| { | |
| "best_metric": 0.8431372549019608, | |
| "best_model_checkpoint": "vit-base-patch16-224-RX1-24\\checkpoint-127", | |
| "epoch": 22.4, | |
| "eval_steps": 500, | |
| "global_step": 168, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.93, | |
| "eval_accuracy": 0.47058823529411764, | |
| "eval_loss": 1.3484954833984375, | |
| "eval_runtime": 0.828, | |
| "eval_samples_per_second": 61.591, | |
| "eval_steps_per_second": 2.415, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 5.4654088050314465e-05, | |
| "loss": 1.3674, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.5490196078431373, | |
| "eval_loss": 1.2284042835235596, | |
| "eval_runtime": 1.1035, | |
| "eval_samples_per_second": 46.217, | |
| "eval_steps_per_second": 1.812, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.119496855345912e-05, | |
| "loss": 1.2414, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "eval_accuracy": 0.6470588235294118, | |
| "eval_loss": 1.130693793296814, | |
| "eval_runtime": 0.8296, | |
| "eval_samples_per_second": 61.476, | |
| "eval_steps_per_second": 2.411, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 4.7735849056603775e-05, | |
| "loss": 1.1146, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.6470588235294118, | |
| "eval_loss": 1.0230462551116943, | |
| "eval_runtime": 0.8185, | |
| "eval_samples_per_second": 62.312, | |
| "eval_steps_per_second": 2.444, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "eval_accuracy": 0.6862745098039216, | |
| "eval_loss": 0.9250622391700745, | |
| "eval_runtime": 0.8344, | |
| "eval_samples_per_second": 61.119, | |
| "eval_steps_per_second": 2.397, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 4.427672955974843e-05, | |
| "loss": 0.9522, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.6470588235294118, | |
| "eval_loss": 0.9121541976928711, | |
| "eval_runtime": 0.8143, | |
| "eval_samples_per_second": 62.627, | |
| "eval_steps_per_second": 2.456, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 4.081761006289308e-05, | |
| "loss": 0.8247, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "eval_accuracy": 0.6274509803921569, | |
| "eval_loss": 0.9373806118965149, | |
| "eval_runtime": 0.8134, | |
| "eval_samples_per_second": 62.702, | |
| "eval_steps_per_second": 2.459, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 3.735849056603773e-05, | |
| "loss": 0.6825, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.6862745098039216, | |
| "eval_loss": 0.8320279121398926, | |
| "eval_runtime": 0.812, | |
| "eval_samples_per_second": 62.807, | |
| "eval_steps_per_second": 2.463, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 8.93, | |
| "eval_accuracy": 0.6666666666666666, | |
| "eval_loss": 0.8286137580871582, | |
| "eval_runtime": 0.8144, | |
| "eval_samples_per_second": 62.624, | |
| "eval_steps_per_second": 2.456, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 3.389937106918239e-05, | |
| "loss": 0.6191, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.6666666666666666, | |
| "eval_loss": 0.841847836971283, | |
| "eval_runtime": 0.8329, | |
| "eval_samples_per_second": 61.231, | |
| "eval_steps_per_second": 2.401, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 10.67, | |
| "learning_rate": 3.0440251572327043e-05, | |
| "loss": 0.5312, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 10.93, | |
| "eval_accuracy": 0.8235294117647058, | |
| "eval_loss": 0.7835917472839355, | |
| "eval_runtime": 0.8003, | |
| "eval_samples_per_second": 63.726, | |
| "eval_steps_per_second": 2.499, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 2.69811320754717e-05, | |
| "loss": 0.454, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.803921568627451, | |
| "eval_loss": 0.7356413006782532, | |
| "eval_runtime": 0.8146, | |
| "eval_samples_per_second": 62.606, | |
| "eval_steps_per_second": 2.455, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 12.93, | |
| "eval_accuracy": 0.8235294117647058, | |
| "eval_loss": 0.6117302775382996, | |
| "eval_runtime": 0.8494, | |
| "eval_samples_per_second": 60.045, | |
| "eval_steps_per_second": 2.355, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 13.33, | |
| "learning_rate": 2.3522012578616354e-05, | |
| "loss": 0.3752, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_accuracy": 0.8235294117647058, | |
| "eval_loss": 0.6014484167098999, | |
| "eval_runtime": 0.983, | |
| "eval_samples_per_second": 51.882, | |
| "eval_steps_per_second": 2.035, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 14.67, | |
| "learning_rate": 2.0062893081761005e-05, | |
| "loss": 0.3269, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 14.93, | |
| "eval_accuracy": 0.803921568627451, | |
| "eval_loss": 0.610191822052002, | |
| "eval_runtime": 0.9673, | |
| "eval_samples_per_second": 52.721, | |
| "eval_steps_per_second": 2.068, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "learning_rate": 1.660377358490566e-05, | |
| "loss": 0.2733, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_accuracy": 0.803921568627451, | |
| "eval_loss": 0.6404368877410889, | |
| "eval_runtime": 0.9418, | |
| "eval_samples_per_second": 54.15, | |
| "eval_steps_per_second": 2.124, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 16.93, | |
| "eval_accuracy": 0.8431372549019608, | |
| "eval_loss": 0.5686964988708496, | |
| "eval_runtime": 0.9812, | |
| "eval_samples_per_second": 51.975, | |
| "eval_steps_per_second": 2.038, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 17.33, | |
| "learning_rate": 1.3144654088050315e-05, | |
| "loss": 0.2711, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_accuracy": 0.8235294117647058, | |
| "eval_loss": 0.6119526624679565, | |
| "eval_runtime": 0.8276, | |
| "eval_samples_per_second": 61.624, | |
| "eval_steps_per_second": 2.417, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 18.67, | |
| "learning_rate": 9.685534591194969e-06, | |
| "loss": 0.2519, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 18.93, | |
| "eval_accuracy": 0.8431372549019608, | |
| "eval_loss": 0.6250138282775879, | |
| "eval_runtime": 0.828, | |
| "eval_samples_per_second": 61.593, | |
| "eval_steps_per_second": 2.415, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 6.226415094339623e-06, | |
| "loss": 0.2484, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_accuracy": 0.7843137254901961, | |
| "eval_loss": 0.608646035194397, | |
| "eval_runtime": 0.8743, | |
| "eval_samples_per_second": 58.331, | |
| "eval_steps_per_second": 2.287, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 20.93, | |
| "eval_accuracy": 0.8235294117647058, | |
| "eval_loss": 0.6228971481323242, | |
| "eval_runtime": 0.9312, | |
| "eval_samples_per_second": 54.768, | |
| "eval_steps_per_second": 2.148, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 21.33, | |
| "learning_rate": 2.767295597484277e-06, | |
| "loss": 0.2258, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_accuracy": 0.7843137254901961, | |
| "eval_loss": 0.6390360593795776, | |
| "eval_runtime": 0.8254, | |
| "eval_samples_per_second": 61.789, | |
| "eval_steps_per_second": 2.423, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 22.4, | |
| "eval_accuracy": 0.803921568627451, | |
| "eval_loss": 0.633681058883667, | |
| "eval_runtime": 0.9585, | |
| "eval_samples_per_second": 53.211, | |
| "eval_steps_per_second": 2.087, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 22.4, | |
| "step": 168, | |
| "total_flos": 1.627205965669122e+18, | |
| "train_loss": 0.5914713264930815, | |
| "train_runtime": 386.7057, | |
| "train_samples_per_second": 58.153, | |
| "train_steps_per_second": 0.434 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 168, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 24, | |
| "save_steps": 500, | |
| "total_flos": 1.627205965669122e+18, | |
| "train_batch_size": 32, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |