Skip to content

修改关于DQN 的代码 #1

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
xieyj17 wants to merge 7 commits into
base: master
Choose a base branch
from
Open

修改关于DQN 的代码 #1

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
7cd5aa3
fix codes for dqn
Feb 24, 2023
16a5336
fix typo
xieyj17 Feb 24, 2023
c5a5174
dueling network
xieyj17 Feb 27, 2023
46830ea
Update 04_dqn.py
xieyj17 Mar 30, 2023
c937acb
Update 06_doubledqn.py
xieyj17 Mar 30, 2023
4495b6c
Update 06_dueling_network.py
xieyj17 Mar 30, 2023
e9a3880
Update 06_target_network.py
xieyj17 Mar 30, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 9 additions & 22 deletions 04_dqn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,8 +36,6 @@ class DQN:
def __init__(self, dim_state=None, num_action=None, discount=0.9):
self.discount = discount
self.Q = QNet(dim_state, num_action)
self.target_Q = QNet(dim_state, num_action)
self.target_Q.load_state_dict(self.Q.state_dict())

def get_action(self, state):
qvals = self.Q(state)
Expand All @@ -46,21 +44,13 @@ def get_action(self, state):
def compute_loss(self, s_batch, a_batch, r_batch, d_batch, next_s_batch):
# 计算s_batch，a_batch对应的值。
qvals = self.Q(s_batch).gather(1, a_batch.unsqueeze(1)).squeeze()
# 使用target Q网络计算next_s_batch对应的值。
next_qvals, _ = self.target_Q(next_s_batch).detach().max(dim=1)
# 使用原始的 Q网络计算next_s_batch对应的值。
next_qvals, _ = self.Q(next_s_batch).detach().max(dim=1)
# 使用MSE计算loss。
loss = F.mse_loss(r_batch + self.discount * next_qvals * (1 - d_batch), qvals)
return loss


def soft_update(target, source, tau=0.01):
"""
update target by target = tau * source + (1 - tau) * target.
"""
for target_param, param in zip(target.parameters(), source.parameters()):
target_param.data.copy_(target_param.data * (1.0 - tau) + param.data * tau)


@dataclass
class ReplayBuffer:
maxsize: int
Expand Down Expand Up @@ -126,7 +116,7 @@ def train(args, env, agent):
if np.random.rand() < epsilon or i < args.warmup_steps:
action = env.action_space.sample()
else:
action = agent.get_action(torch.from_numpy(state))
action = agent.get_action(torch.from_numpy(state).to(args.device))
action = action.item()
next_state, reward, terminated, truncated, _ = env.step(action)
done = terminated or truncated
Expand Down Expand Up @@ -155,11 +145,11 @@ def train(args, env, agent):

if i > args.warmup_steps:
bs, ba, br, bd, bns = replay_buffer.sample(n=args.batch_size)
bs = torch.tensor(bs, dtype=torch.float32)
ba = torch.tensor(ba, dtype=torch.long)
br = torch.tensor(br, dtype=torch.float32)
bd = torch.tensor(bd, dtype=torch.float32)
bns = torch.tensor(bns, dtype=torch.float32)
bs = torch.tensor(bs, dtype=torch.float32).to(args.device)
ba = torch.tensor(ba, dtype=torch.long).to(args.device)
br = torch.tensor(br, dtype=torch.float32).to(args.device)
bd = torch.tensor(bd, dtype=torch.float32).to(args.device)
bns = torch.tensor(bns, dtype=torch.float32).to(args.device)

loss = agent.compute_loss(bs, ba, br, bd, bns)
loss.backward()
Expand All @@ -168,8 +158,6 @@ def train(args, env, agent):

log["loss"].append(loss.item())

soft_update(agent.target_Q, agent.Q)

# 3. 画图。
plt.plot(log["loss"])
plt.yscale("log")
Expand All @@ -191,7 +179,7 @@ def eval(args, env, agent):
state, _ = env.reset()
for i in range(5000):
episode_length += 1
action = agent.get_action(torch.from_numpy(state)).item()
action = agent.get_action(torch.from_numpy(state).to(args.device)).item()
next_state, reward, terminated, truncated, _ = env.step(action)
done = terminated or truncated
env.render()
Expand Down Expand Up @@ -229,7 +217,6 @@ def main():
set_seed(args)
agent = DQN(dim_state=args.dim_state, num_action=args.num_action, discount=args.discount)
agent.Q.to(args.device)
agent.target_Q.to(args.device)

if args.do_train:
train(args, env, agent)
Expand Down
51 changes: 27 additions & 24 deletions 06_doubledqn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,19 +32,21 @@ def forward(self, obs):
class DoubleDQN:
def __init__(self, dim_obs=None, num_act=None, discount=0.9):
self.discount = discount
self.model = QNet(dim_obs, num_act)
self.target_model = QNet(dim_obs, num_act)
self.target_model.load_state_dict(self.model.state_dict())
self.Q = QNet(dim_obs, num_act)
self.target_Q = QNet(dim_obs, num_act)
self.target_Q.load_state_dict(self.Q.state_dict())

def get_action(self, obs):
qvals = self.model(obs)
qvals = self.Q(obs)
return qvals.argmax()

def compute_loss(self, s_batch, a_batch, r_batch, d_batch, next_s_batch):
# Compute current Q value based on current states and actions.
qvals = self.model(s_batch).gather(1, a_batch.unsqueeze(1)).squeeze()
qvals = self.Q(s_batch).gather(1, a_batch.unsqueeze(1)).squeeze()
# 选择行动
next_a_batch = self.Q(next_s_batch).argmax(dim=1)
# next state的value不参与导数计算，避免不收敛。
next_qvals, _ = self.target_model(next_s_batch).detach().max(dim=1)
next_qvals = self.target_Q(next_s_batch).gather(1, next_a_batch.unsqueeze(1)).squeeze().detach()
loss = F.mse_loss(r_batch + self.discount * next_qvals * (1 - d_batch), qvals)
return loss

Expand Down Expand Up @@ -96,7 +98,7 @@ def set_seed(args):

def train(args, env, agent):
replay_buffer = ReplayBuffer(100_000)
optimizer = torch.optim.Adam(agent.model.parameters(), lr=args.lr)
optimizer = torch.optim.Adam(agent.Q.parameters(), lr=args.lr)
optimizer.zero_grad()

epsilon = 1
Expand All @@ -107,16 +109,16 @@ def train(args, env, agent):
log_ep_rewards = []
log_losses = [0]

agent.model.train()
agent.target_model.train()
agent.model.zero_grad()
agent.target_model.zero_grad()
agent.Q.train()
agent.target_Q.train()
agent.Q.zero_grad()
agent.target_Q.zero_grad()
state, _ = env.reset()
for i in range(args.max_steps):
if np.random.rand() < epsilon or i < args.warmup_steps:
action = env.action_space.sample()
else:
action = agent.get_action(torch.from_numpy(state))
action = agent.get_action(torch.from_numpy(state).to(args.device))
action = action.item()
next_state, reward, terminated, truncated, _ = env.step(action)
done = terminated or truncated
Expand All @@ -140,8 +142,8 @@ def train(args, env, agent):
print(f"i={i}, reward={episode_reward:.0f}, length={episode_length}, max_reward={max_episode_reward}, loss={log_losses[-1]:.1e}, epsilon={epsilon:.3f}")

if episode_length < 180 and episode_reward > max_episode_reward:
save_path = os.path.join(args.output_dir, "model.bin")
torch.save(agent.model.state_dict(), save_path)
save_path = os.path.join(args.output_dir, "Q.bin")
torch.save(agent.Q.state_dict(), save_path)
max_episode_reward = episode_reward

episode_reward = 0
Expand All @@ -150,11 +152,11 @@ def train(args, env, agent):

if i > args.warmup_steps:
bs, ba, br, bd, bns = replay_buffer.sample(n=args.batch_size)
bs = torch.tensor(bs, dtype=torch.float32)
ba = torch.tensor(ba, dtype=torch.long)
br = torch.tensor(br, dtype=torch.float32)
bd = torch.tensor(bd, dtype=torch.float32)
bns = torch.tensor(bns, dtype=torch.float32)
bs = torch.tensor(bs, dtype=torch.float32).to(args.device)
ba = torch.tensor(ba, dtype=torch.long).to(args.device)
br = torch.tensor(br, dtype=torch.float32).to(args.device)
bd = torch.tensor(bd, dtype=torch.float32).to(args.device)
bns = torch.tensor(bns, dtype=torch.float32).to(args.device)

loss = agent.compute_loss(bs, ba, br, bd, bns)
loss.backward()
Expand All @@ -164,7 +166,7 @@ def train(args, env, agent):
log_losses.append(loss.item())

# 更新目标网络。
for target_param, param in zip(agent.target_model.parameters(), agent.model.parameters()):
for target_param, param in zip(agent.target_Q.parameters(), agent.Q.parameters()):
target_param.data.copy_(args.lr_target * param.data + (1 - args.lr_target) * target_param.data)

plt.plot(log_losses)
Expand All @@ -179,16 +181,16 @@ def train(args, env, agent):

def eval(args, env, agent):
model_path = os.path.join(args.output_dir, "model.bin")
agent.model.load_state_dict(torch.load(model_path))
agent.Q.load_state_dict(torch.load(model_path))

episode_length = 0
episode_reward = 0

agent.model.eval()
agent.Q.eval()
state, _ = env.reset()
for i in range(5000):
episode_length += 1
action = agent.get_action(torch.from_numpy(state)).item()
action = agent.get_action(torch.from_numpy(state).to(args.device)).item()
next_state, reward, terminated, truncated, _ = env.step(action)
done = terminated or truncated
episode_reward += reward
Expand Down Expand Up @@ -225,7 +227,8 @@ def main():
env = gym.make(args.env)
set_seed(args)
agent = DoubleDQN(dim_obs=args.dim_obs, num_act=args.num_act, discount=args.discount)
agent.model.to(args.device)
agent.Q.to(args.device)
agent.target_Q.to(args.device)

if args.do_train:
train(args, env, agent)
Expand Down
Loading