批量输入显示 3d，但得到 2d，2d 张量

Question

我有这个训练循环

def train(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset)
    model.train()
    for batch, (X, y) in enumerate(dataloader):
        X, y = torch.stack(X).to(device), torch.stack(y).to(device)

        # Compute prediction error
        pred = model(X)
        loss = loss_fn(pred, y)

        # Backpropagation
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        if batch % 100 == 0:
            loss, current = loss.item(), batch * len(X)
            print(f"loss: {loss:>7f}  [{current:>5d}/{size:>5d}]")

和这个 lstm：

import torch
import torch.nn as nn
import pandas as pd
import numpy as np
class BELT_LSTM(nn.Module):

    def __init__(self, input_size, hidden_size, num_layers):
        super (BELT_LSTM, self).__init__()
        self.hidden_size = hidden_size
        self.num_layers = num_layers
        self.input_size = input_size
        self.BELT_LSTM   = nn.LSTM(input_size, hidden_size, num_layers)  

    def forward(self, x):
        # receive an input, create a new hidden state, return output?
        # reset the hidden state?
        hidden = (torch.zeros(self.num_layers, self.hidden_size), torch.zeros(self.num_layers, self.hidden_size))
        x, _ = self.BELT_LSTM(x, hidden)
        
        #since our observation has several sequences, we only want the output after the last sequence of the observation'''
        x = x[:, -1]     
        return x

这是数据集 class:

from __future__ import print_function, division
import os
import torch
import pandas as pd
import numpy as np
import math
from torch.utils.data import Dataset, DataLoader
class rcvLSTMDataSet(Dataset):
    """rcv dataset."""
    TIMESTEPS = 10

    def __init__(self, csv_data_file, annotations_file):
        """
        Args:
            csv_data_file (string): Path to the csv file with the training data
            annotations_file (string): Path to the file with the annotations
            
        """
        
        self.csv_data_file = csv_data_file
        self.annotations_file = annotations_file
        self.labels = pd.read_csv(annotations_file)
        self.data = pd.read_csv(csv_data_file)
        

    def __len__(self):
        return math.floor(len(self.labels) / 10)

    def __getitem__(self, idx):
        """
        pytorch expects whatever data is returned is in the form of a tensor.  Included, it expects the label for the data.
        Together, they make a tuple.          
        """
        
        # convert every ten indexes and label into one observation
        Observation = []
        counter = 0
        start_pos = self.TIMESTEPS *idx
        avg_avg_1 = 0
        avg_avg_2 = 0
        avg_avg_3 = 0
        while counter < self.TIMESTEPS:
            Observation.append(self.data.iloc[idx + counter].values)            
            avg_avg_1 += self.labels.iloc[idx + counter][2]
            avg_avg_2 += self.labels.iloc[idx + counter][1]
            avg_avg_3 += self.labels.iloc[idx + counter][0]
            counter += 1        
        #average the avg_1, avg_2, avg_3 for TIMESTEPS length
        avg_avg_1 = avg_avg_1 / self.TIMESTEPS
        avg_avg_2 = avg_avg_2 / self.TIMESTEPS
        avg_avg_3 = avg_avg_3 / self.TIMESTEPS
        current_labels = [avg_avg_1, avg_avg_2, avg_avg_3]
        print(current_labels)
        return Observation, current_labels
        

def main():
        loader = rcvDataSet('foo','foo2.csv')
        j = 0
        while j < len(loader.data % loader.TIMESTEPS):
            print(loader.__getitem__(j))
            j += 1

if "__main__" == __name__:
    main()

当运行这个时，我得到：

  File "module.py", line 1110, in _call_impl
    return forward_call(*input, **kwargs)
  File "lstm.py", line 21, in forward
    x, _ = self.BELT_LSTM(x, hidden)
  File "module.py", line 1110, in _call_impl
    return forward_call(*input, **kwargs)
  File "rnn.py", line 747, in forward
    raise RuntimeError(msg)
RuntimeError: For batched 3-D input, hx and cx should also be 3-D but got (2-D, 2-D) tensors

但据我所知，我已按照 nn.LSTM 说明设置图层并正确调整数据。我做错了什么？

作为参考，传入数据是来自 csv 文件的行，12 列宽，我每次观察提供 10 行

谢谢

Answer 1

您的代码：

hidden = (torch.zeros(self.num_layers, self.hidden_size), torch.zeros(self.num_layers, self.hidden_size))
x, _ = self.BELT_LSTM(x, hidden)

这里hx和cx都是二维张量。正确的方法应该是：

h_0 = torch.randn(self.num_directions * self.num_layers, self.batch_size, self.hidden_size)
c_0 = torch.randn(self.num_directions * self.num_layers, self.batch_size, self.hidden_size)
x, _ = self.BELT_LSTM(x, (h_0, c_0))

批量输入显示 3d，但得到 2d，2d 张量

Batched input shows 3d, but got 2d, 2d tensor

pytorch

tensor