feat: Implement outlines converter model for structured output (#1211) #1318
Conversation
There was a problem hiding this comment.
Thanks @MuggleJinx! I have added two comments for reference, please have a look when possible.
Co-authored-by: Harry Ye <116691547+harryeqs@users.noreply.github.com>
|
Thanks @harryeqs, updated! |
Thanks @MuggleJinx ! |
| # 1. Using a Pydantic model | ||
| class Temperature(BaseModel): | ||
| location: str | ||
| date: str | ||
| temperature: float | ||
|
|
||
|
|
||
| output = model.convert_json( | ||
| "Today is 2023-09-01, the temperature in Beijing is 30 degrees.", | ||
| output_schema=Temperature, | ||
| ) | ||
|
|
||
| print(output) | ||
| ''' | ||
| location='Beijing' date='2023-09-01' temperature=30.0 | ||
| ''' |
There was a problem hiding this comment.
the output format use '=' @Wendong-Fan , not {'location':'Beijing'}
There was a problem hiding this comment.
Thanks, I split the it into 2 functions, one returns pydantic object, one returns the internal dict object.
There was a problem hiding this comment.
Thanks @MuggleJinx ! Overall LGTM, left some comments below
camel/schemas/outlines_converter.py
Outdated
|
|
||
| from typing import Any, Callable, List, Literal, Type, Union | ||
|
|
||
| import outlines |
There was a problem hiding this comment.
outlines is not necessary packages
There was a problem hiding this comment.
sry, it is needed in outlines.generate.regex for example.
There was a problem hiding this comment.
we can move import outlines within the class to make it optional
There was a problem hiding this comment.
thanks @MuggleJinx ! left some comments below and added one commit here:87ccb62 feel free to review and check, let me know if there's any issue~
| @@ -67,6 +67,7 @@ sentencepiece = { version = "^0", optional = true } | |||
| opencv-python = { version = "^4", optional = true } | |||
|
|
|||
| # tools | |||
| outlines = { version = "^0.1.7", optional = true } | |||
There was a problem hiding this comment.
also add it to tool and all list
| ) | ||
|
|
||
| print(output) | ||
| ''' |
There was a problem hiding this comment.
unify by using " instead of '
camel/schemas/outlines_converter.py
Outdated
|
|
||
| from typing import Any, Callable, List, Literal, Type, Union | ||
|
|
||
| import outlines |
There was a problem hiding this comment.
we can move import outlines within the class to make it optional
camel/schemas/outlines_converter.py
Outdated
| case _: | ||
| raise ValueError(f"Unsupported platform: {platform}") | ||
|
|
||
| def convert_regex(self, content: str, regex_pattern: str): |
There was a problem hiding this comment.
missing return type hint, even it could be Any
camel/schemas/outlines_converter.py
Outdated
| ) | ||
| return json_generator(content) | ||
|
|
||
| def convert_type(self, content: str, type_name: type): |
There was a problem hiding this comment.
missing return type hint, same for other methods
There was a problem hiding this comment.
file naming, use _ instead of -
Description
Integrate outlines library to produce the structured output.
Motivation and Context
Close #1211.
Types of changes
What types of changes does your code introduce? Put an
xin all the boxes that apply:Implemented Tasks
Checklist
Go over all the following points, and put an
xin all the boxes that apply.If you are unsure about any of these, don't hesitate to ask. We are here to help!