https://github.com/mudler/dockerfiles
Various app dockerfiles
https://github.com/mudler/dockerfiles
Last synced: 3 months ago
JSON representation
Various app dockerfiles
- Host: GitHub
- URL: https://github.com/mudler/dockerfiles
- Owner: mudler
- Created: 2015-12-07T15:41:18.000Z (over 9 years ago)
- Default Branch: master
- Last Pushed: 2025-02-08T09:02:37.000Z (4 months ago)
- Last Synced: 2025-02-08T09:25:29.220Z (4 months ago)
- Language: Python
- Size: 8.99 MB
- Stars: 1
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Dockerfiles
Yet another stack of Dockerfiles!
## Images
| Image | Description | Usage |
| --- | --- | --- |
| [quay.io/mudler/nvidia-kairos](https://quay.io/repository/mudler/nvidia-kairos) | Nvidia l4t kairos image to run dockerized workload | [Nvidia Kairos](#nvidia-kairos) |
| [quay.io/mudler/nvidia-l4t-unsloth](https://quay.io/repository/mudler/nvidia-l4t-unsloth) | A ready to use image for fine-tuning | [Unsloth image](#unsloth-image) |
## Usage
### Nvidia Kairos
The image's purpose is to be used as an OS image for [Kairos](https://kairos.io). If you have for instance an Nvidia device running with Kairos (e.g. [Nvidia AGX Orin](https://kairos.io/docs/installation/nvidia_agx_orin/)), all you have to do is to upgrade the OS image:
```bash
kairos-agent upgrade --source oci:quay.io/mudler/nvidia-kairos:master
```
the OS now will have docker installed and ready to use the GPU.
Additional tools added to the image:
- `docker`
- `jtop` (for stats monitoring)
- `tmux`
- `vim`
### Unsloth image
The image is a ready to use image for fine-tuning models which works with Nvidia L4T devices (tested with Nvidia AGX Orin). It contains the following tools:
- `unsloth` (current)
- `torch` (2.5)
- `torchvision` (2.5)
- `torchaudio` (0.20)
- `xformers` (current)
- `jupyterlab`
The image is ready to use with the following command to run a console:
```bash
docker run -ti --runtime nvidia --entrypoint /bin/bash --gpus all --rm -v $PWD/huggingface_cache:/root/.cache/huggingface/hub quay.io/mudler/nvidia-l4t-unsloth:latest
```
You can find training examples in `/work/examples`.
To start it with jupyter lab:
```bash
docker run -ti --runtime nvidia -p 9090:9090 --gpus all --rm -v $PWD/huggingface_cache:/root/.cache/huggingface/hub quay.io/mudler/nvidia-l4t-unsloth:latest
```
It automatically starts jupyter lab, you can access it by opening the browser and going to `http://localhost:9090`.