petals/.github/workflows/run-tests.yaml
justheuristic 5af04524dd
Split long sequences into chunks (#403)
This PR is designed to avoid OOMs when processing long sequences that happen due to the huge attention logits matrices.

Co-authored-by: Alexander Borzunov <borzunov.alexander@gmail.com>
2023-07-22 23:10:46 +04:00

75 lines
2.9 KiB
YAML

name: Tests
on:
push:
branches: [ main ]
pull_request:
jobs:
run-tests:
runs-on: ubuntu-latest
strategy:
matrix:
python-version: [ '3.8', '3.9', '3.10', '3.11' ]
fail-fast: false
timeout-minutes: 15
steps:
- name: Checkout
uses: actions/checkout@v3
- name: Set up Python
uses: actions/setup-python@v3
with:
python-version: ${{ matrix.python-version }}
- name: Cache dependencies
uses: actions/cache@v3
with:
path: ~/.cache/pip
key: Key-v1-${{ matrix.python-version }}-${{ hashFiles('setup.cfg') }}
- name: Install dependencies
run: |
python -m pip install --upgrade pip
pip install .[dev]
- name: Test
run: |
export MODEL_NAME=bigscience/bloom-560m
export REF_NAME=bigscience/bloom-560m
export ADAPTER_NAME=artek0chumak/bloom-560m-safe-peft
python -m petals.cli.run_server --converted_model_name_or_path $MODEL_NAME --block_indices 0:12 \
--new_swarm --identity tests/test.id --host_maddrs /ip4/127.0.0.1/tcp/31337 --throughput 1 \
--torch_dtype float32 --compression NONE --attn_cache_tokens 2048 --max_chunk_size_bytes 1024 \
--adapters $ADAPTER_NAME &> server1.log &
SERVER1_PID=$!
sleep 5 # wait for the first server to initialize DHT
export INITIAL_PEERS=/ip4/127.0.0.1/tcp/31337/p2p/QmS9KwZptnVdB9FFV7uGgaTq4sEKBwcYeKZDfSpyKDUd1g
# ^-- server 1 multiaddr is determined by --identity and --host_maddrs
python -m petals.cli.run_server --converted_model_name_or_path $MODEL_NAME --block_indices 12:22 \
--initial_peers $INITIAL_PEERS --throughput 1 --torch_dtype float32 --adapters $ADAPTER_NAME &> server2.log &
SERVER2_PID=$!
sleep 10 # wait for initial servers to declare blocks, then let server decide which blocks to serve
python -m petals.cli.run_server --converted_model_name_or_path $MODEL_NAME --block_indices 12:15 \
--initial_peers $INITIAL_PEERS --throughput 1 --torch_dtype float32 --tensor_parallel_devices cpu cpu &> server3.log &
SERVER3_PID=$!
python -m petals.cli.run_server --converted_model_name_or_path $MODEL_NAME --num_blocks 3 \
--initial_peers $INITIAL_PEERS --throughput 1 --torch_dtype float32 --adapters $ADAPTER_NAME &> server4.log &
SERVER4_PID=$!
tail -n 100 -f server*.log &
LOGGER_PID=$!
sleep 30 # wait for servers to download layers
kill -0 $SERVER1_PID $SERVER2_PID $SERVER3_PID $SERVER4_PID # ensure all servers survived init
pytest tests --durations=0 --durations-min=1.0 -v
kill -0 $SERVER1_PID $SERVER2_PID $SERVER3_PID $SERVER4_PID # ensure all servers survived tests
kill -s SIGINT $SERVER1_PID $SERVER2_PID $SERVER3_PID $SERVER4_PID $LOGGER_PID
echo "Done!"