Nous-Hermes-Llama2-13b
| Property | Value |
|---|---|
| Parameter Count | 13B |
| License | MIT |
| Training Data | 300,000+ instructions |
| Architecture | Llama-2 Based |
| Sequence Length | 4096 tokens |
What is Nous-Hermes-Llama2-13b?
Nous-Hermes-Llama2-13b is a state-of-the-art language model that represents a significant advancement in AI text generation. Developed by Nous Research, it's built upon the Llama-2 architecture and fine-tuned using over 300,000 carefully curated instructions, primarily derived from GPT-4 outputs. The model was trained on an 8x A100 80GB DGX machine, emphasizing quality and performance.
Implementation Details
The model leverages synthetic GPT-4 outputs and diverse datasets including GPTeacher, roleplay datasets, code instruct datasets, and Nous Instruct & PDACTL. It follows the Alpaca prompt format and has achieved impressive benchmark results, including top positions in ARC-c, ARC-e, Hellaswag, and OpenBookQA tests.
- Extensive fine-tuning on 300k+ instructions
- 4096 token sequence length capability
- Supports both instruction-only and instruction-with-input formats
- Built using the Axolotl framework
Core Capabilities
- Enhanced response length and detail
- Reduced hallucination rate compared to similar models
- Strong performance in reasoning and comprehension tasks
- Versatile application from creative writing to technical instruction
- Benchmark scores averaging 70.0 on GPT4All tests
Frequently Asked Questions
Q: What makes this model unique?
The model stands out for its combination of long-form responses, reduced hallucination rate, and absence of traditional censorship mechanisms. It achieves state-of-the-art performance on multiple benchmarks while maintaining versatility across different use cases.
Q: What are the recommended use cases?
The model excels in various applications including creative writing, technical documentation, coding assistance, and complex instruction following. It's particularly well-suited for tasks requiring detailed, accurate responses and can be integrated into chat interfaces through platforms like LM Studio.
