CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

[SYCL]set context default value to avoid memory issue, update guide (#9476)

Browse Source 
* set context default to avoid memory issue, update guide

* Update docs/backend/SYCL.md

Co-authored-by: Meng, Hengyu <hengyu.meng@intel.com>

---------

Co-authored-by: arthw <14088817+arthw@users.noreply.github.com>
Co-authored-by: Meng, Hengyu <hengyu.meng@intel.com>
This commit is contained in:
Neo Zhang Jianyu
2024-09-18 08:30:31 +08:00
committed by GitHub
parent 7be099fa81
commit faf67b3de4
2 changed files with 12 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
docs/backend/SYCL.md
View File

@@ -636,6 +636,14 @@ use 1 SYCL GPUs: [0] with Max compute units:512
It's same for other projects including llama.cpp SYCL backend. It's same for other projects including llama.cpp SYCL backend.
- Meet issue: `Native API failed. Native API returns: -6 (PI_ERROR_OUT_OF_HOST_MEMORY) -6 (PI_ERROR_OUT_OF_HOST_MEMORY) -999 (UNKNOWN PI error)` or `failed to allocate SYCL0 buffer`
Device Memory is not enough.
|Reason|Solution|
|-|-|
|Default Context is too big. It leads to more memory usage.|Set `-c 8192` or smaller value.|
|Model is big and require more memory than device's.|Choose smaller quantized model, like Q5 -> Q4;<br>Use more than one devices to load model.|
### **GitHub contribution**: ### **GitHub contribution**:
Please add the **[SYCL]** prefix/tag in issues/PRs titles to help the SYCL-team check/address them without delay. Please add the **[SYCL]** prefix/tag in issues/PRs titles to help the SYCL-team check/address them without delay.

7
examples/sycl/run-llama2.sh
View File

@@ -11,16 +11,17 @@ source /opt/intel/oneapi/setvars.sh
#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. #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.
INPUT_PROMPT="Building a website can be done in 10 simple steps:\nStep 1:" INPUT_PROMPT="Building a website can be done in 10 simple steps:\nStep 1:"
MODEL_FILE=llama-2-7b.Q4_0.gguf MODEL_FILE=models/llama-2-7b.Q4_0.gguf
NGL=33 NGL=33
CONEXT=8192
if [ $# -gt 0 ]; then if [ $# -gt 0 ]; then
GGML_SYCL_DEVICE=$1 GGML_SYCL_DEVICE=$1
echo "use $GGML_SYCL_DEVICE as main GPU" echo "use $GGML_SYCL_DEVICE as main GPU"
#use signle GPU only #use signle GPU only
ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m models/${MODEL_FILE} -p "${INPUT_PROMPT}" -n 400 -e -ngl ${NGL} -s 0 -mg $GGML_SYCL_DEVICE -sm none ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m ${MODEL_FILE} -p "${INPUT_PROMPT}" -n 400 -e -ngl ${NGL} -s 0 -c ${CONEXT} -mg $GGML_SYCL_DEVICE -sm none
else else
#use multiple GPUs with same max compute units #use multiple GPUs with same max compute units
ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m models/${MODEL_FILE} -p "${INPUT_PROMPT}" -n 400 -e -ngl ${NGL} -s 0 ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m ${MODEL_FILE} -p "${INPUT_PROMPT}" -n 400 -e -ngl ${NGL} -s 0 -c ${CONEXT}
fi fi