DeepSpeed
Fix sparse attention for small block-sizes
#1545
Merged

Commits
  • fixing the softmax masking when using triangular masking
    Reza Yazdani committed 4 years ago
  • Merge branch 'master' of github.com:microsoft/DeepSpeed
    Reza Yazdani committed 4 years ago
  • Merge branch 'master' of github.com:microsoft/DeepSpeed
    Reza Yazdani committed 4 years ago
  • Merge branch 'master' of github.com:microsoft/DeepSpeed
    Reza Yazdani committed 4 years ago
  • Merge branch 'master' of github.com:microsoft/DeepSpeed
    Reza Yazdani committed 4 years ago
  • fixing the sparse attention for low block-size
    Reza Yazdani committed 4 years ago
  • remove attn
    Reza Yazdani committed 4 years ago
  • Merge branch 'master' into fix-sparse-attn
    RezaYazdaniAminabadi committed 4 years ago
  • Merge branch 'master' into fix-sparse-attn
    jeffra committed 4 years ago
Loading