evaluate_trained_models.py

import glob
import pickle
import matplotlib.pyplot as plt


# Find all date_times.
all_paths = glob.glob("models/*")
date_times = []
for path in all_paths:
    split = path.split("/")[-1].split("-")
    date = split[0]
    time = split[1]
    date_time = date + "-" + time
    date_times.append(date_time)
date_times = sorted(list(set(date_times)))

# Process histories of date-times.
def plot_date_times(date_times, start_index):
    for date_time in date_times:

        # Load all histories for date-time.
        history_paths = [path for path in all_paths if "history" in path and date_time in path]
        histories = []
        for history_path in history_paths:
            history = pickle.load(open(history_path, "rb"))
            histories.append(history)

        # Plot the histories.
        for history, history_path in zip(histories, history_paths):
            split = history_path.split("/")[-1].split("-")
            for key in history.keys():
                plt.plot(history[key][start_index:], label=key + " " + split[2] + " " + date_time)
    plt.legend()
    plt.show()
    plt.close()

plot_date_times(date_times, start_index=0)