vit-cifar10 / train_results.json
jamescalam's picture
first model
b363d25
{
"epoch": 4.0,
"total_flos": 1.54995091808256e+19,
"train_loss": 0.08309991182349008,
"train_runtime": 5212.8854,
"train_samples_per_second": 38.366,
"train_steps_per_second": 1.199
}