Pytorch RuntimeError: Expected tensor for argument #1 'indices' to have one of the following scalar types: Long, Int; but got torch.FloatTenso instead

Question

Pytorch summary caused runtimer error. Following is the model. Runtime error occured while printing the summary of the model. I used torchsummary.summary in order to get the model summary.

Full Error message: RuntimeError: Expected tensor for argument #1 'indices' to have one of the following scalar types: Long, Int; but got torch.FloatTensor instead (while checking arguments for embedding)

class SentimentalModelV3(nn.Module):
  def __init__(self, output_size, vocab_size, embedding_dim = 128, hidden_dim = 64, batch_size = 32, padded_seq_len = 10, n_layers = 1, drop_prob = 0.3, bidirectional = False):
    super().__init__()
    self.batch_size = batch_size
    self.output_size = output_size
    self.n_layers = n_layers
    self.hidden_dim = hidden_dim
    self.padded_seq_len = padded_seq_len

    self.embedding = nn.Embedding(num_embeddings = vocab_size, embedding_dim = embedding_dim)
    self.lstm = nn.LSTM(input_size = embedding_dim, hidden_size = hidden_dim, num_layers = n_layers, dropout = drop_prob, batch_first = True, bidirectional = bidirectional)

    self.dropout = nn.Dropout(0.3)

    #Linear and activation layer
    self.fc1=nn.Linear(self.hidden_dim * self.padded_seq_len, 64)
    self.fc2=nn.Linear(64, 16)
    self.fc3=nn.Linear(16,output_size)
    self.Relu = nn.ReLU()

  def forward(self, one_hot, hn, cn):
    embed = self.embedding(one_hot)
    lstm_out, hidden = self.lstm(embed)

    #stack up the lstm output
    lstm_out = lstm_out.reshape(shape = (lstm_out.shape[0], lstm_out.shape[1] * lstm_out.shape[2]))
    # dropout and fully connected layers
    out = self.dropout(lstm_out)
    out = self.Relu(out)
    out = self.Relu(self.fc1(out))
    out = self.Relu(self.fc2(out))
    out = self.fc3(out)

    return out

  def initCellState(self):
    h =  torch.zeros(self.n_layers , self.batch_size , self.hidden_dim).to(device)
    c =  torch.zeros(self.n_layers , self.batch_size , self.hidden_dim).to(device)
    return h, c

While, printing the summary, i got the error as shown below in the picture

from torchsummary import summary

n_layers = 5
batch_size = 32
hidden_dim = 64

model_v3 = SentimentalModelV3(output_size = 3, 
                              vocab_size = UNIQUE_WORD_COUNT, 
                              embedding_dim = 128, 
                              hidden_dim = hidden_dim, 
                              n_layers = n_layers, 
                              drop_prob = 0.3, 
                              padded_seq_len = 10, 
                              batch_size = batch_size,
                              bidirectional = False).to(device)

hn, cn = model_v3.initCellState()
summary(model_v3, [(1, 10), (n_layers, batch_size, hidden_dim), (n_layers, batch_size, hidden_dim)])

Answer 1

By default, summary runs model_v3 with FloatTensor of given shapes as input. But your code expects one_hot to be a LongTensor. The fix is to specify the correct tensor type to summary. Thankfully, summary accepts a sequence of tensors for this purpose:

summary(model_v3, [
    torch.zeros(1, 10).long(),  # ensure one_hot is a LongTensor
    torch.zeros(n_layers, batch_size, hidden_dim),
    torch.zeros(n_layers, batch_size, hidden_dim),
])