NotImplementedError: Could not run 'aten::_sparse_coo_tensor_with_dims_and_tensors' with arguments from the 'SparseHPU' backend

Training PyTorch
1

We encounter NotImplementedError: Could not run 'aten::_sparse_coo_tensor_with_dims_and_tensors' with arguments from the 'SparseHPU' backend. when using sparse Metapath2vec with sparseAdam optimizer.

We are able to adapt the code without sparse computation and run it.

  • I use the docker image vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest and then pip install torch_geometric
  • Specifically, torch-geometric==2.6.1

See the error messages here and see the code here.

2

Hi,

There is no plan to support ‘aten::_sparse_coo_tensor_with_dims_and_tensors’ at the moment

Please take a look at the below document for supported ops.
https://docs.habana.ai/en/latest/PyTorch/Reference/Pytorch_Operators/Pytorch_Operators.html