| { |
| "best_metric": 0.27537301182746887, |
| "best_model_checkpoint": "faces_age_detection/checkpoint-13224", |
| "epoch": 29.0, |
| "eval_steps": 500, |
| "global_step": 13224, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 1.0, |
| "eval_accuracy": 0.8806292412091302, |
| "eval_loss": 0.3289386034011841, |
| "eval_runtime": 29.1731, |
| "eval_samples_per_second": 111.13, |
| "eval_steps_per_second": 3.496, |
| "step": 456 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 1.9339691856199557e-06, |
| "loss": 0.3578, |
| "step": 500 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_accuracy": 0.879086983343615, |
| "eval_loss": 0.3256358802318573, |
| "eval_runtime": 29.8484, |
| "eval_samples_per_second": 108.615, |
| "eval_steps_per_second": 3.417, |
| "step": 912 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 1.8606016140865737e-06, |
| "loss": 0.2998, |
| "step": 1000 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_accuracy": 0.8834053053670574, |
| "eval_loss": 0.3201422393321991, |
| "eval_runtime": 29.3669, |
| "eval_samples_per_second": 110.396, |
| "eval_steps_per_second": 3.473, |
| "step": 1368 |
| }, |
| { |
| "epoch": 3.29, |
| "learning_rate": 1.7872340425531913e-06, |
| "loss": 0.2666, |
| "step": 1500 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_accuracy": 0.8867982726711906, |
| "eval_loss": 0.31441426277160645, |
| "eval_runtime": 29.5521, |
| "eval_samples_per_second": 109.705, |
| "eval_steps_per_second": 3.452, |
| "step": 1824 |
| }, |
| { |
| "epoch": 4.39, |
| "learning_rate": 1.7138664710198092e-06, |
| "loss": 0.2445, |
| "step": 2000 |
| }, |
| { |
| "epoch": 5.0, |
| "eval_accuracy": 0.8840222085132634, |
| "eval_loss": 0.31040704250335693, |
| "eval_runtime": 29.3243, |
| "eval_samples_per_second": 110.557, |
| "eval_steps_per_second": 3.478, |
| "step": 2280 |
| }, |
| { |
| "epoch": 5.48, |
| "learning_rate": 1.6404988994864268e-06, |
| "loss": 0.2307, |
| "step": 2500 |
| }, |
| { |
| "epoch": 6.0, |
| "eval_accuracy": 0.8756940160394818, |
| "eval_loss": 0.31916671991348267, |
| "eval_runtime": 29.1138, |
| "eval_samples_per_second": 111.356, |
| "eval_steps_per_second": 3.503, |
| "step": 2736 |
| }, |
| { |
| "epoch": 6.58, |
| "learning_rate": 1.5671313279530447e-06, |
| "loss": 0.2218, |
| "step": 3000 |
| }, |
| { |
| "epoch": 7.0, |
| "eval_accuracy": 0.8821714990746453, |
| "eval_loss": 0.30659398436546326, |
| "eval_runtime": 29.0887, |
| "eval_samples_per_second": 111.452, |
| "eval_steps_per_second": 3.507, |
| "step": 3192 |
| }, |
| { |
| "epoch": 7.68, |
| "learning_rate": 1.4937637564196623e-06, |
| "loss": 0.2007, |
| "step": 3500 |
| }, |
| { |
| "epoch": 8.0, |
| "eval_accuracy": 0.8793954349167181, |
| "eval_loss": 0.3129652738571167, |
| "eval_runtime": 28.3857, |
| "eval_samples_per_second": 114.212, |
| "eval_steps_per_second": 3.593, |
| "step": 3648 |
| }, |
| { |
| "epoch": 8.77, |
| "learning_rate": 1.4203961848862801e-06, |
| "loss": 0.192, |
| "step": 4000 |
| }, |
| { |
| "epoch": 9.0, |
| "eval_accuracy": 0.8809376927822332, |
| "eval_loss": 0.31223684549331665, |
| "eval_runtime": 28.6423, |
| "eval_samples_per_second": 113.189, |
| "eval_steps_per_second": 3.561, |
| "step": 4104 |
| }, |
| { |
| "epoch": 9.87, |
| "learning_rate": 1.3470286133528978e-06, |
| "loss": 0.1832, |
| "step": 4500 |
| }, |
| { |
| "epoch": 10.0, |
| "eval_accuracy": 0.8883405305367057, |
| "eval_loss": 0.29440754652023315, |
| "eval_runtime": 29.0492, |
| "eval_samples_per_second": 111.604, |
| "eval_steps_per_second": 3.511, |
| "step": 4560 |
| }, |
| { |
| "epoch": 10.96, |
| "learning_rate": 1.2736610418195158e-06, |
| "loss": 0.1743, |
| "step": 5000 |
| }, |
| { |
| "epoch": 11.0, |
| "eval_accuracy": 0.8815545959284392, |
| "eval_loss": 0.3053087294101715, |
| "eval_runtime": 29.3331, |
| "eval_samples_per_second": 110.523, |
| "eval_steps_per_second": 3.477, |
| "step": 5016 |
| }, |
| { |
| "epoch": 12.0, |
| "eval_accuracy": 0.8858729179518815, |
| "eval_loss": 0.2922166883945465, |
| "eval_runtime": 29.148, |
| "eval_samples_per_second": 111.225, |
| "eval_steps_per_second": 3.499, |
| "step": 5472 |
| }, |
| { |
| "epoch": 12.06, |
| "learning_rate": 1.2002934702861334e-06, |
| "loss": 0.1668, |
| "step": 5500 |
| }, |
| { |
| "epoch": 13.0, |
| "eval_accuracy": 0.8815545959284392, |
| "eval_loss": 0.3020872473716736, |
| "eval_runtime": 29.262, |
| "eval_samples_per_second": 110.792, |
| "eval_steps_per_second": 3.486, |
| "step": 5928 |
| }, |
| { |
| "epoch": 13.16, |
| "learning_rate": 1.1269258987527513e-06, |
| "loss": 0.1608, |
| "step": 6000 |
| }, |
| { |
| "epoch": 14.0, |
| "eval_accuracy": 0.8830968537939543, |
| "eval_loss": 0.3037404716014862, |
| "eval_runtime": 29.0749, |
| "eval_samples_per_second": 111.505, |
| "eval_steps_per_second": 3.508, |
| "step": 6384 |
| }, |
| { |
| "epoch": 14.25, |
| "learning_rate": 1.053558327219369e-06, |
| "loss": 0.1563, |
| "step": 6500 |
| }, |
| { |
| "epoch": 15.0, |
| "eval_accuracy": 0.8843306600863664, |
| "eval_loss": 0.29694926738739014, |
| "eval_runtime": 29.4263, |
| "eval_samples_per_second": 110.173, |
| "eval_steps_per_second": 3.466, |
| "step": 6840 |
| }, |
| { |
| "epoch": 15.35, |
| "learning_rate": 9.801907556859868e-07, |
| "loss": 0.1467, |
| "step": 7000 |
| }, |
| { |
| "epoch": 16.0, |
| "eval_accuracy": 0.8877236273904997, |
| "eval_loss": 0.29596608877182007, |
| "eval_runtime": 29.2838, |
| "eval_samples_per_second": 110.71, |
| "eval_steps_per_second": 3.483, |
| "step": 7296 |
| }, |
| { |
| "epoch": 16.45, |
| "learning_rate": 9.068231841526045e-07, |
| "loss": 0.139, |
| "step": 7500 |
| }, |
| { |
| "epoch": 17.0, |
| "eval_accuracy": 0.8864898210980876, |
| "eval_loss": 0.2912845313549042, |
| "eval_runtime": 29.0512, |
| "eval_samples_per_second": 111.596, |
| "eval_steps_per_second": 3.511, |
| "step": 7752 |
| }, |
| { |
| "epoch": 17.54, |
| "learning_rate": 8.334556126192222e-07, |
| "loss": 0.1363, |
| "step": 8000 |
| }, |
| { |
| "epoch": 18.0, |
| "eval_accuracy": 0.891425046267736, |
| "eval_loss": 0.2851228415966034, |
| "eval_runtime": 29.8791, |
| "eval_samples_per_second": 108.504, |
| "eval_steps_per_second": 3.414, |
| "step": 8208 |
| }, |
| { |
| "epoch": 18.64, |
| "learning_rate": 7.6008804108584e-07, |
| "loss": 0.1283, |
| "step": 8500 |
| }, |
| { |
| "epoch": 19.0, |
| "eval_accuracy": 0.892350400987045, |
| "eval_loss": 0.28452929854393005, |
| "eval_runtime": 29.4229, |
| "eval_samples_per_second": 110.186, |
| "eval_steps_per_second": 3.467, |
| "step": 8664 |
| }, |
| { |
| "epoch": 19.74, |
| "learning_rate": 6.867204695524578e-07, |
| "loss": 0.1302, |
| "step": 9000 |
| }, |
| { |
| "epoch": 20.0, |
| "eval_accuracy": 0.8901912399753239, |
| "eval_loss": 0.28534042835235596, |
| "eval_runtime": 29.1008, |
| "eval_samples_per_second": 111.406, |
| "eval_steps_per_second": 3.505, |
| "step": 9120 |
| }, |
| { |
| "epoch": 20.83, |
| "learning_rate": 6.133528980190756e-07, |
| "loss": 0.1269, |
| "step": 9500 |
| }, |
| { |
| "epoch": 21.0, |
| "eval_accuracy": 0.892041949413942, |
| "eval_loss": 0.28531745076179504, |
| "eval_runtime": 28.8168, |
| "eval_samples_per_second": 112.504, |
| "eval_steps_per_second": 3.54, |
| "step": 9576 |
| }, |
| { |
| "epoch": 21.93, |
| "learning_rate": 5.399853264856933e-07, |
| "loss": 0.1211, |
| "step": 10000 |
| }, |
| { |
| "epoch": 22.0, |
| "eval_accuracy": 0.892041949413942, |
| "eval_loss": 0.2809159457683563, |
| "eval_runtime": 28.7701, |
| "eval_samples_per_second": 112.686, |
| "eval_steps_per_second": 3.545, |
| "step": 10032 |
| }, |
| { |
| "epoch": 23.0, |
| "eval_accuracy": 0.8892658852560148, |
| "eval_loss": 0.28494611382484436, |
| "eval_runtime": 28.5362, |
| "eval_samples_per_second": 113.61, |
| "eval_steps_per_second": 3.574, |
| "step": 10488 |
| }, |
| { |
| "epoch": 23.03, |
| "learning_rate": 4.6661775495231103e-07, |
| "loss": 0.1199, |
| "step": 10500 |
| }, |
| { |
| "epoch": 24.0, |
| "eval_accuracy": 0.8942011104256632, |
| "eval_loss": 0.2793309986591339, |
| "eval_runtime": 28.7961, |
| "eval_samples_per_second": 112.585, |
| "eval_steps_per_second": 3.542, |
| "step": 10944 |
| }, |
| { |
| "epoch": 24.12, |
| "learning_rate": 3.932501834189288e-07, |
| "loss": 0.1197, |
| "step": 11000 |
| }, |
| { |
| "epoch": 25.0, |
| "eval_accuracy": 0.8948180135718692, |
| "eval_loss": 0.2778417766094208, |
| "eval_runtime": 27.4503, |
| "eval_samples_per_second": 118.105, |
| "eval_steps_per_second": 3.716, |
| "step": 11400 |
| }, |
| { |
| "epoch": 25.22, |
| "learning_rate": 3.1988261188554655e-07, |
| "loss": 0.1163, |
| "step": 11500 |
| }, |
| { |
| "epoch": 26.0, |
| "eval_accuracy": 0.8960518198642813, |
| "eval_loss": 0.27623382210731506, |
| "eval_runtime": 27.6199, |
| "eval_samples_per_second": 117.379, |
| "eval_steps_per_second": 3.693, |
| "step": 11856 |
| }, |
| { |
| "epoch": 26.32, |
| "learning_rate": 2.4651504035216434e-07, |
| "loss": 0.1152, |
| "step": 12000 |
| }, |
| { |
| "epoch": 27.0, |
| "eval_accuracy": 0.8963602714373843, |
| "eval_loss": 0.2759145498275757, |
| "eval_runtime": 29.0519, |
| "eval_samples_per_second": 111.594, |
| "eval_steps_per_second": 3.511, |
| "step": 12312 |
| }, |
| { |
| "epoch": 27.41, |
| "learning_rate": 1.7314746881878208e-07, |
| "loss": 0.1105, |
| "step": 12500 |
| }, |
| { |
| "epoch": 28.0, |
| "eval_accuracy": 0.8969771745835904, |
| "eval_loss": 0.27611085772514343, |
| "eval_runtime": 27.509, |
| "eval_samples_per_second": 117.852, |
| "eval_steps_per_second": 3.708, |
| "step": 12768 |
| }, |
| { |
| "epoch": 28.51, |
| "learning_rate": 9.977989728539984e-08, |
| "loss": 0.11, |
| "step": 13000 |
| }, |
| { |
| "epoch": 29.0, |
| "eval_accuracy": 0.8969771745835904, |
| "eval_loss": 0.27537301182746887, |
| "eval_runtime": 28.0691, |
| "eval_samples_per_second": 115.501, |
| "eval_steps_per_second": 3.634, |
| "step": 13224 |
| } |
| ], |
| "logging_steps": 500, |
| "max_steps": 13680, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 30, |
| "save_steps": 500, |
| "total_flos": 6.555338630367276e+19, |
| "train_batch_size": 64, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
