train_results.json

{
    "MSE": 0.0,
    "MSE/layer0": 0.0,
    "dead_code_fraction": 1.0,
    "dead_code_fraction/layer0": 1.0,
    "epoch": 1.04,
    "input_norm": 0.0,
    "input_norm/layer0": 0.0,
    "loss": 2.0762174885749816,
    "max_norm": 79.63946533203125,
    "max_norm/layer0": 79.63946533203125,
    "mean_norm": 68.03947448730469,
    "mean_norm/layer0": 68.03947448730469,
    "multicode_k": 1,
    "output_norm": 0.0,
    "output_norm/layer0": 0.0,
    "runtime": 12054.7701,
    "samples_per_second": 39.818,
    "steps_per_second": 0.83,
    "train_samples": 459760
}