Version 0.10

Version 0.10.7

Trains

Features and Bug Fixes

  • Artifacts support.
  • Removed apache-libcloud from requirements.
  • trains-init now verifies credentials against the trains-server installation.

Version 0.10.6

Trains

Features and Bug Fixes

  • Fix broken (v0.10.5) Keras Binding support.

Version 0.10.5

Trains

Features and Bug Fixes

  • Add GPU monitoring support (add gpustat package to extras_require).
  • Install with GPU monitoring support:

    pip install trains[gpu]
    
  • Move all cloud storage package requirements to extras_require. Install with specific cloud provider support:

    • Microsoft Azure support:

      pip install trains[azure]
      
    • Google Storage support:

      pip install trains[gs]
      
    • Amazon S3 support:

      pip install trains[s3]
      
    • Combine Cloud support with GPU monitoring. For example, install S3 and GPU using the following command:

      pip install trains[s3,gpu]
      
  • Improve stability with intermittent network connection.

  • Support upgrading trains-server while running training jobs without losing log data.

Version 0.10.4

Trains

Features and Bug Fixes

  • Replace opencv-python with the more standard Pillow package.
  • Improve matplotlib support (custom axis ticks).
  • Improve Python package detection.

Version 0.10.3

Trains

Full feature overview here

Features and Bug Fixes

Version 0.10.2

Trains

Features and Bug Fixes

  • Add Matplotlib SVG support.
  • Add Seaborn support.
  • Add TRAINS_LOG_ENVIRONMENT environment logging (GitHub trains Issue 17).
  • Add Microsoft Azure notebook support.
  • Add Google Colab support.
  • Fix TensorBoard RGB channel order.

Version 0.10.1

Trains

Features and Bug Fixes

  • Fix Jenkins CI/CD support.

Trains Server

  • Add simple username/password login scheme, see (GitHub trains-server Issue 3

  • Replace the Python web-server with NGINX.

  • Improve sub-domain support (GitHub trains-server Issue 9).

  • Extended configuration options

    • Environment variables for overriding mongoDB and Elastic hosts URLs.
    • File server disabling browser cache.
    • File server compression support.
    • File server CORS support.

Version 0.10.0

Trains

  • Experiment code execution detection

  • Automatically create package requirements section (including used versions).

  • Automatically detect and store source code uncommitted changes.
  • Jupyter Notebook support, automatically convert notebook to Python script (stored under uncommitted changes).
  • Jupyter Notebook support, automatically update used packages in Jupyter Notebook (including used versions).

  • Add resource monitoring to experiment metrics

  • Sampled every 500ms, averaged over 30 seconds.

  • CPU, network, I/O, memory, and other resources.
  • For GPU support please install gpustat. (currently not part of the requirements due to gpustat compatibility issues with Windows). pip install gpustat

  • Automatically stop inactive experiments (default: 2 hours)

  • Improve visibility

  • Finer status definitions: Identify successful completion vs. user aborted.

  • Experiment plot comparison: Ensure different colors for different experiments.
  • Parse newline character in experiment description.
  • Show experiment start time in table display.
  • Add vertical guide in scalar plots.
  • Move hyperparameters to the designated tab.
  • "Admin" section now named "Profile".

Trains Server

  • Experiment code execution detection:

  • Add source-control diff section.

  • Add code requirements section.

  • Automatically stop inactive experiments (default: 2 hours).

  • Configuration details in readme.md.
    • Improve visibility:
    • Finer status definitions: Identify successful completion vs. user aborted.
    • Experiment plot comparison: Ensure different colors for different experiments.
    • Parse newline character in experiment description.
    • Show experiment start time in table display.
  • Add vertical guide in scalar plots.
  • Move hyperparameters to the designated tab.
  • "Admin" section now named "Profile".