V10-40G / final /all_results.json
gotzmann's picture
..
d1ae03b
raw
history blame
223 Bytes
{
"epoch": 0.9996918335901387,
"total_flos": 2.085852585718815e+19,
"train_loss": 1.4798993557213855,
"train_runtime": 22030.8952,
"train_samples_per_second": 2.357,
"train_steps_per_second": 0.074
}