icefog72/IceWhiskeyRP-7b

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7BQuant:FP8Context Size:4kTool Calling:SupportedPublished:Oct 2, 2024License:cc-by-nc-4.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Icefog72/IceWhiskeyRP-7b is a 7 billion parameter language model merged using the SLERP method from Ice0.14-02.10-RP and Ice0.13-02.10-RP. This model is designed for roleplay applications, with an expected context window of 16-25k tokens, potentially up to 32k. It is optimized for handling extended conversational contexts in roleplaying scenarios.

Loading preview...

Overview

IceWhiskeyRP-7b is a 7 billion parameter language model developed by icefog72, created through a SLERP merge of two pre-trained models: icefog72/Ice0.14-02.10-RP and icefog72/Ice0.13-02.10-RP. This model is specifically designed and optimized for roleplaying (RP) applications.

Key Capabilities & Features

  • Extended Context Window: Expected to handle a context window of 16,000 to 25,000 tokens, with potential for up to 32,000 tokens, making it suitable for long-form roleplay scenarios.
  • Roleplay Optimization: Fine-tuned for roleplay, suggesting proficiency in generating coherent and engaging narrative responses within character and setting.
  • Merge Method: Utilizes the SLERP (Spherical Linear Interpolation) merge method, combining the strengths of its constituent models.
  • Quantized Versions Available: Provides Exl2 quantizations (4.2bpw, 6.5bpw, 8bpw) and GGUF formats for efficient deployment and inference.

Performance Metrics

Based on Open LLM Leaderboard evaluations, IceWhiskeyRP-7b (referred to as Ice0.15-02.10-RP in the leaderboard) shows an average score of 21.40. Specific scores include:

  • IFEval (0-Shot): 53.43
  • BBH (3-Shot): 30.13
  • MMLU-PRO (5-shot): 22.96

Good For

  • Roleplaying Applications: Ideal for developers building interactive storytelling, character-driven chatbots, or other roleplay-focused AI experiences.
  • Long Context Scenarios: Suitable for use cases requiring the model to maintain context over extended conversations or narratives.
  • Resource-Efficient Deployment: The availability of Exl2 and GGUF quantizations allows for deployment on systems with varying hardware constraints.

Finetunes
5 models