Seeking pre-built flash_attn wheel for Jetso Thor

Robotics & Edge Computing Jetson Systems Jetson Thor
1

I am working on the Jetson Thor platform and trying to set up flash_attn (Flash Attention) for my projects.

The Issue:
When I attempt to install flash_attn by compiling it from source (e.g., via pip install flash-attn --no-binary flash-attn), the system becomes unstable during the build process and ultimately restarts without completing the installation.

I suspect this is due to the high memory pressure or thermal constraints during the heavy compilation workload on the device.

Request:

  1. Does anyone have a pre-compiled wheel (.whl) for flash_attn specifically built for the Jetson Thor?

Any guidance or shared binaries would be greatly appreciated. I am happy to test any community builds.

2

https://pypi.jetson-ai-lab.io/sbsa/cu130

And if you want to experiment with flash_attn 4

https://github.com/Dao-AILab/flash-attention/releases/tag/fa4-v4.0.0.beta4

3

Thank you