I've been doing some vision model interpretability work using the original Lucent library. This fork contains a bunch of updates / bug fixes that I've found useful for my own projects. The changes I've made are:

• add an L2 objective.
• update the default direction_neuron objective's behaviour to use dot product which is more reflective of the original Lucid library.
• add capacity to visualise a single branch of an inception module.
• if torch.cuda.is_available(), don't use cuda:0 by default. Instead, allow for specific GPU usage via torch.cuda.set_device(device) or setting the CUDA_VISIBLE_DEVICES environmental variable.

PyTorch + Lucid = Lucent

The wonderful Lucid library adapted for the wonderful PyTorch!

Lucent is not affiliated with Lucid or OpenAI's Clarity team, although we would love to be! Credit is due to the original Lucid authors, we merely adapted the code for PyTorch and we take the blame for all issues and bugs found here.

Lucent is still in pre-alpha phase and can be installed locally with the following command:

pip install torch-lucent

In the spirit of Lucid, get up and running with Lucent immediately, thanks to Google's Colab!

You can also clone this repository and run the notebooks locally with Jupyter.

import torch

from lucent.optvis import render
from lucent.modelzoo import inceptionv1

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model = inceptionv1(pretrained=True)
model.to(device).eval()

render.render_vis(model, "mixed4a:476")

Here, we have tried to recreate some of the Lucid notebooks! You can also check out the lucent-notebooks repo to clone all the notebooks.

• Feature Visualization
• The Building Blocks of Interpretability
• Using Artificial Intelligence to Augment Human Intelligence
• Visualizing Representations: Deep Learning and Human Beings
• Differentiable Image Parameterizations
• Activation Atlas

• Lessons from a year of Distill ML Research (Shan Carter, OpenVisConf)
• Machine Learning for Visualization (Ian Johnson, OpenVisConf)

Check out #proj-lucid and #circuits on the Distill slack!

You may use this software under the Apache 2.0 License. See LICENSE.