- 03 Feb, 2026 1 commit
-
-
wanglong3 authored
-
- 28 Jan, 2026 1 commit
-
-
laibao authored
增加环境变量开关,禁用 Marlin W16A16 MoE 路径 强制 Triton 且权重已是 Marlin packed 时给出明确报错 Marlin 支持探测改为 best-effort(不再依赖 VLLM_USE_LIGHTOP)
-
- 27 Jan, 2026 1 commit
-
-
SAC_fanth authored
-
- 23 Jan, 2026 1 commit
-
-
zhuwenwen authored
-
- 21 Jan, 2026 5 commits
- 20 Jan, 2026 5 commits
-
-
laibao authored
- 使用 get_moe_cuda_marlin_config_w16a16(status) 判断 W16A16 Marlin MoE 是否可用 - 在 FusedMoE 初始化阶段计算并缓存 _marlin_w16a16_moe_enabled,满足条件时强制 use_nn_moe=False - 权重加载后按缓存结果进行一次性 Marlin pack;运行时按 packed 标记走 Marlin fast path - 删除 envs.py 中 VLLM_USE_MARLIN_W16A16_MOE 环境变量定义与解析逻辑
-
zhuwenwen authored
-
王敏 authored
-
zhuwenwen authored
-
laibao authored
-
- 19 Jan, 2026 2 commits
- 14 Jan, 2026 3 commits
- 10 Jan, 2026 2 commits
- 07 Jan, 2026 3 commits
- 06 Jan, 2026 3 commits
- 05 Jan, 2026 6 commits
- 04 Jan, 2026 1 commit
-
-
laibao authored
实现了用于优化张量计算的 rms_mrope_fuse 和 rms_mrope_fuse_fake 方法 更新了 forward:在满足条件时走新的 M-RoPE 融合路径 增强了 Qwen3MoeModel 对动态参数维度的支持,以适配该功能
-
- 24 Dec, 2025 1 commit
-
-
yangql authored
-
- 23 Dec, 2025 1 commit
-
-
yangql authored
-
- 22 Dec, 2025 3 commits
- 20 Dec, 2025 1 commit
-
-
zhuwenwen authored
-