base_model: meta-llama/Llama-3.2-3B-Instruct
languages:
- en
- es
- pt
license: apache-2.0
tags:
- text-generation-inference
- transformers
- facebook
- meta
- pytorch
- llama
- gguf
- CreativeWorksAi
- NeuraLake
- 256k
- 🇧🇷
model_creator: Celso H A Diniz
model_name: iSA-01-Mini-3B-GGUF
Note: This is our first public release on Hugging Face, and the Model Card is still a work in progress. Further improvements and updates will follow.
CreativeWorksAi + NeuraLake: Designed by Earth's Creatives, Assembled by AI
Model Description
The iSA-01-Mini-3B-GGUF is a small yet advanced language model developed by CreativeWorksAi, designed to enhance text generation and reasoning capabilities. It extends the context window from 128K to 256K tokens, effectively doubling its information retention and significantly improving performance compared to its base model, meta-llama/Llama-3.2-3B-Instruct.
Hardware Requirements Estimate
| Name | Quant method | Size | Memory (RAM, vRAM) required |
|---|---|---|---|
| iSA-01-Mini-3B.F16.gguf | F16 | 6.43 GB | 12.86 GB |
| iSA-01-Mini-3B.Q4_K_M.gguf | Q4_K_M | 2.02 GB | 4.04 GB |
| iSA-01-Mini-3B.Q5_K_M.gguf | Q5_K_M | 2.32 GB | 4.64 GB |
Key Features
- Extended Context Window: The model's context window has been expanded from 128K to 256K tokens, enabling it to retain more information for better reasoning and logical deductions.
- Enhanced Reasoning: The increased context size leads to superior performance in complex tasks like Retrieval-Augmented Generation (RAG), resulting in more precise and context-aware outputs.
- Improved Information Integration: With a larger context window, the model integrates external information more effectively, producing accurate and contextually relevant responses.
- Fine-tuned with NeuraLake/Megalodon: The model was fine-tuned using synthetic data generated by the state-of-the-art NeuraLake/Megalodon, enhancing its ability to process and analyze complex scenarios.
- NeuraLake/Megalodon Model: This proprietary, closed-source LLM has been developed by NeuraLake over the past three years to enhance reasoning capabilities, especially for small models and agents.
Training Data
The iSA-01-Mini-3B-GGUF was trained using synthetic data generated by NeuraLake/Megalodon, focused on realistic scenarios to improve reasoning and performance in RAG tasks.
Model Details
- Developed by: CreativeWorksAi
- License: Apache License 2.0
- Fine-tuned from: unsloth/llama-3.2-3b-instruct-bnb-4bit
- NeuraLake: Learn more about NeuraLake and its advanced AI projects at NeuraLake Cloud.
- NeuraLake/Megalodon Model: Discover more about the model used for synthetic data generation at NeuraLake.
- Model Card Contact: Celso H A Diniz LinkedIn profile.
Usage
CreativeWorksAi's Intelligence System for Advanced Dialogue and Organized Responses Assistance (i.S.A.D.O.R.A. architecture) is designed to offer users a sophisticated tool for generating coherent, contextually rich text, making it ideal for applications that require advanced natural language understanding and generation.