Model-J__DINO__model_idx_0309 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
ebd4b82 verified
{
"best_metric": 0.9109333333333334,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0309/checkpoints/checkpoint-1332",
"epoch": 4.0,
"eval_steps": 500,
"global_step": 1332,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 1.0,
"grad_norm": 14.708928108215332,
"learning_rate": 2.2522522522522523e-05,
"loss": 0.8249,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.8608,
"eval_loss": 0.4072793424129486,
"eval_runtime": 6.194,
"eval_samples_per_second": 605.428,
"eval_steps_per_second": 9.525,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": 2.797719955444336,
"learning_rate": 1.5022522522522522e-05,
"loss": 0.1586,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.8874666666666666,
"eval_loss": 0.3532145321369171,
"eval_runtime": 6.6304,
"eval_samples_per_second": 565.58,
"eval_steps_per_second": 8.898,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": 0.10572859644889832,
"learning_rate": 7.522522522522523e-06,
"loss": 0.039,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.8997333333333334,
"eval_loss": 0.3473365008831024,
"eval_runtime": 6.1492,
"eval_samples_per_second": 609.837,
"eval_steps_per_second": 9.595,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 0.18686339259147644,
"learning_rate": 2.2522522522522524e-08,
"loss": 0.0075,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.9109333333333334,
"eval_loss": 0.32793065905570984,
"eval_runtime": 6.8788,
"eval_samples_per_second": 545.151,
"eval_steps_per_second": 8.577,
"step": 1332
}
],
"logging_steps": 500,
"max_steps": 1332,
"num_input_tokens_seen": 0,
"num_train_epochs": 4,
"save_steps": 500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 6.58965282490368e+18,
"train_batch_size": 64,
"trial_name": null,
"trial_params": null
}