dylanyunlong/project_6_89d52222
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

fix yaml

Browse Source
This commit is contained in:
root
2026-09-02 02:44:22 +00:00
parent 27878c8689
commit 7f16b711a7
1 changed files with 38 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

71
computility-run.yaml
View File

@@ -1,48 +1,53 @@
concurrency: 1 concurrency: 1
command: command:
- bash - python3
- -c - -m
- >- - vllm.entrypoints.openai.api_server
python3 /workspace/qwen3_6_scripts/patch_chat_template.py /model 2>&1 || echo '[runtime] chat template patch failed'; - --model
exec python3 -m vllm.entrypoints.openai.api_server - /model
--model /model - --served-model-name
--served-model-name llm - llm
--max-model-len 131072 - --max-model-len
--gpu-memory-utilization 0.92 - '131072'
--trust-remote-code - --gpu-memory-utilization
-tp 4 - '0.92'
--max-num-seqs 2 - --trust-remote-code
--disable-log-requests - -tp
--disable-frontend-multiprocessing - '4'
--max-num-batched-tokens 4096 - --max-num-seqs
--enable-chunked-prefill - '2'
--max-seq-len-to-capture 32768 - --disable-log-requests
--enable-auto-tool-choice - --disable-frontend-multiprocessing
--tool-call-parser qwen3_coder - --max-num-batched-tokens
--reasoning-parser qwen3 - '4096'
--enable-prefix-caching - --enable-chunked-prefill
--enforce-eager - --max-seq-len-to-capture
--dtype half - '32768'
- --enable-auto-tool-choice
- --tool-call-parser
- qwen3_coder
- --reasoning-parser
- qwen3
- --enable-prefix-caching
- --enforce-eager
- --dtype
- half
env: env:
- name: VLLM_ENGINE_ITERATION_TIMEOUT_S - name: VLLM_ENGINE_ITERATION_TIMEOUT_S
value: '3600' value: 3600
- name: BI100_MAX_NUM_SEQS - name: BI100_MAX_NUM_SEQS
value: '2' value: 2
# --- MoE kernel selection ---
- name: BI100_MOE_COREX_DIRECT_ROUTED - name: BI100_MOE_COREX_DIRECT_ROUTED
value: '1' value: 1
- name: BI100_MOE_COREX_TOPK_SOFTMAX - name: BI100_MOE_COREX_TOPK_SOFTMAX
value: '1' value: 1
# --- GDN kernel selection ---
- name: BI100_GDN_COREX_PACKED_DECODE - name: BI100_GDN_COREX_PACKED_DECODE
value: '1' value: 1
# --- Hybrid KV/GDN cache ---
- name: BI100_HYBRID_KV_ACCOUNTING - name: BI100_HYBRID_KV_ACCOUNTING
value: full_attention value: full_attention
- name: BI100_GDN_CACHE_POLICY - name: BI100_GDN_CACHE_POLICY
value: admission64 value: admission64
- name: BI100_GDN_RESTORE_MODE - name: BI100_GDN_RESTORE_MODE
value: hybrid64 value: hybrid64
# --- Image fetch timeout (container network) ---
- name: VLLM_IMAGE_FETCH_TIMEOUT - name: VLLM_IMAGE_FETCH_TIMEOUT
value: '10' value: 10