Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: creativeml-openrail-m
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
thumbnail:
|
6 |
+
tags:
|
7 |
+
- text generation
|
8 |
+
- conversational
|
9 |
+
- ggml
|
10 |
+
inference: false
|
11 |
+
---
|
12 |
+
|
13 |
+
# Pygmalion 6B GGML (main branch)
|
14 |
+
### This repository contains quantized conversions of the "main" Pygmalion 6B checkpoint.
|
15 |
+
*For use with frontends that support GGML quantized GPT-J models, such as KoboldCpp and Oobabooga (with the CTransformers loader).*
|
16 |
+
|
17 |
+
The original model card can be found below.
|
18 |
+
|
19 |
+
* * *
|
20 |
+
|
21 |
+
# Pygmalion 6B
|
22 |
+
|
23 |
+
## Model description
|
24 |
+
|
25 |
+
Pymalion 6B is a proof-of-concept dialogue model based on EleutherAI's [GPT-J-6B](https://huggingface.co/EleutherAI/gpt-j-6B).
|
26 |
+
|
27 |
+
**Warning:** This model is **NOT** suitable for use by minors. It **will** output X-rated content under certain circumstances.
|
28 |
+
|
29 |
+
## Training data
|
30 |
+
|
31 |
+
The fine-tuning dataset consisted of 56MB of dialogue data gathered from multiple sources, which includes both real _and_ partially machine-generated conversations.
|
32 |
+
|
33 |
+
## Training procedure
|
34 |
+
|
35 |
+
Model weights were initialized from the `uft-6b` ConvoGPT model made available in [this commit](https://huggingface.co/hakurei/convogpt/tree/41b67bfddb6cd97070ffddf708e9720c9cb8d224/6b-uft).
|
36 |
+
|
37 |
+
The model was then further fine-tuned on ~48.5 million tokens for ~5k steps on 4 NVIDIA A40s using DeepSpeed.
|
38 |
+
|
39 |
+
## Intended use
|
40 |
+
|
41 |
+
### The easy way
|
42 |
+
|
43 |
+
We provide a notebook with a Gradio UI for playing around with the model without having to manually format inputs. This notebook can be found [here](https://github.com/PygmalionAI/gradio-ui/blob/master/notebooks/GPU.ipynb).
|
44 |
+
|
45 |
+
### The manual way
|
46 |
+
|
47 |
+
The model can be used as a regular text generation model, but it'll perform best if the input prompt adheres to the following format:
|
48 |
+
|
49 |
+
```
|
50 |
+
[CHARACTER]'s Persona: [A few sentences about the character you want the model to play]
|
51 |
+
<START>
|
52 |
+
[DIALOGUE HISTORY]
|
53 |
+
You: [Your input message here]
|
54 |
+
[CHARACTER]:
|
55 |
+
```
|
56 |
+
|
57 |
+
Where `[CHARACTER]` is, as you can probably guess, the name of the character you want the model to portray, `<START>` should be used verbatim as a delimiter token to separate persona and scenario data from the dialogue, and `[DIALOGUE HISTORY]` is chat history so the model can have some conversational context to draw from. Ideally it'll be pairs of messages like:
|
58 |
+
|
59 |
+
```
|
60 |
+
[CHARACTER]: [some dialogue here]
|
61 |
+
You: [your response to the dialogue above]
|
62 |
+
```
|
63 |
+
|
64 |
+
Apart from chat history, you can also just add example conversations in `[DIALOGUE HISTORY]` to show how the character should speak - ideally at the beginning, so it doesn't get confused as to what's conversation history vs. character definition.
|
65 |
+
|
66 |
+
## Known issues
|
67 |
+
|
68 |
+
We haven't played around with the model enough to enumerate them. Feel free to give us some feedback!
|