nlp

AI-powered text compression library for RAG systems and API calls. Reduce token usage by up to 50-60% while preserving semantic meaning with advanced compression strategies.

openai/whisper + extra features

classy is a simple-to-use library for building high-performance Machine Learning models in NLP.

Self-hosted AI-powered transcription platform with speaker diarization, search, and collaboration features. Built with Svelte, FastAPI, and Docker for easy deployment.

🛠️ Tools for Transformers compression using PyTorch Lightning ⚡

Qualcomm Cloud AI SDK (Platform and Apps) enable high performance deep learning inference on Qualcomm Cloud AI platforms delivering high throughput and low latency across Computer Vision, Object Detection, Natural Language Processing and Generative AI models.

Bangla-Bert is a pretrained bert model for Bengali language

Official implementation of the paper "IteraTeR: Understanding Iterative Revision from Human-Written Text" (ACL 2022)

easy-bert是一个中文NLP工具，提供诸多bert变体调用和调参方法，极速上手；清晰的设计和代码注释，也很适合学习

📲 Transformers android examples (Tensorflow Lite & Pytorch Mobile)

SDK for using LLM

SCOPE: Self-evolving Context Optimization via Prompt Evolution - A framework for automatic prompt optimization

VerifAI initiative to build open-source easy-to-deploy generative question-answering engine that can reference and verify answers for correctness (using posteriori model)

🤖 language modeling sft demos

Tensorflow, Pytorch, Huggingface Transformer, Fastai, etc. tutorial Colab Notebooks.

Roy: A lightweight, model-agnostic framework for crafting advanced multi-agent systems using large language models.

Transformer-powered hashtag segmentation with beam search, RRF ensembles, and agent-ready MCP workflows. A scalable alternative to heuristic splitters and massive LLMs.

Trained T5 and T5-large model for creating keywords from text

A curated list of valuable resources from our studies at the University of Tehran (UT), School of Electrical and Computer Engineering (ECE)

Emotion text classification using Llama3-8b with LoRA and FlashAttention. Based on LLaMA-Factory.