GitLab

update VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT impl

[feat]1.优化ep sequence parallel,区分主模型和mtp逻辑;2.ep sequence parallel添加cudagraph padding到tp_size;3.修复共享专家和deepep combine overlap

- 新增rms_rotary_embedding_fuse自定义操作
- 添加内核配置文件E=160,N=320
- 通过VLLM_USE_FUSED_RMS_ROPE环境变量控制融合路径

add VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT when use USE_FUSED_RMS_QUANT and USE_FUSED_CUSTOM_ALL_REDUCE_RMS_QUANT

restore the default settings of disable_cascade_attn
add VLLM_USE_OPT_ZEROS to replace triton_ (torch.zeros)
set default_max_num_batched_tokens = 10240
update qwen3_moe of layernorm

support prefix cache on kme
fix the error in test_moe caused by moe align not supporting 511 and 211
multi-modal switching to torch implementation on z100l&k100

add VLLM_USE_LIGHTOP_MOE_SUM_MUL_ADD

fix the error in test_moe caused by moe align not supporting 511
multi-modal switching to torch implementation on z100l&k100