Fast and reliable neural text generation.
Install • Guided generation • Prompting primitives • Examples • Stay tuned
Outlines 〰 is a library for neural text generation. You can think of it as a
more flexible replacement for the generate
method in the
transformers library.
Outlines 〰 helps developers guide text generation to build robust interfaces with external systems. Provides generation methods that guarantee that the output will match a regular expressions, or follow a JSON schema.
Outlines 〰 provides robust prompting primitives that separate the prompting from the execution logic and lead to simple implementations of few-shot generations, ReAct, meta-prompting, agents, etc.
Outlines 〰 is designed as a library that is meant to be compatible the broader ecosystem, not to replace it. We use as few abstractions as possible, and generation can be interleaved with control flow, conditionals, custom Python functions and calls to other libraries.
Outlines 〰 is compatible with all models. It only interfaces with models via the next-token logits. It can be used with API-based models as well.
- 🖍️Simple and powerful prompting primitives based on the Jinja templating engine
- 🚄 Guided generation, including multiple choice, type constraints and dynamic stopping
- ⚡ Fast regex-guided generation
- 🔥 Fast JSON generation following a JSON schema or a Pydantic model
- 🐍 Interleave completions with loops, conditionals, and custom Python functions
- 💾 Caching of generations
- 🤗 Integration with HuggingFace's
transformers
models
Outlines 〰 has new releases and features coming every week! Make sure to ⭐ star and 👀 watch this repository to stay up to date.
- Context-Free Grammar guided generation (#178);
- Prompt-token alignment so you don't have to think about tokenization details (#201)
- An infilling DSL (#182)
You can follow @NormalComputing, @remilouf or @BrandonTWillard for regular updates!
Outlines is available on PyPi:
pip install outlines
The dependencies needed to use models are not installed by default. You will need to run:
pip install openai
to be able to use OpenAI models.pip install transformers
to be able to use HuggingFacetransformers
models.
The first step towards reliability of systems that include large language models is to ensure that there is a well-defined interface between their output and user-defined code. Outlines provides ways to control the generation of language models to make their output more predictable.
You can stop the generation after a given sequence has been found:
import outlines.text.generate as generate
import outlines.models as models
model = models.transformers("gpt2")
answer = generate.continuation(model, stop=["."])("Tell me a one-sentence joke.")
You can reduce the completion to a choice between multiple possibilities:
import outlines.text.generate as generate
import outlines.models as models
model = models.transformers("gpt2")
prompt = """You are a sentiment-labelling assistant.
Is the following review positive or negative?
Review: This restaurant is just awesome!
"""
answer = generate.choice(model, ["Positive", "Negative"])(prompt)
You can instruct the model to only return integers or floats:
import outlines.text.generate as generate
import outlines.models as models
model = models.transformers("gpt2")
prompt = "1+1="
answer = generate.integer(model)(prompt)
prompt = "sqrt(2)="
answer = generate.float(model)(prompt)
Outlines also comes with fast regex-guided generation. In fact, the choice
,
integer
and float
functions above all use regex-guided generation under the
hood:
import outlines.models as models
import outlines.text.generate as generate
model = models.transformers("gpt2-medium")
prompt = "Is 1+1=2? "
unguided = generate.continuation(model, max_tokens=30)(prompt)
guided = generate.regex(model, r"\s*([Yy]es|[Nn]o|[Nn]ever|[Aa]lways)", max_tokens=30)(
prompt
)
print(unguided)
# Is 1+1=2?
#
# This is probably the most perplexing question.
# As I said in one of my articles describing how
# I call 2 and 1, there isn't
print(guided)
# Is 1+1=2? Always
import outlines.models as models
import outlines.text.generate as generate
model = models.transformers("gpt2-medium")
prompt = "What is the IP address of the Google DNS servers? "
unguided = generate.continuation(model, max_tokens=30)(prompt)
guided = generate.regex(
model,
r"((25[0-5]|2[0-4]\d|[01]?\d\d?)\.){3}(25[0-5]|2[0-4]\d|[01]?\d\d?)",
max_tokens=30,
)(prompt)
print(unguided)
# What is the IP address of the Google DNS servers?
#
# Passive DNS servers are at DNS servers that are private.
# In other words, both IP servers are private. The database
# does not contain Chelsea Manning
print(guided)
# What is the IP address of the Google DNS servers?
# 2.2.6.1
Unlike other libraries, regex-guided generation in Outlines is almost as fast as non-guided generation.
Outlines 〰 allows to guide the generation process so the output is guaranteed to follow a JSON schema or Pydantic model:
from typing import List
from enum import Enum
from pydantic import BaseModel, constr
import outlines.models as models
import outlines.text.generate as generate
class Weapon(str, Enum):
sword = "sword"
axe = "axe"
mace = "mace"
spear = "spear"
bow = "bow"
crossbow = "crossbow"
class Armor(str, Enum):
leather = "leather"
chainmail = "chainmail"
plate = "plate"
class Character(BaseModel):
name: constr(max_length=10)
age: int
armor: Armor
weapon: Weapon
strength: int
model = models.transformers("gpt2")
sequence = generate.json(model, Character)("Give me a character description")
print(sequence)
# {
# "name": "ranbelt",
# "age": 26,
# "armor": "chainmail",
# "weapon": "bow",
# "strength": 5
# }
parsed = Character.model_validate_json(sequence)
print(parsed)
# name='ranbelt' age=26 armor=<Armor.chainmail: 'chainmail'> weapon=<Weapon.bow: 'bow'> strength=5
The method works with union types, optional types, arrays, nested schemas, etc. Some field constraints are not supported yet, but everything else should work.
Writing prompts by concatenating strings in pure Python quickly becomes cumbersome: the prompt building logic gets entangled with the rest of the program, and the structure of the rendered prompt is obfuscated.Outlines makes it easier to write and manage prompts by encapsulating templates inside "template functions".
These functions make it possible to neatly separate the prompt logic from the general program logic; they can be imported from other modules and libraries.
Template functions require no superfluous abstraction, they use the Jinja2 templating engine to help build complex prompts in a concise manner:
import outlines.text as text
import outlines.models as models
examples = [
("The food was digusting", "Negative"),
("We had a fantastic night", "Positive"),
("Recommended", "Positive"),
("The waiter was rude", "Negative")
]
@text.prompt
def labelling(to_label, examples):
"""You are a sentiment-labelling assistant.
{% for example in examples %}
{{ example[0] }} // {{ example[1] }}
{% endfor %}
{{ to_label }} //
"""
model = models.transformers("gpt2")
prompt = labelling("Just awesome", examples)
answer = text.generate.continuation(model, max_tokens=100)(prompt)
We can teach language models to call external functions to get additional informations or perform tasks, by encoding the functions' description in the prompt. To avoid duplicating information between the function definition and the description passed to the prompt, we define custom Jinja filters that can extract the function's name, description, signature and source:
from typing import Callable, List
import outlines.text as text
def google_search(query: str):
"""Google Search"""
pass
def wikipedia_search(query: str):
"""Wikipedia Search"""
pass
@text.prompt
def agent(tools: List[Callable]):
"""AVAILABLE COMMANDS:
{% for tool in tools %}
TOOL
{{ tool | name }}, {{ tool | description }}, args: {{ tool | signature }}
{{ tool | source }}
{% endfor %}
"""
prompt = my_commands([google_search, wikipedia_search])
We can instruct models to return their output in a pre-defined format, often JSON. To avoid duplicating information between the function definition and the description passed to the prompt we define a custom Jinja filter that can extract the expected response's schema:
from pydantic import BaseModel
import outlines.text as text
class Joke(BaseModel):
joke: str
explanation: str
@text.prompt
def joke_ppt(response_model):
"""Tell a joke and explain why the joke is funny.
RESPONSE FORMAT:
{{ response_model | schema }}
"""
joke_ppt(Joke)
# Tell a joke and explain why the joke is funny.
#
# RESPONSE FORMAT:
# {
# "joke": "The joke"
# "explanation": "The explanation of why the joke is funny"
# }
With these prompting primitives Outlines makes building agents like AutoGPT, BabyAGI, ViperGPT or Transformers Agent easier by removing boilerplate prompting code.
We currently only accept bug fixes and documentation contributions. If you have a feature request, please start a new discussion. The issue tracker is only intended for actionable items.
Run pip install -e .[test]
or conda env create -f environment.yml
. To build the documentation you will also need to run pip install -r requirements-doc.txt
.
Before pushing your code to repository please run pre-commit run --all-files
and pytest
to make sure that the code is formatted correctly and that the tests pass.
Do not hesitate to open a draft PR before your contribution is ready, especially if you have questions and/or need feedback.
- Pick the odd one out
- Meta prompting
- ReAct
- Generate code to solve math problems
- BabyAGI
- Uncertainty
- Simulation-based inference
@article{willard2023efficient,
title={Efficient Guided Generation for LLMs},
author={Willard, Brandon T and Louf, R{\'e}mi},
journal={arXiv preprint arXiv:2307.09702},
year={2023}
}
Outlines is open-source and licensed under the Apache License 2.0.