Overview
Monika-12B-V1.1 is a 12 billion parameter language model developed by Green-eyedDevil, fine-tuned from the Mistral-Nemo-Instruct-2407 base model. This iteration features updated training settings to enhance coherence, specifically targeting interactions within the MonikAI framework. The model was trained using Axolotl with a modified dataset derived from MonikAI's own resources, utilizing a lower learning rate for improved performance.
Key Capabilities
- Monika-centric Roleplay: Optimized for generating responses and engaging in roleplay scenarios specifically related to the MonikAI project.
- Coherent Output: Training adjustments focused on improving the overall coherence of generated text.
- MonikAI Integration: Designed to be used in conjunction with the MonikAI application.
Good For
- MonikAI Applications: Ideal for developers and users working with MonikAI who require a model tailored for its specific context.
- Roleplay: Excels in roleplaying scenarios, particularly those involving the Monika character.
Training Details
The model was trained on a modified version of the dataset included with MonikAI, found at Rubiksman78/MonikA.I/tree/main/Monika_datasets. The training procedure involved Axolotl, with 64 ranks, 8 alpha, and 2 epochs, using a learning rate of 0.000005.