Skip to content

Tool to take your ML model from local to production with one-line of code.

License

Notifications You must be signed in to change notification settings

siftly-ai/airdot-deployer

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

πŸš€ Airdot Deployer

Python Code style: black

Deploy your ML models in minutes, not weeks.

Detailed documentation can be found here

Airdot Deployer will automatically:

  1. Restructure your Python code (from Jupyter Notebook/local IDEs) into modules.
  2. Builds a REST API around your code.
  3. Conterize the app.
  4. Spins up the required hardware (local or K8s or cloud).
  5. Monitors for model/data drift and performance (in development)

Take your ML model from Local to Production with one-line of code

from airdot.deployer import Deployer
deployer_obj = Deployer().run(<your-ml-predictor>)

Once deployed, your model will be up and running on the intra/internet, accessible to your users. No more worrying about complex server setups or manual configuration. Airdot Deployer does all the heavy lifting for you.

curl -XPOST <url> -H 'Content-Type: application/json' -d '{"args": "some-value"}'

Whether you're a data scientist, developer, or tech enthusiast, Airdot Deployer empowers you to showcase your machine learning prowess and share your creations effortlessly.

What does Airdot Deployer supports ?

  • Local Deployment with Docker docker
  • K8s Deployment with seldon core core

Want to try Airdot ? follow setup instructions

πŸ“‹ Setup Instructions

Before we get started, you'll need to have Docker, Docker Compose, and s2i installed on your machine. If you don't have these installed yet, no worries! Follow the steps below to get them set up:

Docker Install

Please visit the appropriate links to install Docker on your machine:

  • For macOS, visit here
  • For Windows, visit here
  • For Linux, visit here

S2I install

For Mac You can either follow the installation instructions for Linux (and use the darwin-amd64 link) or you can just install source-to-image with Homebrew:

$ brew install source-to-image

For Linux just run following command

curl -s https://api.github.com/repos/openshift/source-to-image/releases/latest| grep browser_download_url | grep linux-amd64 | cut -d '"' -f 4  | wget -qi -

For Windows please follow instruction here

πŸ’» Airdot Deployer Installation

Install the Airdot Deployer package using pip:

pip install "git+https://github.com/airdot-io/airdot-deployer.git@main#egg=airdot"

or

pip install airdot

🎯 Let's try out

Local Deployments

Run following in terminal to setup minio and redis on your machine

docker network create minio-network && wget  https://raw.githubusercontent.com/airdot-io/airdot-deployer/main/docker-compose.yaml && docker-compose -p airdot up

Train your model

from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from airdot.deployer import Deployer
from sklearn import datasets
import pandas as pd
import numpy as np

iris = datasets.load_iris()
iris = pd.DataFrame(
    data= np.c_[iris['data'], iris['target']],
    columns= iris['feature_names'] + ['target']
)
X = iris.drop(['target'], axis=1)
X = X.to_numpy()[:, (2,3)]
y = iris['target']
X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.5, random_state=42)
log_reg = LogisticRegression()
log_reg.fit(X_train,y_train)

Test your model

def predict(value):
    return log_reg.predict(value)

Deploy in one step 🀯

deployer_obj = Deployer().run(predict)

Use your deployed Model

curl -XPOST http://127.0.0.1:8000 -H 'Content-Type: application/json' -d '{"value": [[4.7, 1.2]]}'

Want to stop your deployment

deployer.stop('predict') # to stop container

Deployment on k8s using seldon-core deployment

Note - This method will use your current cluster and uses seldon-core to deploy

from airdot import Deployer
import pandas as pd

# this is using default seldon-deployment configuration.
config = {
        'deployment_type':'seldon',
        'bucket_type':'minio',
        'image_uri':'<registry>/get_value_data:latest'
        }
deployer = Deployer(deployment_configuration=config) 


df2 = pd.DataFrame(data=[[10,20],[10,40]], columns=['1', '2'])
def get_value_data(cl_idx='1'):
    return df2[cl_idx].values.tolist()

deployer.run(get_value_data) 

you can also deploy using seldon custom configuration

from airdot import Deployer
import pandas as pd

# this is using default seldon-deployment configuration.
config = {
        'deployment_type':'seldon',
        'bucket_type':'minio',
        'image_uri':'<registry>/get_value_data:latest',
        'seldon_configuration': '' # your custom seldon configuration
        }
deployer = Deployer(deployment_configuration=config) 


df2 = pd.DataFrame(data=[[10,20],[10,40]], columns=['1', '2'])
def get_value_data(cl_idx='1'):
    return df2[cl_idx].values.tolist()

deployer.run(get_value_data)