Skip to content

Commit

Permalink
update README.md
Browse files Browse the repository at this point in the history
Signed-off-by: Molly Sophia <[email protected]>
  • Loading branch information
MollySophia committed Dec 2, 2024
1 parent 0104612 commit 84339c8
Showing 1 changed file with 1 addition and 0 deletions.
1 change: 1 addition & 0 deletions README.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
# Inference RWKV on Qualcomm HTP (Hexagon Tensor Processor) using QNN SDK
## Features
- Support for RWKV v5, v6 and experimentally v7 models(WIP)
- Inference RWKV using QNN SDK, with Qualcomm CPU, GPU or HTP (Hexagon Tensor Processor) as the backend.
- Support for whole-model float16 inference (since Qualcomm HTP cannot do float32 math).
- Support for activation INT16 and weights INT8 quantized inference (with some key operations running with float16).
Expand Down

0 comments on commit 84339c8

Please sign in to comment.