WizardLM-30B-Uncensored-Guanaco-SuperCOT-30b-GGUF
| Property | Value |
|---|---|
| Parameter Count | 32.5B |
| License | Other |
| Author | tensorblock |
| Format | GGUF |
| Base Model | WizardLM-30B-Uncensored-Guanaco |
What is WizardLM-30B-Uncensored-Guanaco-SuperCOT-30b-GGUF?
This is a GGUF-formatted variant of the WizardLM-30B model, specifically optimized for efficient deployment and inference. The model offers multiple quantization options ranging from 11GB to 32GB, making it adaptable to various hardware configurations while maintaining performance.
Implementation Details
The model is available in 12 different quantization variants, from Q2_K (11.221GB) to Q8_0 (32.191GB), each offering different trade-offs between model size and quality. It's trained on seven diverse datasets including WizardLM alpaca evolutional instruction, SuperCOT, and various Alpaca derivatives.
- Compatible with llama.cpp (commit b4011)
- Offers multiple GGUF quantization options
- Trained on comprehensive instruction-following datasets
Core Capabilities
- Uncensored text generation and completion
- Efficient memory usage through various quantization options
- Balanced performance with Q4_K_M (18.274GB) recommended for general use
- Superior quality with Q5_K_M (21.464GB) for demanding applications
Frequently Asked Questions
Q: What makes this model unique?
The model combines the powerful WizardLM architecture with extensive quantization options, making it highly accessible for different hardware configurations while maintaining quality. It's particularly notable for its uncensored nature and comprehensive training on multiple instruction-following datasets.
Q: What are the recommended use cases?
For general use, the Q4_K_M variant (18.274GB) offers a good balance of quality and size. For applications requiring higher quality, the Q5_K_M variant (21.464GB) is recommended. The model is suitable for text generation, completion, and instruction-following tasks where uncensored output is desired.
