empty-michael's picture
End of training
dd10feb verified
raw
history blame
456 Bytes
{
"epoch": 2.04,
"eval_MSE/layer0": 604.5096733395267,
"eval_accuracy": 0.5091345939349958,
"eval_dead_code_fraction/layer0": 0.18795,
"eval_input_norm/layer0": 31.99872850438308,
"eval_loss": 2.132894992828369,
"eval_multicode_k": 1,
"eval_output_norm/layer0": 19.389702240368152,
"eval_runtime": 159.3907,
"eval_samples_per_second": 29.004,
"eval_steps_per_second": 1.813,
"perplexity": 8.439263086993584
}