Feature Request: Support for NVIDIA Jetson (Python 3.10, JetPack 6.x, CUDA)

[peterbb148]

🚩 Current Situation

Running Evo2 on NVIDIA Jetson (Nano, Orin, Xavier, etc.) currently presents several hurdles:
• The official Jetson PyTorch builds by NVIDIA only support Python 3.10 (JetPack 6.x) with CUDA/GPU acceleration.
• Evo2’s Hugging Face integration explicitly requires Python 3.11 due to dependencies on newer libraries and constructs.
• Attempting to run Evo2 on Python 3.10 leads to compatibility issues with libraries (e.g., NumPy 2.x compatibility, transformers’ trust_remote_code=True requirement).

✅ Desired Outcome

Explicitly support NVIDIA Jetson devices (JetPack 6.x) running Python 3.10, ensuring seamless integration without manual workarounds. Specifically:
• Ensure compatibility with Python 3.10 out-of-the-box (e.g., avoid Python 3.11-only dependencies).
• Provide explicit, verified instructions or dependencies that align with NVIDIA Jetson official PyTorch CUDA wheels.
• Officially document and test Evo2 setup and installation steps for Jetson platforms.

[peterbb148]

Maybe this is more a request for Nvidia to update JetPack to support Python 3.11 come to think of it...