Model-J__MAE__model_idx_0976 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
a2ac002 verified
Invalid JSON: Unexpected token 'I', ..."ad_norm": Infinity, "... is not valid JSON
{
"best_metric": 0.49066666666666664,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0976/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": 0.0005,
"loss": 3.1739,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.23013333333333333,
"eval_loss": 2.7883238792419434,
"eval_runtime": 5.8819,
"eval_samples_per_second": 637.554,
"eval_steps_per_second": 10.031,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": Infinity,
"learning_rate": 0.0005,
"loss": 2.4859,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.3816,
"eval_loss": 2.2021303176879883,
"eval_runtime": 6.2269,
"eval_samples_per_second": 602.229,
"eval_steps_per_second": 9.475,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": Infinity,
"learning_rate": 0.0005,
"loss": 1.9424,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.4357333333333333,
"eval_loss": 2.005305051803589,
"eval_runtime": 6.5697,
"eval_samples_per_second": 570.8,
"eval_steps_per_second": 8.981,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 9.052824020385742,
"learning_rate": 0.0005,
"loss": 1.5238,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.49066666666666664,
"eval_loss": 1.7423828840255737,
"eval_runtime": 5.468,
"eval_samples_per_second": 685.808,
"eval_steps_per_second": 10.79,
"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
}