GenVRadmin/AryaBhatta-GemmaUltra-Merged is an 8.5 billion parameter language model, fine-tuned from CorticalStack/gemma-7b-ultrachat-sft. It is specifically optimized for multi-turn chat-based use cases and demonstrates improved performance on Hellaswag datasets compared to its predecessor. This model supports English and nine Indian languages, including Hindi, Tamil, Punjabi, Bengali, Gujarati, Oriya, Telugu, Kannada, and Malayalam.
Loading preview...
AryaBhatta-GemmaUltra-Merged: Multi-Turn Chat and Multilingual Capabilities
GenVRadmin/AryaBhatta-GemmaUltra-Merged is an 8.5 billion parameter language model built upon the CorticalStack/gemma-7b-ultrachat-sft base. This model has been specifically fine-tuned for multi-turn conversational applications, distinguishing it from other models like AryaBhatta-GemmaOrca, which focuses on scientific and literary domains.
Key Capabilities
- Multi-Turn Chat Optimization: Designed and fine-tuned using Ultra-Chat datasets to excel in complex, multi-turn dialogue scenarios.
- Improved Performance: Shows enhanced performance on Hellaswag datasets, particularly in multi-turn conversations, compared to the AryaBhatta-GemmaOrca model.
- Multilingual Support: Offers support for English alongside nine Indian languages: Hindi, Tamil, Punjabi, Bengali, Gujarati, Oriya, Telugu, Kannada, and Malayalam.
Good For
- Conversational AI: Ideal for chatbots and virtual assistants requiring robust multi-turn interaction capabilities.
- Multilingual Applications: Suitable for applications targeting users in India, supporting a broad range of regional languages.
- Benchmarking: Performance can be referenced on the Indic LLM leaderboard.