| Property | Value |
|---|---|
| Parameter Count | 7.24B |
| License | Apache-2.0 |
| Base Model | Mistral-7B-Instruct-v0.2 |
| Format | GGUF |
What is Mistral-Gutenberg-Doppel-7B-FFT-GGUF?
This is a quantized version of the Mistral-Gutenberg-Doppel model, specifically optimized using llama.cpp. The model represents a full finetune of Mistral-7B-Instruct-v0.2, trained on comprehensive literary datasets from Project Gutenberg.
Implementation Details
The model underwent ORPO (Optimal Reward Policy Optimization) training using 4x A100 GPUs for 2 epochs. It incorporates data from two key datasets: jondurbin/gutenberg-dpo-v0.1 and nbeerbower/gutenberg2-dpo, making it particularly well-suited for literary and textual applications.
- Quantized architecture optimized for efficiency
- Full finetune implementation rather than QLoRA
- Trained using ORPO methodology
Core Capabilities
- Enhanced literary text generation
- Improved narrative understanding
- Optimized for conversational applications
- Efficient deployment through GGUF format
Frequently Asked Questions
Q: What makes this model unique?
This model combines the powerful Mistral-7B architecture with specialized training on literary texts, using full finetuning rather than parameter-efficient methods. The GGUF format makes it particularly suitable for deployment in resource-constrained environments.
Q: What are the recommended use cases?
The model is particularly well-suited for literary applications, text generation, and conversational tasks that require understanding of narrative structure and literary style. It's optimized for both performance and efficiency through its GGUF format.