torch.nn 中常用layer

1. BN层

• torch.nn.BatchNorm1d(num_features, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  输入：\((N,C) \ or \ (N, C, L)\), \(\ C\)对应num_features.
• torch.nn.BatchNorm2d(num_features, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  输入：\((N,C, H, W) \ or \ (N, C, H, W)\), \(\ C\)对应num_features.

2. 卷积层

• torch.nn.Conv1d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros')
  输入：\((N, C, L)\), \(\ C\)对应num_features.
• torch.nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros')
  输入：\((N, C, H, W)\), \(\ C\)对应num_features.
• torch.nn.Conv3d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros')
  输入：\((N, C, D, H, W)\), \(\ C\)对应num_features.

3. 激活层

• torch.nn.ReLU(inplace=False)
  输入：\((N,*)\).
• torch.nn.Sigmoid()
  输入：\((N,*)\).

4. 全连接层

• torch.nn.Linear(in_features, out_features, bias=True)
  输入：Input: \((N, *, C_{in})\).

5. LSTM

• torch.nn.LSTM(input_size, hidden_size, num_layers, bias=True, batch_first=False, dropout=0, bidirectional=False, proj_size=0)
  输入：
  \(inputs: (T, N, C)\),\(C\)是输入维度
  \(h_0: (num\_layers * num\_directions,\ N,\ hidden\_size)\)
  \(c_0: (num\_layers * num\_directions,\ N,\ hidden\_size)\)
  输出：
  \(outputs: (T,\ N,\ num\_directions * hidden\_size)\)
  \(h_0: (num\_layers * num\_directions,\ N,\ hidden\_size)\)
  \(c_0: (num\_layers * num\_directions,\ N,\ hidden\_size)\)

6. Transformer