Model-J__MAE__model_idx_0954 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
8e8402a verified
Invalid JSON: Unexpected token 'I', ..."ad_norm": Infinity, "... is not valid JSON
{
"best_metric": 0.8405333333333334,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0954/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": 2.5087,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.6808,
"eval_loss": 1.3544331789016724,
"eval_runtime": 6.6376,
"eval_samples_per_second": 564.959,
"eval_steps_per_second": 8.889,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": Infinity,
"learning_rate": 3e-05,
"loss": 0.9301,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.8146666666666667,
"eval_loss": 0.7189109325408936,
"eval_runtime": 6.7515,
"eval_samples_per_second": 555.43,
"eval_steps_per_second": 8.739,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": 25.245655059814453,
"learning_rate": 3e-05,
"loss": 0.5121,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.8349333333333333,
"eval_loss": 0.584786057472229,
"eval_runtime": 6.738,
"eval_samples_per_second": 556.546,
"eval_steps_per_second": 8.756,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 4.878299713134766,
"learning_rate": 3e-05,
"loss": 0.3335,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.8405333333333334,
"eval_loss": 0.5435130596160889,
"eval_runtime": 6.5761,
"eval_samples_per_second": 570.247,
"eval_steps_per_second": 8.972,
"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
}