Aratako/AntlerStar-RP

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7BQuant:FP8Context Size:4kTool Calling:SupportedPublished:Apr 18, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Aratako/AntlerStar-RP is a 7 billion parameter language model developed by Aratako, created through a complex merge of multiple pre-trained models using mergekit. This model is specifically designed and optimized for roleplay (RP) applications, combining the strengths of its base models and several Chat Vector additions. It leverages advanced merging techniques like Model Stock and DARE TIES to enhance its roleplaying capabilities and conversational coherence. The model is intended for use cases requiring nuanced character interaction and narrative generation.

Loading preview...

AntlerStar-RP: A Merged Model for Roleplay

AntlerStar-RP is a 7 billion parameter language model developed by Aratako, meticulously constructed using the mergekit framework. Its primary purpose is to excel in roleplay (RP) scenarios, building upon the foundations of two core models: Aratako/Antler-7B-RP-v3 and Aratako/Japanese-Starling-ChatV-7B-RP.

Key Capabilities & Merging Process

This model's unique strength comes from its sophisticated merging strategy:

  • Chat Vector Integration: Initially, Chat Vectors from four distinct models—senseable/WestLake-7B-v2, SanjiWatsuki/Kunoichi-DPO-v2-7B, SanjiWatsuki/Silicon-Maid-7B, and SanjiWatsuki/Loyal-MacaroniMaid-7B—were added to the base models. This step aimed to infuse diverse conversational styles and knowledge.
  • Model Stock Merging: The enhanced base models were then merged using the "Model Stock" technique within mergekit, creating two intermediate models. This method combines multiple models based on a base model, allowing for a structured integration of features.
  • DARE TIES Final Merge: The two intermediate models were finally merged using the DARE TIES technique, a method known for effectively combining models while preserving their individual strengths and reducing catastrophic forgetting. This final step consolidates the roleplay-specific enhancements.

Good for

  • Roleplaying Applications: Designed specifically for generating engaging and coherent roleplay narratives and character interactions.
  • Creative Writing: Suitable for tasks requiring imaginative text generation with distinct character voices.
  • Conversational AI: Benefits from the integrated Chat Vectors, potentially offering more nuanced and varied conversational responses.

Finetunes
2 models