Model-J__DINO__model_idx_0902 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
d37d3d5 verified
Invalid JSON: Unexpected token 'I', ..."ad_norm": Infinity, "... is not valid JSON
{
"best_metric": 0.3821333333333333,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0902/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": 0.0003,
"loss": 3.1453,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.2592,
"eval_loss": 2.7599270343780518,
"eval_runtime": 11.806,
"eval_samples_per_second": 317.636,
"eval_steps_per_second": 4.997,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": 19.992555618286133,
"learning_rate": 0.0003,
"loss": 2.5779,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.3168,
"eval_loss": 2.5195701122283936,
"eval_runtime": 11.7814,
"eval_samples_per_second": 318.297,
"eval_steps_per_second": 5.008,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": 26.5197811126709,
"learning_rate": 0.0003,
"loss": 2.3219,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.3608,
"eval_loss": 2.3640687465667725,
"eval_runtime": 11.8537,
"eval_samples_per_second": 316.357,
"eval_steps_per_second": 4.977,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 18.89701271057129,
"learning_rate": 0.0003,
"loss": 2.0986,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.3821333333333333,
"eval_loss": 2.215716600418091,
"eval_runtime": 11.4666,
"eval_samples_per_second": 327.036,
"eval_steps_per_second": 5.145,
"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
}