sskj/deploy/profiles/pro6000/glm53_nvfp4_pro6000_sglang_tp4pp2_indexcache.env
yy-fighting b3165a1d3c feat(pro6000/GLM-5.3): 部署方案入库(TP8+EAGLE 生产标准 / 场景二高并发变体 / TP4PP2+IndexCache / E7b CAR 实验补丁 + deploy profiles)
- deploy_glm53_605.sh:配置 A 生产标准(60.5 在役,全 8 台 md5 fcd9109b 一致),
  支持 MEMFRAC/STEPS/TOPK/DRAFT/CHUNK/EXTRA/RESTART 调参;场景二高并发变体
  仅改 mrr32 + decode 图 bs{4,8,12,16}(KV 池 16.4 驻留上限)
- deploy_glm53_optimal(_s1).sh:配置 B TP4PP2+IndexCache(场景二最优 +41~79%;
  s1 形态唯一差异 radix-on)
- deploy_glm53_607_exp.sh + car_patch/:E7b custom-AR 1stage 补丁(cc1 decode
  每步 -14~-16%,实验性仅 cc1-2 验证;补丁文件与 60.7:/root/patches md5 一致)
- deploy/profiles/pro6000/:两个标准 profile(sskj.deploy 可消费),关键踩坑
  与场景二变体、parser 缺口均在注释中标注
2026-09-08 11:38:06 +08:00

44 lines
2.4 KiB
Bash
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# GLM-5.3-NVFP4 SGLang TP=4 PP=2 + IndexCache deployment profile (single RTX 6000D node).
# 场景二16k 独立输入、cc8-32、prefill 主导)最优配置:吞吐 +41~79%、cc32 TTFT 减半
# (对比 TP8+EAGLE 生产配置2026-09-07 真实语料基线)。
# 可执行部署脚本experiments/pro6000/glm53_nvfp4_pro6000d_sglang_dual_scenario_bench/scripts/deploy_glm53_optimal.sh
#
# 关键点(实测踩坑,勿随意改):
# - index_topk_freq=4 为模型原生默认(层轴索引复用省 75% indexer无质量损失
# - 禁投机解码PP2 与投机框架不兼容(已实测)
# - 本文件为场景二形态(--disable-radix-cache独立输入无前缀复用
# 场景一 90% 命中对比须启用 radix唯一差异去掉 --disable-radix-cache
# 见 deploy_glm53_optimal_s1.sh
# - mem 0.850.90 下 cuda graph 捕获余量不足会运行时 OOMKV 池 569,600 = TP8 的 2.06 倍)
# - --disable-custom-all-reduceTP4 over PCIe 用自定义 AR 在本栈无收益
# - 已知缺口:未带 --tool-call-parser glm47 --reasoning-parser glm45质量门 6/7
# tool call 失败纯属参数缺失,非模型问题);上生产必须补 parser
# - 场景一90% 命中低并发)该配置全面劣于 TP8+EAGLE输出吞吐 25~65%),勿混用选型
PLATFORM=pro6000
EXPERIMENT=glm53_nvfp4_pro6000_sglang_tp4pp2_indexcache
MODEL_NAME=GLM-5.3-NVFP4
ENGINE=sglang
RUNTIME=docker
DOCKER_IMAGE=lmsysorg/sglang:nightly-dev-20260828-daf63171
CONTAINER_NAME=glm53-nvfp4
MODEL_PATH=/data/hf_models/GLM-5.3-NVFP4
SERVED_MODEL_NAME=/data/hf_models/GLM-5.3-NVFP4
PORT=30000
HEALTH_PATH=/health
HEALTH_WAIT_S=2400
CONTAINER_PYTHON=python3
TP=4
DP=1
DEVICE_VARS="CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7"
ENGINE_ENV="PYTHONUNBUFFERED=1 HF_HUB_OFFLINE=1 TRANSFORMERS_OFFLINE=1"
DOCKER_FLAGS="--gpus all --ipc=host --shm-size 64g --entrypoint '' -p ${PORT}:${PORT}"
VOLUMES="/data/hf_models:/data/hf_models:ro"
BOOTSTRAP="exec python3 -m sglang.launch_server ${LAUNCH_ARGS}"
LAUNCH_ARGS="--model-path ${MODEL_PATH} --tp-size ${TP} --pp-size 2 --mem-fraction-static 0.85 --max-running-requests 48 --disable-radix-cache --disable-shared-experts-fusion --moe-runner-backend flashinfer_cutlass --disable-flashinfer-autotune --disable-custom-all-reduce --chunked-prefill-size 16384 --json-model-override-args '{\"index_topk_freq\": 4}' --host 0.0.0.0 --port ${PORT}"