dfurman/Qwen2-72B-Orpo-v0.1

TEXT GENERATIONConcurrency Cost:4Model Size:72.7BQuant:FP8Ctx Length:32kPublished:Jul 5, 2024License:tongyi-qianwenArchitecture:Transformer0.0K Cold

dfurman/Qwen2-72B-Orpo-v0.1 is a 72.7 billion parameter language model, fine-tuned from Qwen/Qwen2-72B-Instruct using the mlabonne/orpo-dpo-mix-40k dataset. This model is designed as a generalist for diverse text generation tasks, including agentic capabilities, roleplaying, reasoning, and multi-turn conversations, with a 32K context length. It aims for long context coherence and broad applicability across various generative AI use cases.

Loading preview...

Model Overview

dfurman/Qwen2-72B-Orpo-v0.1 is a 72.7 billion parameter language model, fine-tuned from the robust Qwen/Qwen2-72B-Instruct base model. The fine-tuning process utilized 1.5k rows from the mlabonne/orpo-dpo-mix-40k dataset, enhancing its capabilities as a generalist language model.

Key Capabilities

  • Generalist Text Generation: Designed to handle a wide array of text generation tasks.
  • Agentic Support: Includes features that support agentic behaviors and interactions.
  • Roleplaying: Capable of engaging in roleplaying scenarios.
  • Reasoning: Demonstrates reasoning abilities for complex problem-solving.
  • Multi-turn Conversations: Maintains coherence and context across extended dialogues.
  • Long Context Coherence: Optimized to manage and understand information over its 32K token context window.

Performance Highlights

Evaluations on the Open LLM Leaderboard show an average score of 43.32. Specific metric scores include:

  • IFEval (0-Shot): 78.80
  • BBH (3-Shot): 57.41
  • MATH Lvl 5 (4-Shot): 35.42
  • MMLU-PRO (5-shot): 49.50

When to Use This Model

This model is suitable for developers seeking a versatile large language model for applications requiring strong general text generation, conversational AI, and reasoning. Its fine-tuning for agentic and roleplaying capabilities makes it particularly useful for interactive and dynamic AI systems.