Skip to content

Request for Pre-compiled QNN Backend Binary for Research #637

@huangzhenhua111

Description

@huangzhenhua111

Hi mllm team,

I'm conducting NPU-aware algorithm optimization research using Qualcomm Device Cloud (QDC) with Snapdragon 8 Gen 3. Due to the lack of Hexagon SDK access and compilation environment constraints on the cloud platform, I'm unable to compile the QNN backend from source.

Could you please provide:

  1. Pre-compiled QNN backend binaries for Android ARM64
  2. Or guidance on using QNN AOT compilation to avoid device-side operator package compilation

My setup:

  • Device: QDC Snapdragon 8 Gen 3
  • QNN SDK: v2.14 (device firmware V73)
  • Goal: Run NPU-aware optimization experiments (token pruning, quantization, KV cache)

Models are already downloaded. I can work with any recent mllm version that's compatible.

Thank you!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions