OccultAI/Musecuilo-12B-Model_Stock
OccultAI/Musecuilo-12B-Model_Stock is a 12 billion parameter language model with a 32768 token context length, created by OccultAI using the 'model_stock' merge method. This model is a merge of Mistral-Nemo-Instruct-2407, Muse-12B, and Tlacuilo-12B, leveraging their combined strengths. It is designed for general language tasks, with a recommendation to use Mistral Tekken or ChatML chat templates for optimal performance.
Loading preview...
Musecuilo-12B-Model_Stock Overview
OccultAI's Musecuilo-12B-Model_Stock is a 12 billion parameter language model built upon the MistralForCausalLM architecture, featuring a substantial 32768 token context length. This model was developed using the model_stock merge method, combining three distinct base models: mistralai/Mistral-Nemo-Instruct-2407, LatitudeGames/Muse-12B, and allura-org/Tlacuilo-12B.
Key Characteristics
- Merge Method: Utilizes the
model_stockmerge technique, which is designed to integrate the strengths of multiple models. - Base Models: Merges Mistral-Nemo-Instruct-2407, Muse-12B, and Tlacuilo-12B.
- Recommended Templates: Optimized for use with Mistral Tekken or ChatML chat templates to achieve the best results.
- Configuration: The merge configuration specifies
filter_wise: trueand usesbfloat16for the output data type.
Usage Considerations
While Musecuilo-12B-Model_Stock is a capable model, users should be aware that it may exhibit some refusals. The README suggests that these can potentially be addressed through jailbreaking or ablation techniques, depending on the specific application requirements.