Mytho Max L2 13b

Introduction

The MythoMax-L2-13B model is an advanced variant of MythoMix, merging the capabilities of MythoLogic-L2 and Huginn using a unique tensor type merge technique. This model excels in roleplaying and storywriting due to its enhanced coherency and functionality.

Architecture

The model is built on the Llama architecture and utilizes a complex tensor merging method, where each layer comprises multiple tensors with specific functions. The integration of MythoLogic-L2's understanding and Huginn's writing ability results in a model proficient in both areas. The model's structure cannot be easily visualized due to the unique ratios applied to its 363 tensors, refined with gradients for finetuning.

Training

The MythoMax-L2-13B model was developed using a novel tensor type merge technique, allowing more intermingling of Huginn within the model's structure. This approach enhances model coherency and performance. The model uses Alpaca formatting for optimal performance, with specific instructions for roleplay scenarios.

Guide: Running Locally

1. Setup Environment: Install required libraries such as PyTorch and Transformers.
2. Download Model: Access quantized model versions from TheBloke on Hugging Face for GGUF, GPTQ, or AWQ formats.
3. Run Inference: Utilize the text-generation-inference tool for model deployment.
4. Hardware Recommendation: For optimal performance, consider using cloud GPUs from providers like AWS, GCP, or Azure.

License

The MythoMax-L2-13B model is released under a non-standard license, classified as "other." Users should refer to the specific license terms for usage guidelines.