Convert keras model architecture to Pytorch

I am trying to convert a keras model to pytorch for human activity recognition. The keras model could achieve up to 98% accuracy, while Pytorch model could achieve only ~60% accuracy. I couldn’t figure out the problem, first I was thinking of the padding=‘same’ of keras, but I have already adjusted the padding of pytorch already. Can you check what is wrong?

The keras code is as above

model = keras.Sequential()
model.add(layers.Conv1D(filters=32, kernel_size=3, padding="same"))
model.add(layers.Conv1D(filters=64, kernel_size=3, padding="same"))
model.add(layers.Dense(units=128, activation='relu'))
model.add(layers.Dense(13, activation='softmax'))

and my pytorch model code is as below

class CNN(nn.Module):
    def __init__(self):
        super(CNN, self).__init__()
        seq_len = 100
        # output: [, 32, 100]
        self.conv1 = nn.Conv1d(seq_len, 32, kernel_size=3, stride=1, padding=1)
        self.bn1 = nn.BatchNorm1d(32)
        # output: [, 64, 100]
        self.conv2 = nn.Conv1d(32, 64, kernel_size=3, padding=1)
        self.bn2 = nn.BatchNorm1d(64)
        # output: [, 64, 50] = nn.MaxPool1d(kernel_size=2, stride=2)
        # output: [, 64]
        self.lstm = nn.LSTM(6, 64, 1)
        # output: [, 128]
        self.fc1 = nn.Linear(64, 128)
        # output: [, 13]
        self.fc2 = nn.Linear(128, 13)
    def forward(self, x):
        x = self.conv1(x) 
        x = self.bn1(x)
        x = F.relu(x)

        x = self.conv2(x)
        x = self.bn2(x)
        x = F.relu(x)
        x =
        out, _ = self.lstm(x)
        x = out[:, -1, :]
        x = self.fc1(x)
        x = F.relu(x)
        x = self.fc2(x)
        return x

And this is the model summary of the Keras model, the output size of each layers do match.

I guess the issue could be created by the nn.LSTM layer in the default setup, which would expect an input in the shape [seq_len, batch_size, nb_features] and could thus process the inputs in a wrong way.
If you want to pass inputs with the batch dimension in dim0, you could create the modules with batch_first=True. The docs explain the input shapes a bit more.