def load_logs()

in plot_log.py [0:0]

• 46 lines of code
• 7 McCabe index (conditional complexity)

def load_logs(experiment_directory, type):

    logs = torch.load(os.path.join(experiment_directory, ws.logs_filename))

    logging.info("latest epoch is {}".format(logs["epoch"]))

    num_iters = len(logs["loss"])
    iters_per_epoch = num_iters / logs["epoch"]

    logging.info("{} iters per epoch".format(iters_per_epoch))

    smoothed_loss_41 = running_mean(logs["loss"], 41)
    smoothed_loss_1601 = running_mean(logs["loss"], 1601)

    fig, ax = plt.subplots()

    if type == "loss":

        ax.plot(
            np.arange(num_iters) / iters_per_epoch,
            logs["loss"],
            "#82c6eb",
            np.arange(20, num_iters - 20) / iters_per_epoch,
            smoothed_loss_41,
            "#2a9edd",
            np.arange(800, num_iters - 800) / iters_per_epoch,
            smoothed_loss_1601,
            "#16628b",
        )

        ax.set(xlabel="Epoch", ylabel="Loss", title="Training Loss")

    elif type == "learning_rate":
        combined_lrs = np.array(logs["learning_rate"])

        ax.plot(
            np.arange(combined_lrs.shape[0]),
            combined_lrs[:, 0],
            np.arange(combined_lrs.shape[0]),
            combined_lrs[:, 1],
        )
        ax.set(xlabel="Epoch", ylabel="Learning Rate", title="Learning Rates")

    elif type == "time":
        ax.plot(logs["timing"], "#833eb7")
        ax.set(xlabel="Epoch", ylabel="Time per Epoch (s)", title="Timing")

    elif type == "lat_mag":
        ax.plot(logs["latent_magnitude"])
        ax.set(xlabel="Epoch", ylabel="Magnitude", title="Latent Vector Magnitude")

    elif type == "param_mag":
        for _name, mags in logs["param_magnitude"].items():
            ax.plot(mags)
        ax.set(xlabel="Epoch", ylabel="Magnitude", title="Parameter Magnitude")
        ax.legend(logs["param_magnitude"].keys())

    else:
        raise Exception('unrecognized plot type "{}"'.format(type))

    ax.grid()
    plt.show()