Giters
huggingface
/
pytorch_block_sparse
Fast Block Sparse Matrices for Pytorch
Geek Repo:
Geek Repo
Github PK Tool:
Github PK Tool
Stargazers:
545
Watchers:
62
Issues:
15
Forks:
34
huggingface/pytorch_block_sparse Issues
Only blockshape=(32, 32) is supported
Updated
a year ago
`bfloat16` support
Updated
2 years ago
Using a smaller block size
Updated
2 years ago
does this package allow multi-gpu training and distributed training?
Updated
2 years ago
Comments count
1
What is "custom" "custom-back" in dispatch_policies.h?
Updated
2 years ago
Incorrect error message (says in_features, should be out_features)
Updated
2 years ago
Comments count
3
OSError: CUDA_HOME environment variable is not set
Updated
2 years ago
Comments count
7
Sparse pattern is not guaranteed to be full rank
Updated
3 years ago
No memory reduction observed in a simple sparse-dense multiplication
Updated
3 years ago
I get error on this simple code
Updated
3 years ago
Comments count
3
Specify the sparsity for the BlockSparseLinear
Closed
4 years ago
Comments count
1
Any way to bypass constraint of input and output features to be a multiple of block size
Updated
4 years ago
can we used pre-defined sparse matrix?
Updated
4 years ago
could you provide some code for loading a saved sparsified model?
Updated
4 years ago
Block Sparse Ax=b Solver?
Updated
4 years ago