Upload README.md
Browse files
README.md
CHANGED
@@ -1,11 +1,20 @@
|
|
1 |
---
|
|
|
|
|
2 |
inference: false
|
|
|
|
|
3 |
license: llama2
|
4 |
model_creator: PygmalionAI
|
5 |
model_link: https://huggingface.co/PygmalionAI/pygmalion-2-7b
|
6 |
model_name: Pygmalion 2 7B
|
7 |
model_type: llama
|
|
|
8 |
quantized_by: TheBloke
|
|
|
|
|
|
|
|
|
9 |
---
|
10 |
|
11 |
<!-- header start -->
|
@@ -53,6 +62,15 @@ The model has been trained on prompts using three different roles, which are den
|
|
53 |
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.
|
54 |
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 form a conversation history.
|
55 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
56 |
|
57 |
<!-- prompt-template end -->
|
58 |
|
@@ -170,6 +188,15 @@ prompt_template=f'''The model has been trained on prompts using three different
|
|
170 |
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.
|
171 |
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 form a conversation history.
|
172 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
173 |
'''
|
174 |
|
175 |
print("\n\n*** Generate:")
|
@@ -239,4 +266,51 @@ And thank you again to a16z for their generous grant.
|
|
239 |
|
240 |
# Original model card: PygmalionAI's Pygmalion 2 7B
|
241 |
|
242 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
datasets:
|
3 |
+
- PygmalionAI/PIPPA
|
4 |
inference: false
|
5 |
+
language:
|
6 |
+
- en
|
7 |
license: llama2
|
8 |
model_creator: PygmalionAI
|
9 |
model_link: https://huggingface.co/PygmalionAI/pygmalion-2-7b
|
10 |
model_name: Pygmalion 2 7B
|
11 |
model_type: llama
|
12 |
+
pipeline_tag: text-generation
|
13 |
quantized_by: TheBloke
|
14 |
+
tags:
|
15 |
+
- text generation
|
16 |
+
- instruct
|
17 |
+
thumbnail: null
|
18 |
---
|
19 |
|
20 |
<!-- header start -->
|
|
|
62 |
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.
|
63 |
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 form a conversation history.
|
64 |
|
65 |
+
The system prompt has been designed to allow the model to "enter" various modes and dictate the reply length. Here's an example:
|
66 |
+
|
67 |
+
```
|
68 |
+
<|system|>Enter RP mode. Pretend to be {{char}} whose persona follows:
|
69 |
+
{{persona}}
|
70 |
+
|
71 |
+
You shall reply to the user while staying in character, and generate long responses.
|
72 |
+
```
|
73 |
+
|
74 |
|
75 |
<!-- prompt-template end -->
|
76 |
|
|
|
188 |
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.
|
189 |
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 form a conversation history.
|
190 |
|
191 |
+
The system prompt has been designed to allow the model to "enter" various modes and dictate the reply length. Here's an example:
|
192 |
+
|
193 |
+
```
|
194 |
+
<|system|>Enter RP mode. Pretend to be {{char}} whose persona follows:
|
195 |
+
{{persona}}
|
196 |
+
|
197 |
+
You shall reply to the user while staying in character, and generate long responses.
|
198 |
+
```
|
199 |
+
|
200 |
'''
|
201 |
|
202 |
print("\n\n*** Generate:")
|
|
|
266 |
|
267 |
# Original model card: PygmalionAI's Pygmalion 2 7B
|
268 |
|
269 |
+
<h1 style="text-align: center">Pygmalion-2 7B</h1>
|
270 |
+
<h2 style="text-align: center">An instruction-tuned Llama-2 biased towards fiction writing and conversation.</h2>
|
271 |
+
|
272 |
+
## Model Details
|
273 |
+
|
274 |
+
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
|
275 |
+
[Llama-2 7B](https://huggingface.co/meta-llama/llama-2-7b-hf) released by Meta AI.
|
276 |
+
|
277 |
+
The Metharme models were an experiment to try and get a model that is usable for conversation, roleplaying and storywriting,
|
278 |
+
but which can be guided using natural language like other instruct models. After much deliberation, we reached the conclusion
|
279 |
+
that the Metharme prompting format is superior (and easier to use) compared to the classic Pygmalion.
|
280 |
+
|
281 |
+
This model was trained by doing supervised fine-tuning over a mixture of regular instruction data alongside roleplay, fictional stories
|
282 |
+
and conversations with synthetically generated instructions attached.
|
283 |
+
|
284 |
+
This model is freely available for both commercial and non-commercial use, as per the Llama-2 license.
|
285 |
+
|
286 |
+
|
287 |
+
## Prompting
|
288 |
+
|
289 |
+
The model has been trained on prompts using three different roles, which are denoted by the following tokens: `<|system|>`, `<|user|>` and `<|model|>`.
|
290 |
+
|
291 |
+
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.
|
292 |
+
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
|
293 |
+
form a conversation history.
|
294 |
+
|
295 |
+
### Prompting example
|
296 |
+
|
297 |
+
The system prompt has been designed to allow the model to "enter" various modes and dictate the reply length. Here's an example:
|
298 |
+
|
299 |
+
```
|
300 |
+
<|system|>Enter RP mode. Pretend to be {{char}} whose persona follows:
|
301 |
+
{{persona}}
|
302 |
+
|
303 |
+
You shall reply to the user while staying in character, and generate long responses.
|
304 |
+
```
|
305 |
+
|
306 |
+
## Dataset
|
307 |
+
The dataset used to fine-tune this model includes our own [PIPPA](https://huggingface.co/datasets/PygmalionAI/PIPPA), along with several other instruction
|
308 |
+
datasets, and datasets acquired from various RP forums.
|
309 |
+
|
310 |
+
## Limitations and biases
|
311 |
+
|
312 |
+
The intended use-case for this model is fictional writing for entertainment purposes. Any other sort of usage is out of scope.
|
313 |
+
|
314 |
+
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
|
315 |
+
are lewd or otherwise offensive. It may produce socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive.
|
316 |
+
Outputs might often be factually wrong or misleading.
|