karths/binary_classification_train_code
031
1{2 "best_metric": null,3 "best_model_checkpoint": null,4 "epoch": 4.999030067895247,5 "eval_steps": 500,6 "global_step": 12885,7 "is_hyper_param_search": false,8 "is_local_process_zero": true,9 "is_world_process_zero": true,10 "log_history": [11 {12 "epoch": 0.19,13 "learning_rate": 4.99e-05,14 "loss": 0.0599,15 "step": 50016 },17 {18 "epoch": 0.39,19 "learning_rate": 4.97999958968678e-05,20 "loss": 0.0763,21 "step": 100022 },23 {24 "epoch": 0.58,25 "learning_rate": 4.920159465099463e-05,26 "loss": 0.0765,27 "step": 150028 },29 {30 "epoch": 0.78,31 "learning_rate": 4.8214408436601485e-05,32 "loss": 0.0868,33 "step": 200034 },35 {36 "epoch": 0.97,37 "learning_rate": 4.685737469208415e-05,38 "loss": 0.0843,39 "step": 250040 },41 {42 "epoch": 1.16,43 "learning_rate": 4.514686169369878e-05,44 "loss": 0.0716,45 "step": 300046 },47 {48 "epoch": 1.36,49 "learning_rate": 4.311270072490833e-05,50 "loss": 0.0691,51 "step": 350052 },53 {54 "epoch": 1.55,55 "learning_rate": 4.0787569434422325e-05,56 "loss": 0.0677,57 "step": 400058 },59 {60 "epoch": 1.75,61 "learning_rate": 3.82142034348649e-05,62 "loss": 0.0695,63 "step": 450064 },65 {66 "epoch": 1.94,67 "learning_rate": 3.5423642164834494e-05,68 "loss": 0.0666,69 "step": 500070 },71 {72 "epoch": 2.13,73 "learning_rate": 3.24656309622951e-05,74 "loss": 0.0555,75 "step": 550076 },77 {78 "epoch": 2.33,79 "learning_rate": 2.9387688608915686e-05,80 "loss": 0.0526,81 "step": 600082 },83 {84 "epoch": 2.52,85 "learning_rate": 2.6245594204007562e-05,86 "loss": 0.0478,87 "step": 650088 },89 {90 "epoch": 2.72,91 "learning_rate": 2.30772470772302e-05,92 "loss": 0.0545,93 "step": 700094 },95 {96 "epoch": 2.91,97 "learning_rate": 1.9939787891736486e-05,98 "loss": 0.0511,99 "step": 7500100 },101 {102 "epoch": 3.1,103 "learning_rate": 1.6883618159833997e-05,104 "loss": 0.0414,105 "step": 8000106 },107 {108 "epoch": 3.3,109 "learning_rate": 1.3963523303957105e-05,110 "loss": 0.0355,111 "step": 8500112 },113 {114 "epoch": 3.49,115 "learning_rate": 1.1214724938519077e-05,116 "loss": 0.0387,117 "step": 9000118 },119 {120 "epoch": 3.69,121 "learning_rate": 8.692185278441231e-06,122 "loss": 0.0376,123 "step": 9500124 },125 {126 "epoch": 3.88,127 "learning_rate": 6.426330704931902e-06,128 "loss": 0.0387,129 "step": 10000130 },131 {132 "epoch": 4.07,133 "learning_rate": 4.458851653214341e-06,134 "loss": 0.0339,135 "step": 10500136 },137 {138 "epoch": 4.27,139 "learning_rate": 2.821354564608486e-06,140 "loss": 0.0337,141 "step": 11000142 },143 {144 "epoch": 4.46,145 "learning_rate": 1.542337081476908e-06,146 "loss": 0.0363,147 "step": 11500148 },149 {150 "epoch": 4.66,151 "learning_rate": 6.372264631526293e-07,152 "loss": 0.0355,153 "step": 12000154 },155 {156 "epoch": 4.85,157 "learning_rate": 1.2349006124565966e-07,158 "loss": 0.0398,159 "step": 12500160 }161 ],162 "logging_steps": 500,163 "max_steps": 12885,164 "num_train_epochs": 5,165 "save_steps": 500,166 "total_flos": 1.0922943771416986e+17,167 "trial_name": null,168 "trial_params": null169}170 