Uh oh!
There was an error while loading. Please reload this page.
Add ITrialResultManager for continue training in AutoML - #6335
Add ITrialResultManager for continue training in AutoML#6335LittleLittleCloud merged 15 commits into
Conversation
| { | ||
| bool IsMaximize { get; } | ||
| string MetricName { get; } |
| "loss", | ||
| "durationInMilliseconds", | ||
| "peakCpu", | ||
| "peakMemoryInMegaByte" |
| { | ||
| if (!File.Exists(filePath)) | ||
| { | ||
| return new TrialResult[0]; |
| _step = _stepUpperBound; | ||
| } | ||
| if (rng != null) |
| { | ||
| // initialize eci with the estimated cost and always start from pipeline which has lowest cost. | ||
| _eci = _pipelineSchemas.ToDictionary(kv => kv, kv => GetEstimatedCostForPipeline(kv, _sweepablePipeline)); | ||
| _initialized = true; |
| public void Update(TrialResult result, string schema) | ||
| { | ||
| _initialized = true; |
| } | ||
| // make sure hash code is greater than 0 | ||
| return (int)(hash >> 1); |
There was a problem hiding this comment.
Is it picked up from Java? :-)
You can simplify that like
https://source.dot.net/#System.Private.CoreLib/src/libraries/System.Private.CoreLib/src/System/Tuple.cs,69
without using checked blocks at all.
There was a problem hiding this comment.
That's cool, thanks!
tarekgh
left a comment
There was a problem hiding this comment.
Added minor comments and questions. LGTM otherwise.
Codecov Report
Additional details and impacted files@@ Coverage Diff @@## main #6335 +/- ##
==========================================
- Coverage 68.58% 68.51% -0.08%
==========================================
Files 1171 1165 -6 Lines 247964 246142 -1822 Branches 25738 25555 -183 ==========================================
- Hits 170076 168635 -1441 + Misses 71129 70834 -295 + Partials 6759 6673 -86
Flags with carried forward coverage won't be shown. Click here to find out more.
|
LittleLittleCloud
commented
Sep 23, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Sep 24, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Sep 28, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Sep 29, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Oct 3, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Oct 4, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Oct 5, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
disable SweepablePipeline_AutoFit_UCI_Adult_CrossValidation_10_Test
LittleLittleCloud
commented
Oct 5, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Oct 6, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Oct 7, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Oct 7, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Oct 10, 2022
/azp run |
LittleLittleCloud
commented
Oct 10, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
LittleLittleCloud
commented
Oct 10, 2022
/azp --help |
|
Command '--help' is not supported by Azure Pipelines. Supported commands
See additional documentation. |
LittleLittleCloud
commented
Oct 10, 2022
/azp run |
|
Azure Pipelines successfully started running 2 pipeline(s). |
We are excited to review your PR.
So we can do the best job, please check:
Fixes #nnnnin your description to cause GitHub to automatically close the issue(s) when your PR is merged.Resolved issue
#5736
What's
ITrialResultManagerITrialResultManageris a component used byAutoMLExperimentto save and loadTrialResult. The default implementation forITrialResultManagerisCsvTrialResultManager, which saves all trial results in csv formatWhat's
ITrialResultManagerused forCurrently, it's used by some tuners (CostFrugalTuner, for example) to recover from previous training, suppose trialResult csv is provided or is available.
How to use it
ITrialResultManageris not directly exposed to public users. User can useSetCheckpoint(string folder)path instead. If a folder is provided viaSetCheckpoint,AutoMLExperimentwill create that folder if not exists, and configuresCsvTrialResultManagerto save trial results tofolder/trialResults.csv.