File size: 1,800 Bytes
54659d8 d40c023 54659d8 1fbd04d 54659d8 153b209 54659d8 153b209 54659d8 6939902 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 |
---
license: apache-2.0
language:
- en
- de
- es
- fr
tags:
- sft
inference: false
datasets:
- OpenAssistant/oasst1
---
# Open-Assistant Llama2 70B SFT OASST
This model is a fine-tuning of [Llama2 70B](https://huggingface.co/meta-llama/Llama-2-70b-hf) LLM.
It was trained on a mixture of OASST top-1 threads.
## Model Details
- **Finetuned from:** [Llama2 70B](https://huggingface.co/meta-llama/Llama-2-70b-hf)
- **Model type:** Causal decoder-only transformer language model
- **Language:** English, German, Spanish, French (and limited capabilities in Italian, Portuguese, Polish, Dutch, Romanian, Czech, Swedish);
- **License:** Apache 2.0
- **Contact:** [Open-Assistant Discord](https://ykilcher.com/open-assistant-discord)
## Prompting
Two special tokens are used to mark the beginning of user and assistant turns:
`<|prompter|>` and `<|assistant|>`. Each turn ends with a `</s>` token.
Input prompt example:
```
<|prompter|>What is a meme, and what's the history behind this word?</s><|assistant|>
```
The input ends with the `<|assistant|>` token to signal that the model should
start generating the assistant reply.
# [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_jordiclive__Llama-2-70b-oasst-1-200)
| Metric | Value |
|-----------------------|---------------------------|
| Avg. | 57.11 |
| ARC (25-shot) | 67.66 |
| HellaSwag (10-shot) | 87.24 |
| MMLU (5-shot) | 69.95 |
| TruthfulQA (0-shot) | 51.28 |
| Winogrande (5-shot) | 84.14 |
| GSM8K (5-shot) | 32.75 |
| DROP (3-shot) | 6.73 |
|