Triangulum 10 B G G U F

Introduction

Triangulum 10B is a collection of pretrained and instruction-tuned generative models optimized for multilingual applications. Built on LLaMA's autoregressive language model, these models leverage advanced transformer architecture to perform complex reasoning tasks effectively.

Architecture

• Foundation Model: Based on LLaMA's autoregressive language model with an optimized transformer architecture.
• Instruction Tuning: Features supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align model outputs with human preferences.
• Multilingual Support: Capable of handling multiple languages, enhancing applicability across diverse linguistic contexts.

Training

The training strategy involves:

1. Synthetic Datasets: Utilizes long chain-of-thought data to boost reasoning capabilities.
2. Supervised Fine-Tuning (SFT): Aligns the model with specific tasks using curated datasets.
3. Reinforcement Learning with Human Feedback (RLHF): Trains the model to adhere to human values and safety through iterative processes.

Guide: Running Locally

How to Run Triangulum-10B on Ollama Locally

1. Download the Model: Use the command ollama run triangulum-10b-f16.gguf to download the Triangulum-10B-F16.gguf model.
2. Model Initialization: Ollama initializes and downloads the model files. Verify the process through the console output.
3. Interact with the Model: Send prompts to the model for interaction. For example, ask "What can you do for me?" and receive a detailed response.
4. Exit the Program: Type /exit to close the program.

Cloud GPUs

For enhanced performance, consider utilizing cloud GPUs available from providers like AWS, Google Cloud, or Azure, which allow for scalable and efficient model inference.

License

The Triangulum-10B model is released under the CreativeML OpenRAIL-M license, allowing for broad usage with specified conditions.