|
{ |
|
"MSE": 0.0, |
|
"MSE/layer0": 0.0, |
|
"dead_code_fraction": 1.0, |
|
"dead_code_fraction/layer0": 1.0, |
|
"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, |
|
"input_norm": 0.0, |
|
"input_norm/layer0": 0.0, |
|
"loss": 2.325971780395508, |
|
"max_norm": 76.40007019042969, |
|
"max_norm/layer0": 76.40007019042969, |
|
"mean_norm": 45.70543670654297, |
|
"mean_norm/layer0": 45.70543670654297, |
|
"multicode_k": 1, |
|
"output_norm": 0.0, |
|
"output_norm/layer0": 0.0, |
|
"perplexity": 8.439263086993584, |
|
"runtime": 15639.0026, |
|
"samples_per_second": 61.385, |
|
"steps_per_second": 0.639 |
|
} |