Issues

• [Feature]: 请问什么时候能够支持lora微调

  enhancement
  UbiquitousLearning/mllm#705 · Filter by author virus188 opened on Sep 3, 2026

• [Feature]: Support for IBM granite vision 4.1 4b

  enhancement
  UbiquitousLearning/mllm#689 · Filter by author Dineshkumar-Anandan-ZS0367 opened on Jul 13, 2026

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

  UbiquitousLearning/mllm#684 · Filter by author 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

  UbiquitousLearning/mllm#678 · Filter by author Lucyliu1234 opened on May 26, 2026
  2

• [Bug]: QNN AOT Qwen2系列模型输出乱码

  bug
  UbiquitousLearning/mllm#665 · Filter by author TRM-coding opened on Apr 7, 2026
  3

• QNN AOT编译错误

  UbiquitousLearning/mllm#654 · Filter by author zjj6084bty opened on Mar 19, 2026
  13

• Development Roadmap (2026 H1)

  UbiquitousLearning/mllm#651 · Filter by author chenghuaWang opened on Mar 11, 2026
  9

• 推测解码的实现

  UbiquitousLearning/mllm#649 · Filter by author zjj6084bty opened on Mar 10, 2026
  3

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

  UbiquitousLearning/mllm#648 · Filter by author ThinkOffApp opened on Mar 8, 2026
  1

• 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

  UbiquitousLearning/mllm#647 · Filter by author zjj6084bty opened on Mar 6, 2026
  3

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

  bug
  UbiquitousLearning/mllm#646 · Filter by author rururush opened on Mar 5, 2026
  1

• [Model]: Add support for qwen3-VL

  enhancement
  UbiquitousLearning/mllm#644 · Filter by author anxiaohuai opened on Feb 28, 2026
  1

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

  UbiquitousLearning/mllm#642 · Filter by author LHB116 opened on Feb 26, 2026
  2

• Request for Pre-compiled QNN Backend Binary for Research

  UbiquitousLearning/mllm#637 · Filter by author huangzhenhua111 opened on Feb 9, 2026
  2

• How to use the qwen-1.5-0.5b model?

  UbiquitousLearning/mllm#635 · Filter by author zjj6084bty opened on Feb 4, 2026
  1

• ddr memory requrement in long-context (8k or more)

  UbiquitousLearning/mllm#632 · Filter by author sigpro opened on Feb 3, 2026
  2

• Support for Non-Qwen Models on Hexagon NPU (e.g., LLaMA, Gemma, Mistral)?

  UbiquitousLearning/mllm#598 · Filter by author kimminsu38oo opened on Jan 10, 2026
  1

• [Bug]: random characters as response and crash opencl backend demo

  bug
  UbiquitousLearning/mllm#590 · Filter by author ramcn opened on Jan 8, 2026
  1

• run_qwen_qnn.sh failed to run on the Xiaomi 14.

  UbiquitousLearning/mllm#574 · Filter by author singlecatlmx opened on Dec 25, 2025
  2

• The obtained qwen1.5-1.8b-chat-rot_q4_0.mllm file size does not match the one in the ModelScope repository.

  UbiquitousLearning/mllm#573 · Filter by author singlecatlmx opened on Dec 24, 2025
  4

• [v2] Missing demo scripts for Android QNN backend execution (was available in v1)

  UbiquitousLearning/mllm#560 · Filter by author OujaKivi opened on Dec 5, 2025
  34

• The QNN_README link is missing in README.md

  UbiquitousLearning/mllm#559 · Filter by author OujaKivi opened on Dec 5, 2025
  3

• 请问v2现在支持npu(qualcomm)qwen2-vl-2b吗？

  UbiquitousLearning/mllm#550 · Filter by author 970814 opened on Nov 24, 2025
  2

• 视觉权重无法找到，请问 demo_qwen2_vl_npu.cpp 中的模型文件如何获取

  UbiquitousLearning/mllm#535 · Filter by author 970814 opened on Nov 19, 2025
  2

  

• QNN NPU on Snapdragon 8 Gen3 (RedMagic Pad) with Qwen2-VL-2B: gibberish output then Aborted, CPU path OK

  UbiquitousLearning/mllm#528 · Filter by author 970814 opened on Nov 18, 2025