ADD A NEW IDEA

All ideas

Showing 1983

Spark Driver slot allocation

Currently spark executors and drivers are allocated same number of slots based on a configuration at the resource group level. This request is provide a separate configuration for driver slot allocation.
about 4 years ago in Spectrum Conductor / Version 2.4.1 0 Future consideration

Functionality/API to undeploy SIGs on hosts

Spectrum conductor provides APIs and UI functionality to deploy SIGs to all or selected cluster members. There are at times corruption on the member nodes due to disk out of space or misconfiguration, or state issues. This feature should clean up ...
about 4 years ago in Spectrum Conductor / Version 2.4.1 0 Future consideration

Support RBFOpt as one of the hyperparameter optimization algorithms in WMLA

Watson Studio supports RBFOpt off the shelf (see https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/ml_dlaas_hpo.html), but WMLA doesn't. The inconsistency between both platforms is surprising, and taking into consideration that RBFO...
about 4 years ago in Openscale 0 Future consideration

Add configurable per session/per PDQ thread log cache similar to Sybase user log cache.

For high volume environments OLTP environments e.g. thousands of sessions and HDR/RSS nodes add a per session/PDQ thread log cache. This is flushed to the physical and logical log buffers either when full or at transaction commit time.
about 4 years ago in Informix / Informix Server 0 Future consideration

Additional Implementations of Hyperparameter Optimization Algorithms

The current methods for hyperparameter optimization available in WMLA are limited. While the new BYOF framework allows end users to wrap implementations of additional methods, it would be desirable to have a few more ones supported by default: - A...
about 4 years ago in Openscale 0 Future consideration

Make Bayesian Optimization parallelize more

Currently the implementation of Bayesian Optimization in WMLA is completely sequential. This renders it essentially useless for models with long training times. It would be ideal to parallelize it more. There are two angles to this: first, at its ...
about 4 years ago in Openscale 0 Future consideration

Include all metrics (not just the one being optimized) in the output of GET hypersearch

currently when one queries the results of a hyperparameter optimization round in Watson Machine Learning Accelerator, e.g. with https://<>:9243/platform/rest/deeplearning/v1/hypersearch/username-hpo-860740667802465 ... only the metric being ...
about 4 years ago in Openscale 0 Future consideration

dynamically move nodes from a resource group to another by the means of updating node attributes

Background: today, it's hard to move nodes from one RG to another one using dynamically selection of hosts based on arttributes. We can tag a node with an attribute, but changing its value requires one to change ego.conf, remove the host from the ...
about 4 years ago in Spectrum Conductor / Version 2.4.1 0 Future consideration

MultiCluster RestAPI

Could you please provide samples and docs for SMC REST API ? Samples should ideally cover logon, service add, service sync, profile registration, profile fragment update, retrieve profile.As Ajith mentioned, please deliver samples before docs if t...
about 4 years ago in Spectrum Symphony / Version 7.3 0 Future consideration

SMC Changing session priority

Add new functionality in SMC to support changing session priority (RFE).this can be done using the Symphony WSDL API, but you would need to create a WSDL client and connect to each Symphony cluster in turn... but it should be done by SMC who are m...
about 4 years ago in Spectrum Symphony / Version 7.3 0 Future consideration

By clicking the "Post Comment" or "Submit Idea" button, you are agreeing to the IBM Ideas Portal Terms of Use.
Do not place IBM confidential, company confidential, or personal information into any field.