Jiahao Li

chatglm.cpp. C++ implementation of ChatGLM-6B & ChatGLM2-6B & ChatGLM3 & GLM4(V)

chat4u. 用微信聊天记录训练一个你专属的聊天机器人

DSNet. DSNet: A Flexible Detect-to-Summarize Network for Video Summarization

seam-carving. A super-fast Python implementation of seam carving algorithm for intelligent image resizing.

redbase-cpp. A simple relational database based on Stanford CS346 RedBase, implemented in elegant modern C++14.

flash-preference. Accelerate LLM preference tuning via prefix sharing with a single line of code

rouge-metric. A Python wrapper of the official ROUGE-1.5.5.pl script and a re-implementation of full ROUGE metrics.

nanoRLHF. Train a tiny LLaMA model from scratch to repeat your words using Reinforcement Learning from Human Feedback (RLHF)

undergrad. My undergraduate projects at THU-CST

socks5-proxy. A minimal SOCKS5 proxy written in C.

tinypt. A tiny path tracer accelerated by OpenMP & CUDA.

csapp. My CSAPP Lab Solutions

SSM-VOS. Separable Structure Modeling for Semi-supervised Video Object Segmentation

secure-socks. A mini encrypted socks tunnel to bypass the firewall.

playground. Cuda

GPA-Calculator. A GPA calculator for undergraduate transcripts of Tsinghua University.

mini-router. A mini-router supporting basic ARP protocol, IP forwarding, ICMP echo / reply, and RIP routing.

echo-server. A simple echo server using event-driven programming.

OpenRLHF. A Ray-based High-performance RLHF framework (for large models)

DeepSpeed. DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.

KernelBench. KernelBench: Can LLMs Write GPU Kernels? - Benchmark with Torch -> CUDA problems

ucore_os_lab. Operating System - 2020 Spring - THU CST

SPIM23. Simple MIPS32 Implementation

llama.cpp. Port of Facebook's LLaMA model in C/C++

Router-Lab. 2019-Fall Computer Networks

ChatGLM-Tuning. 一种平价的chatgpt实现方案, 基于ChatGLM-6B + LoRA

li-plus.

Decaf-Compiler. 2019-Fall Compiler Construction.

homepage. My homepage using starter-academic template, powered by netlify

cs144. An elegant implementation for Stanford CS144 TCP Lab