GPU Setup

Baseline checklist for local GPU readiness.

This guide tracks a stable local GPU setup process for AI experimentation.

Driver and Runtime

1. Install the latest stable GPU driver.
2. Confirm CUDA toolkit and runtime compatibility with your framework.
3. Verify support with:

nvidia-smi

Python Package Alignment

1. Pin PyTorch build for your CUDA version.
2. Pin critical libraries (transformers, accelerate, bitsandbytes) per project.
3. Save versions in a project requirements.txt.

Validation Script

Run a quick tensor operation on GPU before any training session.