Summary with embeddings

[siddBanPsu]

class TextCNN(nn.Module):
    def __init__(self, nb_words, embed_dim, embedding_matrix, max_seq_len, num_filters, num_classes):
        super(TextCNN, self).__init__()
        self.num_filters=num_filters
        self.embed = nn.Embedding(nb_words, embed_dim)
        self.dropout = nn.Dropout(0.3)
        self.conv = nn.Conv1d(embed_dim, num_filters, kernel_size=2, stride=1)
        self.fc1 = nn.Linear(num_filters, 32)
        self.fc2 = nn.Linear(32, num_classes)
        self.logsigmoid = nn.Sigmoid()

    def forward(self, x):
        x = self.embed(x)
        x = x.permute(0, 2, 1)
        x = self.dropout(x)
        x = self.conv(x).permute(0, 2, 1).max(1)[0]
        x = self.fc1(x)
        x = F.relu(x)
        x = self.dropout(x)
        x = self.fc2(x)
        return self.logsigmoid(x)

model = TextCNN(1000, 100, [], 10, 20, 3)
data = torch.from_numpy(np.array([[1,4,5], [7,7,9]]))
output = model(data)
print(output)
print(model)
print(data[0].shape)
print(summary(model, (1, 10)))

Gives error:

RuntimeError: Expected tensor for argument #1 'indices' to have scalar type Long; but got CPUFloatTensor instead (while checking arguments for embedding)

[meshiguge]

same problem

[BerenLuthien]

RuntimeErrorTraceback (most recent call last)
<ipython-input-243-70bbb59e2bd0> in <module>()
     26 print(model)
     27 print(data[0].shape)
---> 28 print(summary(model, (1, 10)))
...
RuntimeError: Expected tensor for argument #1 'indices' to have scalar type Long; but got CUDAFloatTensor instead (while checking arguments for embedding)

[LittleStars666]

I also encountered this problem. Has anyone solved it？

[alexandru-dinu]

One quick-fix is to change the dtype from FloatTensor to LongTensor, in the source file where the package is installed:

if device == "cuda" and torch.cuda.is_available():
    # dtype = torch.cuda.FloatTensor
    dtype = torch.cuda.LongTensor
else:
    # dtype = torch.FloatTensor
    dtype = torch.LongTensor

[alfaijmansuri]

same problem and also changing the data types from floattensor to long is not workng

[alexandru-dinu]

Not working is not satisfactory. What is the error message?

[wassimseif]

Hey I'm having the same issue

class BiLSTMBaseline(nn.Module):
    
    def __init__(self,hidden_dim,emb_dim=300,
                recurrent_dropout=0.1,num_linear=1):
        super().__init__()
        self.embedding = nn.Embedding(len(TEXT.vocab),emb_dim)
        self.encoder = nn.LSTM(emb_dim,hidden_dim,num_linear,dropout = recurrent_dropout)
        self.linear_layers = []
        for _ in range(num_linear - 1):
            self.linear_layers.append(nn.Linear(hidden_dim,hidden_dim))
        self.linear_layers = nn.ModuleList(self.linear_layers)
        self.predictor =   nn.Linear(hidden_dim,6)
    
    def forward(self,seq):
        embeddings = self.embedding(seq)
        hdn, _  = self.encoder(embeddings)
        feature = hdn[-1,:,:]
        for layer in self.linear_layers:
            feature = layer(feature)
        preds = self.predictor(feature)
        return preds
        

summary(model,input_size=(402, 64),device='cpu')

Error :

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-44-ad1f81b1c254> in <module>
----> 1 summary(model,input_size=(402, 64),device='cpu')

<ipython-input-43-4e60852cb827> in summary(model, input_size, batch_size, device)
     72 #     print(x.shape)
     73 
---> 74     model(*x)
     75 
     76     # remove these hooks

~/anaconda3/envs/sent-analy/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
    487             result = self._slow_forward(*input, **kwargs)
    488         else:
--> 489             result = self.forward(*input, **kwargs)
    490         for hook in self._forward_hooks.values():
    491             hook_result = hook(self, input, result)

<ipython-input-12-e2f879788e90> in forward(self, seq)
     13 
     14     def forward(self,seq):
---> 15         embeddings = self.embedding(seq)
     16         hdn, _  = self.encoder(embeddings)
     17         feature = hdn[-1,:,:]

~/anaconda3/envs/sent-analy/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
    487             result = self._slow_forward(*input, **kwargs)
    488         else:
--> 489             result = self.forward(*input, **kwargs)
    490         for hook in self._forward_hooks.values():
    491             hook_result = hook(self, input, result)

~/anaconda3/envs/sent-analy/lib/python3.6/site-packages/torch/nn/modules/sparse.py in forward(self, input)
    116         return F.embedding(
    117             input, self.weight, self.padding_idx, self.max_norm,
--> 118             self.norm_type, self.scale_grad_by_freq, self.sparse)
    119 
    120     def extra_repr(self):

~/anaconda3/envs/sent-analy/lib/python3.6/site-packages/torch/nn/functional.py in embedding(input, weight, padding_idx, max_norm, norm_type, scale_grad_by_freq, sparse)
   1452         # remove once script supports set_grad_enabled
   1453         _no_grad_embedding_renorm_(weight, input, max_norm, norm_type)
-> 1454     return torch.embedding(weight, input, padding_idx, scale_grad_by_freq, sparse)
   1455 
   1456 

RuntimeError: Expected tensor for argument #1 'indices' to have scalar type Long; but got torch.FloatTensor instead (while checking arguments for embedding)

[ruathudo]

Is there any fix for this? I'm still getting this error.

[RaffaeleGalliera]

One quick-fix is to change the dtype from FloatTensor to LongTensor, in the source file where the package is installed:

    if device == "cuda" and torch.cuda.is_available():
        # dtype = torch.cuda.FloatTensor
        dtype = torch.cuda.LongTensor
    else:
        # dtype = torch.FloatTensor
        dtype = torch.LongTensor

I was having the same issue and this quick fix worked for me

[turbolt]

Hey @siddBanPsu, thanks for starting this thread - I was wondering if you've fixed your issue - if yes, do you mind sharing how you fix it? thanks a lot!

[nikhilkarnwal]

Hi, I also encountered same issue and as a hack, I changed the data type of tensor in the forward method where I am calling embedding module to long just for testing purpose. It worked for me.

[santteegt]

to solve this I used the torchsummaryX package that allows you to specify a sample input tensor:

!pip install torchsummaryX

Example:

from torchsummaryX import summary as summaryx

# sample input tensor
input_size = (1, 80, 108)
x_sample = torch.zeros(input_size, dtype=torch.long, device=torch.device('cuda'))

print(summaryx(model, x_sample))