CodeRM-SFT-Warmup-Selection-4B / all_results.json
t2ance's picture
SFT warmup LoRA for 4B judge (9367 samples, 1 epoch)
bc9cb33 verified
{
"epoch": 1.0,
"total_flos": 7.995150581897871e+17,
"train_loss": 0.5977290161287891,
"train_runtime": 6942.3995,
"train_samples_per_second": 1.349,
"train_steps_per_second": 0.112
}