-
The paper and project look super interesting to me, but there are several questions that confused me and I listed those questions below.
### Questions
1. In example/spmm folder, the Python c…
-
This is not a priority feature, but efficient SPMMs would allow to compute memory-efficient message-passing in cases where we have very large graphs. The sparse matrix would be the `edge_index` repres…
-
源码编译,然后在运行时,发现没有调用cogdl/cogdl/operators/中相关的算子,例如GCN算法中,使用的spmm_scatter,而不是使用spmm/spmm_kernel.cu中的算子
pdb调试发现cogdl/operators/spmm.py:12,有个try: except;try失败,从而csrspmm = None
此处try的报错为:
RuntimeError…
-
Hi, does [this line](https://github.com/marsupialtail/gpu-sparsert/blob/97540e4b03735463a52365eaa5c5217000b88413/autotune_float.sh#L75) would need to be uncommented if I wanted to autotune all example…
-
It seems that I cannot set device to 'cuda:1' for using MinkowskiEngine.
For example, I changed the device to `cuda:1` in this [line](https://github.com/NVIDIA/MinkowskiEngine/blob/master/examples/…
-
When I run `"python3.8 gcn_custom_2layer.py --n-hidden=32"`, the following situation occurred:
```
Using /tmp/torch_extensions as PyTorch extensions root...
Detected CUDA files, patching ldflags
E…
-
Hello, I am reading your paper and code and I have some problems about the merge-path SpMM.
(1) In my understanding, `limit[i]` meaning paritition i should start from row limit[i] of A, and end in r…
-
## 🚀 Feature
Deprecate `torch.spmm` and `torch.dsmm` functions.
## Motivation
These functions are aliases to `torch.mm`, undocumented, unused within PyTorch, and it is not clear why the…
pearu updated
2 years ago
-
### Description
From cuda 11.2, cusparse SpMM can be used for performing sparse matrix - dense matrix multiplication, where the sparse matrix is represented in Blocked-ELL format (see https://github.…
-
I was just running the documented benchmarks for csrmv and csrmm and realized that csrmm is not responsive to changes to number of threads (while csrmv was); Then looking into the [code](https://githu…