|
|
|
@ -72,15 +72,15 @@ jobs:
|
|
|
|
|
export REF_NAME=bigscience/bloom-560m
|
|
|
|
|
|
|
|
|
|
python -m cli.run_server --converted_model_name_or_path $MODEL_NAME --block_indices 0:12 \
|
|
|
|
|
--identity tests/test.id --host_maddrs /ip4/127.0.0.1/tcp/31337 --throughput 1 \
|
|
|
|
|
--new_swarm --identity tests/test.id --host_maddrs /ip4/127.0.0.1/tcp/31337 --throughput 1 \
|
|
|
|
|
--torch_dtype float32 --compression NONE --attn_cache_size 0.2GiB &> server1.log &
|
|
|
|
|
SERVER1_PID=$!
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 cli.run_server --converted_model_name_or_path $MODEL_NAME --block_indices 12:22 \
|
|
|
|
|
--initial_peers $INITIAL_PEERS --throughput 1 --torch_dtype float32 &> server2.log &
|
|
|
|
|
SERVER2_PID=$!
|
|
|
|
@ -94,20 +94,20 @@ jobs:
|
|
|
|
|
python -m cli.run_server --converted_model_name_or_path $MODEL_NAME --block_indices 4:16 \
|
|
|
|
|
--torch_dtype float32 --initial_peers $INITIAL_PEERS --throughput 1 &> server4.log &
|
|
|
|
|
SERVER4_PID=$!
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
python -m cli.run_server --converted_model_name_or_path $MODEL_NAME --num_blocks 3 \
|
|
|
|
|
--initial_peers $INITIAL_PEERS --throughput 1 --torch_dtype float32 &> server5.log &
|
|
|
|
|
SERVER5_PID=$!
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 $SERVER5_PID # ensure all servers survived init
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
PYTHONPATH=. pytest tests --durations=0 --durations-min=1.0 -v
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
kill -0 $SERVER1_PID $SERVER2_PID $SERVER3_PID $SERVER4_PID $SERVER5_PID # ensure all servers survived tests
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
kill -s SIGINT $SERVER1_PID $SERVER2_PID $SERVER3_PID $SERVER4_PID $SERVER5_PID $LOGGER_PID
|
|
|
|
|
echo "Done!"
|
|
|
|
|