yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated

TEXT GENERATIONConcurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 4, 2025Architecture:Transformer0.0K Featherless Exclusive Cold

yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated is a 12 billion parameter language model, combining the shisa-ai/shisa-v2-mistral-nemo-12b base model with the nbeerbower/Mistral-Nemo-12B-abliterated-LORA adapter. This merged model is provided in bfloat16 format, optimized for immediate deployment or further fine-tuning. Its primary characteristic is being a composite model, leveraging a LoRA adapter to potentially enhance or specialize the base model's capabilities.

Loading preview...

Model Overview

This model, named yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated, is a 12 billion parameter language model created by merging a base model with a LoRA adapter. It is provided in bfloat16 format, making it suitable for direct deployment or subsequent fine-tuning tasks.

Key Components

  • Base Model: The foundation of this model is shisa-ai/shisa-v2-mistral-nemo-12b.
  • LoRA Adapter: The base model has been enhanced using the nbeerbower/Mistral-Nemo-12B-abliterated-LORA adapter.

Merging Process

The model was created by loading the base model in bfloat16, then applying and merging the LoRA adapter's weights into the base model. This process results in a unified model that integrates the specialized knowledge or fine-tuning from the LoRA adapter directly into the larger base model.

Use Cases

This model is ready for:

  • Deployment: Can be used directly for inference in applications.
  • Further Fine-tuning: Serves as a strong starting point for additional domain-specific or task-specific fine-tuning, building upon the combined strengths of its base and LoRA components.