Skip to main content

A deep learning package to inquire intelligence

Project description

Purpose

Make it easy to create, to trace, to debug, and to recreate models/algorithms.

What does that means:

  • Incorporation of common components (res-block, seblock,...) that can easily be used and constructed (which is good for model creations and modifications).
  • Data processing in functional form, which allows clear knowledge of what happens to the data before it is fed to the algorithms/models.
  • A tool to view and debug:
    • view model statistics
    • tinkering with the data and see what happens at the result (top-k result)
  • Inference must be constructed inside the model, with the input is the most basic data point (thinking of a complete stranger who needs to use your model, that person will not know anything about the nit-picks of your models and your data, they only have a data point and want to see the result coming out of your model)

Roadmap to usability

  • Session must work
  • Ability to get batch of data
  • Summarize session and model information
  • Implement mixup
  • Test all convs architecture

Model

  • A model should have the evaluate method ready (this part should be abstracted away from the progress / training procedure)
  • A model should have the load method ready (only for inference, because continual training requires knowledge about optimizer, training iteration)

Data

Model is not the only part in creating intelligent system. Data plays a vital role in this process too. A lot of time, playing around with data, seeing how the model behaves when data is tweak a little bit can provide crucial insights for model improvement. Hence, data manipulation must be made easy to use.

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

Dawnet-0.0.1.tar.gz (9.9 kB view hashes)

Uploaded Source

Built Distribution

Dawnet-0.0.1-py3-none-any.whl (12.3 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