Today's Open Source (2024-09-02): Cohere Launches Command R and R+ for RAG and Tool Optimization
Discover the latest in AI open-source projects, including Cohere's Command R models, AWPortrait-FL, minusx, How_Much_VRAM, Spider 2.0, and WilmerAI.
Here are some interesting AI open-source models and frameworks I wanted to share today:
Project: C4AI Command-R-08-24
C4AI Command-R is a large language model developed by Cohere. It is optimized for RAG and tool-usage scenarios.
Recently, the new Command R and Command R+ versions were released. Command R has 35 billion parameters, while Command R+ boasts 104 billion parameters. These models are highly advanced, supporting retrieval-augmented generation (RAG) and tool usage, allowing automation of complex tasks and multi-step tool operations.
https://huggingface.co/CohereForAI/c4ai-command-r-v01
https://huggingface.co/CohereForAI/c4ai-command-r-08-2024
https://huggingface.co/CohereForAI/c4ai-command-r-plus-08-2024
Project: AWPortrait-FL
AWPortrait-FL is a portrait model based on the Flux architecture, fine-tuned on FLUX.1-dev. It was trained with AWPortrait-XL's dataset and nearly 2,000 high-quality fashion photos.
The model shows significant improvements in composition and detail, making skin and textures more refined and realistic.
https://huggingface.co/Shakker-Labs/AWPortrait-FL
Project: minusx
MinusX is an open-source project by AI data scientists. It allows users to analyze data and answer queries by entering prompts. MinusX can add a chat feature to applications. Currently, it supports Jupyter and Metabase, with plans to add more tools.
https://github.com/minusxai/minusx
Project: How_Much_VRAM
How Much VRAM is an open-source project that estimates the memory needed for model training or inference.
This tool helps developers choose the right hardware without testing multiple configurations.
https://github.com/AlexBodner/How_Much_VRAM
Project: Spider2
Spider 2.0 is a new benchmark for evaluating LLMs in enterprise-level text-to-SQL workflows.
The project tests LLM performance in complex data environments, multiple SQL dialects, and diverse operations.
Compared to Spider 1.0, Spider 2.0 offers more challenging tests, involving over 3,000 columns and SQL dialects like BigQuery and Snowflake.
https://github.com/xlang-ai/Spider2
Project: WilmerAI
WilmerAI is an advanced middleware system that processes incoming prompts before sending them to LLM APIs.
It uses LLMs to classify and route prompts to the right workflows or handle large contexts to create smaller prompts suitable for most local models. WilmerAI supports multiple backend LLM connections and offers OpenAI API-compatible chat/completion endpoints.
