Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

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

Merged
merged 2 commits into from
Sep 18, 2024
Merged

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

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8241151
set context default to avoid memory issue, update guide
arthw Sep 14, 2024
a6a8f8d
Update docs/backend/SYCL.md
NeoZhangJianyu Sep 17, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions docs/backend/SYCL.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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.

- Meet issue: `Native API failed. Native API returns: -6 (PI_ERROR_OUT_OF_HOST_MEMORY) -6 (PI_ERROR_OUT_OF_HOST_MEMORY)` 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**:
Please add the **[SYCL]** prefix/tag in issues/PRs titles to help the SYCL-team check/address them without delay.
Expand Down
7 changes: 4 additions & 3 deletions examples/sycl/run-llama2.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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.

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
CONEXT=8192

if [ $# -gt 0 ]; then
GGML_SYCL_DEVICE=$1
echo "use $GGML_SYCL_DEVICE as main GPU"
#use signle GPU only
ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m models/${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
#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
Loading