1. 10 Feb, 2026 1 commit
    • 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
  2. 19 Aug, 2025 1 commit
  3. 24 Jul, 2025 1 commit
  4. 27 Jun, 2025 1 commit
  5. 26 Jun, 2025 1 commit
  6. 03 Jun, 2025 1 commit
  7. 28 Apr, 2025 1 commit