lewtun
/

gemma-7b-dpo-full-mix1-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-mix1-beta-0.1 / eval_results.json

lewtun's picture

lewtun HF staff

End of training

f9e1bdc verified 3 months ago

raw history blame contribute delete

No virus

564 Bytes

	{
	"epoch": 0.99,
	"eval_logits/chosen": 105.48260498046875,
	"eval_logits/rejected": 99.1939697265625,
	"eval_logps/chosen": -473.2679748535156,
	"eval_logps/rejected": -486.23388671875,
	"eval_loss": 0.6159864068031311,
	"eval_rewards/accuracies": 0.7083333134651184,
	"eval_rewards/chosen": -1.986350178718567,
	"eval_rewards/margins": 1.4820828437805176,
	"eval_rewards/rejected": -3.468433141708374,
	"eval_runtime": 50.9713,
	"eval_samples": 750,
	"eval_samples_per_second": 14.714,
	"eval_steps_per_second": 0.471
	}