Help me... How to resolve error due to tensor size ? Plz

import gym
import collections
import random

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim

learning_rate = 0.0005
gamma = 0.98
buffer_limit = 50000
batch_size = 32

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 Qnet(nn.Module):
def init(self):
super(Qnet, self).init()
self.fc1 = nn.Linear(2, 128)
self.fc2 = nn.Linear(128, 128)
self.fc3 = nn.Linear(128, 1)

def forward(self, x):
    x = F.relu(self.fc1(x)) 
    x = F.relu(self.fc2(x)) 
    x = self.fc3(x)

    return x
def sample_action(self, obs, epsilon):
    out = self.forward(obs)
    coin = random.random()
    if coin < epsilon:
        return random.randint(0,1)
    else : 
        return out.argmax().item()

def train(q, q_target, memory, optimizer):
for i in range(10):=
s,a,r,s_prime,done_mask = memory.sample(batch_size)
q_out = q(s)
q_a = q_out.gather(1,a)
max_q_prime = q_target(s_prime).max(1)[0].unsqueeze(1)
target = r + gamma * max_q_prime * done_mask
loss = F.smooth_l1_loss(q_a, target)
optimizer.zero_grad()
loss.backward()
optimizer.step()

def main():
env = gym.make(‘Pendulum-v0’)
q = Qnet()
q_target = Qnet() #fix
q_target.load_state_dict(q.state_dict())

memory = ReplayBuffer()
print_interval = 20
score = 0.0  
optimizer = optim.Adam(q.parameters(), lr=learning_rate)
render = False

for n_epi in range(10000):
    epsilon = max(0.01, 0.08 - 0.01*(n_epi/200)) 
    s = env.reset()
    done = False

    while not done: 
        a = q.sample_action(torch.from_numpy(s).float(), epsilon)
        s_prime, r, done, info = env.step(a)
        done_mask = 0.0 if done else 1.0
        memory.put((s,a,r/100.0,s_prime, done_mask))
    
        s = s_prime

        score += r
        
        if render:
            env.render()
            
        if done:
            break
            
     
        
    if score/print_interval > 300: 
        render = True

    if memory.size()>2000: 
        train(q, q_target, memory, optimizer)

    if n_epi%print_interval==0 and n_epi!=0:

        q_target.load_state_dict(q.state_dict())
        print("n_episode :{}, score : {:.1f}, n_buffer : {}, eps : {:.1f}%".format(
                                                        n_epi, score/print_interval, memory.size(), epsilon*100))
        
        score = 0.0
env.close()

if name == ‘main’:
main()

RuntimeError: size mismatch, m1: [1 x 3], m2: [2 x 128] at C:\w\b\windows\pytorch\aten\src\TH/generic/THTensorMath.cpp:41

Hi, your self.fc1 connects from 2 units to 128 units.
But your input seems to 3 be units.
Either change input to 2, or change self.fc1 to
nn.Linear(3,128)

1 Like

Thank you

So, I changed it to 3. It looks like this.

“IndexError: invalid index to scalar variable.”

What’s the problem?.. Help me…

Can you share the error completely ?
What did you chage to 3 ? nn.Linear(3,128) ?
If so, verify if your input is of shape (N, 3) .