[Kernel][RFC] Refactor the punica kernel based on Triton (#5036)

7ecee343 · Jee Jee Li · GitHub · 7eb0cb4a · 7ecee343 · 7ecee343
Unverified Commit 7ecee343 authored Aug 01, 2024 by Jee Jee Li Committed by GitHub Jul 31, 2024
7 changed files
--- a/vllm/lora/ops/sgmv_expand_slice.py
+++ b/vllm/lora/ops/sgmv_expand_slice.py
--- a/vllm/lora/ops/sgmv_shrink.py
+++ b/vllm/lora/ops/sgmv_shrink.py
--- a/vllm/lora/ops/utils.py
+++ b/vllm/lora/ops/utils.py
--- a/vllm/lora/punica.py
+++ b/vllm/lora/punica.py
--- a/vllm/triton_utils/__init__.py
+++ b/vllm/triton_utils/__init__.py
--- a/vllm/triton_utils/libentry.py
+++ b/vllm/triton_utils/libentry.py
--- a/vllm/worker/model_runner.py
+++ b/vllm/worker/model_runner.py