lewtun
/

gemma-7b-dpo-full-mix2-beta-0.1

Text Generation

alignment-handbook

generated_from_trainer

Inference Endpoints

text-generation-inference

Model card Files Files and versions Metrics Training metrics Community

gemma-7b-dpo-full-mix2-beta-0.1 / all_results.json

lewtun's picture

lewtun HF staff

End of training

4b03d26 verified 3 months ago

raw history blame contribute delete

No virus

741 Bytes

	{
	"epoch": 1.0,
	"eval_logits/chosen": 83.16413879394531,
	"eval_logits/rejected": 83.84245300292969,
	"eval_logps/chosen": -392.341552734375,
	"eval_logps/rejected": -414.519775390625,
	"eval_loss": 0.40556877851486206,
	"eval_rewards/accuracies": 0.792553186416626,
	"eval_rewards/chosen": -0.3995126485824585,
	"eval_rewards/margins": 3.1725716590881348,
	"eval_rewards/rejected": -3.5720841884613037,
	"eval_runtime": 140.3273,
	"eval_samples": 3000,
	"eval_samples_per_second": 21.379,
	"eval_steps_per_second": 0.67,
	"train_loss": 0.5623811487565961,
	"train_runtime": 6612.2068,
	"train_samples": 72994,
	"train_samples_per_second": 11.039,
	"train_steps_per_second": 0.086
	}