# Error about type **URL:** \ **Category:** Uncategorized\ **Created:** [June 30, 2020, 7:49am UTC](https://discuss.pytorch.org/t/error-about-type/87409 "2020-06-30T07:49:28Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![111348](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/111348/32/25811_2.png) [@111348](https://discuss.pytorch.org/u/111348)\ **Post date:** [June 30, 2020, 7:49am UTC](https://discuss.pytorch.org/t/error-about-type/87409/1 "2020-06-30T07:49:28Z")
import gym import random import collections import numpy as np import torch import torch.nn as nn import torch.nn.functional as F import torch.optim as optim #Hyperparameters lr\_mu = 0.0005 # learning rate of Actor lr\_q = 0.001 # learning rate of Critic gamma = 0.99 batch\_size = 32 buffer\_limit = 50000 tau = 0.005 # for target network soft update class ReplayBuffer(): ``` def __init__ (self): self.buffer = collections.deque(maxlen=buffer_limit) def put(self, data): self.buffer.append(data) def sample(self, n): mini_batch = random.sample(self.buffer, n) s_lst, a_lst, r_lst, s_prime_lst, done_mask_lst = [], [], [], [], [] for transition in mini_batch: s, a, r, s_prime, done_mask = transition s_lst.append(s) a_lst.append([a]) r_lst.append([r]) s_prime_lst.append(s_prime) done_mask_lst.append([done_mask]) return torch.tensor(s_lst, dtype=torch.float), torch.tensor(a_lst), \ torch.tensor(r_lst), torch.tensor(s_prime_lst, dtype=torch.float), \ torch.tensor(done_mask_lst) def size(self): return len(self.buffer) ``` class MuNet(nn.Module): # ``` def __init__ (self): super(MuNet, self). __init__ () self.fc1 = nn.Linear(3, 128) self.fc2 = nn.Linear(128, 64) self.fc_mu = nn.Linear(64, 1) def forward(self, x): x = F.relu(self.fc1(x)) x = F.relu(self.fc2(x)) mu = torch.tanh(self.fc_mu(x))*2 # Multipled by 2 because the action space of the Pendulum-v0 is [-2,2] return mu ``` class QNet(nn.Module): #score(state, action) # enters the network in middle layer ``` def __init__ (self): super(QNet, self). __init__ () self.fc_s = nn.Linear(3, 64) #encodes the state self.fc_a = nn.Linear(1,64) #encodes the actions self.fc_q = nn.Linear(128, 32) #s+a = 128 self.fc_3 = nn.Linear(32,1) # output def forward(self, x, a): h1 = F.relu(self.fc_s(x)) h2 = F.relu(self.fc_a(a)) cat = torch.cat([h1,h2], dim=1) q = F.relu(self.fc_q(cat)) q = self.fc_3(q) return q ``` class OrnsteinUhlenbeckNoise: ``` def __init__ (self, mu): self.theta, self.dt, self.sigma = 0.1, 0.01, 0.1 self.mu = mu self.x_prev = np.zeros_like(self.mu) def __call__ (self): x = self.x_prev + self.theta * (self.mu - self.x_prev) * self.dt + \ self.sigma * np.sqrt(self.dt) * np.random.normal(size=self.mu.shape) self.x_prev = x return x ``` def train(mu, mu\_target, q, q\_target, memory, q\_optimizer, mu\_optimizer): ``` s,a,r,s_prime,done_mask = memory.sample(batch_size) target = r + gamma * q_target(s_prime, mu_target(s_prime)) q_loss = F.smooth_l1_loss(q(s,a), target.detach()) q_optimizer.zero_grad() q_loss.backward() q_optimizer.step() mu_loss = -q(s,mu(s)).mean() # That's all for the policy loss. mu_optimizer.zero_grad() mu_loss.backward() mu_optimizer.step() ``` def soft\_update(net, net\_target): ``` for param_target, param in zip(net_target.parameters(), net.parameters()): param_target.data.copy_(param_target.data * (1.0 - tau) + param.data * tau) ``` def main(): env = gym.make(‘Pendulum-v0’) memory = ReplayBuffer() q, q\_target = QNet(), QNet() q\_target.load\_state\_dict(q.state\_dict()) mu, mu\_target = MuNet(), MuNet() mu\_target.load\_state\_dict(mu.state\_dict()) score = 0.0 print\_interval = 20 min\_play\_reward = 0 iterations = 100000 mu\_optimizer = optim.Adam(mu.parameters(), lr=lr\_mu) q\_optimizer = optim.Adam(q.parameters(), lr=lr\_q) ou\_noise = OrnsteinUhlenbeckNoise(mu=np.zeros(1)) def play\_game(): ``` done = False state = env.reset() while(not done): a = mu(torch.from_numpy(s).float()) a = a.item() + ou_noise()[0] s_prime, r, done, info = env.step([a]) env.render() state = s_prime ``` for iteration in range(iterations): ``` s = env.reset() for t in range(300): # maximum length of episode is 200 for Pendulum-v0 a = mu(torch.from_numpy(s).float()) a = a.item() + ou_noise()[0] s_prime, r, done, info = env.step([a]) memory.put((s,a,r/100.0,s_prime,done)) score +=r s = s_prime if done: if score/print_interval > min_play_reward: play_game() break if memory.size()>2000: train(mu, mu_target, q, q_target, memory, q_optimizer, mu_optimizer) soft_update(mu, mu_target) soft_update(q, q_target) if iteration%print_interval==0 and iteration!=0: print("# of episode :{}, avg score : {:.1f}".format(iteration, score/print_interval)) score = 0.0 ``` env.close() if **name** == ‘ **main** ’: ``` main() ``` This is Pendelum project code. It has some error about type. Could you let me know how can i solve? RuntimeError Traceback (most recent call last) in () 150 151 if memory.size()\>2000: –\> 152 train(mu, mu\_target, q, q\_target, memory, q\_optimizer, mu\_optimizer) 153 soft\_update(mu, mu\_target) 154 soft\_update(q, q\_target) 5 frames in train(mu, mu\_target, q, q\_target, memory, q\_optimizer, mu\_optimizer) 86 87 target = r + gamma \* q\_target(s\_prime, mu\_target(s\_prime)) —\> 88 q\_loss = F.smooth\_l1\_loss(q(s,a), target.detach()) 89 90 q\_optimizer.zero\_grad() /usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py in **call** (self, \*input, \*\*kwargs) 548 result = self.\_slow\_forward(\*input, \*\*kwargs) 549 else: –\> 550 result = self.forward(\*input, \*\*kwargs) 551 for hook in self.\_forward\_hooks.values(): 552 hook\_result = hook(self, input, result) in forward(self, x, a) 65 def forward(self, x, a): 66 h1 = F.relu(self.fc\_s(x)) —\> 67 h2 = F.relu(self.fc\_a(a)) 68 cat = torch.cat([h1,h2], dim=1) 69 q = F.relu(self.fc\_q(cat)) /usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py in **call** (self, \*input, \*\*kwargs) 548 result = self.\_slow\_forward(\*input, \*\*kwargs) 549 else: –\> 550 result = self.forward(\*input, \*\*kwargs) 551 for hook in self.\_forward\_hooks.values(): 552 hook\_result = hook(self, input, result) /usr/local/lib/python3.6/dist-packages/torch/nn/modules/linear.py in forward(self, input) 85 86 def forward(self, input): —\> 87 return F.linear(input, self.weight, self.bias) 88 89 def extra\_repr(self): /usr/local/lib/python3.6/dist-packages/torch/nn/functional.py in linear(input, weight, bias) 1608 if input.dim() == 2 and bias is not None: 1609 # fused op is marginally faster -\> 1610 ret = torch.addmm(bias, input, weight.t()) 1611 else: 1612 output = input.matmul(weight.t()) RuntimeError: Expected object of scalar type Float but got scalar type Double for argument #2 ‘mat1’ in call to \_th\_addmm above sentence ---
**Author:** ![shlomish3](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/s/41988e/32.png) [@shlomish3](https://discuss.pytorch.org/u/shlomish3)\ **Post date:** [June 30, 2020, 9:19am UTC](https://discuss.pytorch.org/t/error-about-type/87409/2 "2020-06-30T09:19:02Z")
Apperently you need to type case (or initialize) one of your variables to another, shorter data type. As the error indicates, try casting it [torch.float32](https://pytorch.org/docs/stable/tensors.html): ![image](https://discuss.pytorch.org/uploads/default/original/3X/3/d/3da688298305c1d719fc91e732cc1335dc424ef0.png) It’s a bit difficult to follow your code example, so I can’t really tell what variable causes the problem.