You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm Pin Fang, a software engineer at Tiktok, Shanghai. Nice to meet you!
Building & Built
Siamese PyTorch - Siamese Networks for one-shot image recognition on Omniglot, connected to my ICME 2019 paper.
LLM from Scratch - A modern decoder-only Transformer stack in PyTorch, including BPE, RoPE, RMSNorm, SwiGLU, Flash Attention 2, DDP, SFT, and RL fine-tuning.
miniDFS - A C++ mini distributed file system with file striping, replicated chunks, name server metadata, and data server threads.
redis-rs - Redis server implementation in Rust, covering RESP, RDB persistence, replication, streams, and transactions.
Remote Nvim - Remote development in Neovim for devboxes and SSH-backed workflows.
dl-framework - A from-scratch deep learning framework project with autograd, tensor operations, and CUDA-backed systems pieces.
simple_json - A small C++17 JSON library focused on parsing and lightweight header-only usage.
ThreadPool - A compact C++11 thread pool implementation for practicing concurrency primitives and task execution.
dcgan - Deep Convolutional GAN implementation in TensorFlow and TensorLayer, trained on CelebA-style image generation.
gobang - A Java Gobang game with a GUI, minimax search, and alpha-beta pruning.
Lark Agent Bridge - Bridge coding agents into Lark chats, cloud docs, worktree groups, and collaborative agent workflows.
yolo-caffe - A Caffe-based YOLO v1 implementation connecting Darknet-style object detection ideas to the Caffe ecosystem.
GPU HPC Book - A hands-on book about GPU high-performance computing, CUDA, Tensor Cores, TMA, WGMMA, TPU collectives, and TIRx.
Contributing to
CowAgent - Added Azure-hosted ChatGPT service support and Railway deployment support.
GPU High Performance Computing - A hands-on online book about GPU programming, performance modeling, CUDA, Tensor Cores, TMA, WGMMA, TPU collectives, and TIRx.
Sponsorships
If you like my projects and want to support me, you can use the following payment methods:
- Siamese Networks for one-shot image recognition on Omniglot, connected to my ICME 2019 paper.
- A modern decoder-only Transformer stack in PyTorch, including BPE, RoPE, RMSNorm, SwiGLU, Flash Attention 2, DDP, SFT, and RL fine-tuning.
- A C++ mini distributed file system with file striping, replicated chunks, name server metadata, and data server threads.
- Redis server implementation in Rust, covering RESP, RDB persistence, replication, streams, and transactions.
- Remote development in Neovim for devboxes and SSH-backed workflows.
- A from-scratch deep learning framework project with autograd, tensor operations, and CUDA-backed systems pieces.
- A small C++17 JSON library focused on parsing and lightweight header-only usage.
- A compact C++11 thread pool implementation for practicing concurrency primitives and task execution.
- Deep Convolutional GAN implementation in TensorFlow and TensorLayer, trained on CelebA-style image generation.
- A Java Gobang game with a GUI, minimax search, and alpha-beta pruning.
- Bridge coding agents into Lark chats, cloud docs, worktree groups, and collaborative agent workflows.
- A Caffe-based YOLO v1 implementation connecting Darknet-style object detection ideas to the Caffe ecosystem.
- A hands-on book about GPU high-performance computing, CUDA, Tensor Cores, TMA, WGMMA, TPU collectives, and TIRx.
- Added Azure-hosted ChatGPT service support and Railway deployment support.