76 lines
1.9 KiB
JSON
76 lines
1.9 KiB
JSON
|
{
|
||
|
|
"best_global_step": 1268,
|
||
|
|
"best_metric": 73.68802166810151,
|
||
|
|
"best_model_checkpoint": "mbert-loraxs-qa-vanilla/run-17/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": 5.264754295349121,
|
||
|
|
"learning_rate": 0.0004209817281796039,
|
||
|
|
"loss": 2.7526726467000198,
|
||
|
|
"step": 634
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.0,
|
||
|
|
"eval_exact_match": 55.6,
|
||
|
|
"eval_f1": 70.7495450084009,
|
||
|
|
"eval_loss": 1.3779001235961914,
|
||
|
|
"eval_runtime": 7.7718,
|
||
|
|
"eval_samples_per_second": 162.768,
|
||
|
|
"eval_steps_per_second": 10.294,
|
||
|
|
"step": 634
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.0,
|
||
|
|
"grad_norm": 5.9264044761657715,
|
||
|
|
"learning_rate": 0.0003157777803412638,
|
||
|
|
"loss": 1.3429681916342173,
|
||
|
|
"step": 1268
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.0,
|
||
|
|
"eval_exact_match": 58.96,
|
||
|
|
"eval_f1": 73.68802166810151,
|
||
|
|
"eval_loss": 1.2954691648483276,
|
||
|
|
"eval_runtime": 7.8505,
|
||
|
|
"eval_samples_per_second": 161.136,
|
||
|
|
"eval_steps_per_second": 10.19,
|
||
|
|
"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": 0.0004734177652725306,
|
||
|
|
"lora_alpha": 64,
|
||
|
|
"lora_dropout": 0.052068221604058385,
|
||
|
|
"lora_r": 256,
|
||
|
|
"num_train_epochs": 5,
|
||
|
|
"weight_decay": 0.017014537049436163
|
||
|
|
}
|
||
|
|
}
|