Aratako/Antler-7B-RP
Aratako/Antler-7B-RP is a 7 billion parameter language model, fine-tuned from Elizezen/Antler-7B, specifically optimized for role-playing tasks. It leverages a 4096-token context length and is designed to generate character-specific dialogue and narrative descriptions based on detailed persona and scenario prompts. This model excels at maintaining consistent character voices and handling complex role-play interactions.
Loading preview...
Aratako/Antler-7B-RP: Role-Play Optimized Language Model
Aratako/Antler-7B-RP is a 7 billion parameter model, fine-tuned from the Elizezen/Antler-7B base model, with a primary focus on enhancing role-playing capabilities. It was trained using LoRA with specialized role-play datasets, including grimulkan/LimaRP-augmented and Aratako/Rosebleu-1on1-Dialogues-RP.
Key Capabilities
- Specialized Role-Play: Designed to generate character dialogue and narrative descriptions consistent with provided persona and scenario settings.
- Mistral Chat Template: Utilizes the Mistral chat template for prompt formatting, with specific recommendations for structuring role-play instructions, world-building, and character settings.
- Contextual Generation: Supports detailed input formats for character speech (
キャラ名「発話」) and internal thoughts/scene descriptions (（）). - Multiturn Conversation: Handles ongoing role-play sessions, requiring an
eos_token(</s>) at the end of each assistant response in multiturn dialogues.
Training Details
The model was trained for 10 epochs with a batch size of 64 and a maximum sequence length of 8192, using a learning rate of 2e-5. LoRA parameters included lora_r: 128 and lisa_alpha: 256.
Known Issues
- Occasionally, the model may output an EOS token prematurely without generating content. This can often be resolved by continuing the generation or retrying the prompt.
Licensing
Released under the Apache-2.0 license. Note that the license may be subject to change depending on clarification regarding the base model's licensing.