- 13 Mar, 2026 1 commit
-
-
yangshj1 authored
-
- 09 Mar, 2026 2 commits
- 08 Feb, 2026 1 commit
-
-
王敏 authored
-
- 28 Jan, 2026 1 commit
-
-
laibao authored
- 缺失请求统一延后重新加入,避免同一轮重复写入 - 抢占恢复用覆盖写入(add_row),正常运行用增量追加(append_row) - 保持正常请求的追加语义不变
-
- 17 Jan, 2026 2 commits
- 13 Jan, 2026 3 commits
- 12 Jan, 2026 2 commits
- 09 Jan, 2026 1 commit
-
-
jujl1 authored
-
- 07 Jan, 2026 1 commit
-
-
laibao authored
- repeat_counts/CPU 元数据为 numpy/array-like 时会在 repeat_interleave/.to() 崩溃 - 统一转换为 CPU torch.Tensor 后再扩展并拷到 GPU
-
- 06 Jan, 2026 1 commit
-
-
jujl1 authored
-
- 04 Jan, 2026 1 commit
-
-
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)
-
- 29 Dec, 2025 1 commit
-
-
yangql authored
-
- 25 Dec, 2025 2 commits
- 24 Dec, 2025 1 commit
-
-
laibao authored
新增环境变量 VLLM_V1_USE_REDUCED_TOPK_TOPP_SAMPLER 用于开关控制 扩展 SamplingMetadata,增加 max_top_k 与 has_any_no_top_k 在 InputBatch 侧计算 top-k 的主机端汇总信息,避免 device 同步 更新 Sampler/TopKTopPSampler 传递并使用新参数以启用优化采样
-
- 23 Dec, 2025 1 commit
-
-
yangql authored
-
- 22 Dec, 2025 2 commits
- 21 Dec, 2025 1 commit
-
-
yangql authored
-
- 18 Dec, 2025 3 commits
- 09 Dec, 2025 1 commit
-
-
王敏 authored
-
- 08 Dec, 2025 2 commits
- 05 Dec, 2025 1 commit
-
-
laibao authored
- 在 InputBatch.refresh_metadata 中为展开后的采样元数据引入 repeat_count 记录重复次数 - 完善元数据刷新逻辑以适配 reject-sampling 优化路径 - 更新 GPUModelRunnerBase,在 batch 处理阶段正确消费新的采样元数据与重复计数
-
- 28 Nov, 2025 1 commit
-
-
jujl1 authored
-
- 27 Nov, 2025 1 commit
-
-
王敏 authored
-
- 26 Nov, 2025 2 commits
- 21 Nov, 2025 1 commit
-
-
jujl1 authored
-
- 15 Nov, 2025 1 commit
-
-
王敏 authored
-
- 13 Nov, 2025 1 commit
-
-
王敏 authored
-
- 27 Oct, 2025 1 commit
-
-
王敏 authored
-
- 17 Oct, 2025 1 commit
-
-
zhuwenwen authored
-