train_results.json

{
    "MSE": 0.0,
    "MSE/layer0": 0.0,
    "dead_code_fraction": 1.0,
    "dead_code_fraction/layer0": 1.0,
    "epoch": 2.04,
    "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,
    "runtime": 15639.0026,
    "samples_per_second": 61.385,
    "steps_per_second": 0.639
}