Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-11-22 08:07:56 +01:00
Code Issues Packages Projects Releases Wiki Activity

Update Chat-mode.md

Browse Source
This commit is contained in:
oobabooga 2023-08-01 07:57:47 -07:00
parent 959feba602
commit 9773534181
1 changed files with 8 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
docs/Chat-mode.md
View File

@ -1,36 +1,30 @@
## Chat characters
Custom chat mode characters are defined by `.yaml` files inside the `characters` folder. An example is included: [Example.yaml](https://github.com/oobabooga/text-generation-webui/blob/main/characters/Example.yaml)
Custom chat mode characters are defined by `.yaml` files inside the `characters` folder. An example is included: [Example.yaml](https://github.com/oobabooga/text-generation-webui/blob/main/characters/Example.yaml).
The following fields may be defined:
| Field | Description |
|-------|-------------|
| `name` or `bot` | The character's name. |
| `context` | A string that appears at the top of the prompt. It usually contains a description of the character's personality and a few example messages. |
| `greeting` (optional) | The character's opening message. It appears when the character is first loaded or when the history is cleared. |
| `your_name` or `user` (optional) | Your name. This overwrites what you had previously written in the `Your name` field in the interface. |
| `context` | A string that appears at the top of the prompt. It usually contains a description of the character's personality. |
| `greeting` (optional) | The character's opening message when a new conversation is started. |
| `example_dialogue` (optional) | A few example messages to guide the model. |
| `turn_template` (optional) | Used to define where the spaces and new line characters should be in Instruct mode. See the characters in `characters/instruction-following` for examples. |
#### Special tokens
* `{{char}}` or `<BOT>`: are replaced with the character's name
* `{{user}}` or `<USER>`: are replaced with your name
The following replacements happen when the prompt is generated, and they apply to the `context` and `greeting` fields:
These replacements happen when the character is loaded, and they apply to the `context`, `greeting`, and `example_dialogue` fields.
* `{{char}}` and `<BOT>` get replaced with the character's name.
* `{{user}}` and `<USER>` get replaced with your name.
#### How do I add a profile picture for my character?
Put an image with the same name as your character's yaml file into the `characters` folder. For example, if your bot is `Character.yaml`, add `Character.jpg` or `Character.png` to the folder.
Put an image with the same name as your character's `.yaml` file into the `characters` folder. For example, if your bot is `Character.yaml`, add `Character.jpg` or `Character.png` to the folder.
#### Is the chat history truncated in the prompt?
Once your prompt reaches the 2048 token limit, old messages will be removed one at a time. The context string will always stay at the top of the prompt and will never get truncated.
#### Pygmalion format characters
These are also supported out of the box. Simply put the JSON file in the `characters` folder, or upload it directly from the web UI by clicking on the "Upload character" tab at the bottom.
Once your prompt reaches the `truncation_length` parameter (2048 by default), old messages will be removed one at a time. The context string will always stay at the top of the prompt and will never get truncated.
## Chat styles