vocab_size = 4008
x_train_input_shape = (9611, 9200)
max_len = 7200
model = Sequential()
model.add(Embedding(input_dim= vocab_size,
output_dim= 100,
input_length = max_len))
model.add(Dropout(0.5))
model.add(Conv1D(32, 5, kernel_regularizer=regularizers.l1(0.001), activation='relu'))
model.add(MaxPooling1D(pool_size=4))
model.add(LSTM(100))
model.add(Dense(1,kernel_regularizer=regularizers.l1(0.001), activation='sigmoid'))
print('Ready to compile')
So I’m not looking for a complete Pytorch code but how would I add a MaxPooling1D , Embedding and LSTM layers in a Pytorch Neural Net? Also how would I add regularization to certain layers?
This is what I have so far. I feel like the numbers that I provided are off.
x_train_input_shape = (9611, 9200)
vocab_size = 4008
max_len = 7200
class NeuralNet(nn.Module):
def __init__(self, max_len, vocab_size, hidden_dim = 100, num_layers = 1):
super(NeuralNet, self).__init__()
self.hidden_dim = hidden_dim
self.batch_size = 4008
self.num_layers = num_layers
self.emb = nn.Embedding(vocab_size, max_len)
self.drop = nn.Dropout(0.5)
self.conv = nn.Conv1d(max_len, 32, kernel_size = 5)
self.pool = nn.MaxPool1d(2)
self.lstm = nn.LSTM(32, hidden_dim)
self.out = nn.Linear(hidden_dim, 1)
self.rel = nn.ReLU()
self.sig = nn.Sigmoid()
def init_hidden(self):
return (torch.zeros(self.num_layers, self.batch_size, self.hidden_dim),
torch.zeros(self.num_layers, self.batch_size, self.hidden_dim))
def forward(self,x):
x = self.emb(x)
x = self.drop(x)
x = self.rel(self.conv(x))
x = self.pool(x)
lstm_out, _ = self.lstm(x.view(len(x), self.batch_size, -1))
output = self.sig(self.out(lstm_out[:, -1, :]))
return output