mistral-sft-7b-dpo-qlora / trainer_state.json
MichaelR207's picture
Model save
53d9518 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 5.0,
"eval_steps": 100,
"global_step": 5,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 1.0,
"learning_rate": 5e-06,
"logits/chosen": -2.7210071086883545,
"logits/rejected": -2.6927733421325684,
"logps/chosen": -90.6630859375,
"logps/rejected": -77.25151062011719,
"loss": 0.1733,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/rejected": 0.0,
"step": 1
},
{
"epoch": 5.0,
"step": 5,
"total_flos": 0.0,
"train_loss": 0.17302747964859008,
"train_runtime": 39.8084,
"train_samples_per_second": 0.628,
"train_steps_per_second": 0.126
}
],
"logging_steps": 10,
"max_steps": 5,
"num_input_tokens_seen": 0,
"num_train_epochs": 5,
"save_steps": 100,
"total_flos": 0.0,
"train_batch_size": 4,
"trial_name": null,
"trial_params": null
}