CS-479-Assignment-2

Report

Running Code

Create a Python virtual environment:

python -m venv .venv

Activate it:

• Windows (PowerShell)

.venv\Scripts\Activate.ps1

• Windows (bash)

source .venv/Scripts/activate

• macOS/Linux

source .venv/bin/activate

GPU Requirements

GPU acceleration requires a CUDA-compatible NVIDIA GPU with the matching CUDA Toolkit installed.

First, check your CUDA version:

nvcc --version

Update requirements.txt file to reflect the version you have installed.

Verify CuPy can see your GPU:

python gpu_check.py

If you see a DLL load failed error, the installed cupy version does not match your CUDA toolkit — uninstall cupy and reinstall with the correct version above, or use --cpu.

Running the Experiments

Install dependencies:

pip install -r requirements.txt

By default, the experiment auto-detects GPU availability and falls back to CPU if unavailable.

• Default:

python experiment_1.py

• Force CPU (NumPy):

python experiment_1.py --cpu

• Force GPU (CuPy) — exits with an error if unavailable:

python experiment_1.py --gpu