GitHubHelp

Hu Yunhai

@Yunhai-Hu · User

GitHub profile ↗ · Compare

Hi, I'm Hu Yunhai! I'm a developer passionate about technologies. Currently diving into Java and Python. Click links below for more info.

@New York University, Courant11 followers30 repositories

Repositories

Yunhai-Hu/GB10-3.8-Flash-Next

A script to launch Qwen3.8-Flash-Next in a single DGX-Spark (or its variants) at 40+ tok/s inference speed.

★ 0ShellForks 0

Yunhai-Hu/EAGLE-Qwen3

Official Implementation of EAGLE-1 (ICML'24), EAGLE-2 (EMNLP'24), and EAGLE-3.

★ 1Forks 0

Yunhai-Hu/AntiEcho

AntiEcho: agentic browser system to reduce information echo chambers using Playwright and multimodal LLMs (Qwen3-VL compatible).

★ 1PythonForks 0

Yunhai-Hu/Memento-S

Memento-Skills: Making Prompts Stateful and Agents Continually Learnable

★ 1PythonForks 0

Yunhai-Hu/verl

verl: Volcano Engine Reinforcement Learning for LLMs

★ 0PythonForks 0

Yunhai-Hu/Lfm2-VL-MobileNet

Lfm2-VL-MobileNet: A lightweight vision-language model with MobileNet as the vision encoder for efficient multimodal understanding.

★ 0PythonForks 0

Yunhai-Hu/stark

(NeurIPS D&B 2024) STaRK: Benchmarking LLM Retrieval on Textual and Relational Knowledge Bases

★ 0PythonForks 0

Yunhai-Hu/verl-agent

verl-agent is an extension of veRL, designed for training LLM/VLM agents via RL. verl-agent is also the official code for paper "Group-in-Group Policy Optimization for LLM Agent Training"

★ 0PythonForks 0

Yunhai-Hu/O1-24-Game

This project fine-tunes a small language model (Qwen2.5-0.5B) using the O1-CoT paradigm to solve the classic 24-point game with high accuracy and efficiency.

★ 1PythonForks 0

Yunhai-Hu/MultiDeviceSpeculativeDecoding

Achieve Draft User Client and Verification Server Architecture. Draft Models can be installed in user devices and use Verification Server to verify their draft tokens.

★ 1PythonForks 1