mat1 and mat2 shapes cannot be multiplied (360x1 and 32x4),原因是输入参数不对

登录后复制

# Traceback (most recent call last):
# 
#     output = model(x)
#   File "K:\ProgramData\Anaconda3\envs\py37\lib\site-packages\torch\nn\modules\module.py", line 1194, in _call_impl
#     return forward_call(*input, **kwargs)
#   File "K:/download/文档/bnn_pytorch.py", line 76, in forward
#     x = F.relu(self.fc1(x))
#   File "K:\ProgramData\Anaconda3\envs\py37\lib\site-packages\torch\nn\modules\module.py", line 1194, in _call_impl
#     return forward_call(*input, **kwargs)
#   File "K:/download/文档/bnn_pytorch.py", line 65, in forward
#     return F.linear(input, weight_samples, bias_samples)
# RuntimeError: mat1 and mat2 shapes cannot be multiplied (360x1 and 32x4)

• 1.
• 2.
• 3.
• 4.
• 5.
• 6.
• 7.
• 8.
• 9.
• 10.
• 11.
• 12.

登录后复制

# 输入层
your_input_size = 32

# 隐藏层
your_hidden_size = 4

your_num_layers = 32

# 输出层
your_output_size = 32

your_learning_rate = 0.001

• 1.
• 2.
• 3.
• 4.
• 5.
• 6.
• 7.
• 8.
• 9.
• 10.
• 11.
• 12.

登录后复制

# 实例化模型
model = BNN(input_size=your_input_size, hidden_size=your_hidden_size, output_size=your_output_size)
 
 
x_train = torch.randn(100, 1)  # x_train 的形状是 [batch_size, in_features]，
y_train = torch.randn(100, 1)  # x_train 的形状是 [batch_size, in_features]，

• 1.
• 2.
• 3.
• 4.
• 5.
• 6.

改写成为

登录后复制

x_train = torch.randn(100, 32)  # x_train 的形状是 [batch_size, in_features]，
y_train = torch.randn(100, 32)  # x_train 的形状是 [batch_size, in_features]，

• 1.
• 2.

原创作者: u_16120231 转载于: https://blog.51cto.com/u_16120231/11633013