Neuro Platform: ML Developers Forum

[01.09.2019] Paperspace Gradient

This overview describes the state of the platform on September 1, 2019.

Overview procedure description

Product website

Gradient CI is a killer feature!

Developer experience - 2

  1. Project template - YES
  2. Tutorials - YES
  3. Remote debugging - NO
  4. Hyper-parameter search - YES
  5. CLI, API, Browser - YES

ML environments extensibility - 3

  1. Pre-defined environments - YES
  2. Bring your frameworks - YES (docker)
  3. Integrate open source and commercial ML tools - YES

Data Ingestion - 2

  1. Uploading data to the platform - jupyter for datasets, VM with ssh, Machine Learning in a Box / cli for artifacts
  2. Making data available for training - auto

AI starter kits - 3

  1. Vision - YES (fast ai)
  2. Language - YES (fast ai)
  3. Recommendation - YES (fast ai)
  4. Speech - YES (fast ai)

Collaboration - 3

  1. …with engineers - Share checkpoint
  2. …with non-engineers - Jupyter notebook
  3. Dataset and environments sharing - public dataset, environments are docker files

Bring your own cloud - 3

  1. Bare metal or Kubernetes - keep
  2. AWS - YES
  3. GCP - YES
  4. Azure - YES

Enterprise-ready - 1

  1. Audit logs - cannot find
  2. Single sign-on - cannot find
  3. Reporting - cannot find
  4. Role-based security - cannot find

Final score - 17