Update README.md
Browse files
README.md
CHANGED
@@ -1,70 +0,0 @@
|
|
1 |
-
---
|
2 |
-
language:
|
3 |
-
- en
|
4 |
-
thumbnail: null
|
5 |
-
tags:
|
6 |
-
- text generation
|
7 |
-
- instruct
|
8 |
-
pipeline_tag: text-generation
|
9 |
-
inference: false
|
10 |
-
license: llama2
|
11 |
-
datasets:
|
12 |
-
- PygmalionAI/PIPPA
|
13 |
-
- Open-Orca/OpenOrca
|
14 |
-
- Norquinal/claude_multiround_chat_30k
|
15 |
-
- jondurbin/airoboros-gpt4-1.4.1
|
16 |
-
- databricks/databricks-dolly-15k
|
17 |
-
---
|
18 |
-
<h1 style="text-align: center">Pygmalion-2 7B</h1>
|
19 |
-
<h2 style="text-align: center">An instruction-tuned Llama-2 biased towards fiction writing and conversation.</h2>
|
20 |
-
|
21 |
-
## Model Details
|
22 |
-
|
23 |
-
The long-awaited release of our new models based on Llama-2 is finally here. Pygmalion-2 7B (formerly known as Metharme) is based on
|
24 |
-
[Llama-2 7B](https://huggingface.co/meta-llama/llama-2-7b-hf) released by Meta AI.
|
25 |
-
|
26 |
-
The Metharme models were an experiment to try and get a model that is usable for conversation, roleplaying and storywriting,
|
27 |
-
but which can be guided using natural language like other instruct models. After much deliberation, we reached the conclusion
|
28 |
-
that the Metharme prompting format is superior (and easier to use) compared to the classic Pygmalion.
|
29 |
-
|
30 |
-
This model was trained by doing supervised fine-tuning over a mixture of regular instruction data alongside roleplay, fictional stories
|
31 |
-
and conversations with synthetically generated instructions attached.
|
32 |
-
|
33 |
-
This model is freely available for both commercial and non-commercial use, as per the Llama-2 license.
|
34 |
-
|
35 |
-
|
36 |
-
## Prompting
|
37 |
-
|
38 |
-
The model has been trained on prompts using three different roles, which are denoted by the following tokens: `<|system|>`, `<|user|>` and `<|model|>`.
|
39 |
-
|
40 |
-
The `<|system|>` prompt can be used to inject out-of-channel information behind the scenes, while the `<|user|>` prompt should be used to indicate user input.
|
41 |
-
The `<|model|>` token should then be used to indicate that the model should generate a response. These tokens can happen multiple times and be chained up to
|
42 |
-
form a conversation history.
|
43 |
-
|
44 |
-
### Prompting example
|
45 |
-
|
46 |
-
The system prompt has been designed to allow the model to "enter" various modes and dictate the reply length. Here's an example:
|
47 |
-
|
48 |
-
```
|
49 |
-
<|system|>Enter RP mode. Pretend to be {{char}} whose persona follows:
|
50 |
-
{{persona}}
|
51 |
-
|
52 |
-
You shall reply to the user while staying in character, and generate long responses.
|
53 |
-
```
|
54 |
-
|
55 |
-
## Dataset
|
56 |
-
The dataset used to fine-tune this model includes our own [PIPPA](https://huggingface.co/datasets/PygmalionAI/PIPPA), along with several other instruction
|
57 |
-
datasets, and datasets acquired from various RP forums.
|
58 |
-
|
59 |
-
## Limitations and biases
|
60 |
-
|
61 |
-
The intended use-case for this model is fictional writing for entertainment purposes. Any other sort of usage is out of scope.
|
62 |
-
|
63 |
-
As such, it was **not** fine-tuned to be safe and harmless: the base model _and_ this fine-tune have been trained on data known to contain profanity and texts that
|
64 |
-
are lewd or otherwise offensive. It may produce socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive.
|
65 |
-
Outputs might often be factually wrong or misleading.
|
66 |
-
|
67 |
-
## Acknowledgements
|
68 |
-
We would like to thank [SpicyChat](https://spicychat.ai/) for sponsoring the training for this model.
|
69 |
-
|
70 |
-
[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|