def learn_model_fn(model, timesteps, save, period):
return learn(model,
model_name=model_name,
model_path=model_path,
timesteps=timesteps,
save=save,
period=period)
def learn_model_fn(model, timesteps, save, period):
save_thresh = 19.2 if args.selfplay else None
return learn(model,
model_name=model_name,
model_path=model_path,
timesteps=timesteps,
save=save,
period=period,
save_thresh=save_thresh)
optim_param=[alpha],
loss_function=nn.MSELoss(),
tau=1,
device=device)
buffer = QBuffer(memory_size, batch_size, device)
learning_policy = EpsDecay(eps_start, eps_min, eps_decay, env.action_space.n)
playing_policy = Greedy()
agent = Agent(model=model,
buffer=buffer,
learn_every=4,
update_every=4,
policy_learning=learning_policy,
policy_playing=playing_policy)
scores = util.learn(env, goal_size, average_goal, agent, max_step, nb_epi_max,
gamma, learning_policy)
print(len(buffer))
fig = plt.figure()
ax = fig.add_subplot(111)
plt.plot(np.arange(len(scores)), scores)
plt.ylabel('Score')
plt.xlabel('Episode #')
plt.show()
for i in range(10):
state = env.reset()
score = 0
env.render()
for j in range(max_step):
action = agent.act(state)