Play bipedal walker

play_bipedal.py

import numpy as np
from tinynet import RNN1L
import gym

pre_trained_weights = np.array(
       [[-6.42131474e-02,  1.18558117e+00, -8.65697398e-02,
        -7.25508614e-01, -1.64208890e+00,  2.36031412e-01,
        -2.30686547e-01,  1.01184284e-02,  9.14193609e-02,
         4.09360317e-01,  5.28414810e-01,  3.01208252e+00,
        -9.78603700e-02,  3.53038690e-02,  8.94094020e-01,
         2.90466615e-01, -1.79276182e-01, -2.17272103e-02,
        -2.13970051e-01, -1.67635995e+00, -6.02311909e-02,
        -1.96442401e+00,  1.19746455e-01,  1.79969903e-01,
         1.05914865e+00, -6.51820003e-01,  1.37623963e-01,
        -2.71994849e-01,  6.59062949e-01],
       [ 2.02307561e+00,  1.20449209e+00,  6.05621780e-01,
         6.98229811e-01,  2.14649694e+00, -5.37398203e-01,
        -3.97149533e+00, -2.12493526e-01, -4.23998290e-02,
         9.53130511e-01,  4.16741016e-01, -1.54060458e+00,
         5.21017760e-01, -2.45511887e-01,  4.14342494e-01,
         9.14911539e-01, -1.23614789e+00,  7.33691074e-01,
         5.45596174e-01, -4.17843212e-02, -8.70911009e-02,
         9.51670845e-01,  3.85517330e-02,  8.25852210e-03,
        -1.90154830e+00, -8.03895467e-01, -4.54610968e-01,
         5.96982247e-02, -4.74628318e-01],
       [ 1.37118663e+00,  1.27937587e+00, -1.71887878e+00,
        -1.94443529e+00,  8.90548993e-01, -7.23938737e-01,
         1.79740914e-01, -5.48218269e-04,  6.50140805e-02,
        -1.47901011e+00, -7.39203613e-01, -4.17242880e-01,
        -4.00900026e-01,  2.04101041e-01,  5.31842291e-01,
        -1.53630768e+00,  3.99904773e-01, -7.84735830e-01,
        -1.62958331e+00, -5.39509441e-01,  1.39857350e+00,
        -7.35478300e-01,  3.25341560e+00,  4.71190751e-01,
        -5.17805666e-03, -1.22670637e+00, -2.69994440e-01,
        -5.59723682e-01,  3.75500268e-01],
       [ 3.47982673e-01,  5.91370880e-01,  2.42805457e+00,
         8.28226086e-01, -1.08526012e+00,  1.99388318e-01,
        -3.05907307e-01,  4.35067255e-02, -2.76692831e-02,
         1.59416946e-01, -2.39363737e-01, -1.70776626e+00,
        -2.36775784e-01, -4.18781827e-02, -9.31394093e-01,
        -7.61417665e-01, -2.36382980e-01, -1.27730737e-01,
        -5.48062621e-01,  5.19030962e-01, -3.08506491e-01,
        -3.01197781e-01, -8.71729552e-02,  2.38021638e-01,
         2.54997762e-01, -4.54133943e-01, -3.76671113e-02,
        -9.02776638e-02,  3.64898985e-02]])

# Environment setup
env = gym.make("BipedalWalker-v2")
# env = gym.wrappers.Monitor(env, 'video', force = True) # Uncomment to save video
nactions = env.action_space.shape[0]
ninputs = env.reset().size

# Network setup
net = RNN1L(ninputs, nactions)
net.set_weights(pre_trained_weights)

# Gameplay loop
obs = env.reset()
score = 0
done = False
while not done:
    env.render()
    action = net.activate(obs)
    obs, rew, done, info = env.step(action)
    score += rew
print(f"Fitness: {score}")
env.close()

# import IPython; IPython.embed() # Uncomment to drop into iPython for analysis