vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Support multiple GPUs (split mode) on SYCL backend (#5806)

Browse source 
* suport multiple cards: split-mode - layer|row

* rm warning

* rebase with master, support tow new OPs, close feature for -sm=row, fix for unit test

* update news

* fix merge error

* update according to review comments
This commit is contained in:
Neo Zhang Jianyu 2024-03-02 19:49:30 +08:00 • committed by GitHub
parent 9bf297a02b
commit 715641391d
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
8 changed files with 1506 additions and 814 deletions
Download patch file Download diff file Expand all files Collapse all files

17
examples/sycl/run-llama2.sh
View file

@ -8,12 +8,19 @@ INPUT2="Building a website can be done in 10 simple steps:\nStep 1:"
source /opt/intel/oneapi/setvars.sh
if [ $# -gt 0 ]; then
export GGML_SYCL_DEVICE=$1
GGML_SYCL_DEVICE=$1
else
export GGML_SYCL_DEVICE=0
GGML_SYCL_DEVICE=0
fi
echo GGML_SYCL_DEVICE=$GGML_SYCL_DEVICE
echo "use $GGML_SYCL_DEVICE as main GPU"
#export GGML_SYCL_DEBUG=1
./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33 -s 0
#./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 5 -e -ngl 33 -t 1 -s 0
#ZES_ENABLE_SYSMAN=1, Support to get free memory of GPU by sycl::aspect::ext_intel_free_memory. Recommended to use when --split-mode = layer.
#use all GPUs with same max compute units
ZES_ENABLE_SYSMAN=1 ./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33 -s 0
#use main GPU only
#ZES_ENABLE_SYSMAN=1 ./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33 -s 0 -mg $GGML_SYCL_DEVICE -sm none