Model-J__DINO__model_idx_0048 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
e894b99 verified
{
"best_metric": 0.9178666666666667,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0048/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": 2.4079174995422363,
"learning_rate": 8.543862808774192e-05,
"loss": 0.9753,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.8093333333333333,
"eval_loss": 0.6363093256950378,
"eval_runtime": 11.2989,
"eval_samples_per_second": 331.89,
"eval_steps_per_second": 5.222,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": 0.04762320965528488,
"learning_rate": 5.011792754282769e-05,
"loss": 0.3161,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.8776,
"eval_loss": 0.41732093691825867,
"eval_runtime": 11.7787,
"eval_samples_per_second": 318.372,
"eval_steps_per_second": 5.009,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": 58.21407699584961,
"learning_rate": 1.472814664270234e-05,
"loss": 0.0887,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.9096,
"eval_loss": 0.3170669972896576,
"eval_runtime": 11.3089,
"eval_samples_per_second": 331.596,
"eval_steps_per_second": 5.217,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 0.15726087987422943,
"learning_rate": 1.3906924697382196e-10,
"loss": 0.0153,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.9178666666666667,
"eval_loss": 0.2888360321521759,
"eval_runtime": 11.1346,
"eval_samples_per_second": 336.788,
"eval_steps_per_second": 5.299,
"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
}