onnxruntime
[webgpu] Use components for VxAttentionScore
#23726
Merged
Overview
Commits
10
Changes
View On GitHub
Commits
  • Use components for VxAttentionScore
    qjia7 committed 1 year ago
  • use flash attention to merge softmax to the last stage
    qjia7 committed 1 year ago
  • use f32 instead of f16 for max/sum
    qjia7 committed 1 year ago
  • make it work with GQA
    qjia7 committed 1 year ago
  • recover ApplyFlashAttention
    qjia7 committed 1 year ago
  • Revert "recover ApplyFlashAttention"
    qjia7 committed 1 year ago
  • Revert "make it work with GQA"
    qjia7 committed 1 year ago
  • Revert "use f32 instead of f16 for max/sum"
    qjia7 committed 1 year ago
  • Revert "use flash attention to merge softmax to the last stage"
    qjia7 committed 1 year ago
  • Merge branch 'main' into attention_opt
    qjia7 committed 1 year ago
FAQ Terms Privacy Refunds Impressum
Loading