Model-J__DINO__model_idx_0133 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
59c2fe2 verified
{
"best_metric": 0.3506666666666667,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0133/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": 15.687226295471191,
"learning_rate": 0.00042760859989611786,
"loss": 3.3847,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.19093333333333334,
"eval_loss": 2.9884166717529297,
"eval_runtime": 11.6316,
"eval_samples_per_second": 322.397,
"eval_steps_per_second": 5.072,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": 13.836655616760254,
"learning_rate": 0.00025117927214825806,
"loss": 2.806,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.2712,
"eval_loss": 2.6617624759674072,
"eval_runtime": 11.304,
"eval_samples_per_second": 331.741,
"eval_steps_per_second": 5.219,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": 18.025524139404297,
"learning_rate": 7.405914276967749e-05,
"loss": 2.4874,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.32266666666666666,
"eval_loss": 2.4096028804779053,
"eval_runtime": 11.6501,
"eval_samples_per_second": 321.885,
"eval_steps_per_second": 5.064,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 24.447607040405273,
"learning_rate": 2.7813810714871766e-09,
"loss": 2.2282,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.3506666666666667,
"eval_loss": 2.302328586578369,
"eval_runtime": 11.2407,
"eval_samples_per_second": 333.61,
"eval_steps_per_second": 5.249,
"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
}