主角torch.nn.LSTM()

初始化時要傳入的引數

 |  Args:
 |      input_size: The number of expected features in the input `x`
 |      hidden_size: The number of features in the hidden state `h`
 |      num_layers: Number of recurrent layers. E.g., setting ``num_layers=2``
 |          would mean stacking two LSTMs together to form a `stacked LSTM`,
 |          with the second LSTM taking in outputs of the first LSTM and
 |          computing the final results. Default: 1
 |      bias: If ``False``, then the layer does not use bias weights `b_ih` and `b_hh`.
 |          Default: ``True``
 |      batch_first: If ``True``, then the input and output tensors are provided
 |          as `(batch, seq, feature)` instead of `(seq, batch, feature)`.
 |          Note that this does not apply to hidden or cell states. See the
 |          Inputs/Outputs sections below for details.  Default: ``False``
 |      dropout: If non-zero, introduces a `Dropout` layer on the outputs of each
 |          LSTM layer except the last layer, with dropout probability equal to
 |          :attr:`dropout`. Default: 0
 |      bidirectional: If ``True``, becomes a bidirectional LSTM. Default: ``False``
 |      proj_size: If ``> 0``, will use LSTM with projections of corresponding size. Default: 0
 

input_size：一般是詞嵌入的大小
hidden_size：隱含層的維度
num_layers：預設是1，單層LSTM
bias：是否使用bias
batch_first：預設為False，如果設定為True，則表示第一個維度表示的是batch_size
dropout：直接看英文吧
bidirectional：預設為False，表示單向LSTM，當設定為True，表示為雙向LSTM，一般和num_layers配合使用（需要注意的是當該項設定為True時，將num_layers設定為1，表示由1個雙向LSTM構成）

模型輸入輸出-單向LSTM

import torch
import torch.nn as nn
import numpy as np

inputs_numpy = np.random.random((64,32,300))
inputs = torch.from_numpy(inputs_numpy).to(torch.float32)
inputs.shape
 

torch.Size([64, 32, 300])：表示[batchsize, max_length, embedding_size]

hidden_size = 128
lstm = nn.LSTM(300, 128, batch_first=True, num_layers=1)
output, (hn, cn) = lstm(inputs)
print(output.shape)
print(hn.shape)
print(cn.shape)

torch.Size([64, 32, 128])
torch.Size([1, 64, 128])
torch.Size([1, 64, 128])
說明：
output：儲存了每個時間步的輸出，如果想要獲取最後一個時間步的輸出，則可以這麼獲取：output_last = output[:,-1,:]

output_last = output[:,-1,:]
hn_last = hn[-1]
print(output_last.eq(hn_last))

模型輸入輸出-雙向LSTM

首先我們要明確：
output ：（seq_len, batch, num_directions * hidden_size）
h_n：(num_layers * num_directions, batch, hidden_size)
c_n ：（num_layers * num_directions, batch, hidden_size）
其中num_layers表示層數，這裡是1，num_directions表示方向數，由於是雙向的，這裡是2，也是，我們就有下面的結果：

import torch
import torch.nn as nn
import numpy as np

inputs_numpy = np.random.random((64,32,300))
inputs = torch.from_numpy(inputs_numpy).to(torch.float32)
inputs.shape
hidden_size = 128
lstm = nn.LSTM(300, 128, batch_first=True, num_layers=1, bidirectional=True)
output, (hn, cn) = lstm(inputs)
print(output.shape)
print(hn.shape)
print(cn.shape)

torch.Size([64, 32, 256])
torch.Size([2, 64, 128])
torch.Size([2, 64, 128])
這裡面的hn包含兩個元素，一個是正向的隱含層輸出，一個是方向的隱含層輸出。

#獲取反向的最後一個output
output_last_backward = output[:,0,-hidden_size:]
#獲反向最後一層的hn
hn_last_backward = hn[-1]
 
#反向最後的output等於最後一層的hn
print(output_last_backward.eq(hn_last_backward))
 
#獲取正向的最後一個output
output_last_forward = output[:,-1,:hidden_size]
#獲取正向最後一層的hn
hn_last_forward = hn[-2]
# 反向最後的output等於最後一層的hn
print(output_last_forward.eq(hn_last_forward))

https://www.cnblogs.com/LiuXinyu12378/p/12322993.html
https://blog.csdn.net/m0_45478865/article/details/104455978
https://blog.csdn.net/foneone/article/details/104002372