Model-J__MAE__model_idx_0670 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
a7a91e1 verified
Invalid JSON: Unexpected token 'I', ..."ad_norm": Infinity, "... is not valid JSON
{
"best_metric": 0.8490666666666666,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0670/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": Infinity,
"learning_rate": 3e-05,
"loss": 2.4573,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.7176,
"eval_loss": 1.191417932510376,
"eval_runtime": 3.9385,
"eval_samples_per_second": 952.137,
"eval_steps_per_second": 14.98,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": Infinity,
"learning_rate": 3e-05,
"loss": 0.8148,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.8133333333333334,
"eval_loss": 0.6759256720542908,
"eval_runtime": 3.9602,
"eval_samples_per_second": 946.91,
"eval_steps_per_second": 14.898,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": 10.80749225616455,
"learning_rate": 3e-05,
"loss": 0.4135,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.8344,
"eval_loss": 0.5511858463287354,
"eval_runtime": 3.9768,
"eval_samples_per_second": 942.977,
"eval_steps_per_second": 14.836,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 1.73154616355896,
"learning_rate": 3e-05,
"loss": 0.2363,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.8490666666666666,
"eval_loss": 0.5005912780761719,
"eval_runtime": 3.9135,
"eval_samples_per_second": 958.219,
"eval_steps_per_second": 15.076,
"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
}