Dolphin3.0 Llama3.2 3 B LLM Model

Introduction

Dolphin 3.0 is an instruct-tuned model developed as part of Cognitive Computations' Dolphin series. It is designed to be a versatile, general-purpose AI model with applications in coding, math, agentic tasks, function calling, and more. Unlike some AI models that impose control over system prompts and model versions, Dolphin 3.0 provides users with autonomy over these aspects, ensuring data privacy and alignment customization.

Architecture

Dolphin 3.0 is based on the Llama 3.2 architecture, with a 3 billion parameter base model. This version aims to offer a high degree of flexibility and control to its users, allowing for custom system prompts and behavior settings.

Training

The model was trained using a diverse set of datasets, including:

OpenCoder-LLM/opc-sft-stage1
OpenCoder-LLM/opc-sft-stage2
Microsoft/orca-agentinstruct-1M-v1
Microsoft/orca-math-word-problems-200k
NousResearch/hermes-function-calling-v1
AI-MO/NuminaMath-CoT
AI-MO/NuminaMath-TIR
AllenAI/tulu-3-sft-mixture
CognitiveComputations/dolphin-coder
HuggingFaceTB/smoltalk
CognitiveComputations/samantha-data
M-a-p/CodeFeedback-Filtered-Instruction
M-a-p/Code-Feedback

The model was supported by various sponsors providing cloud resources and grants for efficient training and evaluation.

Guide: Running Locally

To run Dolphin 3.0 locally, follow these steps:

Set Up Environment: Ensure you have Python and the Hugging Face Transformers library installed.
Download Model: Clone the Dolphin 3.0 model from the Hugging Face Model Hub.
Configure System Prompt: Customize the system prompt to set the desired tone and behavior.
Run Inference: Use the Hugging Face Transformers library to load the model and generate outputs.

Cloud GPUs

For optimal performance, consider using cloud GPUs such as those from AWS, Google Cloud, or Azure. These platforms offer on-demand GPU instances that can significantly speed up model training and inference.

License

Dolphin 3.0 is released under the Llama3.2 license. Users should review the license terms for specific usage and distribution rights.

More Related APIs