experiment.ipynb

Kernel: Python 3 (ipykernel)

Denoising Diffusion Probabilistic Models (DDPM)

This notebook trains a DDPM based model on MNIST digits dataset.

Install the packages

!pip install labml-nn --quiet

Imports

from labml import experiment
from labml_nn.diffusion.ddpm.experiment import Configs

Create an experiment

experiment.create(name="diffuse", writers={'screen'})

Configurations

configs = Configs()

Set experiment configurations and assign a configurations dictionary to override configurations

experiment.configs(configs, {
    'dataset': 'MNIST',
    'image_channels': 1,
    'epochs': 5,
})

Initializ

configs.init()

Set PyTorch models for loading and saving

experiment.add_pytorch_models({'eps_model': configs.eps_model})

Start the experiment and run the training loop.

# Start the experiment
with experiment.start():
    configs.run()