Model-J__MAE__model_idx_0707 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
c0c915f verified
Invalid JSON: Unexpected token 'I', ..."ad_norm": Infinity, "... is not valid JSON
{
"best_metric": 0.8762666666666666,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0707/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": 7e-05,
"loss": 1.7129,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.824,
"eval_loss": 0.6835076808929443,
"eval_runtime": 6.4042,
"eval_samples_per_second": 585.549,
"eval_steps_per_second": 9.213,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": 13.127287864685059,
"learning_rate": 7e-05,
"loss": 0.4604,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.8514666666666667,
"eval_loss": 0.5301175713539124,
"eval_runtime": 6.1534,
"eval_samples_per_second": 609.419,
"eval_steps_per_second": 9.588,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": Infinity,
"learning_rate": 7e-05,
"loss": 0.2366,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.8477333333333333,
"eval_loss": 0.5319370031356812,
"eval_runtime": 6.4351,
"eval_samples_per_second": 582.742,
"eval_steps_per_second": 9.168,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 4.227926254272461,
"learning_rate": 7e-05,
"loss": 0.1304,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.8762666666666666,
"eval_loss": 0.44424715638160706,
"eval_runtime": 6.208,
"eval_samples_per_second": 604.062,
"eval_steps_per_second": 9.504,
"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
}