japanese-mistral-300m-base / train_results.json
ce-lery's picture
feat: pretrained by recipe v0.1.0
c783850
{
"epoch": 1.0,
"train_loss": 3.89913355111991,
"train_runtime": 393554.9634,
"train_samples": 10794765,
"train_samples_per_second": 27.429,
"train_steps_per_second": 0.107
}