xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-25 07:45:46 +08:00
Code Issues Packages Projects Releases Wiki Activity

updated

Browse Source 
Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com>
This commit is contained in:
rshaw@neuralmagic.com 2025-06-19 13:16:50 +00:00
parent 87bf6812b2
commit 852ee4b132
1 changed files with 4 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
pd_justfile/Justfile
View File

@ -13,20 +13,10 @@ port PORT:
prefill:
VLLM_NIXL_SIDE_CHANNEL_PORT=$(just port 5557) \
CUDA_VISIBLE_DEVICES=0,1 \
CUDA_VISIBLE_DEVICES=0 \
vllm serve {{MODEL}} \
--port $(just port 8100) \
--tensor-parallel-size 2 \
--enforce-eager \
--disable-log-requests \
--block-size 128 \
--kv-transfer-config '{"kv_connector":"NixlConnector","kv_role":"kv_both"}'
prefill_b:
VLLM_NIXL_SIDE_CHANNEL_PORT=$(just port 5558) \
CUDA_VISIBLE_DEVICES=6 \
vllm serve {{MODEL}} \
--port $(just port 8200) \
--tensor-parallel-size 1 \
--enforce-eager \
--disable-log-requests \
--block-size 128 \
@ -34,22 +24,15 @@ prefill_b:
decode:
VLLM_NIXL_SIDE_CHANNEL_PORT=$(just port 5559) \
CUDA_VISIBLE_DEVICES=2,3,4,5 \
CUDA_VISIBLE_DEVICES=1 \
vllm serve {{MODEL}} \
--port $(just port 8300) \
--tensor-parallel-size 2 \
--tensor-parallel-size 1 \
--enforce-eager \
--disable-log-requests \
--block-size 128 \
--kv-transfer-config '{"kv_connector":"NixlConnector","kv_role":"kv_both"}'
# proxy:
# python "{{vllm-directory}}tests/v1/kv_connector/nixl_integration/toy_proxy_server.py" \
# --port $(just port 8192) \
# --prefiller-port $(just port 8100) $(just port 8200) \
# --prefiller-host localhost localhost \
# --decoder-port $(just port 8300)
proxy:
python "{{vllm-directory}}tests/v1/kv_connector/nixl_integration/toy_proxy_server.py" \
--port $(just port 8192) \