Efficient GPU kernels for block-sparse matrix multiplication
The blocksparse repository provides efficient GPU kernels (TensorFlow custom ops) for block-sparse matrix multiplication and convolution operations. The idea is to exploit block-level sparsity — i.e. treat matrices or weight tensors as composed of blocks, many of which may be zero or unused — to save compute and memory when sparsity patterns are structured. This is particularly useful in models like Sparse Transformers, where attention matrices or intermediate layers may adopt block-sparse...
...Update version: 2.11.19
1 - Customizing bash_login and bashrc
2 - Upgrade to version libpri-1.4.15
3 - Included new module-manager FOP2-1.0.3
4 - Inclusion of new packages and dependencies.
5 - Correction of errors in the installation.
6 - Upgrade to version webmin-1.710
7 - Correction of errors in the pbx-vpn script.
8- Inclusion of new script pbx-status.