Model-J__DINO__model_idx_0588 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
b2ac194 verified
{
"best_metric": 0.3722666666666667,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0588/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": 19.493999481201172,
"learning_rate": 0.00042760859989611786,
"loss": 3.3584,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.19866666666666666,
"eval_loss": 3.046205759048462,
"eval_runtime": 12.7896,
"eval_samples_per_second": 293.207,
"eval_steps_per_second": 4.613,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": 26.901456832885742,
"learning_rate": 0.00025117927214825806,
"loss": 2.7985,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.2733333333333333,
"eval_loss": 2.7168684005737305,
"eval_runtime": 13.6516,
"eval_samples_per_second": 274.692,
"eval_steps_per_second": 4.322,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": 14.65412712097168,
"learning_rate": 7.405914276967749e-05,
"loss": 2.4708,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.35706666666666664,
"eval_loss": 2.366395950317383,
"eval_runtime": 14.4918,
"eval_samples_per_second": 258.767,
"eval_steps_per_second": 4.071,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 24.95326042175293,
"learning_rate": 2.7813810714871766e-09,
"loss": 2.2152,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.3722666666666667,
"eval_loss": 2.27994704246521,
"eval_runtime": 13.9972,
"eval_samples_per_second": 267.91,
"eval_steps_per_second": 4.215,
"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
}