Donate GPU - KellyAI

Setup (One Time)

Never used Python before? No problem. Follow these steps once.

Step 1: Install Python

If you already have Python 3.9+, skip this step.

Windows

Download from python.org/downloads
IMPORTANT: Check "Add Python to PATH" during installation!

Linux (Ubuntu/Debian)

sudo apt install python3 python3-pip

macOS

brew install python3

Note: macOS has no CUDA GPU support. Training will be very slow on CPU.

Step 2: Install PyTorch (with GPU support)

Open a terminal (Command Prompt on Windows, Terminal on Linux/Mac) and run:

pip install torch --index-url https://download.pytorch.org/whl/cu121

This installs PyTorch with CUDA GPU support. ~2.5GB download, only needed once.

"pip" blocked or not found? Use python -m pip instead of pip

Step 3: Install KellyAI Trainer

Then install the trainer:

pip install kellyai-trainer --extra-index-url https://kellyai.site/pip/

Small download (~24KB). Only needed once.

Step 4: Start Training

Run this command and let it do its thing:

kellyai-train --server https://kellyai.site

System: Windows (10) (AMD64)

Python: 3.13.0

GPU: NVIDIA GeForce RTX 3060 (12.0 GB VRAM)

Connecting to KellyAI server...

Assigned: Task #347 (Domain: Python Code)

Downloading training package... done

Training started...

Step 1000/20000 loss=3.21 5% ETA: 3h

# 3. When done, it auto-uploads. That's it!

Training verified and accepted!

KellyAI is now smarter. Thank you!

⚠ Important: Read Before Starting

• The training process takes 10 minutes to 1 hour depending on your GPU. Your PC must stay ON and running the entire time without interruption.
• Do not shut down, restart, or put your PC to sleep while training is in progress. If interrupted, all progress is lost and the task goes back to the queue.
• We recommend starting the training when you have some free time — you can use your PC for light tasks while it runs.
• You can use your PC normally for light tasks (browsing, documents) while training runs in the background, but avoid heavy gaming or other GPU-intensive work.

💡 Don't have a GPU? You can still help! Share this page with friends who have gaming PCs. Every GPU counts.

How Training Works

01

You Get a Unique Block

The server assigns you a unique training task. No two volunteers work on the same thing at the same time.

02

Your GPU Does the Work

Everything runs locally on your machine. Fits on any GPU with 4GB+ VRAM. Your data stays private.

03

Block Gets Verified

Your contribution is tested against quality benchmarks. If it passes, it gets added to KellyAI. If not, the task goes back to the queue.

04

You're on the Leaderboard

Your contribution is recorded forever. See your name on the dashboard, track how many blocks you've trained, and watch KellyAI grow.

GPU Compatibility

Almost any modern GPU works. Here's what to expect:

Best (8GB+ VRAM)

RTX 3060, 3070, 3080, 4060, 4070, 4080, 4090, A4000, A5000
Training in 10-20 minutes.

Good (6GB VRAM)

RTX 2060, 3060 6GB, GTX 1660
Training in 20-40 minutes.

Works (4GB VRAM)

GTX 1650, RX 580, older cards
Training in 30-60 minutes. Every bit helps!

Donate Your GPU Power

Setup (One Time)

Step 1: Install Python

Step 2: Install PyTorch (with GPU support)

Step 3: Install KellyAI Trainer

Step 4: Start Training

How Training Works

You Get a Unique Block

Your GPU Does the Work

Block Gets Verified

You're on the Leaderboard

GPU Compatibility

Best (8GB+ VRAM)

Good (6GB VRAM)

Works (4GB VRAM)

Ready to Help?