Model-J__MAE__model_idx_0584 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
c88b4ef verified
Invalid JSON: Unexpected token 'I', ..."ad_norm": Infinity, "... is not valid JSON
{
"best_metric": 0.41226666666666667,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0584/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": 3.2564,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.20933333333333334,
"eval_loss": 2.9274826049804688,
"eval_runtime": 6.8791,
"eval_samples_per_second": 545.128,
"eval_steps_per_second": 8.577,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": Infinity,
"learning_rate": 0.0005,
"loss": 2.6874,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.3,
"eval_loss": 2.533846378326416,
"eval_runtime": 7.1266,
"eval_samples_per_second": 526.195,
"eval_steps_per_second": 8.279,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": 12.39351749420166,
"learning_rate": 0.0005,
"loss": 2.3342,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.3552,
"eval_loss": 2.3192245960235596,
"eval_runtime": 8.7675,
"eval_samples_per_second": 427.716,
"eval_steps_per_second": 6.729,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 10.72602367401123,
"learning_rate": 0.0005,
"loss": 2.0661,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.41226666666666667,
"eval_loss": 2.0998599529266357,
"eval_runtime": 7.2743,
"eval_samples_per_second": 515.511,
"eval_steps_per_second": 8.111,
"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
}