Mirage-Studio/llama-gaan-2-7b-chat-hf-dutch-epoch-5
TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kArchitecture:Transformer0.0K Cold
Mirage-Studio/llama-gaan-2-7b-chat-hf-dutch-epoch-5 is a 7 billion parameter Llama 2 Chat model, fine-tuned by Mirage Studio for Dutch language support. This model serves as a direct replacement for existing Llama 2 7B Chat models, specifically optimized for generating responses in Dutch. It is designed for applications requiring a Dutch-speaking conversational AI, offering improved performance in Dutch compared to its base model.
Loading preview...
Overview
Mirage-Studio/llama-gaan-2-7b-chat-hf-dutch-epoch-5 is a 7 billion parameter Llama 2 Chat model, developed by Mirage Studio. It is an epoch 5 checkpoint of a fine-tuned version of daryl149/llama-2-7b-chat-hf, with the primary goal of enhancing Dutch language capabilities.
Key Capabilities
- Dutch Language Support: Specifically fine-tuned to speak Dutch, making it suitable for Dutch-centric conversational AI applications.
- Llama 2 Chat Compatibility: Designed as a drop-in replacement for
meta-llama/Llama-2-7b-chat-hfanddaryl149/llama-2-7b-chat-hffor Dutch language tasks. - Conversational AI: Capable of engaging in helpful, respectful, and honest conversations, adhering to safety guidelines.
Usage Notes
- Prompt Template: Utilizes the standard Llama 2 chat prompt template with
[INST]and<<SYS>>tags. pad_token_id: Users must setpad_token_id=18610in their generator to avoid gibberish output.- Performance: Achieved 32 tokens/second on a V100S during training without advanced optimizations.
Limitations
- The model's Dutch proficiency is noted as "not quite perfect yet," indicating ongoing development and potential for further refinement.