Python 使用Adam优化器对时装设计师进行逻辑回归训练时出错

数据集是FashionList（784输入，10输出）。我正在尝试使用Adam optimizer（也对其进行编码）训练逻辑回归：

列车功能为：

def train_logistic_regression(weights, bias, batch, loss, optimizer):

    inputs, labels = batch

    inputs = inputs.view(inputs.shape[0], -1)

    optimizer.zero_grad()
    y_pred = torch.sigmoid(weights@inputs + bias) # there must be the problem
    loss = criterion(y_pred, labels)
    loss.backward()
    optimizer.step()


from IPython.display import clear_output


for epoch in range(1, 5):

    for batch in train_dataloader: # have to go with batches
      metrics = train_logistic_regression(weights, bias, batch, criterion, optimizer)

每次出现错误时：

RuntimeError                              Traceback (most recent call last)
<ipython-input-161-408b80d71db1> in <module>()
      5 
      6     for batch in train_dataloader:
----> 7       metrics = train_logistic_regression(weights, bias, batch, criterion, optimizer)
      8 
      9 

<ipython-input-160-9c2f95ee56ee> in train_logistic_regression(weights, bias, batch, loss, optimizer)
      6 
      7     optimizer.zero_grad()
----> 8     y_pred = torch.sigmoid(weights@inputs + bias)
      9     # y_pred = model(inputs)
     10     loss = criterion(y_pred, labels)

RuntimeError: size mismatch, m1: [784 x 10], m2: [128 x 784] at /pytorch/aten/src/TH/generic/THTensorMath.cpp:41

运行时错误回溯（最近一次调用）
在（）
5.
6对于批量入列数据加载器：
---->7个指标=训练逻辑回归（权重、偏差、批次、标准、优化器）
8.
9
序列内逻辑回归（权重、偏差、批次、损失、优化器）
6.
7优化器0_grad（）
---->8 y_pred=火炬乙状体(weights@inputs+偏差）
9#y#u pred=模型（输入）
10损失=标准（y_pred，标签）
运行时错误：大小不匹配，m1:[784 x 10]，m2:[128 x 784]at/pytorch/aten/src/TH/generic/THTensorMath.cpp:41

---

如果有人能帮助我，我将不胜感激。

而不是

y\u pred=torch.sigmoid(weights@inputs+偏差）

应为

y\u pred=torch.sigmoid（输入.mm（权重）+偏差）

RuntimeError                              Traceback (most recent call last)
<ipython-input-161-408b80d71db1> in <module>()
      5 
      6     for batch in train_dataloader:
----> 7       metrics = train_logistic_regression(weights, bias, batch, criterion, optimizer)
      8 
      9 

<ipython-input-160-9c2f95ee56ee> in train_logistic_regression(weights, bias, batch, loss, optimizer)
      6 
      7     optimizer.zero_grad()
----> 8     y_pred = torch.sigmoid(weights@inputs + bias)
      9     # y_pred = model(inputs)
     10     loss = criterion(y_pred, labels)

RuntimeError: size mismatch, m1: [784 x 10], m2: [128 x 784] at /pytorch/aten/src/TH/generic/THTensorMath.cpp:41