简体ä¸ć–‡ 🚀 Apart from the general language model (LLM), we also need more domain specific GPTs to assist us in handling downstream tasks or driving industry development.
roomGPT - redesign your room with AI
It uses an ML model called ControlNet to generate variations of rooms. This application gives you the ability to upload a photo of any room, which will send it through this ML Model using a Next.js API route, and return your generated room. The ML Model is hosted on Replicate and Upload is used for image storage.
BloombergGPT: A Large Language Model for Finance - (didn't open source)
LLM specialized for the financial domain. BloombergGPT, a 50 billion parameter language model that is trained on a wide range of financial data. The authors construct a 363 billion token dataset based on Bloomberg's extensive data sources, perhaps the largest domain-specific dataset yet, augmented with 345 billion tokens from general purpose datasets.
FinGPT - AI for Finance
Wall Street didn't open source LLMs or API, so FinGPT is coming out! Unlike proprietary models, FinGPT takes a data-centric approach, providing researchers and practitioners with accessible and transparent resources to develop their FinLLMs. They highlight the importance of an automatic data curation pipeline and the lightweight low-rank adaptation technique in building FinGPT.
LaWGPT - LLM based on Chinese legal knowledge
The series of models are based on the general Chinese language base models such as Chinese-LLaMA and ChatGLM, with the addition of an exclusive legal vocabulary and large-scale Chinese legal corpus pre-training to enhance the basic semantic understanding capability of large models in the legal field. Based on this, a legal domain dialogue question answering dataset and a Chinese judicial examination dataset were constructed for fine-tuning, which improved the model's understanding and execution capabilities of legal content.
FoodGPT - (didn't opensource)
A “specialized” billion-parameter FoodUDT-1B model from ThirdAI that is pre-trained on 500K recipes. An expert network, FoodUDT-1B, trained only on recipes, views “meat” and “sugar” as non-complementary flavors. Meat and sugar usually don’t go together in a recipe. For a food expert, everything is about the pairing, the taste, enhancement, etc.
WriteGPT - help you write
WriteGPT is a generative AI writing framework based on the state of the art OCR and NLP models. The first version of the finetune model is aimed at Chinese college entrance examination compositions (mainly argumentative essays), and can effectively generate articles that conform to human cognition. Most of the articles can reach the passing composition level of normal high school students after testing.
ChatLaw - Legal LLM
An open-source legal large language model named ChatLaw. Due to the importance of data quality, we carefully designed a legal domain fine-tuning dataset. Additionally, to overcome the problem of model hallucinations in legal data screening during reference data retrieval, we introduce a method that combines vector database retrieval with keyword retrieval to effectively reduce the inaccuracy of relying solely on vector database retrieval. Furthermore, we propose a self-attention method to enhance the ability of large models to overcome errors present in reference data, further optimizing the issue of model hallucinations at the model level and improving the problem-solving capabilities of large models.
We have open-sourced a specific LLaMA-7B model, which has been fine-tuned using the Instruct-tuning technique with Chinese medical instructions. To achieve this, we constructed a Chinese medical instruction dataset based on a medical knowledge graph and GPT3.5 API. By leveraging this dataset, we were able to fine-tune LLaMA and significantly improve its question-answering performance in the medical domain.