-
Are Block Sparse Tensors already supported or are there plan on supporting this?
Comming from Quantum Chemistry we often deal with sparsity limited to a specific range rather then a randomly equal di…
-
### 🚀 The feature, motivation and pitch
I am working with weight matrices that look as follows:
As you can tell, considerable parts of the tensor are 0 and do not influence the output. Typical…
-
As talked about in #90025, there might be use-cases where a stable ABI is required for sparse tensors with dynamic shapes. cc @aartbik
As a consequence, `sparse_tensor.assemble` should also suppor…
-
I want to use my model in double precision, so I need to pass in data in double precision, but I will have error, here is I reproduced the error with toy example by changing precision:
```Python
# m…
-
### Is there an existing issue for this?
- [X] I have searched the existing issues for a bug report that matches the one I want to file, without success.
### Problem Description
Sparse tensors are …
-
Hi!
I'm trying to reshape CSR/CSC tensors. While reshaping CSR tensors works, running the `PassManager` on CSC reshape fails with:
```
Traceback (most recent call last):
File "/home/mtsokol…
-
-
At the moment, it can't be really done. See
https://github.com/pytorch/pytorch/issues/20248
We may have to switch to https://github.com/rusty1s/pytorch_sparse
in order to really scale the spar…
-
for small rank tensors with small bond dimensions, block-sparse tensor contractions are currently not very fast. We should change this. Possible options:
* port code to cython
* use dense encoding …
-
the ConvNets are currently using strides, which is not supported by by sparse tensors.
We can either extract the values manually (as done in [spconv](https://github.com/traveller59/spconv/blob/master…