🦉 Data Versioning and ML Experiments

Autonomously train research-agent LLMs on custom data using reinforcement learning and self-verification.

My learning notes/codes for ML SYS.

FlashMLA: Efficient MLA decoding kernels

A very simple GRPO implement for reproducing r1-like LLM thinking.

Fully open data curation for reasoning models

An Easy-to-use, Scalable and High-performance RLHF Framework based on Ray (PPO & GRPO & REINFORCE++ & LoRA & vLLM & RFT)

Code for BLT research paper

Token Omission Via Attention

A financial agent for investment research

Pretraining code for a large-scale depth-recurrent language model

A lightweight reproduction of DeepSeek-R1-Zero with indepth analysis of self-reflection behavior.

s1: Simple test-time scaling

Fully open reproduction of DeepSeek-R1

Clean, minimal, accessible reproduction of DeepSeek R1-Zero

Reproduce R1 Zero on Logic Puzzle

Data processing for and with foundation models! 🍎 🍋 🌽 ➡️ ➡️🍸 🍹 🍷

LLMs build upon Evol Insturct: WizardLM, WizardCoder, WizardMath

Code implementation of synthetic continued pretraining

SGLang is a fast serving framework for large language models and vision language models.

Free, simple, fast interactive diagrams for any GitHub repository

O1 Replication Journey

Materials for EACL2024 tutorial: Transformer-specific Interpretability

Official implementation for "GLaPE: Gold Label-agnostic Prompt Evaluation and Optimization for Large Language Models" (stay tuned & more will be updated)

About The official GitHub page for ''Unleashing the Potential of Large Language Models as Prompt Optimizers: An Analogical Analysis with Gradient-based Model Optimizers'' Resources