Minbyul
/

mistral-7b-dpo-full-sft-wo-healthsearch_qa

Text Generation

alignment-handbook

Generated from Trainer

Inference Endpoints

text-generation-inference

Model card Files Files and versions Community

mistral-7b-dpo-full-sft-wo-healthsearch_qa / eval_results.json

Minbyul's picture

End of training

2d20acd verified about 2 months ago

raw history blame contribute delete

No virus

575 Bytes

	{
	"epoch": 1.0,
	"eval_logits/chosen": -2.9083194732666016,
	"eval_logits/rejected": -2.835282325744629,
	"eval_logps/chosen": -817.455078125,
	"eval_logps/rejected": -1091.8406982421875,
	"eval_loss": 0.6746304631233215,
	"eval_rewards/accuracies": 0.6612318754196167,
	"eval_rewards/chosen": -0.020445356145501137,
	"eval_rewards/margins": 0.039513371884822845,
	"eval_rewards/rejected": -0.059958722442388535,
	"eval_runtime": 2163.8157,
	"eval_samples": 19864,
	"eval_samples_per_second": 9.18,
	"eval_steps_per_second": 0.287
	}