EleutherAI has a significant public presence on GitHub, featuring a wide range of repositories primarily written in Python, Jupyter Notebook, C++, JavaScript, Rust, and Cuda. Notable projects include lm-evaluation-harness, gpt-neox, and pythia, which focus on language model evaluation, model parallel transformers, and interpretability research, respectively.
Top languages
Public repositories
lm-evaluation-harness
A framework for few-shot evaluation of language models.
gpt-neox
An implementation of model parallel autoregressive transformers on GPUs, based on the Megatron and DeepSpeed libraries
pythia
The hub for EleutherAI's work on interpretability and learning dynamics
math-lm
No description provided for this repository.
cookbook
Deep learning for dummies. All the practical details and useful utilities that go into working with real models.
sparsify
Sparsify transformers with SAEs and transcoders
polyglot
Polyglot: Large Language Models of Well-balanced Competence in Multi-languages
vqgan-clip
No description provided for this repository.
concept-erasure
Erasing concepts from neural representations with provable guarantees
elk
Keeping language models honest by directly eliciting knowledge encoded in their activations.
nanoGPT-mup
The simplest, fastest repository for training/finetuning medium-sized GPTs.
oslo
OSLO: Open Source for Large-scale Optimization
aria
Official repository for the paper: Scaling Self-Supervised Representation Learning for Symbolic Piano Performance (ISMIR 2025)
dps
Data processing system for polyglot
improved-t5
Experiments for efforts to train a new and improved t5
minetest
Minetest is an open source voxel game engine with easy modding and game creation
aria-amt
Efficient and robust implementation of seq-to-seq automatic piano transcription.
bergson
Mapping out the "memory" of neural nets with data attribution
magiCARP
One stop shop for all things carp
semantic-memorization
No description provided for this repository.
features-across-time
Understanding how features learned by neural networks evolve throughout training
hae-rae
No description provided for this repository.
rnngineering
Engineering the state of RNN language models (Mamba, RWKV, etc.)
elk-generalization
Investigating the generalization behavior of LM probes trained to predict truth labels: (1) from one annotator to another, and (2) from easy questions to hard
steering-llama3
No description provided for this repository.
tokengrams
Efficiently computing & storing token n-grams from large corpora
training-jacobian
No description provided for this repository.
w2s
No description provided for this repository.
deep-ignorance
No description provided for this repository.
polyglot-data
data related codebase for polyglot project
pile_dedupe
Pile Deduplication Code
latent-video-diffusion
Latent video diffusion
NeMo
NeMo: a toolkit for conversational AI
attribute
No description provided for this repository.
exploring-contrastive-topology
No description provided for this repository.
polyapprox
Closed-form polynomial approximations to neural networks
pilev2
No description provided for this repository.
lm_dataformat
No description provided for this repository.
transformer-reasoning
Experiments in transformer knowledge and reasoning
architecture-objective
No description provided for this repository.
attention-probes
Linear probes with attention weighting
equinox-llama
Equinox implementation of llama3 and llama3.1
GPTeacher
A collection of modular datasets generated by GPT-4, General-Instruct - Roleplay-Instruct - Code-Instruct - and Toolformer
minetest-baselines
Baseline agents for Minetest tasks.
aria-utils
MIDI tokenizers and pre-processing utils.
cupbearer
A library for mechanistic anomaly detection
weak-to-strong
No description provided for this repository.
trlx
A repo for distributed training of language models with Reinforcement Learning via Human Feedback (RLHF)
minetest-interpretabilty-notebook
Jupyter notebook for the interpretablity section of the minetester blog post
CodeCARP
Data collection pipeline for CodeCARP. Includes PyCharm plugins.
clearnets
No description provided for this repository.
optax-galore
Adds GaLore style projection wrappers to optax optimizers
architecture-experiments
Repository to host architecture experiments and development using Paxml and Praxis
FLAN
No description provided for this repository.
thonkenizers
yes
scalable-elicitation
The code used in "Balancing Label Quantity and Quality for Scalable Elicitation"
monkfish
No description provided for this repository.
alignment-handbook
Robust recipes for to align language models with human and AI preferences
Unpaired-Image-Generation
Project Repo for Unpaired Image Generation project
lm-scope
No description provided for this repository.
sae_overlap
Acompanying code for our research on SAE feature overlap when trained on different seeds.
variance-across-time
Studying the variance in neural net predictions across training time
EvilModel
A replication of "EvilModel 2.0: Bringing Neural Network Models into Malware Attacks"
eai-prompt-gallery
Library of interesting prompt generations
gamescope
Can interpretability methods confer an advantage in competitive games?
fmri
Analogue of fMRI on artificial neural networks
rtopk
https://github.com/xiexi51/RTopK PyTorch wrapper
pd-books
No description provided for this repository.
tuned-lens
Tools for understanding how transformer predictions are built layer-by-layer
tinydpo
No description provided for this repository.
eleutherai-instruct-dataset
A large instruct dataset for open-source models (WIP).
examples
Mosaicml example benchmarks + LLM scripts
minetest_game
Minetest Game - The default game for the Minetest engine [https://github.com/minetest/minetest/]
groupoid-rl
No description provided for this repository.
truffaldino
Investigating goal instability in RL
rllm
Democratizing Reinforcement Learning for LLMs
bayesian-adam
Exactly what it says on the tin
RWKV-LM
RWKV is an RNN with transformer-level LLM performance. It can be directly trained like a GPT (parallelizable). So it's combining the best of RNN and transformer - great performance, fast inference, saves VRAM, fast training, "infinite" ctx_len, and free sentence embedding.
conceptual-constraints
Applying LEACE to models during training
aria.cpp
GGML implementation of https://github.com/EleutherAI/aria
classifier-latent-diffusion
No description provided for this repository.
language-adaptation
No description provided for this repository.
maxtext
A simple, performant and scalable Jax LLM!
irrlicht
Minetest's fork of Irrlicht
lm-evaulation-ui
App for generating html table from LM evaluation JSONs
gradient-routing
No description provided for this repository.
rh-indicators
No description provided for this repository.
hackable-bergson
Simplified library for mapping out the "memory" of neural nets with data attribution
vllm
A high-throughput and memory-efficient inference and serving engine for LLMs
verifiers
Verifiers for LLM Reinforcement Learning
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.
Megatron-LM
Ongoing research training transformer models at scale
mixture-of-depths
No description provided for this repository.
llm-score-behavior
No description provided for this repository.
TransformerEngine
A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit floating point (FP8) precision on Hopper and Ada GPUs, to provide better performance with lower memory utilization in both training and inference.
Plenoxels_FreeNerf
implmentation of Plenoxels radiance fields without neural networks, with free nerf strategy
oslo-1
OSLO: Open Source for Large-scale Optimization
t-zero
Reproduce results and replicate training fo T0 (Multitask Prompted Training Enables Zero-Shot Task Generalization)
CommonLoopUtils
[WIP] a version of CLU with WandB logging added.
pytorch-fid
Compute FID scores with PyTorch.
Frequently asked questions
What does EleutherAI build on GitHub?
EleutherAI builds several projects on GitHub that focus on language models and machine learning. Key repositories include lm-evaluation-harness for model evaluation and gpt-neox for implementing autoregressive transformers.
Which programming languages does EleutherAI use?
EleutherAI primarily uses Python and Jupyter Notebook for its projects, along with C++, JavaScript, Rust, and Cuda. This diverse language usage supports a variety of applications in AI and machine learning.
Are EleutherAI's repositories public?
Yes, all of EleutherAI's repositories are public on GitHub. This transparency allows the community to access, use, and contribute to their work on language models and related technologies.
Is this exposure intended?
Monitor EleutherAI with RepoGuard and get alerted the moment a new public repository appears.
Monitor this account