UX4567/Khushi-Business-AI-4bit
Khushi-Business-AI-4bit is a 7-billion parameter, 4-bit quantized conversational AI model developed by Kartik Sharma (UX4567). Built on Mistral 7B and optimized with Unsloth, it is designed for low VRAM environments (6GB+) and specializes in Indian business use cases such as customer support, sales chat, and lead handling. The model supports bilingual communication in Hindi and English (Hinglish) and is available in both safetensors and GGUF formats.
Loading preview...
Khushi-Business-AI-4bit (7B) Overview
Khushi-Business-AI-4bit is a 7-billion parameter conversational AI model developed by Kartik Sharma (UX4567), specifically designed for business applications in India. This model is a 4-bit quantized version of Mistral 7B, leveraging Unsloth for efficient fine-tuning and reduced memory footprint, making it suitable for deployment in environments with limited VRAM (6GB+).
Key Capabilities
- Low VRAM Optimization: Runs efficiently on systems with 6GB+ VRAM due to 4-bit quantization (BF16 base).
- Business-Specific Tuning: Optimized for common business scenarios including customer support, sales chat, and lead handling.
- Bilingual Support: Understands and responds in both Hindi and English, catering to the "Hinglish" linguistic context prevalent in India.
- Efficient Fine-tuning: Utilizes Unsloth for 2x faster fine-tuning and quantization processes.
- Flexible Deployment: Provided in both
safetensorsandGGUFformats, supporting platforms like Ollama and LM Studio.
Good for
- Indian Business Applications: Ideal for companies targeting the Indian market that require AI for customer interactions.
- Resource-Constrained Environments: Suitable for developers and businesses with limited GPU resources.
- Conversational AI: Excels in generating human-like responses for business-related queries and interactions.
- Rapid Prototyping: The Unsloth integration allows for quicker experimentation and deployment of fine-tuned models.