Honglu Fan's repositories
mistral_jax
Mistral model in JAX
_diff_model
documenting scripts and workflows for diff model training
fmlang_env
Toy gym env related to formal languages.
hironaka-experiments
Document the experiments of hironaka project
capabilities
Blazon Capabilities SDK
composer
Train neural networks up to 7x faster
examples
Fast and flexible reference benchmarks
gpt-neox
An implementation of model parallel autoregressive transformers on GPUs, based on the DeepSpeed library.
hironaka_v2
This is a clean redo using only JAX and we reconstruct a simpler design.
honglu2875
Config files for my GitHub profile.
jaxformer
Minimal library to train LLMs on TPU in JAX with pjit().
llama
Inference code for LLaMA models
llama.cpp
Port of Facebook's LLaMA model in C/C++
lm-evaluation-harness
A framework for few-shot evaluation of autoregressive language models.
slightly-faster-gpt
A slightly faster GPT-J than Huggingface
tweets
janky twitter replacement
yarn
YaRN: Efficient Context Window Extension of Large Language Models
yarn-patch
This repo exposes simple APIs to patch the YaRN technique to the rotary embeddings of a given Hugging Face model.