Today's Open Source (2024-11-05): Tencent Releases Hunyuan-Large
Discover open-source AI projects like Tencent’s Hunyuan-Large, SAM2Long, LLaMA-O1, In-Context LoRA, FlowToken, and Paper Reviewer.
Here are some interesting AI open-source models and frameworks I wanted to share today:
Project: Hunyuan-Large
Hunyuan-Large is an open-source Transformer model based on Mixture of Experts (MoE) developed by Tencent, with 389B parameters, of which 5.2B are active parameters.
This model excels in fields such as natural language processing and computer vision, aiming to maintain high performance while optimizing resource consumption.
https://github.com/Tencent/Tencent-Hunyuan-Large
Project: SAM2Long
The SAM2Long project aims to enhance the performance of SAM 2 in long video segmentation by introducing an untrained memory tree structure, which solves the common error accumulation problem in long video scenarios.
The project performs excellently in handling object occlusion and reappearance, significantly improving segmentation accuracy and robustness.
https://github.com/Mark12Ding/SAM2Long
Project: LLaMA-O1
LLaMA-O1 is an open-source large inference model framework designed to implement training, inference, and evaluation through Monte Carlo Tree Search (MCTS), Self-Play Reinforcement Learning, PPO, AlphaGo Zero’s dual-strategy paradigm, and large language models.
https://github.com/SimpleBerry/LLaMA-O1
Project: In-Context LoRA
In-Context LoRA is a project for expanding transformers, aimed at generating image sets with customizable internal relationships by fine-tuning text-to-image models.
The project allows for conditioning based on another image set during the generation process, enabling task-agnostic adaptation, making it applicable to a wide range of use cases.
https://github.com/ali-vilab/In-Context-LoRA
Project: FlowToken
FlowToken is a React component library designed to enhance the visual presentation of large language model (LLM) text flow.
The library provides multiple animation effects, allowing text to appear in a smooth and dynamic manner, and offering an engaging user experience.
FlowToken is both responsive and lightweight, optimizing performance and ensuring compatibility with all modern browsers.
https://github.com/data-maki/flowtoken
Project: Paper Reviewer
The Paper Reviewer project leverages large language models (LLMs) and document parsing tools to extract information from academic papers.
This project powers the AI Paper Reviewer website, aiming to automate the parsing of paper content, extracting images, descriptions, chapter details, references, and other information to improve the efficiency and accuracy of paper reviews.
