Dhenu2-In-Llama3.2-3B-Instruct by KissanAI is a 3 billion parameter language model built on the Llama3.2 architecture, specifically optimized for agricultural applications. With a context length of 32768 tokens, it provides reliable conversational capabilities for farmers and agricultural queries. This model excels at delivering accurate information and interactive engagement within the agricultural domain, making it ideal for chatbots and virtual assistants in this sector.
Loading preview...
Dhenu2 India 3B: Agricultural Conversational AI
Dhenu2 India 3B, developed by KissanAI, is a 3 billion parameter instruction-tuned model based on the Llama3.2 architecture. Released on October 24, 2024, it is specifically designed to provide balanced and reliable conversational performance across a wide range of agricultural applications. The model is optimized for knowledge depth and responsiveness, making it a specialized tool for the agricultural sector.
Key Capabilities
- Domain-Specific Accuracy: Tailored to provide precise and relevant information for agricultural queries.
- Conversational Fluency: Engineered for interactive engagement, suitable for chatbots and virtual assistants.
- Optimized Performance: Utilizes full fine-tuning and Low-Rank Adaptation (LoRA) with flash attention for efficient operation.
Good For
- Agricultural Chatbots: Developing interactive tools to assist farmers with daily questions and tasks.
- Virtual Assistants: Creating AI assistants focused on agricultural advice, package of practices, and localized studies.
- Information Retrieval: Extracting accurate information from diverse agricultural data sources, including mobile extension service logs and farmer feedback.
This model was trained on a curated dataset of over 1.5 million instructions, including synthetic conversations covering more than 4,000 diverse agricultural topics. Evaluation involved human review by agricultural experts and synthetic peer assessments to ensure accuracy and consistency. It is important to note that Dhenu2 India 3B is tailored exclusively for agricultural contexts and may not perform optimally outside this domain.