• Joined on 2026-09-24
wanglin pushed to main at wanglin/enginex-bi100-compat 2026-10-03 18:32:51 +08:00
601a11ccf2 enginex-bi100-compat: 补传 preflight/README/shim/自测 + 扁平化 Dockerfile(仓库无子目录,网页上传不再丢相对路径)
wanglin created branch main in wanglin/enginex-bi100-compat 2026-10-03 18:29:27 +08:00
wanglin pushed to main at wanglin/enginex-bi100-compat 2026-10-03 18:29:27 +08:00
391302289d enginex-bi100-compat v1: preflight 兼容增强引擎(R2/R3/R3b/R4/R7),结构与引擎修复模式同型,25 项自测全通过
wanglin created repository wanglin/enginex-bi100-compat 2026-10-03 18:26:53 +08:00
wanglin pushed to main at wanglin/enginex-vllm-bi100-qwen36 2026-10-02 20:45:46 +08:00
8d3ee4a9e3 v3b: A/B对照 — 移除 --hf-overrides 以逐字节对齐成功配置(ouys/sub400)
wanglin pushed to main at wanglin/enginex-vllm-bi100-qwen36 2026-10-02 20:00:21 +08:00
ee1c3c4c9b add --hf-overrides architectures=Qwen3_5MoeForCausalLM (fix model config arch mismatch)
wanglin pushed to main at wanglin/enginex-vllm-bi100-qwen36 2026-10-02 19:53:28 +08:00
941c42a04f v3: 对齐成功配置 — 补 --enforce-eager 与 corex 运行库环境变量, max-model-len 256K / gmu 0.95 / seqs 4 / bt 8192
wanglin pushed tag v2 to wanglin/enginex-vllm-bi100-qwen36 2026-10-02 12:39:56 +08:00
wanglin pushed to main at wanglin/enginex-vllm-bi100-qwen36 2026-10-01 23:29:18 +08:00
d6dfff0918 v2: conservative max-num-seqs=8 + batched-tokens=16384, revert dtype/256K (v1 OOM crash)
wanglin pushed to main at wanglin/enginex-vllm-bi100-qwen36 2026-09-24 07:57:35 +08:00
a56835199e 更新 computility-run.yaml
wanglin created repository wanglin/enginex-vllm-bi100-qwen36 2026-09-24 07:56:05 +08:00