TetrisRL/agent.py

96 lines
2.7 KiB
Python

import torch
from torch.optim import Adam
from torch.distributions import Categorical
from zmq import device
from enviroment import TetrisEnviroment
from network import TetrisRLModel
import torch.nn.functional as F
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print("use:", device)
PATH = "model.h5"
class TetrisRLAgent():
def __init__(self) -> None:
self.model = TetrisRLModel()
self.model = self.model.to(device)
self.optim = Adam(self.model.parameters(), lr=0.001)
def sample(self, observation):
action_prob = self.model(observation)
action_dist = Categorical(action_prob)
action = action_dist.sample()
log_prob = action_dist.log_prob(action)
return action, log_prob
def learn(self, rewards, log_probes):
rewards = rewards.to(device)
loss = ((-log_probes * rewards)).sum()
self.optim.zero_grad()
loss.backward()
self.optim.step()
def save(self, PATH): # You should not revise this
Agent_Dict = {
"network" : self.network.state_dict(),
"optimizer" : self.optimizer.state_dict()
}
torch.save(Agent_Dict, PATH)
def showGame(views:list, score:int) -> None:
for i in range(20):
print(str(i).rjust(2), end=' ')
for j in range(10):
if views[i][j]:
print('', end='')
else:
print('', end='')
print()
print("Score:", score)
print()
env = TetrisEnviroment()
agent = TetrisRLAgent()
avgTotalRewards = []
for batch in range(100000):
log_probs, rewards = [], []
total_rewards = []
for episode in range(5):
total_reward = 0
pixel, reward, done, info = env.reset()
while 1:
showGame(pixel, total_reward)
pixel = torch.tensor(pixel, dtype=torch.float32)
blockType = F.one_hot(torch.tensor(info[2]), 7)
blockLoc = torch.tensor(info[:2], dtype=torch.float32)
observation = torch.cat([pixel.reshape(-1), blockLoc, blockType], dim=0)
observation = observation.to(device)
action, log_prob = agent.sample(observation)
pixel, reward, done, info = env.step(action)
rewards.append(reward)
log_probs.append(log_prob)
total_reward += reward
if done:
total_rewards.append(total_reward)
break
avgTotalReward = sum(total_rewards) / len(total_rewards)
avgTotalRewards.append(avgTotalReward)
rewards = torch.tensor(rewards)
log_probs = torch.stack(log_probs)
print(rewards.shape)
print(log_probs.shape)
agent.learn(rewards, log_probs)
agent.save(PATH)