a1_math_deepmind / train_results.json
ryanmarten's picture
End of training
a8101fc verified
raw
history blame contribute delete
223 Bytes
{
"epoch": 4.988354430379747,
"total_flos": 1.424052118385328e+18,
"train_loss": 0.23391443721768332,
"train_runtime": 42513.0625,
"train_samples_per_second": 3.717,
"train_steps_per_second": 0.029
}