Commit
•
e96aa1e
1
Parent(s):
9b13660
Update README.md
Browse files
README.md
CHANGED
@@ -1,7 +1,53 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
- base model: [andreaskoepf/pythia-12b-pre-2000](https://huggingface.co/andreaskoepf/pythia-12b-pre-2000)
|
6 |
- checkpoint: 4000 steps
|
7 |
- [sampling report](https://open-assistant.github.io/oasst-model-eval/?f=https%3A%2F%2Fraw.githubusercontent.com%2FOpen-Assistant%2Foasst-model-eval%2Fmain%2Fsampling_reports%2Foasst-sft%2F2023-04-03_andreaskoepf_oasst-sft-4-pythia-12b-epoch-3_5_sampling_noprefix_lottery.json%0Ahttps%3A%2F%2Fraw.githubusercontent.com%2FOpen-Assistant%2Foasst-model-eval%2Fmain%2Fsampling_reports%2Fchat-gpt%2F2023-04-11_gpt-3.5-turbo_lottery.json)
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
tags:
|
6 |
+
- sft
|
7 |
+
pipeline_tag: text-generation
|
8 |
+
widget:
|
9 |
+
- text: <|prompter|>What is a meme, and what's the history behind this word?<|endoftext|><|assistant|>
|
10 |
+
- text: <|prompter|>What's the Earth total population<|endoftext|><|assistant|>
|
11 |
+
- text: <|prompter|>Write a story about future of AI development<|endoftext|><|assistant|>
|
12 |
---
|
13 |
+
|
14 |
+
# Open-Assistant SFT-4 12B Model
|
15 |
+
|
16 |
+
|
17 |
+
This is the 4th iteration English supervised-fine-tuning (SFT) model of
|
18 |
+
the [Open-Assistant](https://github.com/LAION-AI/Open-Assistant) project.
|
19 |
+
It is based on a Pythia 12B that was fine-tuned on human demonstrations
|
20 |
+
of assistant conversations collected through the
|
21 |
+
[https://open-assistant.io/](https://open-assistant.io/) human feedback web
|
22 |
+
app before March 25, 2023.
|
23 |
+
|
24 |
+
## Model Details
|
25 |
+
|
26 |
+
- **Developed by:** [Open-Assistant Contributors](https://open-assistant.io/)
|
27 |
+
- **Model type:** Transformer-based Language Model
|
28 |
+
- **Language:** English
|
29 |
+
- **Finetuned from:** [EleutherAI / pythia-12b-deduped](https://huggingface.co/EleutherAI/pythia-12b-deduped)
|
30 |
+
- **Code:** [Open-Assistant/model/model_training](https://github.com/LAION-AI/Open-Assistant/tree/main/model/model_training)
|
31 |
+
- **Demo:** [Continuations for 250 random prompts](https://open-assistant.github.io/oasst-model-eval/?f=https%3A%2F%2Fraw.githubusercontent.com%2FOpen-Assistant%2Foasst-model-eval%2Fmain%2Fsampling_reports%2Foasst-sft%2F2023-03-09_andreaskoepf_oasst-1_12b_7000_sampling_noprefix_lottery.json) ([sampling code](https://github.com/Open-Assistant/oasst-model-eval/blob/3d71f3be100c05cd8ddb568365e036a29fbff8c7/model_eval/manual/sampling_report.py)).
|
32 |
+
- **License:** Apache 2.0
|
33 |
+
- **Contact:** [Open-Assistant Discord](https://ykilcher.com/open-assistant-discord)
|
34 |
+
|
35 |
+
## Prompting
|
36 |
+
|
37 |
+
Two special tokens are used to mark the beginning of user and assistant turns:
|
38 |
+
`<|prompter|>` and `<|assistant|>`. Each turn ends with a `<|endoftext|>` token.
|
39 |
+
|
40 |
+
Input prompt example:
|
41 |
+
```
|
42 |
+
<|prompter|>What is a meme, and what's the history behind this word?<|endoftext|><|assistant|>
|
43 |
+
```
|
44 |
+
The input ends with the `<|assistant|>` token to signal that the model should
|
45 |
+
start generating the assistant reply.
|
46 |
+
|
47 |
+
|
48 |
+
## Dev Details
|
49 |
+
|
50 |
+
- wandb: https://wandb.ai/open-assistant/supervised-finetuning/runs/770a0t41
|
51 |
- base model: [andreaskoepf/pythia-12b-pre-2000](https://huggingface.co/andreaskoepf/pythia-12b-pre-2000)
|
52 |
- checkpoint: 4000 steps
|
53 |
- [sampling report](https://open-assistant.github.io/oasst-model-eval/?f=https%3A%2F%2Fraw.githubusercontent.com%2FOpen-Assistant%2Foasst-model-eval%2Fmain%2Fsampling_reports%2Foasst-sft%2F2023-04-03_andreaskoepf_oasst-sft-4-pythia-12b-epoch-3_5_sampling_noprefix_lottery.json%0Ahttps%3A%2F%2Fraw.githubusercontent.com%2FOpen-Assistant%2Foasst-model-eval%2Fmain%2Fsampling_reports%2Fchat-gpt%2F2023-04-11_gpt-3.5-turbo_lottery.json)
|