Model-J__DINO__model_idx_0286 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
1027fc9 verified
{
"best_metric": 0.9341333333333334,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0286/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": 0.6722883582115173,
"learning_rate": 7.507507507507507e-05,
"loss": 0.7791,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.8437333333333333,
"eval_loss": 0.5169729590415955,
"eval_runtime": 12.8414,
"eval_samples_per_second": 292.025,
"eval_steps_per_second": 4.595,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": 0.18592655658721924,
"learning_rate": 5.0075075075075074e-05,
"loss": 0.2112,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.8858666666666667,
"eval_loss": 0.3689333200454712,
"eval_runtime": 12.4536,
"eval_samples_per_second": 301.119,
"eval_steps_per_second": 4.738,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": 0.9270104765892029,
"learning_rate": 2.5075075075075076e-05,
"loss": 0.0763,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.9096,
"eval_loss": 0.2989353537559509,
"eval_runtime": 12.5657,
"eval_samples_per_second": 298.433,
"eval_steps_per_second": 4.695,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 0.025218656286597252,
"learning_rate": 7.507507507507508e-08,
"loss": 0.0191,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.9341333333333334,
"eval_loss": 0.22753481566905975,
"eval_runtime": 12.7038,
"eval_samples_per_second": 295.188,
"eval_steps_per_second": 4.644,
"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
}