Граф коммитов

  • 58ddb6d099 dasboard_utils.py: skip errors lines Marco Rossi 2019-06-06 14:48:18 -0700
  • 9fa2e78ccb Pass app insights key as parameter rather than environment setting (#135) Sharath Malladi 2019-05-31 15:04:46 -0700
  • e5823ba64f Pass app insights key as parameter rather than environment setting Sharath Malladi 2019-05-29 20:57:29 -0700
  • 19154e0d71 Users/sharathm/avoidemptyfeaturebucket (#134) Sharath Malladi 2019-05-28 14:32:00 -0700
  • e601b53e5c Added app insights telemetry logging Sharath Malladi 2019-05-24 23:48:37 -0700
  • ed8ac48fa2 Avoid empty feature buckets Sharath Malladi 2019-05-24 23:04:44 -0700
  • 010cef1a9a Improve log message Sharath Malladi 2019-05-24 23:04:13 -0700
  • 3171b035de Experimentation.py: Fixed confusing variable name Marco Rossi 2019-01-31 17:49:08 -0800
  • b392bdae41 Azure Batch pipeline for Offline Experimentation (Users/sharathm/azurebatchcf) (#132) Sharath Malladi 2019-05-14 12:25:52 -0700
  • 3dff64278d Rename important features to feature importance Sharath Malladi 2019-05-14 11:09:24 -0700
  • f9dee5fc42 Fixed issue if model file is empty Sharath Malladi 2019-05-13 17:15:48 -0700
  • ae5d3b82a2 Default logdownloader to report progress of downloaded data file Sharath Malladi 2019-05-13 11:39:44 -0700
  • 22c2cc5480 Last bit of feedback from Marco Sharath Malladi 2019-05-10 13:43:32 -0700
  • 34f865fce0 Reacting to feedback Sharath Malladi 2019-05-01 14:42:52 -0700
  • b596035457 Reacting to feedback Sharath Malladi 2019-04-30 18:44:30 -0700
  • 4ad4162cd4 Changed script for to iterate on several l1 values and return the buckets Sharath Malladi 2019-04-30 13:59:21 -0700
  • 6d2d1d7bf7 Copied vw-important-features.py from https://raw.githubusercontent.com/marco-rossi29/vowpal_wabbit/important-features/utl/vw-important-features Sharath Malladi 2019-04-30 13:15:22 -0700
  • b5f790bc97 Change cleanup logic to delete all files Sharath Malladi 2019-04-26 11:51:39 -0700
  • 7ae632eeeb Azure batch setup to call Experimentation Sharath Malladi 2019-04-26 11:01:33 -0700
  • 7abb7bc6b6 update for new summary.json format Dwaipayan Mukherjee 2019-04-18 17:08:12 -0700
  • 3b9abaf693 fix regex Dwaipayan Mukherjee 2019-04-09 18:04:24 -0700
  • 7b3194fa6b add , Dwaipayan Mukherjee 2019-04-09 17:52:28 -0700
  • 81cfce2292 fix summary json structure change and regex -d out Dwaipayan Mukherjee 2019-04-09 17:50:38 -0700
  • 9a1fa0bdd5 fix custom policy run bug Dwaipayan Mukherjee 2019-04-08 11:13:35 -0700
  • 0f3c30784d fix custom policy bug Dwaipayan Mukherjee 2019-04-04 21:29:39 -0700
  • d29220a2ad fix json names Dwaipayan Mukherjee 2019-04-04 19:09:35 -0700
  • 32108cb8de throw ex on critical upload, download jobs failing Dwaipayan Mukherjee 2019-04-04 16:40:35 -0700
  • fef4b38bda update status Dwaipayan Mukherjee 2019-04-04 16:13:10 -0700
  • b60bae3481 update status Dwaipayan Mukherjee 2019-04-04 15:56:33 -0700
  • ff40f5f62a add policy args to summary.json Dwaipayan Mukherjee 2019-04-04 15:14:39 -0700
  • 854fc32d24 add aggregates file as parameter Dwaipayan Mukherjee 2019-04-04 13:08:01 -0700
  • 398836f5b6 evaluate custom policies Dwaipayan Mukherjee 2019-04-03 14:52:27 -0700
  • b46a6ba852 add support for cutom policy run Dwaipayan Mukherjee 2019-04-02 19:58:03 -0700
  • 1dbdf1a944 Update log Dwaipayan Mukherjee 2019-03-28 14:56:47 -0700
  • 2fd5ba086f Add dashboard files Dwaipayan Mukherjee 2019-03-28 14:41:24 -0700
  • c97e91e086 fix issues Dwaipayan Mukherjee 2019-03-27 16:35:21 -0700
  • a8da2aa15f Check system usage Dwaipayan Mukherjee 2019-03-27 15:51:40 -0700
  • b9425fd056 remove connection_string Dwaipayan Mukherjee 2019-03-27 12:46:11 -0700
  • 42b9506e5e Change to run Counter factuals from azure batch Dwaipayan Mukherjee 2019-03-27 12:13:35 -0700
  • 09e215f1fc ds_parse.py: Removed assumption that EventId is at least 16 charachters long Marco Rossi 2019-03-13 17:42:21 -0700
  • 4cdadb5aa2 RankRewardAnalyzer.py: Added progressbar - printing verbose output before summary Marco Rossi 2018-12-31 11:46:12 -0800
  • 50cf03ba88 RankRewardAnalyzer.py: Fixed reward comparison for 0 rewards Marco Rossi 2018-12-31 11:47:09 -0800
  • bc2ee4a569 Experimentation.py: Fixed namespace detection - Marignal ns is also an ActionFeature ns - Marginal ns do not require "id" as feature key Marco Rossi 2018-12-26 17:02:36 -0800
  • 63f78afcc4 dashboard_utils.py: Added update_progress() for gzip files Marco Rossi 2018-12-26 15:48:48 -0800
  • aa5a1983e1 ActionSetVisualization.py: Fix and improvements - Added missing import gzip - Fixed SNIPS - Ask user for JSON path to action_name and string parsing - Improved plots colors, markers, and added plot_rew_zero flag - Added Rewards Counter - Output all actions to console and use ActionId for y-axis label - Avoid to display model preference when uniform probability - Count how many times an action was available - Skip corrupted lines - Added progressbar Marco Rossi 2018-12-18 13:56:07 -0800
  • 9d2648ce15 ds_parse.py: Improvements - Not requiring "Version" field to be present in json_cooked() - Only assuming checkpoint line starts with [ - Reuse extract_field() - Enabled progressbar when total is unknown Marco Rossi 2018-12-17 14:43:34 -0800
  • 1a34f0db70 RankRewardAnalyzer.py: More improvements - Ignoring missing rewards when occuring in duplicates - Using gt datastructure to avoid dup_analysis() - Added elapsed_time printout Marco Rossi 2018-12-18 14:34:31 -0800
  • 9b69254242 RankRewardAnalyzer.py: improvements - Added 3 contexts for rank call and removed --feed as required input - Cosmetic improvements Marco Rossi 2018-12-11 16:18:26 -0800
  • 363c675dde LogDownloader.py: Modifications to enable checkpoint info line flexibility Assumptions: i) checkpoint info is last line; ii) checkpoint info line starts with "[" Marco Rossi 2018-12-12 12:47:46 -0800
  • 5ef4843564 LogDownloader.py: Modifications to enable checkpoint info line flexibility Assumptions: i) checkpoint info is last line; ii) checkpoint info line starts with "[" Marco Rossi 2018-12-12 12:47:46 -0800
  • 997a9f8adb ActionSetVisualization Marco Rossi 2018-11-09 12:17:50 -0500
  • fb1f3a630a Visualization.py: Increased line tickness for last 2 days Marco Rossi 2018-10-10 10:37:04 -0400
  • d80b9a7092 Merge branch 'marossi/activate' Marco Rossi 2018-12-10 15:59:08 -0800
  • e87578b600 _deferred -> _skipLearn Marco Rossi 2018-11-30 15:25:57 -0800
  • 23b843258a ds_parse: support for activate/deactivate dsjson schema - json_cooked() added activate/deactivate fields - process_dsjson_file() added counter of skipped lines - process_files() added count of # of rews in addition to # of obs and sum of rews Marco Rossi 2018-10-26 13:27:53 -0400
  • 224e1e94d7 LogDownloader.py: Fixed bug when erasing checkpoint info line Marco Rossi 2018-12-10 15:33:02 -0800
  • 70ac7af754 pretty print of vw audit (#129) Jayant Gupchup 2018-11-06 10:13:07 -0800
  • 3f067e86bd
    Fixes and small improvements Marco Rossi 2018-11-05 15:41:09 -0500
  • 7a167d28d8 Dashboard: Added gaussian confidence intervals and improvements - dashboard.html: use ips instead of snips - dashboard.html: simplified tot creation - dashboard.html: removed interpolation type drop-down - dashboard_utils.py: added total aggregates to output file - dashboard_utils.py: added aggregates explanation - dashboard_utils.py: renamed temp variable Marco Rossi 2018-10-30 09:47:26 -0400
  • 2f5640feb2 Merge branch 'audit_pp' of https://github.com/gupchup/mwt-ds into audit_pp Jayant Gupchup 2018-11-02 17:15:23 -0700
  • 4fb83ed735 pretty print of vw audit Jayant Gupchup 2018-11-02 11:37:10 -0700
  • 16f5f110a5 pretty print of vw audit Jayant Gupchup 2018-11-02 11:37:10 -0700
  • 80de0c5a8e Dashboard: Added gaussian confidence intervals and improvements - dashboard.html: use ips instead of snips - dashboard.html: simplified tot creation - dashboard.html: removed interpolation type drop-down - dashboard_utils.py: added total aggregates to output file - dashboard_utils.py: added aggregates explanation - dashboard_utils.py: renamed temp variable Marco Rossi 2018-10-30 09:47:26 -0400
  • 5fc50843e4 Fixes to enable writing checkpoint info in Azure Storage log - LogDownloader.py: Fixed resume downloading from file - ds_parse.py: Ignore checkpoint info line Marco Rossi 2018-10-09 11:00:09 -0400
  • 935e470892 LogDownloader.py: Using os.SEEK_SET and os.SEEK_END Marco Rossi 2018-10-09 10:41:42 -0400
  • a636d2b383 Visualization.py: Changed color of Total Obs Marco Rossi 2018-10-09 11:21:46 -0400
  • 853c6f4304 Experimentation.py: set eps = 0.2 when generating dashboard files Marco Rossi 2018-09-18 10:19:08 -0400
  • 00a1d64833 Merge branch 'marossi/dashboard' Marco Rossi 2018-09-14 17:58:46 -0400
  • dabf6200c0 Added dashboard_readme.md Marco Rossi 2018-09-14 16:03:37 -0400
  • 1ed6cd2411 LogDownloader.py: Fixed small bug in exception handling Marco Rossi 2018-09-05 10:32:03 -0400
  • 9fcc2c6af7 Dashboard.html: Added (optional) graph with total number of rewards/events - Enable using "plot2=true" in url Marco Rossi 2018-05-25 12:15:39 -0400
  • 59aa2bad61 Dashboard.html: cosmetic improvements - Removed hard coded position for axis labels - Correctly positioned labels using middle - Set linear interpolation as default option Marco Rossi 2018-08-06 18:03:50 -0400
  • 1ec801ddc3 dashboard_utils.py: added merging and uniquing dashborad files Marco Rossi 2018-05-07 15:12:29 -0400
  • 2555ce15f6 Added files used for offline experimentation dashboard - dasboard_util.py: create aggregates file for online, baselines, and available prediction files - index.html: parse aggregates file and display using d3js Marco Rossi 2018-04-27 13:35:06 -0400
  • d26827a62f Experimentation.py: Added prediction files generation for best policies Marco Rossi 2018-03-21 10:24:36 -0400
  • 541a0bbce2 LogDownloader.py: Added 2 extra modes for --create_gzip for vw - 0: creates one .gz file per LastConfigurationEditDate - 1: creates a unique .gz file by merging all runs by filename date - 2: creates a unique .gz file by uniquing over EventIds and sorting by Timestamp Marco Rossi 2018-05-03 15:20:52 -0400
  • 19bd15f0b8 ds_parse.py: json_cooked() and json_dangling() return dictionary for readability Marco Rossi 2018-08-28 17:21:22 -0400
  • fe67a4fa91 Experimentation.py: Removed dr and added mtr for cb_type Marco Rossi 2018-08-31 13:26:34 -0400
  • 9471aa42ea ds_parse.py: fixed bug and improvements - fixed missmatch of assumed position of timestamp in tuple of dict e - create_time_hist() added scriptable x and y axis labels - get_e_from_eh_obs() added observation value Marco Rossi 2018-08-17 13:46:05 -0400
  • 02b0fb30c6 Visualization.py: Bug fix and cosmetic improvements Marco Rossi 2018-08-23 17:37:13 -0400
  • a396e1f475 Additional build fixes from bad cleanup Jacob Alber 2018-08-20 13:12:11 -0400
  • 579d53c78f Fix build-break in Crawl's CognitiveServiceEntityLinking function Jacob Alber 2018-08-20 13:05:43 -0400
  • d62c59d5fd Additional build fixes from bad cleanup Jacob Alber 2018-08-20 13:12:11 -0400
  • 34ccdc15a4 Fix build-break in Crawl's CognitiveServiceEntityLinking function Jacob Alber 2018-08-20 13:05:43 -0400
  • 8146305610
    Merge pull request #126 from lokitoth/master Jacob Alber 2018-08-20 11:39:17 -0400
  • 554d0c9f73 Remove redundant check and code style Jacob Alber 2018-08-20 11:37:56 -0400
  • aee6f850af Introduce constants for request size maximum Jacob Alber 2018-08-20 11:26:30 -0400
  • a4e961a482 Remove recursive submodule Jacob Alber 2018-08-14 11:56:56 -0700
  • 78e0f62870 Fix EntityLinking stage (now uses TextAnalytics) Jacob Alber 2018-07-31 16:15:35 -0400
  • 254fcd7c5c
    Merge pull request #125 from ataymano/ds_extractor-test Marco Rossi 2018-08-17 17:11:49 -0400
  • f90565d19c DecisionServiceExtractor: Added flag checking if event HasObservations - added also unit test Marco Rossi 2018-08-17 14:34:09 -0400
  • 4246c18186 Added "o" field in dsjson and added AssertEquals - also added 2 extra elements in "a" and "p" vectors Marco Rossi 2018-08-17 16:52:20 -0400
  • 1d40a60d10 Actions count parsing fix ataymano@microsoft.com 2018-07-24 17:38:31 -0400
  • 7963ce5569 Add test for actions count parsing (failed) ataymano@microsoft.com 2018-07-24 17:37:07 -0400
  • 4b6e0b60bb DecisionServiceExtractor dependencies fix ataymano@microsoft.com 2018-07-24 17:35:12 -0400
  • 375264c7f1 ds_parse.py: create_time_hist improvements - fixed case when e is list of tuples - added eventId in t_vec outputed by create_time_hist() Marco Rossi 2018-06-22 18:55:40 -0400
  • 06a583f6b3 ds_parse.py: Count observations rather than non-zero rewards this allows to catch observations with zero rewards by comparing with sum rewards Marco Rossi 2018-06-12 10:41:03 -0400
  • 3bcec881c4 Experimentation.py: Improvements - Using physical processors rather than logical for default n_proc - Allow 0 for hyperparamters values when only 1 step is used Marco Rossi 2018-05-24 08:40:11 -0400
  • c2a0380884 LogDownloader.py: Adding --confirm to always ask the user before downloading Marco Rossi 2018-05-08 18:28:15 -0400
  • 620c6f8475 added Azure Data Lake process to extract data from Decision Service style logs marcozo 2018-05-15 16:48:50 -0400