Skip to content

Latest commit

 

History

History
78 lines (65 loc) · 3.66 KB

CHANGELOG.md

File metadata and controls

78 lines (65 loc) · 3.66 KB

Change Log

All notable changes to this project will be documented in this file. This project adheres to Semantic Versioning.

[1.6.0] 2017-09-29

Added

Changed

  • Internal restructuring of CLI commands

[1.5.0] 2017-07-19

Added

  • New pairargs keyword for experiment specification which allows argument combinations to be pairwise instead of a cross-product.
  • New prologue keyword for experiment specification which can run arbitrary commands when a job starts
  • User-defined parsing can now be objects with various helper functions (onInit, onNewFile, parseLine).
  • New regex parser that can use command line options to specify what to parse as a regular expression
  • The output of dimebox parse can be specified to be either TSV or JSON

Changed

  • Improved SVN version control extraction
  • watch will no longer open result files that do not exist
  • The weakargs specification uses the same eval code as filter
  • The weakargs specification can now use the $depth variable

Bugfixes

  • Fixed issue where batch size was not taken into account for stagger
  • Local machine now uses bash instead of sh
  • Fixed width issue with dimebox summary

[1.4.0] 2016-09-20

Added

  • Support for raw flags in experiment files that can be passed to job headers or run flags directly.

Changed

  • Richer interface for watch that is more interactive.
  • Fixed bug with job creation on edison machine
  • Fixed bug where walltime in files were not correctly interpreted as times

[1.3.0] 2016-07-21

Added

  • Subcommand kill, which allows a user to cancel all submitted jobs for an epoch
  • User defaults for all commands can now be read in from $HOME/.dimebox/defaults.yml
  • parse now has options to select specific columns, sort rows or filter rows
  • Custom machine configurations can now be created and placed in $HOME/.dimebox/machines
  • Support for machine Edison (edison.nersc.gov)
  • Bash completition script in tools/
  • Dump of machine information when generating jobs in jobs/$epoch/machine.yml

Changed

  • Result files can contain more than one observation with the dbx.obs keyword, allowing for tidy data when using parse.
  • Cleaner output for watch subcommand which includes time of last status update
  • Turned off sample job display in summary by default to cut down on noise

[1.2.0] - 2016-06-16

Added

  • Subcommand watch, which allows for a user to monitor the status of an epoch
  • parse now allows for aggregation of numeric values with the same key into mean, stddev and other statistics with the --agg flag. Combined with the trials parameter in the expfile, this allows for multiple runs to be aggregated and parsed into one output.
  • Adding richer symbolic epoch syntax (e.g., HEAD~~ is the third to last experiment).
  • Version control information is now captured for SVN repos as well as git
  • Metadata for result files are now stored in `$epoch/.md$ as structured YAML data.

Changed

  • Failures during runs (non-zero exit codes) are now recorded by touching a file in the a $epoch/.fail directory in results. This replaces the need for searching all files for failure strings. All previous experiment sets will have incorrect counts for number of failed files.
  • optarg values that have spaces no longer cause problems with generated filenames with spaces

[1.1.0] - 2016-06-04

Added

  • The commands generate, summary, init, submit, parse, rm, and resolve.
  • Support for the 'rain' and 'vulcan' platforms.
  • Basic support for key-value extraction from result files

[1.0.0] - 2016-02-24

Added

  • Initial version.