[Inference] Adapt to Fused rotary (#5348)

* revise rotary embedding

* remove useless print

* adapt

* fix

* add

* fix

* modeling

* fix

* fix

* fix
This commit is contained in:
Jianghai
2024-02-07 11:36:04 +08:00
committed by GitHub
parent 35382a7fbf
commit 9f4ab2eb92
5 changed files with 161 additions and 22 deletions

View File

@@ -1,4 +1,5 @@
ROOT=$(realpath $(dirname $0))
echo $ROOT
PY_SCRIPT=${ROOT}/benchmark_llama.py
GPU=$(nvidia-smi -L | head -1 | cut -d' ' -f4 | cut -d'-' -f1)
mode=$1