pavanyellow
Follow
-
alpha-zero-general Public
Forked from suragnair/alpha-zero-generalSparse Autoencoders for extracting new superhuman concepts from AlphaZero and Pluribus
Python UpdatedJun 19, 2024 -
feature-steering Public
Controlling LLM outputs by activating/suppressing feature vectors
Jupyter Notebook Apache License 2.0 UpdatedJun 14, 2024 -
-
TransformerLens Public
Forked from TransformerLensOrg/TransformerLensA library for mechanistic interpretability of GPT-style language models
Python MIT License UpdatedApr 5, 2024 -
sparse_autoencoder Public
Forked from Alignment-Lab-AI/sparse_autoencoderJupyter Notebook MIT License UpdatedFeb 8, 2024 -
-
toy-counter-model Public
Interpretability of 2*2*2 layer model calculating the number of non zero entries in the input
Jupyter Notebook UpdatedFeb 8, 2024 -
phase-change-temperature Public
Exploring the sudden incoherence of language models at higher sampling temperatures
Jupyter Notebook Apache License 2.0 UpdatedFeb 2, 2024 -
-
sparse-autoencoder Public
Interpreting the ultra-low density cluster in sparse autoencoders from Anthropic's Towards Monosemanticity work
Jupyter Notebook UpdatedDec 6, 2023 -
mlx-examples Public
Forked from ml-explore/mlx-examplesExamples in the MLX framework
Python MIT License UpdatedDec 6, 2023 -
Attention Public
A simple, pure Python implementation of the original attention mechanism with no PyTorch or NumPy dependencies
Python UpdatedDec 6, 2023 -
-
Transformers-From-Scratch Public
A simple Transformer trained on Tinystories Dataset for 5min on M1 Air that can produce coherant looking language
Jupyter Notebook UpdatedDec 4, 2023 -
Interpretability Public
My Working codebase for exploratory Interpretability work and replicating important papers in the field
Jupyter Notebook UpdatedDec 2, 2023 -
alphazero-general Public
Forked from kevaday/alphazero-generalA fast, generalized, and modified implementation of Deepmind's distinguished AlphaZero in PyTorch.
Python MIT License UpdatedFeb 22, 2023