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

Fix N150 llama3 demo CI tests to properly save perf information to superset #15220

Merged
merged 1 commit into from
Nov 20, 2024
+21 −15
Merged

Fix N150 llama3 demo CI tests to properly save perf information to superset #15220

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
36 changes: 21 additions & 15 deletions tests/scripts/single_card/run_single_card_demo_tests.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,21 +42,6 @@ run_common_perf_tests(){
# Skipping: kills WH cards, check issue #14440
# WH_ARCH_YAML=wormhole_b0_80_arch_eth_dispatch.yaml pytest -n auto models/demos/wormhole/mistral7b/demo/demo_with_prefill.py --timeout 420; fail+=$?

# Llama3.1-8B
llama8b=/mnt/MLPerf/tt_dnn-models/llama/Meta-Llama-3.1-8B-Instruct/
# Llama3.2-1B
llama1b=/mnt/MLPerf/tt_dnn-models/llama/Llama3.2-1B-Instruct/
# Llama3.2-3B
llama3b=/mnt/MLPerf/tt_dnn-models/llama/Llama3.2-3B-Instruct/
# Llama3.2-11B
llama11b=/mnt/MLPerf/tt_dnn-models/llama/Llama3.2-11B-Vision-Instruct/

# Run all Llama3 tests for 8B, 1B, and 3B weights
for llama_dir in "$llama1b" "$llama3b" "$llama8b" "$llama11b"; do
LLAMA_DIR=$llama_dir WH_ARCH_YAML=wormhole_b0_80_arch_eth_dispatch.yaml pytest -n auto models/demos/llama3/demo/demo.py --timeout 600; fail+=$?
echo "LOG_METAL: Llama3 tests for $llama_dir completed"
done

# Mamba
WH_ARCH_YAML=wormhole_b0_80_arch_eth_dispatch.yaml pytest -n auto --disable-warnings -q -s --input-method=json --input-path='models/demos/wormhole/mamba/demo/prompts.json' models/demos/wormhole/mamba/demo/demo.py --timeout 420; fail+=$?

Expand Down Expand Up @@ -91,6 +76,27 @@ run_n300_perf_tests(){

run_common_perf_tests; fail+=$?

# Llama3.1-8B
llama8b=/mnt/MLPerf/tt_dnn-models/llama/Meta-Llama-3.1-8B-Instruct/
# Llama3.2-1B
llama1b=/mnt/MLPerf/tt_dnn-models/llama/Llama3.2-1B-Instruct/
# Llama3.2-3B
llama3b=/mnt/MLPerf/tt_dnn-models/llama/Llama3.2-3B-Instruct/
# Llama3.2-11B
llama11b=/mnt/MLPerf/tt_dnn-models/llama/Llama3.2-11B-Vision-Instruct/

# Run all Llama3 tests for 1B, 3B, 8B and 11B weights for N150
# To ensure a proper perf measurement and dashboard upload of the Llama3 models on a N150, we have to run them on the N300 perf pipeline for now
for llama_dir in "$llama1b" "$llama3b" "$llama8b" "$llama11b"; do
FAKE_DEVICE=N150 LLAMA_DIR=$llama_dir WH_ARCH_YAML=wormhole_b0_80_arch_eth_dispatch.yaml pytest -n auto models/demos/llama3/demo/demo.py --timeout 600; fail+=$?
echo "LOG_METAL: Llama3 tests for $llama_dir completed on N150"
done
# Run all Llama3 tests for 1B, 3B, 8B and 11B weights
for llama_dir in "$llama1b" "$llama3b" "$llama8b" "$llama11b"; do
LLAMA_DIR=$llama_dir WH_ARCH_YAML=wormhole_b0_80_arch_eth_dispatch.yaml pytest -n auto models/demos/llama3/demo/demo.py --timeout 600; fail+=$?
echo "LOG_METAL: Llama3 tests for $llama_dir completed"
done

# Falcon7b (perf verification for 128/1024/2048 seq lens and output token verification)
WH_ARCH_YAML=wormhole_b0_80_arch_eth_dispatch.yaml pytest -n auto --disable-warnings -q -s --input-method=json --input-path='models/demos/falcon7b_common/demo/input_data.json' models/demos/wormhole/falcon7b/demo_wormhole.py; fail+=$?

Expand Down
Loading