manifest/README.md

Manifest

How to make prompt programming with FMs a little easier.

Install

Download the code:

git clone git@github.com:HazyResearch/manifest.git
cd manifest

Install:

pip install -e .

Dev Install:

make dev

Getting Started

Running is simple to get started. If using OpenAI, set export OPENAI_API_KEY=<OPENAIKEY> then run

from manifest import Manifest

# Start a manifest session
manifest = Manifest(
    client_name = "openai",
)
manifest.run("Why is the grass green?")

Manifest Components

Manifest is meant to be a very light weight package to help with prompt iteration. When a user starts a Manifest session, we start to record user query history for that session. This is saved locally and is user specific. We also optionally cache all model results globally so that queries can be shared across users.

Three key design decisions of Manifest are

• Prompt are functional -- they can take an input example and dynamically change
• All models are behind API calls (e.g., OpenAI)
• Model inputs/outputs are locally cached woth the optional ability to globally cache model results

Prompts

A Manifest prompt is a function that accepts a single input to generate a string prompt to send to a model.

from manifest import Prompt
prompt = Prompt(lambda x: "Hello, my name is {x}")
print(prompt("Laurel"))
>>> "Hello, my name is Laurel"

We also let you use static strings

prompt = Prompt("Hello, my name is static")
print(prompt())
>>> "Hello, my name is static"

Sessions

Each Manifest run is a session that connects to a model endpoint and a local SQLite DB to store user query history.

# Start a manifest session
manifest = Manifest(
    client_name = "openai",
    session_id = "grass_color",
)

will start a Manifest session with the session name grass_color. This can be helpful for a user to logically keep track of sessions and resume them if desired. If no id is provided, we generate a random id for the user.

After a few queries, the user can explore their history

manifest.get_last_queries(4)

will retrieve the last 4 model queries and responses.

We further support having queries and results stored in a global cache (without any unique session information) that can be shared across users. We treat inputs and outputs as key value pairs and support SQLite or Redis backends. To start a session with additional global caching using SQLite, run

manifest = Manifest(
    client_name = "openai",
    session_id = "grass_color",
    cache_name = "sqlite",
    cache_connection = "mycache.sqlite",
)

The cache will be saved in mycache.sqlite.

We also support Redis backend.

manifest = Manifest(
    client_name = "openai",
    cache_name = "redis",
    cache_connection = "localhost:6379"
)

As a hint, if you want to get Redis running, see the docker run command below under development.

We will explain below how to use Manifest for a locally hosted HuggingFace model.

Running Queries

Once you have a session open, you can write and develop prompts.

prompt = Prompt(lambda x: "Hello, my name is {x}")
result = manifest.run(prompt, "Laurel")

You can also run over multiple examples.

results = manifest.batch_run(prompt, ["Laurel", "Avanika"])

If something doesn't go right, you can also ask to get a raw manifest Response.

result_object = manifest.batch_run(prompt, ["Laurel", "Avanika"], return_response=True)
print(result_object.get_request())
print(result_object.is_cached())
print(result_object.get_json_response())

By default, we do not truncate results based on a stop token. You can change this by either passing a new stop token to a Manifest session or to a run or batch_run. If you set the stop token to "", we will not truncate the model output.

result = manifest.run(prompt, "Laurel", stop_token="and")

If you want to change default parameters to a model, we pass those as kwargs to the client.

result = manifest.run(prompt, "Laurel", max_tokens=50)

Huggingface Models

To use a HuggingFace generative model, in manifest/api we have a Falsk application that hosts the models for you.

In a separate terminal or Tmux/Screen session, run

python3 manifest/api/app.py --model_type huggingface --model_name EleutherAI/gpt-j-6B --device 0

You will see the Flask session start and output a URL http://127.0.0.1:5000. Pass this in to Manifest. If you want to use a different port, set the FLASK_PORT environment variable.

manifest = Manifest(
    client_name = "huggingface",
    client_connection = "http://127.0.0.1:5000",
)

If you have a custom model you trained, pass the model path to --model_name.

Development

Before submitting a PR, run

export REDIS_PORT="6380"  # or whatever PORT local redis is running for those tests
cd <REDIS_PATH>
docker run -d -p 127.0.0.1:${REDIS_PORT}:6379 -v `pwd`:`pwd` -w `pwd` --name manifest_redis_test redis
make test

To use our development Redis database, email Laurel. If you have access to our GCP account, in a separate terminal, run

gcloud compute ssh "manifest-connect" --zone "europe-west4-a" --project "hai-gcp-head-models" -- -N -L 6379:10.152.93.107:6379

Then if you issue

redis-cli ping

You should see a PONG response from our database.