这是indexloc提供的服务，不要输入任何密码
Skip to content

hotfix：fix test failure in cuda environment #289

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

Merged
merged 5 commits into from
Feb 9, 2021
Merged

hotfix：fix test failure in cuda environment #289

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
c37bf24
fix cuda bug
ChenDRAG Feb 9, 2021
9656a1e
another fix
ChenDRAG Feb 9, 2021
15bd6e7
Apply suggestions from code review
Trinkle23897 Feb 9, 2021
38496cf
last fix
ChenDRAG Feb 9, 2021
0575136
Merge branch 'cudafix' of github.com:ChenDRAG/tianshou into cudafix
ChenDRAG Feb 9, 2021
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
4 changes: 2 additions & 2 deletions examples/atari/runnable/pong_a2c.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,8 +65,8 @@ def test_a2c(args=get_args()):
# model
net = Net(args.state_shape, hidden_sizes=args.hidden_sizes,
device=args.device)
actor = Actor(net, args.action_shape).to(args.device)
critic = Critic(net).to(args.device)
actor = Actor(net, args.action_shape, device=args.device).to(args.device)
critic = Critic(net, device=args.device).to(args.device)
optim = torch.optim.Adam(set(
actor.parameters()).union(critic.parameters()), lr=args.lr)
dist = torch.distributions.Categorical
Expand Down
4 changes: 2 additions & 2 deletions examples/atari/runnable/pong_ppo.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,8 +65,8 @@ def test_ppo(args=get_args()):
# model
net = Net(args.state_shape, hidden_sizes=args.hidden_sizes,
device=args.device)
actor = Actor(net, args.action_shape).to(args.device)
critic = Critic(net).to(args.device)
actor = Actor(net, args.action_shape, device=args.device).to(args.device)
critic = Critic(net, device=args.device).to(args.device)
optim = torch.optim.Adam(set(
actor.parameters()).union(critic.parameters()), lr=args.lr)
dist = torch.distributions.Categorical
Expand Down
6 changes: 3 additions & 3 deletions test/discrete/test_a2c_with_il.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,8 +68,8 @@ def test_a2c_with_il(args=get_args()):
# model
net = Net(args.state_shape, hidden_sizes=args.hidden_sizes,
device=args.device)
actor = Actor(net, args.action_shape).to(args.device)
critic = Critic(net).to(args.device)
actor = Actor(net, args.action_shape, device=args.device).to(args.device)
critic = Critic(net, device=args.device).to(args.device)
optim = torch.optim.Adam(set(
actor.parameters()).union(critic.parameters()), lr=args.lr)
dist = torch.distributions.Categorical
Expand Down Expand Up @@ -113,7 +113,7 @@ def stop_fn(mean_rewards):
env.spec.reward_threshold = 190 # lower the goal
net = Net(args.state_shape, hidden_sizes=args.hidden_sizes,
device=args.device)
net = Actor(net, args.action_shape).to(args.device)
net = Actor(net, args.action_shape, device=args.device).to(args.device)
optim = torch.optim.Adam(net.parameters(), lr=args.il_lr)
il_policy = ImitationPolicy(net, optim, mode='discrete')
il_test_collector = Collector(
Expand Down
4 changes: 2 additions & 2 deletions test/discrete/test_ppo.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,8 +68,8 @@ def test_ppo(args=get_args()):
# model
net = Net(args.state_shape, hidden_sizes=args.hidden_sizes,
device=args.device)
actor = Actor(net, args.action_shape).to(args.device)
critic = Critic(net).to(args.device)
actor = Actor(net, args.action_shape, device=args.device).to(args.device)
critic = Critic(net, device=args.device).to(args.device)
# orthogonal initialization
for m in list(actor.modules()) + list(critic.modules()):
if isinstance(m, torch.nn.Linear):
Expand Down
9 changes: 6 additions & 3 deletions test/discrete/test_sac.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,15 +62,18 @@ def test_discrete_sac(args=get_args()):
# model
net = Net(args.state_shape, hidden_sizes=args.hidden_sizes,
device=args.device)
actor = Actor(net, args.action_shape, softmax_output=False).to(args.device)
actor = Actor(net, args.action_shape,
softmax_output=False, device=args.device).to(args.device)
actor_optim = torch.optim.Adam(actor.parameters(), lr=args.actor_lr)
net_c1 = Net(args.state_shape, hidden_sizes=args.hidden_sizes,
device=args.device)
critic1 = Critic(net_c1, last_size=args.action_shape).to(args.device)
critic1 = Critic(net_c1, last_size=args.action_shape,
device=args.device).to(args.device)
critic1_optim = torch.optim.Adam(critic1.parameters(), lr=args.critic_lr)
net_c2 = Net(args.state_shape, hidden_sizes=args.hidden_sizes,
device=args.device)
critic2 = Critic(net_c2, last_size=args.action_shape).to(args.device)
critic2 = Critic(net_c2, last_size=args.action_shape,
device=args.device).to(args.device)
critic2_optim = torch.optim.Adam(critic2.parameters(), lr=args.critic_lr)

# better not to use auto alpha in CartPole
Expand Down
11 changes: 7 additions & 4 deletions tianshou/utils/net/continuous.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,7 +49,8 @@ def __init__(
self.output_dim = np.prod(action_shape)
input_dim = getattr(preprocess_net, "output_dim",
preprocess_net_output_dim)
self.last = MLP(input_dim, self.output_dim, hidden_sizes)
self.last = MLP(input_dim, self.output_dim,
hidden_sizes, device=self.device)
self._max = max_action

def forward(
Expand Down Expand Up @@ -98,7 +99,7 @@ def __init__(
self.output_dim = 1
input_dim = getattr(preprocess_net, "output_dim",
preprocess_net_output_dim)
self.last = MLP(input_dim, 1, hidden_sizes)
self.last = MLP(input_dim, 1, hidden_sizes, device=self.device)

def forward(
self,
Expand Down Expand Up @@ -164,10 +165,12 @@ def __init__(
self.output_dim = np.prod(action_shape)
input_dim = getattr(preprocess_net, "output_dim",
preprocess_net_output_dim)
self.mu = MLP(input_dim, self.output_dim, hidden_sizes)
self.mu = MLP(input_dim, self.output_dim,
hidden_sizes, device=self.device)
self._c_sigma = conditioned_sigma
if conditioned_sigma:
self.sigma = MLP(input_dim, self.output_dim, hidden_sizes)
self.sigma = MLP(input_dim, self.output_dim,
hidden_sizes, device=self.device)
else:
self.sigma_param = nn.Parameter(torch.zeros(self.output_dim, 1))
self._max = max_action
Expand Down
10 changes: 8 additions & 2 deletions tianshou/utils/net/discrete.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,13 +40,16 @@ def __init__(
hidden_sizes: Sequence[int] = (),
softmax_output: bool = True,
preprocess_net_output_dim: Optional[int] = None,
device: Union[str, int, torch.device] = "cpu",
) -> None:
super().__init__()
self.device = device
self.preprocess = preprocess_net
self.output_dim = np.prod(action_shape)
input_dim = getattr(preprocess_net, "output_dim",
preprocess_net_output_dim)
self.last = MLP(input_dim, self.output_dim, hidden_sizes)
self.last = MLP(input_dim, self.output_dim,
hidden_sizes, device=self.device)
self.softmax_output = softmax_output

def forward(
Expand Down Expand Up @@ -91,13 +94,16 @@ def __init__(
hidden_sizes: Sequence[int] = (),
last_size: int = 1,
preprocess_net_output_dim: Optional[int] = None,
device: Union[str, int, torch.device] = "cpu",
) -> None:
super().__init__()
self.device = device
self.preprocess = preprocess_net
self.output_dim = last_size
input_dim = getattr(preprocess_net, "output_dim",
preprocess_net_output_dim)
self.last = MLP(input_dim, last_size, hidden_sizes)
self.last = MLP(input_dim, last_size,
hidden_sizes, device=self.device)

def forward(
self, s: Union[np.ndarray, torch.Tensor], **kwargs: Any
Expand Down