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

Arc Stable version test #10087

Merged
merged 8 commits into from
Feb 6, 2024
Merged

Arc Stable version test #10087

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
42a57ea
add batch_size in stable version test
Ariadne330 Feb 4, 2024
df5f7b6
add batch_size in excludes
Ariadne330 Feb 4, 2024
d49a094
add excludes for batch_size
Ariadne330 Feb 4, 2024
c234743
fix ci
Ariadne330 Feb 5, 2024
de7e9b0
triger regression test
Ariadne330 Feb 5, 2024
dbefe24
fix xpu version
Ariadne330 Feb 5, 2024
b52a120
disable ci
Ariadne330 Feb 5, 2024
e67e871
address kai's comment
Ricky-Ting Feb 6, 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
66 changes: 53 additions & 13 deletions .github/workflows/llm_tests_for_stable_version_on_arc.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,47 +61,87 @@ jobs:
- name: Run LLM install (all) test
uses: ./.github/actions/llm/setup-llm-env
with:
extra-dependency: "xpu"
extra-dependency: "xpu_2.1"

- name: Test installed xpu version
shell: bash
run: |
source /home/arda/intel/oneapi/setvars.sh
source /opt/intel/oneapi/setvars.sh
bash python/llm/test/run-llm-install-tests.sh

- name: Test on xpu (int4)
shell: bash
run: |
source /home/arda/intel/oneapi/setvars.sh
source /opt/intel/oneapi/setvars.sh
export USE_XETLA=OFF
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
mv python/llm/test/benchmark/stable-version-arc-perf-test-sym_int4.yaml python/llm/dev/benchmark/all-in-one/config.yaml
cd python/llm/dev/benchmark/all-in-one
# hide time info
sed -i 's/str(end - st)/"xxxxxx"/g' run.py
# batch_size = 1
sed -i '/batch_size/c\batch_size: 1' config.yaml
python run.py
cp ./*.csv $CSV_SAVE_PATH/int4
cp ./*.csv $CSV_SAVE_PATH/int4/batch_size_1
rm ./*.csv
# batch_size = 2
sed -i '/batch_size/c\batch_size: 2' config.yaml
python run.py
cp ./*.csv $CSV_SAVE_PATH/int4/batch_size_2
rm ./*.csv
# batch_size = 4
sed -i '/batch_size/c\batch_size: 4' config.yaml
python run.py
cp ./*.csv $CSV_SAVE_PATH/int4/batch_size_4
rm ./*.csv
# batch_size = 8
sed -i '/batch_size/c\batch_size: 8' config.yaml
python run.py
cp ./*.csv $CSV_SAVE_PATH/int4/batch_size_8
rm ./*.csv
cd ../../../test/benchmark
python -m pip install pandas==1.5.3
python csv_to_html.py -f $CSV_SAVE_PATH/int4 -b $CSV_SAVE_PATH/int4/transformer_int4_gpu-results-1baseline.csv -t 5.0

python csv_to_html.py -f $CSV_SAVE_PATH/int4/batch_size_1 -b $CSV_SAVE_PATH/int4/batch_size_1/transformer_int4_gpu-results-1baseline.csv -t 5.0
python csv_to_html.py -f $CSV_SAVE_PATH/int4/batch_size_2 -b $CSV_SAVE_PATH/int4/batch_size_2/transformer_int4_gpu-results-1baseline.csv -t 5.0
python csv_to_html.py -f $CSV_SAVE_PATH/int4/batch_size_4 -b $CSV_SAVE_PATH/int4/batch_size_4/transformer_int4_gpu-results-1baseline.csv -t 5.0
python csv_to_html.py -f $CSV_SAVE_PATH/int4/batch_size_8 -b $CSV_SAVE_PATH/int4/batch_size_8/transformer_int4_gpu-results-1baseline.csv -t 5.0

- name: Test on xpu (fp8)
shell: bash
run: |
source /home/arda/intel/oneapi/setvars.sh
source /opt/intel/oneapi/setvars.sh
export USE_XETLA=OFF
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
mv python/llm/test/benchmark/stable-version-arc-perf-test-fp8.yaml python/llm/dev/benchmark/all-in-one/config.yaml
cd python/llm/dev/benchmark/all-in-one
# hide time info
sed -i 's/str(end - st)/"xxxxxx"/g' run.py
# batch_size = 1
sed -i '/batch_size/c\batch_size: 1' config.yaml
python run.py
cp ./*.csv $CSV_SAVE_PATH/fp8
cp ./*.csv $CSV_SAVE_PATH/fp8/batch_size_1
rm ./*.csv
# batch_size = 2
sed -i '/batch_size/c\batch_size: 2' config.yaml
python run.py
cp ./*.csv $CSV_SAVE_PATH/fp8/batch_size_2
rm ./*.csv
# batch_size = 4
sed -i '/batch_size/c\batch_size: 4' config.yaml
python run.py
cp ./*.csv $CSV_SAVE_PATH/fp8/batch_size_4
rm ./*.csv
# batch_size = 8
sed -i '/batch_size/c\batch_size: 8' config.yaml
python run.py
cp ./*.csv $CSV_SAVE_PATH/fp8/batch_size_8
rm ./*.csv
cd ../../../test/benchmark
python -m pip install pandas==1.5.3
python csv_to_html.py -f $CSV_SAVE_PATH/fp8 -b $CSV_SAVE_PATH/fp8/transformer_int4_gpu-results-1baseline.csv -t 5.0
python csv_to_html.py -f $CSV_SAVE_PATH/fp8/batch_size_1 -b $CSV_SAVE_PATH/fp8/batch_size_1/transformer_int4_gpu-results-1baseline.csv -t 5.0
python csv_to_html.py -f $CSV_SAVE_PATH/fp8/batch_size_2 -b $CSV_SAVE_PATH/fp8/batch_size_2/transformer_int4_gpu-results-1baseline.csv -t 5.0
python csv_to_html.py -f $CSV_SAVE_PATH/fp8/batch_size_4 -b $CSV_SAVE_PATH/fp8/batch_size_4/transformer_int4_gpu-results-1baseline.csv -t 5.0
python csv_to_html.py -f $CSV_SAVE_PATH/fp8/batch_size_8 -b $CSV_SAVE_PATH/fp8/batch_size_8/transformer_int4_gpu-results-1baseline.csv -t 5.0

llm-stress-test-on-arc:
needs: llm-perf-regression-test-on-arc
Expand Down Expand Up @@ -143,18 +183,18 @@ jobs:
- name: Run LLM install (all) test
uses: ./.github/actions/llm/setup-llm-env
with:
extra-dependency: "xpu"
extra-dependency: "xpu_2.1"

- name: Test installed xpu version
shell: bash
run: |
source /home/arda/intel/oneapi/setvars.sh
source /opt/intel/oneapi/setvars.sh
bash python/llm/test/run-llm-install-tests.sh

- name: Test on xpu (int4)
shell: bash
run: |
source /home/arda/intel/oneapi/setvars.sh
source /opt/intel/oneapi/setvars.sh
export USE_XETLA=OFF
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
mv python/llm/test/benchmark/stable-version-arc-stress-test-sym_int4.yaml python/llm/dev/benchmark/all-in-one/config.yaml
Expand All @@ -171,7 +211,7 @@ jobs:
- name: Test on xpu (fp8)
shell: bash
run: |
source /home/arda/intel/oneapi/setvars.sh
source /opt/intel/oneapi/setvars.sh
export USE_XETLA=OFF
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
mv python/llm/test/benchmark/stable-version-arc-stress-test-fp8.yaml python/llm/dev/benchmark/all-in-one/config.yaml
Expand Down
3 changes: 2 additions & 1 deletion python/llm/dev/benchmark/all-in-one/run.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -956,7 +956,8 @@ def run_transformer_autocast_bf16( repo_id,
if excludes:
for in_out in conf['in_out_pairs']:
model_id_input = model + ':' + in_out.split('-')[0]
if model_id_input in excludes:
model_id_input_batch_size = model_id_input + ':' + str(conf['batch_size'])
if model_id_input in excludes or model_id_input_batch_size in excludes:
in_out_pairs.remove(in_out)
run_model(model, api, in_out_pairs, conf['local_model_hub'], conf['warm_up'], conf['num_trials'], conf['num_beams'],
conf['low_bit'], conf['cpu_embedding'], conf['batch_size'])
Expand Down
23 changes: 23 additions & 0 deletions python/llm/test/benchmark/stable-version-arc-perf-test-fp8.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,8 +12,31 @@ low_bit: 'fp8' # default to use 'sym_int4' (i.e. symmetric int4)
batch_size: 1 # default to 1
in_out_pairs:
- '32-32'
- '512-256'
- '1024-128'
- '2048-256'
test_api:
- "transformer_int4_gpu" # on Intel GPU
cpu_embedding: False # whether put embedding to CPU (only avaiable now for gpu win related test_api)
exclude:
- 'meta-llama/Llama-2-7b-chat-hf:2048:4'
- 'meta-llama/Llama-2-7b-chat-hf:512:8'
- 'meta-llama/Llama-2-7b-chat-hf:1024:8'
- 'meta-llama/Llama-2-7b-chat-hf:2048:8'
- 'THUDM/chatglm2-6b:2048:8'
- 'THUDM/chatglm3-6b:2048:8'
- 'baichuan-inc/Baichuan2-7B-Chat:2048:2'
- 'baichuan-inc/Baichuan2-7B-Chat:1024:4'
- 'baichuan-inc/Baichuan2-7B-Chat:2048:4'
- 'baichuan-inc/Baichuan2-7B-Chat:512:8'
- 'baichuan-inc/Baichuan2-7B-Chat:1024:8'
- 'baichuan-inc/Baichuan2-7B-Chat:2048:8'
- 'Qwen/Qwen-7B-Chat:2048:1'
- 'Qwen/Qwen-7B-Chat:1024:2'
- 'Qwen/Qwen-7B-Chat:2048:2'
- 'Qwen/Qwen-7B-Chat:512:4'
- 'Qwen/Qwen-7B-Chat:1024:4'
- 'Qwen/Qwen-7B-Chat:2048:4'
- 'Qwen/Qwen-7B-Chat:512:8'
- 'Qwen/Qwen-7B-Chat:1024:8'
- 'Qwen/Qwen-7B-Chat:2048:8'
19 changes: 19 additions & 0 deletions python/llm/test/benchmark/stable-version-arc-perf-test-sym_int4.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,8 +12,27 @@ low_bit: 'sym_int4' # default to use 'sym_int4' (i.e. symmetric int4)
batch_size: 1 # default to 1
in_out_pairs:
- '32-32'
- '512-256'
- '1024-128'
- '2048-256'
test_api:
- "transformer_int4_gpu" # on Intel GPU
cpu_embedding: False # whether put embedding to CPU (only avaiable now for gpu win related test_api)
exclude:
- 'meta-llama/Llama-2-7b-chat-hf:2048:4'
- 'meta-llama/Llama-2-7b-chat-hf:1024:8'
- 'meta-llama/Llama-2-7b-chat-hf:2048:8'
- 'THUDM/chatglm2-6b:2048:8'
- 'THUDM/chatglm3-6b:2048:8'
- 'baichuan-inc/Baichuan2-7B-Chat:2048:2'
- 'baichuan-inc/Baichuan2-7B-Chat:1024:4'
- 'baichuan-inc/Baichuan2-7B-Chat:2048:4'
- 'baichuan-inc/Baichuan2-7B-Chat:512:8'
- 'baichuan-inc/Baichuan2-7B-Chat:1024:8'
- 'baichuan-inc/Baichuan2-7B-Chat:2048:8'
- 'Qwen/Qwen-7B-Chat:2048:2'
- 'Qwen/Qwen-7B-Chat:1024:4'
- 'Qwen/Qwen-7B-Chat:2048:4'
- 'Qwen/Qwen-7B-Chat:512:8'
- 'Qwen/Qwen-7B-Chat:1024:8'
- 'Qwen/Qwen-7B-Chat:2048:8'
Loading