gahow/aituner
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Profile replicated router without TP group

Browse Source
This commit is contained in:
Gahow Wang
2026-07-16 22:49:31 +08:00
parent fcca259475
commit 414838a799
2 changed files with 4 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
runs/frontier-qwen30-vllm020-profile-v1/jobs_router_full.toml
View File

@@ -1,18 +1,18 @@
version = 1 version = 1
[[jobs]] [[jobs]]
name = "qwen30-vllm020-router-full-20260716-v1" name = "qwen30-vllm020-router-full-20260716-v2-disable-tp-init"
gpus = 1 gpus = 1
gpu_model = "H20" gpu_model = "H20"
hosts = ["dash0"] hosts = ["dash0"]
command = "cd /home/admin/cpfs/wjh/aituner/aituner-qwen30-vllm020-profile-v1/runs/frontier-qwen30-vllm020-profile-v1 && timeout --signal=TERM --kill-after=30s 1020 bash run_router_full.sh" command = "cd /home/admin/cpfs/wjh/aituner/aituner-qwen30-vllm020-profile-v1/runs/frontier-qwen30-vllm020-profile-v1 && timeout --signal=TERM --kill-after=30s 1020 bash run_router_full.sh"
artifacts = ["artifacts/router-full-v1"] artifacts = ["artifacts/router-full-v2"]
[jobs.env] [jobs.env]
HOME = "/tmp/wjh" HOME = "/tmp/wjh"
XDG_CACHE_HOME = "/tmp/wjh/.cache" XDG_CACHE_HOME = "/tmp/wjh/.cache"
VLLM_CACHE_ROOT = "/tmp/wjh/.cache/vllm" VLLM_CACHE_ROOT = "/tmp/wjh/.cache/vllm"
OUTPUT_ROOT = "/home/admin/cpfs/wjh/aituner/aituner-qwen30-vllm020-profile-fleet/artifacts/router-full-v1" OUTPUT_ROOT = "/home/admin/cpfs/wjh/aituner/aituner-qwen30-vllm020-profile-fleet/artifacts/router-full-v2"
VENV_ROOT = "/tmp/wjh/venvs/vllm-0.20.0-cu129-profiler-v1" VENV_ROOT = "/tmp/wjh/venvs/vllm-0.20.0-cu129-profiler-v1"
VLLM_SOURCE = "/home/admin/cpfs/wjh/agentic-kv/third_party/vllm_v20_build" VLLM_SOURCE = "/home/admin/cpfs/wjh/agentic-kv/third_party/vllm_v20_build"
MODEL = "/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B" MODEL = "/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B"

1
runs/frontier-qwen30-vllm020-profile-v1/profile_vllm020_router.py
View File

@@ -115,6 +115,7 @@ def main() -> None:
bias=False, bias=False,
quant_config=None, quant_config=None,
prefix="model.layers.0.mlp.gate", prefix="model.layers.0.mlp.gate",
disable_tp=True,
).to(device=device, dtype=torch.bfloat16) ).to(device=device, dtype=torch.bfloat16)
gate.weight.data.uniform_(-0.01, 0.01) gate.weight.data.uniform_(-0.01, 0.01)