Model-J__DINO__model_idx_0233 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
30baed1 verified
Invalid JSON: Unexpected token 'I', ..."ad_norm": Infinity, "... is not valid JSON
{
"best_metric": 0.8794666666666666,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0233/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": 0.7505,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.844,
"eval_loss": 0.5049033761024475,
"eval_runtime": 6.542,
"eval_samples_per_second": 573.218,
"eval_steps_per_second": 9.019,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": 0.7136536836624146,
"learning_rate": 7e-05,
"loss": 0.2415,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.8509333333333333,
"eval_loss": 0.5333887934684753,
"eval_runtime": 6.5949,
"eval_samples_per_second": 568.625,
"eval_steps_per_second": 8.946,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": 285.6513671875,
"learning_rate": 7e-05,
"loss": 0.1457,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.8738666666666667,
"eval_loss": 0.4782179892063141,
"eval_runtime": 6.5607,
"eval_samples_per_second": 571.585,
"eval_steps_per_second": 8.993,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 0.8811736106872559,
"learning_rate": 7e-05,
"loss": 0.1091,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.8794666666666666,
"eval_loss": 0.4526247978210449,
"eval_runtime": 7.4321,
"eval_samples_per_second": 504.565,
"eval_steps_per_second": 7.938,
"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
}