Asa Cooper Stickland's repositories
llama3-jailbreak
A trivial programmatic Llama 3 jailbreak. Sorry Zuck!
FastChat
An open platform for training, serving, and evaluating large language models. Release repo for Vicuna and Chatbot Arena.
representation-engineering
Representation Engineering: A Top-Down Approach to AI Transparency
situational-awareness-evals
Measuring the situational awareness of language models
knowledge-erasure
Removing knowledge from language models
probing_llama
Some probing experiments
memit
Mass-editing thousands of facts into a transformer memory (MEMIT)
inverse-scaling-eval-pipeline
Basic pipeline for running different sized GPT models and plotting the results + calibration
transparent
Exposing transformer language model internals.
FlatMinimaInterpretability
Figuring out connections between loss landscapes and interpretability.
gpt3-arithmetic
Scratchpad/Chain-of-Thought Prompts
hf-sharpness
Simple implementation of flat minima methods (SAM, fisher penalty) for Huggingface trainer.
lm-sandbox
The goal of this project is to enable users to create easy scaling law examples and cool web demos using the OpenAI GPT-3 API or Huggingface compatible LMs with just a few lines of Python.
adapter-transformers
Huggingface Transformers + Adapters = ❤️
ParlAI
A framework for training and evaluating AI models on a variety of openly available dialogue datasets.
Bert-n-Pals
Pytorch implementation of Bert and Pals: Projected Attention Layers for Efficient Adaptation in Multi-Task Learning (https://arxiv.org/abs/1902.02671)
pytorch-pretrained-BERT
A PyTorch implementation of Google AI's BERT model provided with Google's pre-trained models, examples and utilities.
darts
Differentiable architecture search for convolutional and recurrent networks
mf-amazon
Matrix factorization with amazon
Ideas
Added master document and short description of SGHMC stuff
Optimizer_Visualisers
visualise ADAM and other optimzers
Spike_And_Slab
Implementation of Gibbs sampling for spike and slab priors
Spatial_Transformer
Added Adam optimiser and relu units to Lenet. Now trying to make Spatial Transformer layer work
LBM2
initlal