e-05, | |
| "loss": 0.6813, | |
| "step": 2162 | |
| }, | |
| { | |
| "epoch": 16.52, | |
| "learning_rate": 2.2519083969465647e-05, | |
| "loss": 0.7177, | |
| "step": 2164 | |
| }, | |
| { | |
| "epoch": 16.53, | |
| "learning_rate": 2.2493638676844785e-05, | |
| "loss": 0.5161, | |
| "step": 2166 | |
| }, | |
| { | |
| "epoch": 16.55, | |
| "learning_rate": 2.246819338422392e-05, | |
| "loss": 0.5074, | |
| "step": 2168 | |
| }, | |
| { | |
| "epoch": 16.56, | |
| "learning_rate": 2.2442748091603055e-05, | |
| "loss": 0.5322, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 16.58, | |
| "learning_rate": 2.241730279898219e-05, | |
| "loss": 0.6602, | |
| "step": 2172 | |
| }, | |
| { | |
| "epoch": 16.6, | |
| "learning_rate": 2.2391857506361324e-05, | |
| "loss": 0.6103, | |
| "step": 2174 | |
| }, | |
| { | |
| "epoch": 16.61, | |
| "learning_rate": 2.236641221374046e-05, | |
| "loss": 0.6248, | |
| "step": 2176 | |
| }, | |
| { | |
| "epoch": 16.63, | |
| "learning_rate": 2.2340966921119594e-05, | |
| "loss": 0.6812, | |
| "step": 2178 | |
| }, | |
| { | |
| "epoch": 16.64, | |
| "learning_rate": 2.231552162849873e-05, | |
| "loss": 0.6544, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 16.66, | |
| "learning_rate": 2.2290076335877864e-05, | |
| "loss": 0.7274, | |
| "step": 2182 | |
| }, | |
| { | |
| "epoch": 16.67, | |
| "learning_rate": 2.2264631043257e-05, | |
| "loss": 0.6022, | |
| "step": 2184 | |
| }, | |
| { | |
| "epoch": 16.69, | |
| "learning_rate": 2.2239185750636133e-05, | |
| "loss": 0.5973, | |
| "step": 2186 | |
| }, | |
| { | |
| "epoch": 16.7, | |
| "learning_rate": 2.2213740458015268e-05, | |
| "loss": 0.5278, | |
| "step": 2188 | |
| }, | |
| { | |
| "epoch": 16.72, | |
| "learning_rate": 2.2188295165394403e-05, | |
| "loss": 0.6594, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 16.73, | |
| "learning_rate": 2.2162849872773538e-05, | |
| "loss": 0.4612, | |
| "step": 2192 | |
| }, | |
| { | |
| "epoch": 16.75, | |
| "learning_rate": 2.2137404580152673e-05, | |
| "loss": 0.6669, | |
| "step": 2194 | |
| }, | |
| { | |
| "epoch": 16.76, | |
| "learning_rate": 2.2111959287531808e-05, | |
| "loss": 0.5867, | |
| "step": 2196 | |
| }, | |
| { | |
| "epoch": 16.78, | |
| "learning_rate": 2.2086513994910942e-05, | |
| "loss": 0.5023, | |
| "step": 2198 | |
| }, | |
| { | |
| "epoch": 16.79, | |
| "learning_rate": 2.2061068702290077e-05, | |
| "loss": 0.6719, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 16.79, | |
| "eval_cer": 0.35833333333333334, | |
| "eval_loss": 3.5339906215667725, | |
| "eval_runtime": 12.2511, | |
| "eval_samples_per_second": 21.304, | |
| "eval_steps_per_second": 2.694, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 16.81, | |
| "learning_rate": 2.2035623409669212e-05, | |
| "loss": 0.5517, | |
| "step": 2202 | |
| }, | |
| { | |
| "epoch": 16.82, | |
| "learning_rate": 2.2010178117048347e-05, | |
| "loss": 0.6138, | |
| "step": 2204 | |
| }, | |
| { | |
| "epoch": 16.84, | |
| "learning_rate": 2.198473282442748e-05, | |
| "loss": 0.6301, | |
| "step": 2206 | |
| }, | |
| { | |
| "epoch": 16.85, | |
| "learning_rate": 2.1959287531806617e-05, | |
| "loss": 0.4804, | |
| "step": 2208 | |
| }, | |
| { | |
| "epoch": 16.87, | |
| "learning_rate": 2.193384223918575e-05, | |
| "loss": 0.479, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 16.89, | |
| "learning_rate": 2.1908396946564886e-05, | |
| "loss": 0.6461, | |
| "step": 2212 | |
| }, | |
| { | |
| "epoch": 16.9, | |
| "learning_rate": 2.1882951653944024e-05, | |
| "loss": 0.6714, | |
| "step": 2214 | |
| }, | |
| { | |
| "epoch": 16.92, | |
| "learning_rate": 2.1857506361323156e-05, | |
| "loss": 0.6949, | |
| "step": 2216 | |
| }, | |
| { | |
| "epoch": 16.93, | |
| "learning_rate": 2.183206106870229e-05, | |
| "loss": 0.7101, | |
| "step": 2218 | |
| }, | |
| { | |
| "epoch": 16.95, | |
| "learning_rate": 2.1806615776081425e-05, | |
| "loss": 0.5797, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 16.96, | |
| "learning_rate": 2.178117048346056e-05, | |
| "loss": 0.5848, | |
| "step": 2222 | |
| }, | |
| { | |
| "epoch": 16.98, | |
| "learning_rate": 2.1755725190839695e-05, | |
| "loss": 0.5456, | |
| "step": 2224 | |
| }, | |
| { | |
| "epoch": 16.99, | |
| "learning_rate": 2.173027989821883e-05, | |
| "loss": 0.6577, | |
| "step": 2226 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 2.1704834605597965e-05, | |
| "loss": 0.9026, | |
| "step": 2228 | |
| }, | |
| { | |
| "epoch": 17.02, | |
| "learning_rate": 2.16793893129771e-05, | |
| "loss": 0.5493, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 17.04, | |
| "learning_rate": 2.1653944020356234e-05, | |
| "loss": 0.658, | |
| "step": 2232 | |
| }, | |
| { | |
| "epoch": 17.05, | |
| "learning_rate": 2.162849872773537e-05, | |
| "loss": 0.5155, | |
| "step": 2234 | |
| }, | |
| { | |
| "epoch": 17.07, | |
| "learning_rate": 2.1603053435114504e-05, | |
| "loss": 0.4761, | |
| "step": 2236 | |
| }, | |
| { | |
| "epoch": 17.08, | |
| "learning_rate": 2.1577608142493642e-05, | |
| "loss": 0.4809, | |
| "step": 2238 | |
| }, | |
| { | |
| "epoch": 17.1, | |
| "learning_rate": 2.1552162849872777e-05, | |
| "loss": 0.5304, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 17.11, | |
| "learning_rate": 2.1526717557251912e-05, | |
| "loss": 0.5546, | |
| "step": 2242 | |
| }, | |
| { | |
| "epoch": 17.13, | |
| "learning_rate": 2.1501272264631043e-05, | |
| "loss": 0.6821, | |
| "step": 2244 | |
| }, | |
| { | |
| "epoch": 17.15, | |
| "learning_rate": 2.1475826972010178e-05, | |
| "loss": 0.6316, | |
| "step": 2246 | |
| }, | |
| { | |
| "epoch": 17.16, | |
| "learning_rate": 2.1450381679389313e-05, | |
| "loss": 0.6221, | |
| "step": 2248 | |
| }, | |
| { | |
| "epoch": 17.18, | |
| "learning_rate": 2.1424936386768448e-05, | |
| "loss": 0.4736, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 17.19, | |
| "learning_rate": 2.1399491094147583e-05, | |
| "loss": 0.608, | |
| "step": 2252 | |
| }, | |
| { | |
| "epoch": 17.21, | |
| "learning_rate": 2.1374045801526718e-05, | |
| "loss": 0.6325, | |
| "step": 2254 | |
| }, | |
| { | |
| "epoch": 17.22, | |
| "learning_rate": 2.1348600508905852e-05, | |
| "loss": 0.5805, | |
| "step": 2256 | |
| }, | |
| { | |
| "epoch": 17.24, | |
| "learning_rate": 2.1323155216284987e-05, | |
| "loss": 0.6571, | |
| "step": 2258 | |
| }, | |
| { | |
| "epoch": 17.25, | |
| "learning_rate": 2.1297709923664122e-05, | |
| "loss": 0.6797, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 17.27, | |
| "learning_rate": 2.127226463104326e-05, | |
| "loss": 0.4343, | |
| "step": 2262 | |
| }, | |
| { | |
| "epoch": 17.28, | |
| "learning_rate": 2.1246819338422395e-05, | |
| "loss": 0.4951, | |
| "step": 2264 | |
| }, | |
| { | |
| "epoch": 17.3, | |
| "learning_rate": 2.122137404580153e-05, | |
| "loss": 0.5035, | |
| "step": 2266 | |
| }, | |
| { | |
| "epoch": 17.31, | |
| "learning_rate": 2.1195928753180665e-05, | |
| "loss": 0.629, | |
| "step": 2268 | |
| }, | |
| { | |
| "epoch": 17.33, | |
| "learning_rate": 2.11704834605598e-05, | |
| "loss": 0.4054, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 17.34, | |
| "learning_rate": 2.114503816793893e-05, | |
| "loss": 0.6026, | |
| "step": 2272 | |
| }, | |
| { | |
| "epoch": 17.36, | |
| "learning_rate": 2.1119592875318066e-05, | |
| "loss": 0.5428, | |
| "step": 2274 | |
| }, | |
| { | |
| "epoch": 17.37, | |
| "learning_rate": 2.10941475826972e-05, | |
| "loss": 0.4469, | |
| "step": 2276 | |
| }, | |
| { | |
| "epoch": 17.39, | |
| "learning_rate": 2.1068702290076335e-05, | |
| "loss": 0.4458, | |
| "step": 2278 | |
| }, | |
| { | |
| "epoch": 17.4, | |
| "learning_rate": 2.104325699745547e-05, | |
| "loss": 0.5269, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 17.42, | |
| "learning_rate": 2.1017811704834605e-05, | |
| "loss": 0.5308, | |
| "step": 2282 | |
| }, | |
| { | |
| "epoch": 17.44, | |
| "learning_rate": 2.099236641221374e-05, | |
| "loss": 0.6008, | |
| "step": 2284 | |
| }, | |
| { | |
| "epoch": 17.45, | |
| "learning_rate": 2.0966921119592875e-05, | |
| "loss": 0.7206, | |
| "step": 2286 | |
| }, | |
| { | |
| "epoch": 17.47, | |
| "learning_rate": 2.0941475826972013e-05, | |
| "loss": 0.6058, | |
| "step": 2288 | |
| }, | |
| { | |
| "epoch": 17.48, | |
| "learning_rate": 2.0916030534351148e-05, | |
| "loss": 0.6627, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 17.5, | |
| "learning_rate": 2.0890585241730283e-05, | |
| "loss": 0.5918, | |
| "step": 2292 | |
| }, | |
| { | |
| "epoch": 17.51, | |
| "learning_rate": 2.0865139949109417e-05, | |
| "loss": 0.5845, | |
| "step": 2294 | |
| }, | |
| { | |
| "epoch": 17.53, | |
| "learning_rate": 2.0839694656488552e-05, | |
| "loss": 0.5084, | |
| "step": 2296 | |
| }, | |
| { | |
| "epoch": 17.54, | |
| "learning_rate": 2.0814249363867687e-05, | |
| "loss": 0.5598, | |
| "step": 2298 | |
| }, | |
| { | |
| "epoch": 17.56, | |
| "learning_rate": 2.078880407124682e-05, | |
| "loss": 0.5179, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 17.56, | |
| "eval_cer": 0.2816666666666667, | |
| "eval_loss": 6.160032749176025, | |
| "eval_runtime": 13.1562, | |
| "eval_samples_per_second": 19.839, | |
| "eval_steps_per_second": 2.508, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 17.57, | |
| "learning_rate": 2.0763358778625953e-05, | |
| "loss": 0.4853, | |
| "step": 2302 | |
| }, | |
| { | |
| "epoch": 17.59, | |
| "learning_rate": 2.0737913486005088e-05, | |
| "loss": 0.5039, | |
| "step": 2304 | |
| }, | |
| { | |
| "epoch": 17.6, | |
| "learning_rate": 2.0712468193384223e-05, | |
| "loss": 0.6638, | |
| "step": 2306 | |
| }, | |
| { | |
| "epoch": 17.62, | |
| "learning_rate": 2.0687022900763358e-05, | |
| "loss": 0.8646, | |
| "step": 2308 | |
| }, | |
| { | |
| "epoch": 17.63, | |
| "learning_rate": 2.0661577608142493e-05, | |
| "loss": 0.7057, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 17.65, | |
| "learning_rate": 2.063613231552163e-05, | |
| "loss": 0.6036, | |
| "step": 2312 | |
| }, | |
| { | |
| "epoch": 17.66, | |
| "learning_rate": 2.0610687022900766e-05, | |
| "loss": 0.467, | |
| "step": 2314 | |
| }, | |
| { | |
| "epoch": 17.68, | |
| "learning_rate": 2.05852417302799e-05, | |
| "loss": 0.6336, | |
| "step": 2316 | |
| }, | |
| { | |
| "epoch": 17.69, | |
| "learning_rate": 2.0559796437659035e-05, | |
| "loss": 0.4946, | |
| "step": 2318 | |
| }, | |
| { | |
| "epoch": 17.71, | |
| "learning_rate": 2.053435114503817e-05, | |
| "loss": 0.706, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 17.73, | |
| "learning_rate": 2.0508905852417305e-05, | |
| "loss": 1.0023, | |
| "step": 2322 | |
| }, | |
| { | |
| "epoch": 17.74, | |
| "learning_rate": 2.048346055979644e-05, | |
| "loss": 0.5102, | |
| "step": 2324 | |
| }, | |
| { | |
| "epoch": 17.76, | |
| "learning_rate": 2.0458015267175575e-05, | |
| "loss": 0.5677, | |
| "step": 2326 | |
| }, | |
| { | |
| "epoch": 17.77, | |
| "learning_rate": 2.0432569974554706e-05, | |
| "loss": 0.745, | |
| "step": 2328 | |
| }, | |
| { | |
| "epoch": 17.79, | |
| "learning_rate": 2.040712468193384e-05, | |
| "loss": 0.6287, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 17.8, | |
| "learning_rate": 2.0381679389312976e-05, | |
| "loss": 0.5484, | |
| "step": 2332 | |
| }, | |
| { | |
| "epoch": 17.82, | |
| "learning_rate": 2.035623409669211e-05, | |
| "loss": 0.5887, | |
| "step": 2334 | |
| }, | |
| { | |
| "epoch": 17.83, | |
| "learning_rate": 2.033078880407125e-05, | |
| "loss": 0.6744, | |
| "step": 2336 | |
| }, | |
| { | |
| "epoch": 17.85, | |
| "learning_rate": 2.0305343511450384e-05, | |
| "loss": 0.5906, | |
| "step": 2338 | |
| }, | |
| { | |
| "epoch": 17.86, | |
| "learning_rate": 2.027989821882952e-05, | |
| "loss": 0.4837, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 17.88, | |
| "learning_rate": 2.0254452926208653e-05, | |
| "loss": 0.6183, | |
| "step": 2342 | |
| }, | |
| { | |
| "epoch": 17.89, | |
| "learning_rate": 2.0229007633587788e-05, | |
| "loss": 0.575, | |
| "step": 2344 | |
| }, | |
| { | |
| "epoch": 17.91, | |
| "learning_rate": 2.0203562340966923e-05, | |
| "loss": 0.7105, | |
| "step": 2346 | |
| }, | |
| { | |
| "epoch": 17.92, | |
| "learning_rate": 2.0178117048346058e-05, | |
| "loss": 0.6605, | |
| "step": 2348 | |
| }, | |
| { | |
| "epoch": 17.94, | |
| "learning_rate": 2.0152671755725193e-05, | |
| "loss": 0.6811, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 17.95, | |
| "learning_rate": 2.0127226463104327e-05, | |
| "loss": 0.5716, | |
| "step": 2352 | |
| }, | |
| { | |
| "epoch": 17.97, | |
| "learning_rate": 2.0101781170483462e-05, | |
| "loss": 0.6219, | |
| "step": 2354 | |
| }, | |
| { | |
| "epoch": 17.98, | |
| "learning_rate": 2.0076335877862594e-05, | |
| "loss": 0.4857, | |
| "step": 2356 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "learning_rate": 2.005089058524173e-05, | |
| "loss": 0.7918, | |
| "step": 2358 | |
| }, | |
| { | |
| "epoch": 18.02, | |
| "learning_rate": 2.0025445292620867e-05, | |
| "loss": 0.9456, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 18.03, | |
| "learning_rate": 2e-05, | |
| "loss": 0.5068, | |
| "step": 2362 | |
| }, | |
| { | |
| "epoch": 18.05, | |
| "learning_rate": 1.9974554707379136e-05, | |
| "loss": 0.4661, | |
| "step": 2364 | |
| }, | |
| { | |
| "epoch": 18.06, | |
| "learning_rate": 1.994910941475827e-05, | |
| "loss": 0.4722, | |
| "step": 2366 | |
| }, | |
| { | |
| "epoch": 18.08, | |
| "learning_rate": 1.9923664122137406e-05, | |
| "loss": 0.6025, | |
| "step": 2368 | |
| }, | |
| { | |
| "epoch": 18.09, | |
| "learning_rate": 1.989821882951654e-05, | |
| "loss": 0.6247, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 18.11, | |
| "learning_rate": 1.9872773536895676e-05, | |
| "loss": 0.4999, | |
| "step": 2372 | |
| }, | |
| { | |
| "epoch": 18.12, | |
| "learning_rate": 1.984732824427481e-05, | |
| "loss": 0.5055, | |
| "step": 2374 | |
| }, | |
| { | |
| "epoch": 18.14, | |
| "learning_rate": 1.9821882951653945e-05, | |
| "loss": 0.5135, | |
| "step": 2376 | |
| }, | |
| { | |
| "epoch": 18.15, | |
| "learning_rate": 1.979643765903308e-05, | |
| "loss": 0.5206, | |
| "step": 2378 | |
| }, | |
| { | |
| "epoch": 18.17, | |
| "learning_rate": 1.9770992366412215e-05, | |
| "loss": 0.5837, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 18.18, | |
| "learning_rate": 1.974554707379135e-05, | |
| "loss": 0.5924, | |
| "step": 2382 | |
| }, | |
| { | |
| "epoch": 18.2, | |
| "learning_rate": 1.9720101781170485e-05, | |
| "loss": 0.5799, | |
| "step": 2384 | |
| }, | |
| { | |
| "epoch": 18.21, | |
| "learning_rate": 1.969465648854962e-05, | |
| "loss": 0.5315, | |
| "step": 2386 | |
| }, | |
| { | |
| "epoch": 18.23, | |
| "learning_rate": 1.9669211195928754e-05, | |
| "loss": 0.5375, | |
| "step": 2388 | |
| }, | |
| { | |
| "epoch": 18.24, | |
| "learning_rate": 1.964376590330789e-05, | |
| "loss": 0.4988, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 18.26, | |
| "learning_rate": 1.9618320610687024e-05, | |
| "loss": 0.6637, | |
| "step": 2392 | |
| }, | |
| { | |
| "epoch": 18.27, | |
| "learning_rate": 1.959287531806616e-05, | |
| "loss": 0.5626, | |
| "step": 2394 | |
| }, | |
| { | |
| "epoch": 18.29, | |
| "learning_rate": 1.9567430025445294e-05, | |
| "loss": 0.5602, | |
| "step": 2396 | |
| }, | |
| { | |
| "epoch": 18.31, | |
| "learning_rate": 1.954198473282443e-05, | |
| "loss": 0.4954, | |
| "step": 2398 | |
| }, | |
| { | |
| "epoch": 18.32, | |
| "learning_rate": 1.9516539440203563e-05, | |
| "loss": 0.6328, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 18.32, | |
| "eval_cer": 0.7733333333333333, | |
| "eval_loss": 4.303896903991699, | |
| "eval_runtime": 13.5179, | |
| "eval_samples_per_second": 19.308, | |
| "eval_steps_per_second": 2.441, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 18.34, | |
| "learning_rate": 1.9491094147582698e-05, | |
| "loss": 0.4413, | |
| "step": 2402 | |
| }, | |
| { | |
| "epoch": 18.35, | |
| "learning_rate": 1.9465648854961833e-05, | |
| "loss": 0.7464, | |
| "step": 2404 | |
| }, | |
| { | |
| "epoch": 18.37, | |
| "learning_rate": 1.9440203562340968e-05, | |
| "loss": 0.7487, | |
| "step": 2406 | |
| }, | |
| { | |
| "epoch": 18.38, | |
| "learning_rate": 1.9414758269720103e-05, | |
| "loss": 0.9175, | |
| "step": 2408 | |
| }, | |
| { | |
| "epoch": 18.4, | |
| "learning_rate": 1.9389312977099238e-05, | |
| "loss": 0.7402, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 18.41, | |
| "learning_rate": 1.9363867684478372e-05, | |
| "loss": 0.5455, | |
| "step": 2412 | |
| }, | |
| { | |
| "epoch": 18.43, | |
| "learning_rate": 1.9338422391857507e-05, | |
| "loss": 0.5307, | |
| "step": 2414 | |
| }, | |
| { | |
| "epoch": 18.44, | |
| "learning_rate": 1.9312977099236642e-05, | |
| "loss": 0.6569, | |
| "step": 2416 | |
| }, | |
| { | |
| "epoch": 18.46, | |
| "learning_rate": 1.9287531806615777e-05, | |
| "loss": 0.6888, | |
| "step": 2418 | |
| }, | |
| { | |
| "epoch": 18.47, | |
| "learning_rate": 1.926208651399491e-05, | |
| "loss": 0.6493, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 18.49, | |
| "learning_rate": 1.9236641221374046e-05, | |
| "loss": 0.6108, | |
| "step": 2422 | |
| }, | |
| { | |
| "epoch": 18.5, | |
| "learning_rate": 1.921119592875318e-05, | |
| "loss": 0.558, | |
| "step": 2424 | |
| }, | |
| { | |
| "epoch": 18.52, | |
| "learning_rate": 1.9185750636132316e-05, | |
| "loss": 0.5925, | |
| "step": 2426 | |
| }, | |
| { | |
| "epoch": 18.53, | |
| "learning_rate": 1.916030534351145e-05, | |
| "loss": 0.4832, | |
| "step": 2428 | |
| }, | |
| { | |
| "epoch": 18.55, | |
| "learning_rate": 1.9134860050890586e-05, | |
| "loss": 0.4764, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 18.56, | |
| "learning_rate": 1.910941475826972e-05, | |
| "loss": 0.4937, | |
| "step": 2432 | |
| }, | |
| { | |
| "epoch": 18.58, | |
| "learning_rate": 1.9083969465648855e-05, | |
| "loss": 0.5787, | |
| "step": 2434 | |
| }, | |
| { | |
| "epoch": 18.6, | |
| "learning_rate": 1.905852417302799e-05, | |
| "loss": 0.5666, | |
| "step": 2436 | |
| }, | |
| { | |
| "epoch": 18.61, | |
| "learning_rate": 1.9033078880407125e-05, | |
| "loss": 0.5288, | |
| "step": 2438 | |
| }, | |
| { | |
| "epoch": 18.63, | |
| "learning_rate": 1.900763358778626e-05, | |
| "loss": 0.4817, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 18.64, | |
| "learning_rate": 1.8982188295165395e-05, | |
| "loss": 0.6006, | |
| "step": 2442 | |
| }, | |
| { | |
| "epoch": 18.66, | |
| "learning_rate": 1.895674300254453e-05, | |
| "loss": 0.5381, | |
| "step": 2444 | |
| }, | |
| { | |
| "epoch": 18.67, | |
| "learning_rate": 1.8931297709923664e-05, | |
| "loss": 0.4153, | |
| "step": 2446 | |
| }, | |
| { | |
| "epoch": 18.69, | |
| "learning_rate": 1.89058524173028e-05, | |
| "loss": 0.5422, | |
| "step": 2448 | |
| }, | |
| { | |
| "epoch": 18.7, | |
| "learning_rate": 1.8880407124681934e-05, | |
| "loss": 0.7129, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 18.72, | |
| "learning_rate": 1.885496183206107e-05, | |
| "loss": 0.7264, | |
| "step": 2452 | |
| }, | |
| { | |
| "epoch": 18.73, | |
| "learning_rate": 1.8829516539440204e-05, | |
| "loss": 0.507, | |
| "step": 2454 | |
| }, | |
| { | |
| "epoch": 18.75, | |
| "learning_rate": 1.880407124681934e-05, | |
| "loss": 0.483, | |
| "step": 2456 | |
| }, | |
| { | |
| "epoch": 18.76, | |
| "learning_rate": 1.8778625954198477e-05, | |
| "loss": 0.4916, | |
| "step": 2458 | |
| }, | |
| { | |
| "epoch": 18.78, | |
| "learning_rate": 1.875318066157761e-05, | |
| "loss": 0.4537, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 18.79, | |
| "learning_rate": 1.8727735368956746e-05, | |
| "loss": 0.5219, | |
| "step": 2462 | |
| }, | |
| { | |
| "epoch": 18.81, | |
| "learning_rate": 1.8702290076335878e-05, | |
| "loss": 0.4941, | |
| "step": 2464 | |
| }, | |
| { | |
| "epoch": 18.82, | |
| "learning_rate": 1.8676844783715013e-05, | |
| "loss": 0.5794, | |
| "step": 2466 | |
| }, | |
| { | |
| "epoch": 18.84, | |
| "learning_rate": 1.8651399491094148e-05, | |
| "loss": 0.8022, | |
| "step": 2468 | |
| }, | |
| { | |
| "epoch": 18.85, | |
| "learning_rate": 1.8625954198473282e-05, | |
| "loss": 0.5549, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 18.87, | |
| "learning_rate": 1.8600508905852417e-05, | |
| "loss": 0.6745, | |
| "step": 2472 | |
| }, | |
| { | |
| "epoch": 18.89, | |
| "learning_rate": 1.8575063613231552e-05, | |
| "loss": 0.5353, | |
| "step": 2474 | |
| }, | |
| { | |
| "epoch": 18.9, | |
| "learning_rate": 1.8549618320610687e-05, | |
| "loss": 0.5068, | |
| "step": 2476 | |
| }, | |
| { | |
| "epoch": 18.92, | |
| "learning_rate": 1.852417302798982e-05, | |
| "loss": 0.4106, | |
| "step": 2478 | |
| }, | |
| { | |
| "epoch": 18.93, | |
| "learning_rate": 1.8498727735368956e-05, | |
| "loss": 0.4819, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 18.95, | |
| "learning_rate": 1.8473282442748095e-05, | |
| "loss": 0.6774, | |
| "step": 2482 | |
| }, | |
| { | |
| "epoch": 18.96, | |
| "learning_rate": 1.844783715012723e-05, | |
| "loss": 0.5727, | |
| "step": 2484 | |
| }, | |
| { | |
| "epoch": 18.98, | |
| "learning_rate": 1.8422391857506364e-05, | |
| "loss": 1.0129, | |
| "step": 2486 | |
| }, | |
| { | |
| "epoch": 18.99, | |
| "learning_rate": 1.83969465648855e-05, | |
| "loss": 0.7458, | |
| "step": 2488 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 1.8371501272264634e-05, | |
| "loss": 0.7094, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 19.02, | |
| "learning_rate": 1.8346055979643765e-05, | |
| "loss": 0.9128, | |
| "step": 2492 | |
| }, | |
| { | |
| "epoch": 19.04, | |
| "learning_rate": 1.83206106870229e-05, | |
| "loss": 0.6349, | |
| "step": 2494 | |
| }, | |
| { | |
| "epoch": 19.05, | |
| "learning_rate": 1.8295165394402035e-05, | |
| "loss": 0.4806, | |
| "step": 2496 | |
| }, | |
| { | |
| "epoch": 19.07, | |
| "learning_rate": 1.826972010178117e-05, | |
| "loss": 0.5949, | |
| "step": 2498 | |
| }, | |
| { | |
| "epoch": 19.08, | |
| "learning_rate": 1.8244274809160305e-05, | |
| "loss": 0.5064, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 19.08, | |
| "eval_cer": 1.6666666666666667, | |
| "eval_loss": 3.5248308181762695, | |
| "eval_runtime": 15.663, | |
| "eval_samples_per_second": 16.663, | |
| "eval_steps_per_second": 2.107, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 19.1, | |
| "learning_rate": 1.821882951653944e-05, | |
| "loss": 0.5567, | |
| "step": 2502 | |
| }, | |
| { | |
| "epoch": 19.11, | |
| "learning_rate": 1.8193384223918574e-05, | |
| "loss": 0.4422, | |
| "step": 2504 | |
| }, | |
| { | |
| "epoch": 19.13, | |
| "learning_rate": 1.816793893129771e-05, | |
| "loss": 0.6828, | |
| "step": 2506 | |
| }, | |
| { | |
| "epoch": 19.15, | |
| "learning_rate": 1.8142493638676847e-05, | |
| "loss": 0.4808, | |
| "step": 2508 | |
| }, | |
| { | |
| "epoch": 19.16, | |
| "learning_rate": 1.8117048346055982e-05, | |
| "loss": 0.6971, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 19.18, | |
| "learning_rate": 1.8091603053435117e-05, | |
| "loss": 0.7127, | |
| "step": 2512 | |
| }, | |
| { | |
| "epoch": 19.19, | |
| "learning_rate": 1.8066157760814252e-05, | |
| "loss": 0.7504, | |
| "step": 2514 | |
| }, | |
| { | |
| "epoch": 19.21, | |
| "learning_rate": 1.8040712468193387e-05, | |
| "loss": 0.6705, | |
| "step": 2516 | |
| }, | |
| { | |
| "epoch": 19.22, | |
| "learning_rate": 1.801526717557252e-05, | |
| "loss": 0.6327, | |
| "step": 2518 | |
| }, | |
| { | |
| "epoch": 19.24, | |
| "learning_rate": 1.7989821882951653e-05, | |
| "loss": 0.4723, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 19.25, | |
| "learning_rate": 1.7964376590330788e-05, | |
| "loss": 0.6639, | |
| "step": 2522 | |
| }, | |
| { | |
| "epoch": 19.27, | |
| "learning_rate": 1.7938931297709923e-05, | |
| "loss": 0.5713, | |
| "step": 2524 | |
| }, | |
| { | |
| "epoch": 19.28, | |
| "learning_rate": 1.7913486005089058e-05, | |
| "loss": 0.6523, | |
| "step": 2526 | |
| }, | |
| { | |
| "epoch": 19.3, | |
| "learning_rate": 1.7888040712468192e-05, | |
| "loss": 0.436, | |
| "step": 2528 | |
| }, | |
| { | |
| "epoch": 19.31, | |
| "learning_rate": 1.7862595419847327e-05, | |
| "loss": 0.5954, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 19.33, | |
| "learning_rate": 1.7837150127226465e-05, | |
| "loss": 0.5139, | |
| "step": 2532 | |
| }, | |
| { | |
| "epoch": 19.34, | |
| "learning_rate": 1.78117048346056e-05, | |
| "loss": 0.7467, | |
| "step": 2534 | |
| }, | |
| { | |
| "epoch": 19.36, | |
| "learning_rate": 1.7786259541984735e-05, | |
| "loss": 0.515, | |
| "step": 2536 | |
| }, | |
| { | |
| "epoch": 19.37, | |
| "learning_rate": 1.776081424936387e-05, | |
| "loss": 0.4295, | |
| "step": 2538 | |
| }, | |
| { | |
| "epoch": 19.39, | |
| "learning_rate": 1.7735368956743005e-05, | |
| "loss": 0.5099, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 19.4, | |
| "learning_rate": 1.770992366412214e-05, | |
| "loss": 0.5244, | |
| "step": 2542 | |
| }, | |
| { | |
| "epoch": 19.42, | |
| "learning_rate": 1.7684478371501274e-05, | |
| "loss": 0.4796, | |
| "step": 2544 | |
| }, | |
| { | |
| "epoch": 19.44, | |
| "learning_rate": 1.765903307888041e-05, | |
| "loss": 0.5541, | |
| "step": 2546 | |
| }, | |
| { | |
| "epoch": 19.45, | |
| "learning_rate": 1.763358778625954e-05, | |
| "loss": 0.6066, | |
| "step": 2548 | |
| }, | |
| { | |
| "epoch": 19.47, | |
| "learning_rate": 1.7608142493638675e-05, | |
| "loss": 0.5268, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 19.48, | |
| "learning_rate": 1.758269720101781e-05, | |
| "loss": 0.475, | |
| "step": 2552 | |
| }, | |
| { | |
| "epoch": 19.5, | |
| "learning_rate": 1.7557251908396945e-05, | |
| "loss": 0.5441, | |
| "step": 2554 | |
| }, | |
| { | |
| "epoch": 19.51, | |
| "learning_rate": 1.7531806615776083e-05, | |
| "loss": 0.5939, | |
| "step": 2556 | |
| }, | |
| { | |
| "epoch": 19.53, | |
| "learning_rate": 1.7506361323155218e-05, | |
| "loss": 0.4331, | |
| "step": 2558 | |
| }, | |
| { | |
| "epoch": 19.54, | |
| "learning_rate": 1.7480916030534353e-05, | |
| "loss": 0.64, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 19.56, | |
| "learning_rate": 1.7455470737913488e-05, | |
| "loss": 0.6079, | |
| "step": 2562 | |
| }, | |
| { | |
| "epoch": 19.57, | |
| "learning_rate": 1.7430025445292623e-05, | |
| "loss": 0.549, | |
| "step": 2564 | |
| }, | |
| { | |
| "epoch": 19.59, | |
| "learning_rate": 1.7404580152671757e-05, | |
| "loss": 0.5349, | |
| "step": 2566 | |
| }, | |
| { | |
| "epoch": 19.6, | |
| "learning_rate": 1.7379134860050892e-05, | |
| "loss": 0.8238, | |
| "step": 2568 | |
| }, | |
| { | |
| "epoch": 19.62, | |
| "learning_rate": 1.7353689567430027e-05, | |
| "loss": 0.5387, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 19.63, | |
| "learning_rate": 1.7328244274809162e-05, | |
| "loss": 0.3779, | |
| "step": 2572 | |
| }, | |
| { | |
| "epoch": 19.65, | |
| "learning_rate": 1.7302798982188297e-05, | |
| "loss": 0.5164, | |
| "step": 2574 | |
| }, | |
| { | |
| "epoch": 19.66, | |
| "learning_rate": 1.7277353689567428e-05, | |
| "loss": 0.5982, | |
| "step": 2576 | |
| }, | |
| { | |
| "epoch": 19.68, | |
| "learning_rate": 1.7251908396946563e-05, | |
| "loss": 0.5311, | |
| "step": 2578 | |
| }, | |
| { | |
| "epoch": 19.69, | |
| "learning_rate": 1.72264631043257e-05, | |
| "loss": 0.5054, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 19.71, | |
| "learning_rate": 1.7201017811704836e-05, | |
| "loss": 0.5302, | |
| "step": 2582 | |
| }, | |
| { | |
| "epoch": 19.73, | |
| "learning_rate": 1.717557251908397e-05, | |
| "loss": 0.6818, | |
| "step": 2584 | |
| }, | |
| { | |
| "epoch": 19.74, | |
| "learning_rate": 1.7150127226463106e-05, | |
| "loss": 0.471, | |
| "step": 2586 | |
| }, | |
| { | |
| "epoch": 19.76, | |
| "learning_rate": 1.712468193384224e-05, | |
| "loss": 0.5866, | |
| "step": 2588 | |
| }, | |
| { | |
| "epoch": 19.77, | |
| "learning_rate": 1.7099236641221375e-05, | |
| "loss": 0.5497, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 19.79, | |
| "learning_rate": 1.707379134860051e-05, | |
| "loss": 0.6197, | |
| "step": 2592 | |
| }, | |
| { | |
| "epoch": 19.8, | |
| "learning_rate": 1.7048346055979645e-05, | |
| "loss": 0.5797, | |
| "step": 2594 | |
| }, | |
| { | |
| "epoch": 19.82, | |
| "learning_rate": 1.702290076335878e-05, | |
| "loss": 0.4112, | |
| "step": 2596 | |
| }, | |
| { | |
| "epoch": 19.83, | |
| "learning_rate": 1.6997455470737915e-05, | |
| "loss": 0.6759, | |
| "step": 2598 | |
| }, | |
| { | |
| "epoch": 19.85, | |
| "learning_rate": 1.697201017811705e-05, | |
| "loss": 0.4158, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 19.85, | |
| "eval_cer": 0.43833333333333335, | |
| "eval_loss": 3.8013916015625, | |
| "eval_runtime": 12.6815, | |
| "eval_samples_per_second": 20.581, | |
| "eval_steps_per_second": 2.602, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 19.86, | |
| "learning_rate": 1.6946564885496184e-05, | |
| "loss": 0.5147, | |
| "step": 2602 | |
| }, | |
| { | |
| "epoch": 19.88, | |
| "learning_rate": 1.692111959287532e-05, | |
| "loss": 0.443, | |
| "step": 2604 | |
| }, | |
| { | |
| "epoch": 19.89, | |
| "learning_rate": 1.6895674300254454e-05, | |
| "loss": 0.6891, | |
| "step": 2606 | |
| }, | |
| { | |
| "epoch": 19.91, | |
| "learning_rate": 1.687022900763359e-05, | |
| "loss": 0.6668, | |
| "step": 2608 | |
| }, | |
| { | |
| "epoch": 19.92, | |
| "learning_rate": 1.6844783715012724e-05, | |
| "loss": 0.3972, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 19.94, | |
| "learning_rate": 1.681933842239186e-05, | |
| "loss": 0.5349, | |
| "step": 2612 | |
| }, | |
| { | |
| "epoch": 19.95, | |
| "learning_rate": 1.6793893129770993e-05, | |
| "loss": 0.5152, | |
| "step": 2614 | |
| }, | |
| { | |
| "epoch": 19.97, | |
| "learning_rate": 1.6768447837150128e-05, | |
| "loss": 0.4808, | |
| "step": 2616 | |
| }, | |
| { | |
| "epoch": 19.98, | |
| "learning_rate": 1.6743002544529263e-05, | |
| "loss": 0.4996, | |
| "step": 2618 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 1.6717557251908398e-05, | |
| "loss": 0.5772, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 20.02, | |
| "learning_rate": 1.6692111959287533e-05, | |
| "loss": 0.6665, | |
| "step": 2622 | |
| }, | |
| { | |
| "epoch": 20.03, | |
| "learning_rate": 1.6666666666666667e-05, | |
| "loss": 0.5312, | |
| "step": 2624 | |
| }, | |
| { | |
| "epoch": 20.05, | |
| "learning_rate": 1.6641221374045802e-05, | |
| "loss": 0.497, | |
| "step": 2626 | |
| }, | |
| { | |
| "epoch": 20.06, | |
| "learning_rate": 1.6615776081424937e-05, | |
| "loss": 0.6712, | |
| "step": 2628 | |
| }, | |
| { | |
| "epoch": 20.08, | |
| "learning_rate": 1.6590330788804072e-05, | |
| "loss": 0.4766, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 20.09, | |
| "learning_rate": 1.6564885496183207e-05, | |
| "loss": 0.5655, | |
| "step": 2632 | |
| }, | |
| { | |
| "epoch": 20.11, | |
| "learning_rate": 1.653944020356234e-05, | |
| "loss": 0.5094, | |
| "step": 2634 | |
| }, | |
| { | |
| "epoch": 20.12, | |
| "learning_rate": 1.6513994910941476e-05, | |
| "loss": 0.5919, | |
| "step": 2636 | |
| }, | |
| { | |
| "epoch": 20.14, | |
| "learning_rate": 1.648854961832061e-05, | |
| "loss": 0.5843, | |
| "step": 2638 | |
| }, | |
| { | |
| "epoch": 20.15, | |
| "learning_rate": 1.6463104325699746e-05, | |
| "loss": 0.4332, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 20.17, | |
| "learning_rate": 1.643765903307888e-05, | |
| "loss": 0.5958, | |
| "step": 2642 | |
| }, | |
| { | |
| "epoch": 20.18, | |
| "learning_rate": 1.6412213740458016e-05, | |
| "loss": 0.5763, | |
| "step": 2644 | |
| }, | |
| { | |
| "epoch": 20.2, | |
| "learning_rate": 1.638676844783715e-05, | |
| "loss": 0.5673, | |
| "step": 2646 | |
| }, | |
| { | |
| "epoch": 20.21, | |
| "learning_rate": 1.6361323155216285e-05, | |
| "loss": 0.4903, | |
| "step": 2648 | |
| }, | |
| { | |
| "epoch": 20.23, | |
| "learning_rate": 1.633587786259542e-05, | |
| "loss": 0.414, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 20.24, | |
| "learning_rate": 1.6310432569974555e-05, | |
| "loss": 0.5613, | |
| "step": 2652 | |
| }, | |
| { | |
| "epoch": 20.26, | |
| "learning_rate": 1.628498727735369e-05, | |
| "loss": 0.5852, | |
| "step": 2654 | |
| }, | |
| { | |
| "epoch": 20.27, | |
| "learning_rate": 1.6259541984732825e-05, | |
| "loss": 0.5332, | |
| "step": 2656 | |
| }, | |
| { | |
| "epoch": 20.29, | |
| "learning_rate": 1.623409669211196e-05, | |
| "loss": 0.6841, | |
| "step": 2658 | |
| }, | |
| { | |
| "epoch": 20.31, | |
| "learning_rate": 1.6208651399491094e-05, | |
| "loss": 0.4666, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 20.32, | |
| "learning_rate": 1.618320610687023e-05, | |
| "loss": 0.5628, | |
| "step": 2662 | |
| }, | |
| { | |
| "epoch": 20.34, | |
| "learning_rate": 1.6157760814249364e-05, | |
| "loss": 0.5494, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 20.35, | |
| "learning_rate": 1.61323155216285e-05, | |
| "loss": 0.6103, | |
| "step": 2666 | |
| }, | |
| { | |
| "epoch": 20.37, | |
| "learning_rate": 1.6106870229007634e-05, | |
| "loss": 0.4859, | |
| "step": 2668 | |
| }, | |
| { | |
| "epoch": 20.38, | |
| "learning_rate": 1.608142493638677e-05, | |
| "loss": 0.5271, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 20.4, | |
| "learning_rate": 1.6055979643765903e-05, | |
| "loss": 0.4128, | |
| "step": 2672 | |
| }, | |
| { | |
| "epoch": 20.41, | |
| "learning_rate": 1.6030534351145038e-05, | |
| "loss": 0.5364, | |
| "step": 2674 | |
| }, | |
| { | |
| "epoch": 20.43, | |
| "learning_rate": 1.6005089058524173e-05, | |
| "loss": 0.4226, | |
| "step": 2676 | |
| }, | |
| { | |
| "epoch": 20.44, | |
| "learning_rate": 1.597964376590331e-05, | |
| "loss": 0.629, | |
| "step": 2678 | |
| }, | |
| { | |
| "epoch": 20.46, | |
| "learning_rate": 1.5954198473282446e-05, | |
| "loss": 0.6607, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 20.47, | |
| "learning_rate": 1.592875318066158e-05, | |
| "loss": 0.554, | |
| "step": 2682 | |
| }, | |
| { | |
| "epoch": 20.49, | |
| "learning_rate": 1.5903307888040712e-05, | |
| "loss": 0.6167, | |
| "step": 2684 | |
| }, | |
| { | |
| "epoch": 20.5, | |
| "learning_rate": 1.5877862595419847e-05, | |
| "loss": 0.4845, | |
| "step": 2686 | |
| }, | |
| { | |
| "epoch": 20.52, | |
| "learning_rate": 1.5852417302798982e-05, | |
| "loss": 0.5157, | |
| "step": 2688 | |
| }, | |
| { | |
| "epoch": 20.53, | |
| "learning_rate": 1.5826972010178117e-05, | |
| "loss": 0.4487, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 20.55, | |
| "learning_rate": 1.580152671755725e-05, | |
| "loss": 0.6119, | |
| "step": 2692 | |
| }, | |
| { | |
| "epoch": 20.56, | |
| "learning_rate": 1.5776081424936386e-05, | |
| "loss": 0.4845, | |
| "step": 2694 | |
| }, | |
| { | |
| "epoch": 20.58, | |
| "learning_rate": 1.575063613231552e-05, | |
| "loss": 0.6076, | |
| "step": 2696 | |
| }, | |
| { | |
| "epoch": 20.6, | |
| "learning_rate": 1.5725190839694656e-05, | |
| "loss": 0.4932, | |
| "step": 2698 | |
| }, | |
| { | |
| "epoch": 20.61, | |
| "learning_rate": 1.569974554707379e-05, | |
| "loss": 0.5817, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 20.61, | |
| "eval_cer": 1.2683333333333333, | |
| "eval_loss": 3.7289490699768066, | |
| "eval_runtime": 14.8824, | |
| "eval_samples_per_second": 17.538, | |
| "eval_steps_per_second": 2.217, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 20.63, | |
| "learning_rate": 1.567430025445293e-05, | |
| "loss": 0.5956, | |
| "step": 2702 | |
| }, | |
| { | |
| "epoch": 20.64, | |
| "learning_rate": 1.5648854961832064e-05, | |
| "loss": 0.7098, | |
| "step": 2704 | |
| }, | |
| { | |
| "epoch": 20.66, | |
| "learning_rate": 1.56234096692112e-05, | |
| "loss": 0.5259, | |
| "step": 2706 | |
| }, | |
| { | |
| "epoch": 20.67, | |
| "learning_rate": 1.5597964376590334e-05, | |
| "loss": 0.5975, | |
| "step": 2708 | |
| }, | |
| { | |
| "epoch": 20.69, | |
| "learning_rate": 1.557251908396947e-05, | |
| "loss": 0.4992, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 20.7, | |
| "learning_rate": 1.55470737913486e-05, | |
| "loss": 0.5587, | |
| "step": 2712 | |
| }, | |
| { | |
| "epoch": 20.72, | |
| "learning_rate": 1.5521628498727735e-05, | |
| "loss": 0.5982, | |
| "step": 2714 | |
| }, | |
| { | |
| "epoch": 20.73, | |
| "learning_rate": 1.549618320610687e-05, | |
| "loss": 0.4909, | |
| "step": 2716 | |
| }, | |
| { | |
| "epoch": 20.75, | |
| "learning_rate": 1.5470737913486004e-05, | |
| "loss": 0.4484, | |
| "step": 2718 | |
| }, | |
| { | |
| "epoch": 20.76, | |
| "learning_rate": 1.544529262086514e-05, | |
| "loss": 0.4779, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 20.78, | |
| "learning_rate": 1.5419847328244274e-05, | |
| "loss": 0.6331, | |
| "step": 2722 | |
| }, | |
| { | |
| "epoch": 20.79, | |
| "learning_rate": 1.539440203562341e-05, | |
| "loss": 0.4751, | |
| "step": 2724 | |
| }, | |
| { | |
| "epoch": 20.81, | |
| "learning_rate": 1.5368956743002547e-05, | |
| "loss": 0.4198, | |
| "step": 2726 | |
| }, | |
| { | |
| "epoch": 20.82, | |
| "learning_rate": 1.5343511450381682e-05, | |
| "loss": 0.6411, | |
| "step": 2728 | |
| }, | |
| { | |
| "epoch": 20.84, | |
| "learning_rate": 1.5318066157760817e-05, | |
| "loss": 0.4794, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 20.85, | |
| "learning_rate": 1.529262086513995e-05, | |
| "loss": 0.8884, | |
| "step": 2732 | |
| }, | |
| { | |
| "epoch": 20.87, | |
| "learning_rate": 1.5267175572519086e-05, | |
| "loss": 0.6413, | |
| "step": 2734 | |
| }, | |
| { | |
| "epoch": 20.89, | |
| "learning_rate": 1.524173027989822e-05, | |
| "loss": 0.5209, | |
| "step": 2736 | |
| }, | |
| { | |
| "epoch": 20.9, | |
| "learning_rate": 1.5216284987277354e-05, | |
| "loss": 0.5557, | |
| "step": 2738 | |
| }, | |
| { | |
| "epoch": 20.92, | |
| "learning_rate": 1.519083969465649e-05, | |
| "loss": 0.5334, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 20.93, | |
| "learning_rate": 1.5165394402035624e-05, | |
| "loss": 0.5286, | |
| "step": 2742 | |
| }, | |
| { | |
| "epoch": 20.95, | |
| "learning_rate": 1.5139949109414759e-05, | |
| "loss": 0.4974, | |
| "step": 2744 | |
| }, | |
| { | |
| "epoch": 20.96, | |
| "learning_rate": 1.5114503816793894e-05, | |
| "loss": 0.4457, | |
| "step": 2746 | |
| }, | |
| { | |
| "epoch": 20.98, | |
| "learning_rate": 1.5089058524173027e-05, | |
| "loss": 0.6571, | |
| "step": 2748 | |
| }, | |
| { | |
| "epoch": 20.99, | |
| "learning_rate": 1.5063613231552162e-05, | |
| "loss": 0.4064, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 1.50381679389313e-05, | |
| "loss": 0.6135, | |
| "step": 2752 | |
| }, | |
| { | |
| "epoch": 21.02, | |
| "learning_rate": 1.5012722646310435e-05, | |
| "loss": 0.4958, | |
| "step": 2754 | |
| }, | |
| { | |
| "epoch": 21.04, | |
| "learning_rate": 1.498727735368957e-05, | |
| "loss": 0.449, | |
| "step": 2756 | |
| }, | |
| { | |
| "epoch": 21.05, | |
| "learning_rate": 1.4961832061068704e-05, | |
| "loss": 0.4873, | |
| "step": 2758 | |
| }, | |
| { | |
| "epoch": 21.07, | |
| "learning_rate": 1.4936386768447837e-05, | |
| "loss": 0.5427, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 21.08, | |
| "learning_rate": 1.4910941475826972e-05, | |
| "loss": 0.5974, | |
| "step": 2762 | |
| }, | |
| { | |
| "epoch": 21.1, | |
| "learning_rate": 1.4885496183206107e-05, | |
| "loss": 0.5277, | |
| "step": 2764 | |
| }, | |
| { | |
| "epoch": 21.11, | |
| "learning_rate": 1.4860050890585242e-05, | |
| "loss": 0.4605, | |
| "step": 2766 | |
| }, | |
| { | |
| "epoch": 21.13, | |
| "learning_rate": 1.4834605597964377e-05, | |
| "loss": 0.489, | |
| "step": 2768 | |
| }, | |
| { | |
| "epoch": 21.15, | |
| "learning_rate": 1.4809160305343512e-05, | |
| "loss": 0.3826, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 21.16, | |
| "learning_rate": 1.4783715012722646e-05, | |
| "loss": 0.7243, | |
| "step": 2772 | |
| }, | |
| { | |
| "epoch": 21.18, | |
| "learning_rate": 1.4758269720101781e-05, | |
| "loss": 0.4984, | |
| "step": 2774 | |
| }, | |
| { | |
| "epoch": 21.19, | |
| "learning_rate": 1.4732824427480918e-05, | |
| "loss": 0.4792, | |
| "step": 2776 | |
| }, | |
| { | |
| "epoch": 21.21, | |
| "learning_rate": 1.4707379134860053e-05, | |
| "loss": 0.5277, | |
| "step": 2778 | |
| }, | |
| { | |
| "epoch": 21.22, | |
| "learning_rate": 1.4681933842239187e-05, | |
| "loss": 0.5729, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 21.24, | |
| "learning_rate": 1.4656488549618322e-05, | |
| "loss": 0.5802, | |
| "step": 2782 | |
| }, | |
| { | |
| "epoch": 21.25, | |
| "learning_rate": 1.4631043256997457e-05, | |
| "loss": 0.5329, | |
| "step": 2784 | |
| }, | |
| { | |
| "epoch": 21.27, | |
| "learning_rate": 1.4605597964376592e-05, | |
| "loss": 0.4111, | |
| "step": 2786 | |
| }, | |
| { | |
| "epoch": 21.28, | |
| "learning_rate": 1.4580152671755725e-05, | |
| "loss": 0.5507, | |
| "step": 2788 | |
| }, | |
| { | |
| "epoch": 21.3, | |
| "learning_rate": 1.455470737913486e-05, | |
| "loss": 0.5248, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 21.31, | |
| "learning_rate": 1.4529262086513995e-05, | |
| "loss": 0.4178, | |
| "step": 2792 | |
| }, | |
| { | |
| "epoch": 21.33, | |
| "learning_rate": 1.450381679389313e-05, | |
| "loss": 0.512, | |
| "step": 2794 | |
| }, | |
| { | |
| "epoch": 21.34, | |
| "learning_rate": 1.4478371501272264e-05, | |
| "loss": 0.4594, | |
| "step": 2796 | |
| }, | |
| { | |
| "epoch": 21.36, | |
| "learning_rate": 1.44529262086514e-05, | |
| "loss": 0.4647, | |
| "step": 2798 | |
| }, | |
| { | |
| "epoch": 21.37, | |
| "learning_rate": 1.4427480916030536e-05, | |
| "loss": 0.5574, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 21.37, | |
| "eval_cer": 0.45666666666666667, | |
| "eval_loss": 6.2158331871032715, | |
| "eval_runtime": 13.5284, | |
| "eval_samples_per_second": 19.293, | |
| "eval_steps_per_second": 2.439, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 21.39, | |
| "learning_rate": 1.440203562340967e-05, | |
| "loss": 0.5497, | |
| "step": 2802 | |
| }, | |
| { | |
| "epoch": 21.4, | |
| "learning_rate": 1.4376590330788805e-05, | |
| "loss": 0.4333, | |
| "step": 2804 | |
| }, | |
| { | |
| "epoch": 21.42, | |
| "learning_rate": 1.435114503816794e-05, | |
| "loss": 0.6657, | |
| "step": 2806 | |
| }, | |
| { | |
| "epoch": 21.44, | |
| "learning_rate": 1.4325699745547075e-05, | |
| "loss": 0.5721, | |
| "step": 2808 | |
| }, | |
| { | |
| "epoch": 21.45, | |
| "learning_rate": 1.430025445292621e-05, | |
| "loss": 0.6449, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 21.47, | |
| "learning_rate": 1.4274809160305345e-05, | |
| "loss": 0.4825, | |
| "step": 2812 | |
| }, | |
| { | |
| "epoch": 21.48, | |
| "learning_rate": 1.424936386768448e-05, | |
| "loss": 0.5418, | |
| "step": 2814 | |
| }, | |
| { | |
| "epoch": 21.5, | |
| "learning_rate": 1.4223918575063613e-05, | |
| "loss": 0.511, | |
| "step": 2816 | |
| }, | |
| { | |
| "epoch": 21.51, | |
| "learning_rate": 1.4198473282442747e-05, | |
| "loss": 0.3933, | |
| "step": 2818 | |
| }, | |
| { | |
| "epoch": 21.53, | |
| "learning_rate": 1.4173027989821882e-05, | |
| "loss": 0.4775, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 21.54, | |
| "learning_rate": 1.4147582697201017e-05, | |
| "loss": 0.5571, | |
| "step": 2822 | |
| }, | |
| { | |
| "epoch": 21.56, | |
| "learning_rate": 1.4122137404580155e-05, | |
| "loss": 0.6556, | |
| "step": 2824 | |
| }, | |
| { | |
| "epoch": 21.57, | |
| "learning_rate": 1.4096692111959288e-05, | |
| "loss": 0.5842, | |
| "step": 2826 | |
| }, | |
| { | |
| "epoch": 21.59, | |
| "learning_rate": 1.4071246819338423e-05, | |
| "loss": 0.5444, | |
| "step": 2828 | |
| }, | |
| { | |
| "epoch": 21.6, | |
| "learning_rate": 1.4045801526717558e-05, | |
| "loss": 0.5351, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 21.62, | |
| "learning_rate": 1.4020356234096693e-05, | |
| "loss": 0.5468, | |
| "step": 2832 | |
| }, | |
| { | |
| "epoch": 21.63, | |
| "learning_rate": 1.3994910941475828e-05, | |
| "loss": 0.6085, | |
| "step": 2834 | |
| }, | |
| { | |
| "epoch": 21.65, | |
| "learning_rate": 1.3969465648854963e-05, | |
| "loss": 0.4829, | |
| "step": 2836 | |
| }, | |
| { | |
| "epoch": 21.66, | |
| "learning_rate": 1.3944020356234097e-05, | |
| "loss": 0.4787, | |
| "step": 2838 | |
| }, | |
| { | |
| "epoch": 21.68, | |
| "learning_rate": 1.3918575063613232e-05, | |
| "loss": 0.4858, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 21.69, | |
| "learning_rate": 1.3893129770992367e-05, | |
| "loss": 0.4637, | |
| "step": 2842 | |
| }, | |
| { | |
| "epoch": 21.71, | |
| "learning_rate": 1.38676844783715e-05, | |
| "loss": 0.6044, | |
| "step": 2844 | |
| }, | |
| { | |
| "epoch": 21.73, | |
| "learning_rate": 1.3842239185750635e-05, | |
| "loss": 0.478, | |
| "step": 2846 | |
| }, | |
| { | |
| "epoch": 21.74, | |
| "learning_rate": 1.3816793893129773e-05, | |
| "loss": 0.6719, | |
| "step": 2848 | |
| }, | |
| { | |
| "epoch": 21.76, | |
| "learning_rate": 1.3791348600508908e-05, | |
| "loss": 0.4391, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 21.77, | |
| "learning_rate": 1.3765903307888043e-05, | |
| "loss": 0.5353, | |
| "step": 2852 | |
| }, | |
| { | |
| "epoch": 21.79, | |
| "learning_rate": 1.3740458015267178e-05, | |
| "loss": 0.371, | |
| "step": 2854 | |
| }, | |
| { | |
| "epoch": 21.8, | |
| "learning_rate": 1.3715012722646311e-05, | |
| "loss": 0.4898, | |
| "step": 2856 | |
| }, | |
| { | |
| "epoch": 21.82, | |
| "learning_rate": 1.3689567430025446e-05, | |
| "loss": 0.5191, | |
| "step": 2858 | |
| }, | |
| { | |
| "epoch": 21.83, | |
| "learning_rate": 1.366412213740458e-05, | |
| "loss": 0.4718, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 21.85, | |
| "learning_rate": 1.3638676844783715e-05, | |
| "loss": 0.4079, | |
| "step": 2862 | |
| }, | |
| { | |
| "epoch": 21.86, | |
| "learning_rate": 1.361323155216285e-05, | |
| "loss": 0.8327, | |
| "step": 2864 | |
| }, | |
| { | |
| "epoch": 21.88, | |
| "learning_rate": 1.3587786259541985e-05, | |
| "loss": 0.5435, | |
| "step": 2866 | |
| }, | |
| { | |
| "epoch": 21.89, | |
| "learning_rate": 1.356234096692112e-05, | |
| "loss": 0.4663, | |
| "step": 2868 | |
| }, | |
| { | |
| "epoch": 21.91, | |
| "learning_rate": 1.3536895674300255e-05, | |
| "loss": 0.468, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 21.92, | |
| "learning_rate": 1.3511450381679388e-05, | |
| "loss": 0.4782, | |
| "step": 2872 | |
| }, | |
| { | |
| "epoch": 21.94, | |
| "learning_rate": 1.3486005089058526e-05, | |
| "loss": 0.5288, | |
| "step": 2874 | |
| }, | |
| { | |
| "epoch": 21.95, | |
| "learning_rate": 1.3460559796437661e-05, | |
| "loss": 0.5454, | |
| "step": 2876 | |
| }, | |
| { | |
| "epoch": 21.97, | |
| "learning_rate": 1.3435114503816796e-05, | |
| "loss": 0.4063, | |
| "step": 2878 | |
| }, | |
| { | |
| "epoch": 21.98, | |
| "learning_rate": 1.340966921119593e-05, | |
| "loss": 0.5518, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "learning_rate": 1.3384223918575065e-05, | |
| "loss": 0.615, | |
| "step": 2882 | |
| }, | |
| { | |
| "epoch": 22.02, | |
| "learning_rate": 1.3358778625954198e-05, | |
| "loss": 0.4139, | |
| "step": 2884 | |
| }, | |
| { | |
| "epoch": 22.03, | |
| "learning_rate": 1.3333333333333333e-05, | |
| "loss": 0.4635, | |
| "step": 2886 | |
| }, | |
| { | |
| "epoch": 22.05, | |
| "learning_rate": 1.3307888040712468e-05, | |
| "loss": 0.5414, | |
| "step": 2888 | |
| }, | |
| { | |
| "epoch": 22.06, | |
| "learning_rate": 1.3282442748091603e-05, | |
| "loss": 0.4512, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 22.08, | |
| "learning_rate": 1.3256997455470738e-05, | |
| "loss": 0.4532, | |
| "step": 2892 | |
| }, | |
| { | |
| "epoch": 22.09, | |
| "learning_rate": 1.3231552162849873e-05, | |
| "loss": 0.612, | |
| "step": 2894 | |
| }, | |
| { | |
| "epoch": 22.11, | |
| "learning_rate": 1.3206106870229007e-05, | |
| "loss": 0.4963, | |
| "step": 2896 | |
| }, | |
| { | |
| "epoch": 22.12, | |
| "learning_rate": 1.3180661577608144e-05, | |
| "loss": 0.7695, | |
| "step": 2898 | |
| }, | |
| { | |
| "epoch": 22.14, | |
| "learning_rate": 1.3155216284987279e-05, | |
| "loss": 0.3769, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 22.14, | |
| "eval_cer": 1.0666666666666667, | |
| "eval_loss": 3.7434914112091064, | |
| "eval_runtime": 14.3075, | |
| "eval_samples_per_second": 18.242, | |
| "eval_steps_per_second": 2.306, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 22.15, | |
| "learning_rate": 1.3129770992366414e-05, | |
| "loss": 0.5136, | |
| "step": 2902 | |
| }, | |
| { | |
| "epoch": 22.17, | |
| "learning_rate": 1.3104325699745548e-05, | |
| "loss": 0.4998, | |
| "step": 2904 | |
| }, | |
| { | |
| "epoch": 22.18, | |
| "learning_rate": 1.3078880407124683e-05, | |
| "loss": 0.463, | |
| "step": 2906 | |
| }, | |
| { | |
| "epoch": 22.2, | |
| "learning_rate": 1.3053435114503818e-05, | |
| "loss": 0.7542, | |
| "step": 2908 | |
| }, | |
| { | |
| "epoch": 22.21, | |
| "learning_rate": 1.3027989821882953e-05, | |
| "loss": 0.5518, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 22.23, | |
| "learning_rate": 1.3002544529262086e-05, | |
| "loss": 0.4916, | |
| "step": 2912 | |
| }, | |
| { | |
| "epoch": 22.24, | |
| "learning_rate": 1.2977099236641221e-05, | |
| "loss": 0.5718, | |
| "step": 2914 | |
| }, | |
| { | |
| "epoch": 22.26, | |
| "learning_rate": 1.2951653944020356e-05, | |
| "loss": 0.4251, | |
| "step": 2916 | |
| }, | |
| { | |
| "epoch": 22.27, | |
| "learning_rate": 1.292620865139949e-05, | |
| "loss": 0.5434, | |
| "step": 2918 | |
| }, | |
| { | |
| "epoch": 22.29, | |
| "learning_rate": 1.2900763358778625e-05, | |
| "loss": 0.553, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 22.31, | |
| "learning_rate": 1.2875318066157762e-05, | |
| "loss": 0.5578, | |
| "step": 2922 | |
| }, | |
| { | |
| "epoch": 22.32, | |
| "learning_rate": 1.2849872773536897e-05, | |
| "loss": 0.481, | |
| "step": 2924 | |
| }, | |
| { | |
| "epoch": 22.34, | |
| "learning_rate": 1.2824427480916032e-05, | |
| "loss": 0.4623, | |
| "step": 2926 | |
| }, | |
| { | |
| "epoch": 22.35, | |
| "learning_rate": 1.2798982188295166e-05, | |
| "loss": 0.4564, | |
| "step": 2928 | |
| }, | |
| { | |
| "epoch": 22.37, | |
| "learning_rate": 1.2773536895674301e-05, | |
| "loss": 0.4327, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 22.38, | |
| "learning_rate": 1.2748091603053436e-05, | |
| "loss": 0.4521, | |
| "step": 2932 | |
| }, | |
| { | |
| "epoch": 22.4, | |
| "learning_rate": 1.2722646310432571e-05, | |
| "loss": 0.4404, | |
| "step": 2934 | |
| }, | |
| { | |
| "epoch": 22.41, | |
| "learning_rate": 1.2697201017811706e-05, | |
| "loss": 0.5332, | |
| "step": 2936 | |
| }, | |
| { | |
| "epoch": 22.43, | |
| "learning_rate": 1.267175572519084e-05, | |
| "loss": 0.6079, | |
| "step": 2938 | |
| }, | |
| { | |
| "epoch": 22.44, | |
| "learning_rate": 1.2646310432569974e-05, | |
| "loss": 0.4472, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 22.46, | |
| "learning_rate": 1.2620865139949108e-05, | |
| "loss": 0.5221, | |
| "step": 2942 | |
| }, | |
| { | |
| "epoch": 22.47, | |
| "learning_rate": 1.2595419847328243e-05, | |
| "loss": 0.43, | |
| "step": 2944 | |
| }, | |
| { | |
| "epoch": 22.49, | |
| "learning_rate": 1.2569974554707382e-05, | |
| "loss": 0.4141, | |
| "step": 2946 | |
| }, | |
| { | |
| "epoch": 22.5, | |
| "learning_rate": 1.2544529262086516e-05, | |
| "loss": 0.4824, | |
| "step": 2948 | |
| }, | |
| { | |
| "epoch": 22.52, | |
| "learning_rate": 1.251908396946565e-05, | |
| "loss": 0.5227, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 22.53, | |
| "learning_rate": 1.2493638676844784e-05, | |
| "loss": 0.4538, | |
| "step": 2952 | |
| }, | |
| { | |
| "epoch": 22.55, | |
| "learning_rate": 1.2468193384223919e-05, | |
| "loss": 0.4206, | |
| "step": 2954 | |
| }, | |
| { | |
| "epoch": 22.56, | |
| "learning_rate": 1.2442748091603054e-05, | |
| "loss": 0.4709, | |
| "step": 2956 | |
| }, | |
| { | |
| "epoch": 22.58, | |
| "learning_rate": 1.2417302798982189e-05, | |
| "loss": 0.5862, | |
| "step": 2958 | |
| }, | |
| { | |
| "epoch": 22.6, | |
| "learning_rate": 1.2391857506361324e-05, | |
| "loss": 0.6123, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 22.61, | |
| "learning_rate": 1.2366412213740458e-05, | |
| "loss": 0.6074, | |
| "step": 2962 | |
| }, | |
| { | |
| "epoch": 22.63, | |
| "learning_rate": 1.2340966921119595e-05, | |
| "loss": 0.5028, | |
| "step": 2964 | |
| }, | |
| { | |
| "epoch": 22.64, | |
| "learning_rate": 1.2315521628498728e-05, | |
| "loss": 0.5615, | |
| "step": 2966 | |
| }, | |
| { | |
| "epoch": 22.66, | |
| "learning_rate": 1.2290076335877863e-05, | |
| "loss": 0.3996, | |
| "step": 2968 | |
| }, | |
| { | |
| "epoch": 22.67, | |
| "learning_rate": 1.2264631043256998e-05, | |
| "loss": 0.6097, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 22.69, | |
| "learning_rate": 1.2239185750636133e-05, | |
| "loss": 0.4233, | |
| "step": 2972 | |
| }, | |
| { | |
| "epoch": 22.7, | |
| "learning_rate": 1.2213740458015267e-05, | |
| "loss": 0.5531, | |
| "step": 2974 | |
| }, | |
| { | |
| "epoch": 22.72, | |
| "learning_rate": 1.2188295165394404e-05, | |
| "loss": 0.5276, | |
| "step": 2976 | |
| }, | |
| { | |
| "epoch": 22.73, | |
| "learning_rate": 1.2162849872773539e-05, | |
| "loss": 0.5667, | |
| "step": 2978 | |
| }, | |
| { | |
| "epoch": 22.75, | |
| "learning_rate": 1.2137404580152672e-05, | |
| "loss": 0.421, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 22.76, | |
| "learning_rate": 1.2111959287531807e-05, | |
| "loss": 0.3711, | |
| "step": 2982 | |
| }, | |
| { | |
| "epoch": 22.78, | |
| "learning_rate": 1.2086513994910942e-05, | |
| "loss": 0.6067, | |
| "step": 2984 | |
| }, | |
| { | |
| "epoch": 22.79, | |
| "learning_rate": 1.2061068702290076e-05, | |
| "loss": 0.3717, | |
| "step": 2986 | |
| }, | |
| { | |
| "epoch": 22.81, | |
| "learning_rate": 1.2035623409669211e-05, | |
| "loss": 0.6554, | |
| "step": 2988 | |
| }, | |
| { | |
| "epoch": 22.82, | |
| "learning_rate": 1.2010178117048348e-05, | |
| "loss": 0.6248, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 22.84, | |
| "learning_rate": 1.1984732824427483e-05, | |
| "loss": 0.4462, | |
| "step": 2992 | |
| }, | |
| { | |
| "epoch": 22.85, | |
| "learning_rate": 1.1959287531806616e-05, | |
| "loss": 0.5547, | |
| "step": 2994 | |
| }, | |
| { | |
| "epoch": 22.87, | |
| "learning_rate": 1.193384223918575e-05, | |
| "loss": 0.3156, | |
| "step": 2996 | |
| }, | |
| { | |
| "epoch": 22.89, | |
| "learning_rate": 1.1908396946564885e-05, | |
| "loss": 0.413, | |
| "step": 2998 | |
| }, | |
| { | |
| "epoch": 22.9, | |
| "learning_rate": 1.188295165394402e-05, | |
| "loss": 0.5049, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 22.9, | |
| "eval_cer": 0.655, | |
| "eval_loss": 4.91603422164917, | |
| "eval_runtime": 13.887, | |
| "eval_samples_per_second": 18.794, | |
| "eval_steps_per_second": 2.376, | |
| "step": 3000 | |
| } | |
| ], | |
| "logging_steps": 2, | |
| "max_steps": 3930, | |
| "num_train_epochs": 30, | |
| "save_steps": 1000, | |
| "total_flos": 1.7876533212893676e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |