unslothai/unsloth

OOM-ing on Nvidia Jetson Orin Nano

Open

#3443 opened on Oct 13, 2025

View on GitHub
 (12 comments) (0 reactions) (0 assignees)Python (64,271 stars) (5,658 forks)batch import
help wanted

Description

I've installed unsloth from this index: https://pypi.jetson-ai-lab.io/jp6/cu126

The device has 8 GB of RAM/GPU, which should be enough for the following models: Gemma3 variants, 0.3B, 1B & 3B.

With ollama, I can easily do inference with all the above-mentioned models.

Contributor guide

Tech stack
pythonpytorch
Domain
performancemachine learning
Issue type
bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
clear
Prerequisites
Python knowledgePyTorch basicsUnderstanding of GPU memory
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
50
Research direction
Investigate the model loading process in unsloth to identify where excess memory is allocated. Compare memory usage profiles between unsloth and ollama for the same model. Check if there are any memory saving options like offloading or quantization that can be enabled. Look at the issue comments for any suggestions from maintainers or other users.