Skip to content

GitLab

Switch branch/tag
  1. 10 Feb, 2026 4 commits
    • zhuwenwen's avatar
      support fa kvcache fp8, add VLLM_USE_QUERY_QUANT to not use q quant(todo) · 0dfb30d5
      zhuwenwen authored 
      [opt] 优化epsp代码, 零消耗添加epsp
update VLLM_USE_FUSED_RMS_ROPE=0 (default). for qwen3, VLLM_USE_FUSED_RMS_ROPE=1 (default)
feat(moe/marlin): Marlin W16A16 MoE 自动探测并预打包（去掉手动开关）
perf(qwen3): 融合 q/k RMSNorm + RoPE
fused_moe_fp8接入lmslim
      0dfb30d5
    • zhuwenwen's avatar
      [qwen3-235b] MoE(TN&NN) configs for nmz TP=8 · 7624bd05
      zhuwenwen authored 
      [qwen3-480b] MoE(TN) configs for nmz TP=8
[opt] 优化deepep相关代码
[fix] 修复deepseek moe模型的awq量化推理bug和精度问题, 修复awq模型的VLLM_USE_LIGHTOP_MOE_SUM_MUL_ADD设置位置, update_state,优化性能，去除冗余操作
pcie 解决custom cudagraph模式需要拷贝的问题，需要配合dtk进行使用
[feat] Switch default w8a8 gemm impl to blaslt. Support w8a8-fp8 GEMM backend.MoE 路由抓取：新增 router_capture 工具链与 envs 统一配置
[envs] set VLLM_CUSTOM_CACHE=1、VLLM_USE_FUSED_RMS_ROPE=1、VLLM_USE_FUSED_FILL_RMS_CAT=1、VLLM_USE_FLASH_ATTN_FP8=1、VLLM_USE_FLASH_MLA_FP8=1、update VLLM_USE_TOPK_RENORM
      7624bd05
    • zhuwenwen's avatar
      Merge branch 'v0.9.2-dev-for-yuanbao' of... · ad1d74cf
      zhuwenwen authored 
      Merge branch 'v0.9.2-dev-for-yuanbao' of http://10.16.6.30/dcutoolkit/deeplearing/vllm into v0.9.2-dev-for-yuanbao
      ad1d74cf
    • zhuwenwen's avatar
      [feat] 适配block和channel fp8, 添加dp attention功能 · 2b7b1a31
      zhuwenwen authored 
      [pd] 修改非堆成切分的判断, pd分离使用默认调度，set VLLM_USE_PD_SPLIT=1
[perf](fused-moe): 预打包 Marlin W16A16 MoE 权重，降低 warmup 显存峰值
[fix]修复丢弃MTP代码报错，pp+chunkprefill多并发input ids更新bug, 修复不开启融合图的断言错误, PP 场景 decode 阶段 token 被误丢弃导致卡住
      2b7b1a31
  3. 09 Feb, 2026 3 commits
  5. 06 Feb, 2026 1 commit
  7. 08 Jan, 2026 5 commits
  9. 07 Jan, 2026 4 commits
  11. 06 Jan, 2026 6 commits
  13. 05 Jan, 2026 5 commits
  15. 04 Jan, 2026 9 commits
  17. 31 Dec, 2025 3 commits