GitLab

set VLLM_USE_FUSED_QA_KVA_GEMM=1
feat:w4a8Linear调用apply_int8_linear，以支持blaslt

fuse_moe_fp8接入marlin算子
fix(v1)：修复抢占恢复时 BlockTable 溢出
feat(moe)：新增 VLLM_USE_MOE_W16A16_TRTION 强制 Triton MoE
fix: 解决原版0消耗chunk-prefill崩溃问题
fp8增加fused_moe_gate参数

[qwen3-480b] MoE(TN) configs for nmz TP=8
[opt] 优化deepep相关代码
[fix] 修复deepseek moe模型的awq量化推理bug和精度问题, 修复awq模型的VLLM_USE_LIGHTOP_MOE_SUM_MUL_ADD设置位置, update_state,优化性能，去除冗余操作
pcie 解决custom cudagraph模式需要拷贝的问题，需要配合dtk进行使用
[feat] Switch default w8a8 gemm impl to blaslt. Support w8a8-fp8 GEMM backend.MoE 路由抓取：新增 router_capture 工具链与 envs 统一配置
[envs] set VLLM_CUSTOM_CACHE=1、VLLM_USE_FUSED_RMS_ROPE=1、VLLM_USE_FUSED_FILL_RMS_CAT=1、VLLM_USE_FLASH_ATTN_FP8=1、VLLM_USE_FLASH_MLA_FP8=1、update VLLM_USE_TOPK_RENORM

Signed-off-by: Isotr0py <2037008807@qq.com>

Signed-off-by: simon-mo <simon.mo@hey.com>

Signed-off-by: Daniele Trifirò <dtrifiro@redhat.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>

Signed-off-by: Crucifixion-Fxl <xmufxl@gmail.com>
Co-authored-by: Crucifixion-Fxl <xmufxl@gmail.com>

Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
Co-authored-by: Aaron Pham <contact@aarnphm.xyz>

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

Signed-off-by: Isotr0py <2037008807@qq.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

Signed-off-by: Yang Chen <yangche@fb.com>

Signed-off-by: yihong0618 <zouzou0208@gmail.com>