Search results

qwen3的模型，请求时候怎么配，能没有think的过程，直接输出结果

Status: Open.

#887 In jd-opensource/xllm;

· w066650 opened on Feb 5, 2026

VLM models crash in NPU Graph Mode during decode phase due to mRoPE position tensor shape mismatch

Status: Open.

#884 In jd-opensource/xllm;

· QwertyJack opened on Feb 4, 2026

[Bug]: early return in prefetch_from_storage

bugSomething isn't workingSomething isn't working

Status: Open.

#863 In jd-opensource/xllm;

· shifengmin opened on Feb 2, 2026

[Roadmap] xLLM Roadmap H1 2026.

Status: Open.

#847 In jd-opensource/xllm;

· XuZhang99 opened on Jan 31, 2026

MTP k>1 crashes during warmup due to uninitialized embed_tokens weights

Status: Open.

#835 In jd-opensource/xllm;

· QwertyJack opened on Jan 29, 2026

[Bug]: 使用xllm-0.7.2-release-hb-rc2-arm镜像PD 分离部署运行 xllm_master_serving 时缺少 librust_tokenizers.so

bugSomething isn't workingSomething isn't working

Status: Open.

#833 In jd-opensource/xllm;

· w1ida opened on Jan 29, 2026

[Question]: xllm 技术报告性能复现问题：与 vllm_ascend 最新版本实测结果不一致

questionFurther information is requestedFurther information is requested

Status: Open.

#832 In jd-opensource/xllm;

· w1ida opened on Jan 29, 2026

KV cache memory leak when prefix cache is enabled

Status: Open.

#828 In jd-opensource/xllm;

· QwertyJack opened on Jan 29, 2026

bugfix: GLM4-MoE decoder layer crashes when falling back to eager mode

Status: Open.

#826 In jd-opensource/xllm;

· QwertyJack opened on Jan 28, 2026

[Feature]: 支持 Suffix Decoding 以提升推理阶段输出吞吐

feature

Status: Open.

#824 In jd-opensource/xllm;

· w1ida opened on Jan 28, 2026

[Bug]: qwen2.5模型npu设备多次推理结果不一致

bugSomething isn't workingSomething isn't working

Status: Open.

#814 In jd-opensource/xllm;

· xiangqianzsh opened on Jan 27, 2026

[Question]: 请求头里Content-Length为空时，异常退出。

questionFurther information is requestedFurther information is requested

Status: Open.

#805 In jd-opensource/xllm;

· dantianji opened on Jan 27, 2026