TheBloke/WizardLM-13B-V1-0-Uncensored-SuperHOT-8K-fp16
TheBloke/WizardLM-13B-V1-0-Uncensored-SuperHOT-8K-fp16 is a 13 billion parameter language model, a merge of Eric Hartford's WizardLM 13B V1.0 Uncensored and Kaio Ken's SuperHOT 8K. This model is designed for extended context understanding, supporting an 8K context length, and aims to reduce refusals and bias compared to its base. It is suitable for applications requiring longer conversational memory and less restrictive content generation.
Loading preview...
Model Overview
This model, TheBloke/WizardLM-13B-V1-0-Uncensored-SuperHOT-8K-fp16, is a 13 billion parameter language model created by merging two distinct models: Eric Hartford's WizardLM 13B V1.0 Uncensored and Kaio Ken's SuperHOT 8K.
Key Capabilities & Features
- Extended Context Window: Leverages Kaio Ken's SuperHOT 8K merge to achieve an 8192-token context length, significantly enhancing its ability to handle longer inputs and maintain conversational coherence over extended interactions.
- Reduced Refusals and Bias: Built upon Eric Hartford's WizardLM 13B V1.0 Uncensored, which was retrained with a filtered dataset to minimize inherent ethical beliefs, refusals, avoidance, and bias present in the base LLaMA model.
- Instruction-Tuned: Follows Vicuna-1.1 style prompts, making it responsive to direct instructions in a helpful AI assistant format.
- fp16 Format: Provided in fp16 pytorch format, suitable for GPU inference and further conversions.
Use Cases
This model is particularly well-suited for applications requiring:
- Long-form content generation: Its 8K context window allows for processing and generating more extensive texts while retaining context.
- Conversational AI: Ideal for chatbots or virtual assistants where maintaining memory over many turns is crucial.
- Less restrictive content generation: The "uncensored" aspect means it will be more compliant with user prompts, reducing instances of refusal or avoidance, though users are responsible for the generated content.