GitLab

add VLLM_USE_FUSED_CACHE_QUANT_BMM_MLA to use fused rmsnorm + contiguous + rope(for dpsk-v3) + concat_and_cache_mla + q quant, control bmm(todo) + cat +mla (fp8)

新增环境变量 VLLM_MOE_ROUTER_CAPTURE / DIR / RANK / MAX_LAYERS / NUM_TOKENS_* 用于开关与过滤控制
新增 router_capture.py，支持按 num_tokens 分桶抓取 router logits 并落盘
在 qwen3_moe 中接入抓取逻辑，默认关闭，仅在开启时记录
固定 skip_profile / skip_stack_funcs 为默认启用，避免抓到 warmup/profile 形状
统一配置入口到 vllm.envs，作为运行时基准

Switch default w8a8 gemm impl to blaslt.

See merge request dcutoolkit/deeplearing/vllm!368

V0.9.2 dev yql 1.15

See merge request dcutoolkit/deeplearing/vllm!367

vllm：export VLLM_CUSTOM_CACHE=1
dtk：export HIP_KERNEL_EVENT_SYSTENFENCE=1

V0.9.2 dev fix pp+cp

See merge request dcutoolkit/deeplearing/vllm!364

fix PP 场景 decode 阶段 token 被误丢弃导致卡住

See merge request dcutoolkit/deeplearing/vllm!363

  - decode 已开始时不再按 partial prefill 丢弃 sampled token，避免 new_token_ids=[] 循环拖尾

修改非堆成切分的判断

See merge request dcutoolkit/deeplearing/vllm!362

nmz适配block和channel fp8

See merge request dcutoolkit/deeplearing/vllm!360

fix: 修复不开启融合图的断言错误。

See merge request dcutoolkit/deeplearing/vllm!359

fix: 修复 expanded sampling metadata 对 numpy/array-like 输入不兼容导致崩溃   perf(fused-moe): 预打包 Marlin W16A16 MoE 权重，降低 warmup 显存峰值

See merge request dcutoolkit/deeplearing/vllm!357

  - 在 post-load hook 中对 w13/w2 做 per-expert Marlin pack，并替换为 packed 参数
  - Marlin fast path 仅接受 packed 权重；未预打包则 fail fast，避免运行时 packing 峰值/不确定性
  - 更新 Marlin wrapper 的入参与 shape 推导（从 packed layout 计算 K/N）

Remove print.

See merge request dcutoolkit/deeplearing/vllm!356

feat: Support enable rms quant and shared expert overlap at same time.

See merge request dcutoolkit/deeplearing/vllm!352

fix：更新 Marlin W16A16 MoE 中支持的断言消息

See merge request dcutoolkit/deeplearing/vllm!349

解决custom allreduce在dp情况下的其服务错误问题

See merge request dcutoolkit/deeplearing/vllm!348

  - repeat_counts/CPU 元数据为 numpy/array-like 时会在 repeat_interleave/.to() 崩溃
  - 统一转换为 CPU torch.Tensor 后再扩展并拷到 GPU