Model-J__DINO__model_idx_0692 / trainer_state.json
Eliahu's picture
Upload folder using huggingface_hub
77acf12 verified
Invalid JSON: Unexpected token 'I', ..."ad_norm": Infinity, "... is not valid JSON
{
"best_metric": 0.3096,
"best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0692/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": 17.119239807128906,
"learning_rate": 0.0003,
"loss": 3.3311,
"step": 333
},
{
"epoch": 1.0,
"eval_accuracy": 0.1768,
"eval_loss": 3.1005797386169434,
"eval_runtime": 11.93,
"eval_samples_per_second": 314.334,
"eval_steps_per_second": 4.946,
"step": 333
},
{
"epoch": 2.0,
"grad_norm": 19.665767669677734,
"learning_rate": 0.0003,
"loss": 2.7965,
"step": 666
},
{
"epoch": 2.0,
"eval_accuracy": 0.26026666666666665,
"eval_loss": 2.6844396591186523,
"eval_runtime": 11.6797,
"eval_samples_per_second": 321.071,
"eval_steps_per_second": 5.052,
"step": 666
},
{
"epoch": 3.0,
"grad_norm": Infinity,
"learning_rate": 0.0003,
"loss": 2.5701,
"step": 999
},
{
"epoch": 3.0,
"eval_accuracy": 0.3061333333333333,
"eval_loss": 2.551422119140625,
"eval_runtime": 12.3296,
"eval_samples_per_second": 304.146,
"eval_steps_per_second": 4.785,
"step": 999
},
{
"epoch": 4.0,
"grad_norm": 18.559833526611328,
"learning_rate": 0.0003,
"loss": 2.4094,
"step": 1332
},
{
"epoch": 4.0,
"eval_accuracy": 0.3096,
"eval_loss": 2.476759910583496,
"eval_runtime": 11.8242,
"eval_samples_per_second": 317.147,
"eval_steps_per_second": 4.99,
"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
}