Commit 5ef7dea

committed

Simplify sparse attention vLLM plugin forward paths

Signed-off-by: Kai Xu <kaix@nvidia.com>

1 parent 6649c28 commit 5ef7deaCopy full SHA for 5ef7dea

7 files changed

modelopt/torch/sparsity/attention_sparsity/plugins
- vllm.py
tests
- gpu_vllm/torch
  - quantization
    - test_vllm_dynamic_modules.py
  - sparsity/attention_sparsity
    - test_sparse_attn_worker.py
- gpu/torch/kernels/common/attention
- unit/torch/kernels/common/attention
  - test_triton_fa.py

Comments

(0)