Transformer Explained Visually: Learn How LLM Transformer Models Work with Interactive Visualization

Traditional Mandarin LLMs for Taiwan

Natural Language Processing (NLP). Covering topics such as Tokenization, Part Of Speech tagging (POS), Machine translation, Named Entity Recognition (NER), Classification, and Sentiment analysis.

Tensorflow implementation of Semi-supervised Sequence Learning (https://arxiv.org/abs/1511.01432)

Source code for our paper: "ARIA: Training Language Agents with Intention-Driven Reward Aggregation".

Python scripts and datasets of the "Extremely Low-Resource Neural Machine Translation: A Case Study of Cantonese" project

Initial Exploratory Works on Knowledge Tracing in Transformer Based Language Models

Minimal RLM implementation

From-scratch 0.15B BitNet 1.58-bit (ternary) RWKV-7 LLM trained on an 8GB GPU — QAT, 2-bit packing, RFT, local chat app.

Worth-reading papers and related resources on pretrained-language models(PLMs). On the Shoulder of Giants!

A small language model built from scratch in PyTorch utilizing transformers. Trained on WikiText-2 with character- and word-level tokenization. Educational project to explore embeddings, positional encodings, multi-head self-attention, and transformer decoder blocks for text generation.

Custom GPT-2 (124M) built from scratch in PyTorch — Multi-Head Attention, Transformer Blocks, SSE streaming, and an OpenAI-compatible REST API. No SDKs. Just raw math.

Redefining prompt engineering as structured cognition.

Official Code for ReactionTeam: Teaming Experts for Divergent Thinking Beyond Typical Reaction Patterns (In IEEE Conference BigData 2025 Oral)

An end-to-end RNN-based text generation pipeline built with PyTorch. Includes web scraping, preprocessing, vocabulary construction, sequence generation, model training, and sample text generation.

Vector Space Model and Language Model for Information Retrieval system based on collections of text documents.

This repository contains the training, evaluation, and inference code for a Sign Language Production (SLP) system based on the principles of the SOKE (Sign as Tokens) architecture. The goal of this system is to translate spoken language (text) into highly realistic, continuous 3D SL motion.

This repository contains the code and resources for the paper "Exploring Explainability in Arabic Language Models: An Empirical Analysis of Techniques," accepted at ACLing 2024.

Experimental end-to-end tiny language models with custom tokenizer, training and inference loop

This repo contains notes of the Spacy Masterclass (NLP Course)