oh-dcft-v3.1-SN-405B-hacky-qwen / train_results.json
sedrickkeh's picture
End of training
9e58634 verified
raw
history blame contribute delete
205 Bytes
{
"epoch": 3.0,
"total_flos": 2108813507297280.0,
"train_loss": 0.4856802497430432,
"train_runtime": 54595.2546,
"train_samples_per_second": 9.429,
"train_steps_per_second": 0.074
}