Skip to main content

A small ML serialization software

Project description

cov-estimator

Simple estimator module for serialization

This is a simple library to build adapters to construct complex estimators using sklearn models as well as keras.

Basic architecture:


    from cov_estimator.nodes import Estimator
    from cov_estimator import Pipeline

    img = "path/to/img"

    img = load_img(img, (180, 180))

    img = np.expand_dims(img, axis=0)

    est = Estimator(
        (
            lambda data: {
                "pneumonia": data["pneumonia"],
                "covid": data["pneumonia"] * data["covid"],
                "normal": 1 - data["pneumonia"],
            }
        ),
        "mult_1",
    )

    covid = tf.keras.models.load_model("path/to/model_1")
    covid_est = Estimator(covid, "covid")

    pneumonia = tf.keras.models.load_model("path/to/model_2")
    pneumonia_est = Estimator(pneumonia, "pneumonia")

    est = est(pneumonia_est, covid_est)
    data = {"covid": img, "pneumonia": img}
    model = Pipeline(data, 'Path/to/save')
    result = model(result)

    print("result: {}".format(result)

Whats happening?

Basically what the library does is create an abstract tree of dependecies which evaluate each node. For it to properly work, you need to pass a dictionary containing as key each nodes which expects some input. You can use, lambda funcions to apply some important transformation. it will always get a single dictionary as input, the dictionary will contain the outputs from each correspondant node. Obs! Each node inside in the graph must have a unique name, it's the developer's responsibility to

Stack:

  • Python 3.8
  • Virtualenv
  • tensorflow > 2.0
  • numpy > 1.6
  • sklearn

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

cov_estimator-0.0.1.tar.gz (6.7 kB view hashes)

Uploaded Source

Built Distribution

cov_estimator-0.0.1-py3-none-any.whl (9.1 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page