pika-gpu v1.0.1

Ultra-light Deployment & Versioning for your AI stack

Set countless GPUs on fire in less than 10 seconds with managed SD dependencies.

npm i -g pika-gpu
pika-gpu start

Or, quick start (this will automatically download 4GB SD1.5 model for default ComfyUI workflow):

npm i -g pika-gpu
pika-gpu start --quick-start

Details

Add your models and their download URLs from huggingface.co to stack.json, or add them using:

# pika-gpu add -- [type] [name] [download-url]

pika-gpu add loras 360-Diffusion-LoRA-sd-v1-5 https://huggingface.co/ProGamerGov/360-Diffusion-LoRA-sd-v1-5/resolve/main/360Diffusion_v1.safetensors?download=true

Commit stack.json.

Prerequisites & Setup

You will need the following on the machine:

- git (any)
- python (python2 or python3)
- node (v18 or newer)

On Windows you can download all of these using GUI and browser, on Linux / Mac consider using brew, apt-get, or your-system equivalent.

Optional: Dependency management scripts install various Python dependencies, so if you want these scoped to a virtual environment - make sure to use conda.

Testing

Responsible for the backend of env-ai.vercel.app.