76 lines
1.9 KiB
JSON
76 lines
1.9 KiB
JSON
|
{
|
||
|
|
"best_global_step": 1268,
|
||
|
|
"best_metric": 72.97295260861294,
|
||
|
|
"best_model_checkpoint": "mbert-loraxs-qa-vanilla/run-8/checkpoint-1268",
|
||
|
|
"epoch": 2.0,
|
||
|
|
"eval_steps": 500,
|
||
|
|
"global_step": 1268,
|
||
|
|
"is_hyper_param_search": true,
|
||
|
|
"is_local_process_zero": true,
|
||
|
|
"is_world_process_zero": true,
|
||
|
|
"log_history": [
|
||
|
|
{
|
||
|
|
"epoch": 1.0,
|
||
|
|
"grad_norm": 13.794865608215332,
|
||
|
|
"learning_rate": 0.00014003511035757225,
|
||
|
|
"loss": 3.036647086639886,
|
||
|
|
"step": 634
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.0,
|
||
|
|
"eval_exact_match": 55.76,
|
||
|
|
"eval_f1": 69.6160406135031,
|
||
|
|
"eval_loss": 1.4716368913650513,
|
||
|
|
"eval_runtime": 7.8535,
|
||
|
|
"eval_samples_per_second": 161.074,
|
||
|
|
"eval_steps_per_second": 10.187,
|
||
|
|
"step": 634
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.0,
|
||
|
|
"grad_norm": 9.303217887878418,
|
||
|
|
"learning_rate": 0.00010504013204984627,
|
||
|
|
"loss": 1.4382544676964215,
|
||
|
|
"step": 1268
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.0,
|
||
|
|
"eval_exact_match": 58.32,
|
||
|
|
"eval_f1": 72.97295260861294,
|
||
|
|
"eval_loss": 1.3276563882827759,
|
||
|
|
"eval_runtime": 7.7377,
|
||
|
|
"eval_samples_per_second": 163.485,
|
||
|
|
"eval_steps_per_second": 10.339,
|
||
|
|
"step": 1268
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"logging_steps": 500,
|
||
|
|
"max_steps": 3170,
|
||
|
|
"num_input_tokens_seen": 0,
|
||
|
|
"num_train_epochs": 5,
|
||
|
|
"save_steps": 500,
|
||
|
|
"stateful_callbacks": {
|
||
|
|
"TrainerControl": {
|
||
|
|
"args": {
|
||
|
|
"should_epoch_stop": false,
|
||
|
|
"should_evaluate": false,
|
||
|
|
"should_log": false,
|
||
|
|
"should_save": true,
|
||
|
|
"should_training_stop": false
|
||
|
|
},
|
||
|
|
"attributes": {}
|
||
|
|
}
|
||
|
|
},
|
||
|
|
"total_flos": 6180991462324224.0,
|
||
|
|
"train_batch_size": 16,
|
||
|
|
"trial_name": null,
|
||
|
|
"trial_params": {
|
||
|
|
"learning_rate": 0.00015747740238476691,
|
||
|
|
"lora_alpha": 128,
|
||
|
|
"lora_dropout": 0.10139328835291189,
|
||
|
|
"lora_r": 256,
|
||
|
|
"num_train_epochs": 5,
|
||
|
|
"weight_decay": 0.05287944657262306
|
||
|
|
}
|
||
|
|
}
|