Search results

Qwen3-0.6B x86 v1 q4_k path fails due to lm_head name mismatch

Status: Open.

#684 In UbiquitousLearning/mllm;

· kkkzbh opened on Jun 13, 2026

Phi-MoE attention 使用 QNN LPBQ/W4A16 Conv2D 编译失败，Qwen3 同路径可以成功，想确认是否有 op/shape 限制

Status: Open.

#678 In UbiquitousLearning/mllm;

· Lucyliu1234 opened on May 26, 2026

[Bug]: QNN AOT Qwen2系列模型输出乱码

bugSomething isn't workingSomething isn't working

Status: Open.

#665 In UbiquitousLearning/mllm;

· TRM-coding opened on Apr 7, 2026

QNN AOT编译错误

Status: Open.

#654 In UbiquitousLearning/mllm;

· zjj6084bty opened on Mar 19, 2026

Development Roadmap (2026 H1)

Status: Open.

#651 In UbiquitousLearning/mllm;

· chenghuaWang opened on Mar 11, 2026

推测解码的实现

Status: Open.

#649 In UbiquitousLearning/mllm;

· zjj6084bty opened on Mar 10, 2026

Related project: AI agent on Galaxy Watch with 2.8 MB binary

Status: Open.

#648 In UbiquitousLearning/mllm;

· ThinkOffApp opened on Mar 8, 2026

MLLM Qnn AOT 的纯NPU方案如何解决静态计算图部署到DSP内存时，DSP内存可能不足的问题

Status: Open.

#647 In UbiquitousLearning/mllm;

· zjj6084bty opened on Mar 6, 2026

[Bug]: ImportError: Dependency conflict between modelscope and datasets (v3.0+) during model.safetensors

bugSomething isn't workingSomething isn't working

Status: Open.

#646 In UbiquitousLearning/mllm;

· rururush opened on Mar 5, 2026

[Model]: Add support for qwen3-VL

enhancementNew feature or requestNew feature or request

Status: Open.

#644 In UbiquitousLearning/mllm;

· anxiaohuai opened on Feb 28, 2026

regarding performance between mllm v1 and v2, with Snapdragon npu backend

Status: Open.

#642 In UbiquitousLearning/mllm;

· LHB116 opened on Feb 26, 2026

Request for Pre-compiled QNN Backend Binary for Research

Status: Open.

#637 In UbiquitousLearning/mllm;

· huangzhenhua111 opened on Feb 9, 2026