OpenLLM-Ro/RoLlama3-8b-Instruct-DPO

TEXT GENERATIONPricing:Input $0.37 / Cached $0.074 / Output $0.38Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Oct 9, 2024License:cc-by-nc-4.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

OpenLLM-Ro/RoLlama3-8b-Instruct-DPO is an 8 billion parameter instruction-tuned generative text model developed by OpenLLM-Ro, built upon the Meta Llama 3 architecture. Specialized for the Romanian language, this model is fine-tuned using Direct Preference Optimization (DPO) on various Romanian datasets to enhance human alignment. It excels in assistant-like chat applications and research use cases within Romanian natural language processing, demonstrating strong performance in Romanian MT-Bench and RoCulturaBench.

Loading preview...

RoLlama3-8b-Instruct-DPO: Romanian-Specialized Llama 3 Model

OpenLLM-Ro/RoLlama3-8b-Instruct-DPO is an 8 billion parameter instruction-tuned model, part of the RoLlama3 family, specifically designed for the Romanian language. Developed by OpenLLM-Ro, this model is built on the Meta Llama 3 architecture and has been fine-tuned using Direct Preference Optimization (DPO) on a collection of Romanian datasets including RoHelpSteer, RoUltraFeedback, RoMagpieDPO, RoArgillaMagpie, and RoHelpSteer2. This DPO fine-tuning aims to enhance human alignment, making it particularly effective for assistant-like chat applications.

Key Capabilities

  • Romanian Language Specialization: Represents the first open-source effort to build a large language model specialized for Romanian, offering strong performance in Romanian NLP tasks.
  • Instruction Following: Optimized for understanding and generating responses based on instructions, suitable for conversational AI.
  • Human Alignment: Fine-tuned with DPO datasets to improve helpfulness and safety, as evidenced by its leading scores on Romanian MT-Bench and RoCulturaBench.
  • Research Tool: Intended for research use in Romanian natural language processing, with base models adaptable for various tasks.

When to Use This Model

  • Assistant-like Chatbots: Ideal for creating conversational agents that interact in Romanian.
  • Romanian NLP Research: A valuable resource for academic and research projects focused on the Romanian language.
  • Content Generation: Suitable for generating text, summaries, or creative content in Romanian.

This model demonstrates competitive performance against its Llama-3-8B-Instruct counterpart in several Romanian-specific benchmarks, notably achieving the highest average scores on Romanian MT-Bench (6.67) and RoCulturaBench (4.83) among the RoLlama3 variants.

Finetunes
3 models