Search results

[Bug]: vllm-ascend 0.13.0 部署qwen3.5-397B-A17B-w8a8报错。 这个版本的vll-ascend可以部署qwen3.5吗？

bugSomething isn't workingSomething isn't working

multimodal-understanding

qwen-3.5

Status: Open.

#8409 In vllm-project/vllm-ascend;

· ageekor opened on Apr 18, 2026

[Misc]: 开启aclgraph，NPU上算子间空泡大

aclgraph

core-features

Status: Open.

#8404 In vllm-project/vllm-ascend;

· ciao0228 opened on Apr 18, 2026

[Bug]: PCP and eagle3 overlaying problem

advanced-features

bugSomething isn't workingSomething isn't working

long-seq

Status: Open.

#8402 In vllm-project/vllm-ascend;

· lilinsiman opened on Apr 17, 2026

[Bug]: A2 8机1P1D大EP，指定FC，P 必现content is not None问题

bugSomething isn't workingSomething isn't working

Status: Open.

#8399 In vllm-project/vllm-ascend;

· Nagaloop opened on Apr 17, 2026

[Bug]: A3官方镜像开启图模式下启动服务化报错没有磁盘空间

bugSomething isn't workingSomething isn't working

Status: Open.

#8393 In vllm-project/vllm-ascend;

· keira2233 opened on Apr 17, 2026

[Bug]: MTP drafter _v_up_proj crashes with "batchIdx out of range" when batch token count exceeds max_cudagraph_capture_size

advanced-features

bugSomething isn't workingSomething isn't working

mtp/speculative-decode

Status: Open.

#8392 In vllm-project/vllm-ascend;

· lingxling opened on Apr 17, 2026

[RFC]: why should model_loader_extra_config be cleared in revert_to_default()?

RFCRequest For CommentsRequest For Comments

Status: Open.

#8385 In vllm-project/vllm-ascend;

· JohnUtah1997 opened on Apr 17, 2026

[Bug][Ascend] Gemma4 FULL_DECODE_ONLY ACL Graph fails during capture with sizes [1,2,4,8] and during replay with size [1]

triaged

Status: Open.

#8370 In vllm-project/vllm-ascend;

· yanceng305-collab opened on Apr 17, 2026

[Bug]: v0.18.0rc1版本部署Qwen3-vl-32B-W8A8量化版本失败，报错KeyError：‘lm_head.weight'

bugSomething isn't workingSomething isn't working

triaged

Status: Open.

#8367 In vllm-project/vllm-ascend;

· sunwenying880211-swy opened on Apr 17, 2026

[Bug]: 开始推理就报错崩溃，GLM-5.1

bugSomething isn't workingSomething isn't working

glm5

llm-model

triaged

Status: Open.

#8360 In vllm-project/vllm-ascend;

· MaoJianwei opened on Apr 17, 2026

[Bug]: [ v0.11,v0.18.rc1]A general accuracy bug report and solution discussion, theoretically involving all MOE models including the Minimax series and Qwen series.

bugSomething isn't workingSomething isn't working

minimax

multi-modality-generate

triaged

Status: Open.

#8359 In vllm-project/vllm-ascend;

· florenceCH opened on Apr 17, 2026

[Bug]: vllm-ascend 使用v0.13.0.tar.gz版本的，搭建的GLM-4.7-W8A8-floatmtp，发现可以调用大模型，但是对接Claude Code报错400

advanced-features

bugSomething isn't workingSomething isn't working

glm-4

llm-model

mtp/speculative-decode

Status: Open.

#8357 In vllm-project/vllm-ascend;

· zhaiqing opened on Apr 17, 2026