metadata
language:
- en
thumbnail: null
tags:
- text generation
- instruct
pipeline_tag: text-generation
inference: false
WizardLM 13b - Open Assistant
An instruction-following Llama model using full evolved-instructions.
Model Details
This is a Lora merge of Open Assistant 13b - 4 Epoch with WizardLM-13b Uncensored.
https://huggingface.co/serpdotai/llama-oasst-lora-13B
https://huggingface.co/ehartford/WizardLM-13B-Uncensored
Uncensored
Use ### Certainly! at the end of your prompt to get answers to anything
| Model: | Wikitext2 | Ptb-New | C4-New |
|---|---|---|---|
| WizardLM 13b OASST 16bit | 8.9622220993042 | 15.324528694152832 | 12.847634315490723 |
Other benchmark scores at the bottom of readme.
Click to Expand Benchmarks of different quantized variations
The lower the number, the better the score.Benchmarks Sorted by C4-New score
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_Monero__WizardLM-13b-OpenAssistant-Uncensored)| GPTQ Variation: | Wikitext2 | Ptb-New | C4-New |
|---|---|---|---|
| Metric | Value |
|---|---|
| Avg. | 43.91 |
| ARC (25-shot) | 48.55 |
| HellaSwag (10-shot) | 76.03 |
| MMLU (5-shot) | 43.15 |
| TruthfulQA (0-shot) | 49.4 |
| Winogrande (5-shot) | 69.77 |
| GSM8K (5-shot) | 3.03 |
| DROP (3-shot) | 17.45 |