ox-llms-sula-10-profiles-sft / eval_results.json
IeBoytsov's picture
End of training
f26207d verified
{
"epoch": 1.0,
"eval_entropy": 0.14896515673398972,
"eval_loss": 0.148485004901886,
"eval_mean_token_accuracy": 0.9394213200569153,
"eval_num_tokens": 24937449.0,
"eval_runtime": 261.3303,
"eval_samples": 5000,
"eval_samples_per_second": 19.133,
"eval_steps_per_second": 2.392
}