Pinned Repositories
cookbook
Deep learning for dummies. All the practical details and useful utilities that go into working with real models.
gpt-neo
An implementation of model parallel GPT-2 and GPT-3-style models using the mesh-tensorflow library.
gpt-neox
An implementation of model parallel autoregressive transformers on GPUs, based on the Megatron and DeepSpeed libraries
lm-evaluation-harness
A framework for few-shot evaluation of language models.
math-lm
minetest
Minetest is an open source voxel game engine with easy modding and game creation
polyglot
Polyglot: Large Language Models of Well-balanced Competence in Multi-languages
pythia
The hub for EleutherAI's work on interpretability and learning dynamics
sparsify
Sparsify transformers with SAEs and transcoders
the-pile
EleutherAI's Repositories
EleutherAI/lm-evaluation-harness
A framework for few-shot evaluation of language models.
EleutherAI/gpt-neox
An implementation of model parallel autoregressive transformers on GPUs, based on the Megatron and DeepSpeed libraries
EleutherAI/pythia
The hub for EleutherAI's work on interpretability and learning dynamics
EleutherAI/cookbook
Deep learning for dummies. All the practical details and useful utilities that go into working with real models.
EleutherAI/sparsify
Sparsify transformers with SAEs and transcoders
EleutherAI/delphi
Delphi was the home of a temple to Phoebus Apollo, which famously had the inscription, 'Know Thyself.' This library lets language models know themselves through automated interpretability.
EleutherAI/elk
Keeping language models honest by directly eliciting knowledge encoded in their activations.
EleutherAI/DeeperSpeed
DeepSpeed is a deep learning optimization library that makes distributed training easy, efficient, and effective.
EleutherAI/nanoGPT-mup
The simplest, fastest repository for training/finetuning medium-sized GPTs.
EleutherAI/aria
Official repository for the paper: Scaling Self-Supervised Representation Learning for Symbolic Piano Performance (ISMIR 2025)
EleutherAI/aria-amt
Efficient and robust implementation of seq-to-seq automatic piano transcription.
EleutherAI/bergson
Mapping out the "memory" of neural nets with data attribution
EleutherAI/clt-training
Sparsify transformers with cross-layer transcoders
EleutherAI/attribute
EleutherAI/deep-ignorance
EleutherAI/tyche
Precisely estimating the volume of basins in neural net parameter space corresponding to interpretable behaviors
EleutherAI/attention-probes
Linear probes with attention weighting
EleutherAI/website
New website for EleutherAI based on Hugo static site generator
EleutherAI/open-r1
Fully open reproduction of DeepSeek-R1
EleutherAI/aria-utils
MIDI tokenizers and pre-processing utils.
EleutherAI/djinn
Provide a lightweight framework for authoring and validating exploitable verifiable coding problems
EleutherAI/POSER
Poser: Unmasking Alignment Faking LLMs by Manipulating Their Internals
EleutherAI/tuned-lens
Tools for understanding how transformer predictions are built layer-by-layer
EleutherAI/emergent-misalignment
EleutherAI/rtopk
https://github.com/xiexi51/RTopK PyTorch wrapper
EleutherAI/SkipTranscoderSAEBench
EleutherAI/truffaldino
Investigating goal instability in RL
EleutherAI/djinn-problems
Problems generated by djinn (exploitably verifiable coding problems)
EleutherAI/verifiers
Verifiers for LLM Reinforcement Learning
EleutherAI/wmdp
WMDP is a LLM proxy benchmark for hazardous knowledge in bio, cyber, and chemical security. We also release code for RMU, an unlearning method which reduces LLM performance on WMDP while retaining general capabilities.