Today's Open Source (2024-08-15): THUDM Tsinghua Releases LongWriter, Capable of Writing 10,000-Word Texts
Explore the latest AI open-source projects, including ultra-long text generation, multimodal models, vector searches, engineering tools, and scientific research assistants.
Here are some interesting AI open-source models and frameworks I wanted to share today:
Project: LongWriter
LongWriter is an open-source ultra-long context model from THUDM, Tsinghua University. It can generate texts over 10,000 words long.
The project offers two models: LongWriter-glm4-9b and LongWriter-llama3.1-8b. They are trained on GLM-4-9B and Meta-Llama-3.1-8B, respectively.
LongWriter is ideal for applications needing lengthy content, such as storytelling, academic writing, and detailed reports.
https://github.com/THUDM/LongWriter
https://arxiv.org/abs/2408.07055
Project: Idefics3
Idefics3 is HuggingFace's third-generation open-source multimodal model. It accepts any sequence of images and text as input and generates text output.
The model can answer questions about images, describe visual content, create stories based on multiple images, or function as a pure language model.
Compared to Idefics1 and Idefics2, Idefics3 shows significant improvements in OCR, document understanding, and visual reasoning.
https://huggingface.co/HuggingFaceM4/Idefics3-8B-Llama3
Project: pgvector
pgvector is an open-source Postgres extension for vector similarity search.
It supports exact and approximate nearest-neighbor searches. It handles single-precision, half-precision, binary, and sparse vectors. It offers various distance metrics like L2, inner product, cosine, L1, Hamming, and Jaccard distances.
The project also includes Postgres's features like ACID compliance, point-in-time recovery, and JOIN operations.
https://github.com/pgvector/pgvector
Project: Omni Engineer
Omni Engineer is an open-source project aiming to provide developers with a versatile set of engineering tools.
Its goal is to simplify the software development process, especially when building and managing complex systems.
Omni Engineer offers tools that enhance developer efficiency at various stages, including code generation, testing, and deployment.
https://github.com/Doriandarko/omni-engineer
Project: OpenResearcher
OpenResearcher is an advanced AI assistant for scientific research. It aims to provide helpful answers to research queries. By accessing the arXiv corpus, it offers users the latest scientific insights.
https://github.com/GAIR-NLP/OpenResearcher
https://arxiv.org/abs/2408.06941
Project: Magic-Doc
Magic-Doc is a lightweight open-source tool that lets users convert various file types (PPT/PPTX/DOC/DOCX/PDF) to Markdown format.
It supports converting both local and S3 files, making it ideal for quick document-to-Markdown conversions.