- 16 Jan, 2026 3 commits
- 15 Jan, 2026 2 commits
- 14 Jan, 2026 4 commits
- 12 Jan, 2026 1 commit
-
-
yangql authored
-
- 10 Jan, 2026 6 commits
-
-
laibao authored
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)
-
zhuwenwen authored
-
zhuwenwen authored
区分pcie和hglink custom allreduce的使用 vllm:export VLLM_CUSTOM_CACHE=1 dtk:export HIP_KERNEL_EVENT_SYSTENFENCE=1
-
zhuwenwen authored
-
zhuwenwen authored
-
zhuwenwen authored
-
- 09 Jan, 2026 4 commits
- 08 Jan, 2026 2 commits
- 07 Jan, 2026 4 commits
-
-
laibao authored
- fused_experts_impl 增加 VLLM_USE_MARLIN_W16A16_MOE fast path:首次对 w1/w2 做 Marlin pack 后缓存,避免重复 reorder;并将原始 权重 offload 到 CPU,降低 GPU 双份驻留 - envs 补齐环境变量 VLLM_USE_LIGHTOP_MOE_SUM_MUL_ADD - 更新 fuse_moe_w16a16_marlin.py 的报错提示为 VLLM_USE_LIGHTOP=1 -
zhuwenwen authored
-
laibao authored
新增环境变量开关 VLLM_V1_FAST_TOKEN_ID_COPY(默认关闭)。开启后在 CachedRequestState 中缓存 prompt_token_ids 的 np.int32,并在 add_request 里用 np.copyto 写入 token_ids_cpu,避免长 prompt 场景反复 list->NumPy 转换开销(尤其是抢占/反复进出 batch 时)
-
zhuwenwen authored
-
- 06 Jan, 2026 3 commits
- 05 Jan, 2026 9 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
-