- 06 Feb, 2026 4 commits
-
-
chenyue3 authored
- 在 C++ 算子中移除 query/key/residual 的 contiguous 强校验 - 对非连续输入使用工作张量执行计算(优化路径与 fallback 路径统一) - 计算完成后将结果 copy_ 回原张量,保持 in-place 语义 - 移除 qwen3 / qwen3_moe Python 前向中的 q、k.contiguous() 预处理
-
chenyue3 authored
-
chenyue3 authored
launch_opt_rms_rope 增加 rot_dim=64/128 分发路径 放宽 supports_qwen3_opt 条件,允许 rot_dim=64 进入优化分支 qwen3 / qwen3_moe 将 q_bias、k_bias 参数统一为 q_residual、k_residual qwen3_moe 的自定义 op 注册增加重复注册保护"`
-
laibao authored
- 在 vLLM _C 扩展中新增 rms_rotary_embedding_fuse(注册 op + CUDA kernel),减少对 LightOp 的硬依赖 - 新增环境变量 VLLM_FUSED_RMS_ROPE_BACKEND=auto|vllm|lightop,auto 优先走 vLLM,缺失时回退 LightOp - 更新 Qwen3 / Qwen3-MoE 的 fused 路径按后端选择执行 - 补充 tc_opt benchmark 结果解析脚本 benchmarks/tc_opt/test/parse_bench_results.py
-
- 03 Feb, 2026 2 commits
- 29 Jan, 2026 1 commit
-
-
zhuwenwen authored
-
- 28 Jan, 2026 3 commits
- 27 Jan, 2026 5 commits
- 26 Jan, 2026 4 commits
- 23 Jan, 2026 4 commits
- 22 Jan, 2026 1 commit
-
-
王敏 authored
-
- 21 Jan, 2026 11 commits
-
-
zhuwenwen authored
新增 VLLM_USE_FUSED_RMS_ROPE 分支,走 fused 路径 注册 torch.ops.vllm.rms_rotary_embedding_fuse(direct_register_custom_op) cos_sin_cache 自动转 device/dtype 并缓存,避免每次重复拷贝
-
zhuwenwen authored
for qwen3, VLLM_USE_FUSED_RMS_ROPE=1 (default)
-
zhuwenwen authored
feat: Supprot fp8 channle-wise matmul. See merge request dcutoolkit/deeplearing/vllm!380
-
wanglong3 authored
-
zhuwenwen authored
V0.11.0 dev kvscale See merge request dcutoolkit/deeplearing/vllm!378
-
zhuwenwen authored
feat: support fp8-blockwise matmul impl. See merge request dcutoolkit/deeplearing/vllm!377
-
xiabo authored
-
lixh6 authored
-
xiabo authored
-
zhuwenwen authored
-
zhuwenwen authored
feat(moe/marlin): 移除 VLLM_USE_MARLIN_W16A16_MOE,改为基于 lightop 探测自动启用并一次性缓存决策 See merge request dcutoolkit/deeplearing/vllm!376
-
- 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
-
zhuwenwen authored
[fix]解决glm4 moe + mtp精度异常 See merge request dcutoolkit/deeplearing/vllm!374
-
王敏 authored
-
王敏 authored
-