pytorchadam

ImportError: cannot import name 'is_sparse_any' from 'torch._subclasses.meta_utils' in PyTorch

I am encountering an import error while trying to define and train a Seq2Seq model using PyTorch. Below is the code snippet I am working with:

import torch
import torch.nn as nn
from torch.nn import TransformerEncoderLayer, TransformerDecoderLayer

class Seq2Seq(nn.Module):
    def __init__(self, encoder, decoder, config):
        super().__init__()
        self.encoder = encoder(config)
        self.decoder = decoder(config)
        
    def forward(self, src, tgt):
        encoder_output = self.encoder(src)
        decoder_output = self.decoder(encoder_output, tgt)
        return decoder_output

model = Seq2Seq(TransformerEncoderLayer, TransformerDecoderLayer, config)
optimizer = torch.optim.Adam(model.parameters())

However, when I run this code, I get the following error at the last line:

ImportError: cannot import name 'is_sparse_any' from 'torch._subclasses.meta_utils'

What I've Tried:

Checked the import statements to ensure they are correct. Verified that my PyTorch installation is up-to-date. Searched online for similar issues but couldn't find a clear solution.

How can I resolve the

ImportError: cannot import name 'is_sparse_any' from 'torch._subclasses.meta_utils' error?

Is it related to an issue with my environment variables or a bug in the PyTorch version I am using?

Additional Information:

I have verified that torch is correctly installed by running other basic PyTorch scripts. This error appears to be related to the environment setup, but I am not sure how to fix it. Thank you for your help!

Solution

  • I dont think there is any issue with your code

    enter image description here

    1. Can you mention the Pytorch version that you are using
    2. Can you also run this piece of code in a new session