huggingface/pytorch_block_sparse

Issues

• I get error on this simple code

  #14 opened 4 years ago by ebrahimHappy
  4

• Only blockshape=(32, 32) is supported

  #20 opened 2 years ago by Routhleck
  0

• `bfloat16` support

  #19 opened 2 years ago by lostmsu
  0

• Using a smaller block size

  #18 opened 2 years ago by pclucas14
  0

• does this package allow multi-gpu training and distributed training?

  #11 opened 4 years ago by DanqingZ
  1

• What is "custom" "custom-back" in dispatch_policies.h?

  #17 opened 2 years ago by ziyuhuang123
  0

• Incorrect error message (says in_features, should be out_features)

  #8 opened 4 years ago by ejohb
  3

• OSError: CUDA_HOME environment variable is not set

  #7 opened 4 years ago by osainz59
  7

• Sparse pattern is not guaranteed to be full rank

  #16 opened 3 years ago by jrapin
  0

• No memory reduction observed in a simple sparse-dense multiplication

  #15 opened 4 years ago by x-zho14
  0

• Specify the sparsity for the BlockSparseLinear

  #10 opened 4 years ago by zheng-ningxin
  1

• Any way to bypass constraint of input and output features to be a multiple of block size

  #9 opened 4 years ago by shreyanshs
  0

• can we used pre-defined sparse matrix?

  #13 opened 4 years ago by hsmills
  0

• could you provide some code for loading a saved sparsified model?

  #12 opened 4 years ago by DanqingZ
  0

• Block Sparse Ax=b Solver?

  #5 opened 4 years ago by twhughes
  0