|Date:||August 22, 2017|
New in this release¶
Nervana Cloud 1.6.0 release hardens our underlying code as well as introduces new enhancements, including initial support for the Tensorflow deep learning framework.
Tensorflow support for interactive sessions - With Nervana Cloud 1.6.0, the user has the ability to select Tensorflow in addition to neon when launching an interactive Jupyter session. From the ncloud CLI this can be specified via the “–environment tf_1.2.1” option to the “ncloud interact start” command. Note that at this time, we only support Tensorflow release 1.2.1 and CPU backend, however we will expand to GPU backends and other job types in the upcoming release.
Full URL returned in model deployments - Prior to this release, when deploying a trained model for streaming inference, only a cryptic pre-signed token value was displayed, making it difficult to infer where to route prediction requests. With Nervana Cloud 1.6.0, a separate url field is included when passing the “–details” flag to stream show, for example:
ncloud stream show stream_id --details shows a full url similar to:
Added data locations to info command - With v1.5.0 release we introduced volumes support for arbitrary file volumes to handle non aeon formatted data such as older datasets, vocabulary files, and output data. In release 1.6.0, when volumes are specified for different job types, the user can now see where they end up getting mounted inside the running container. They can access this information by passing the new “–mounts” flag to the ncloud info command. See Attaching Data for details.
This release of Nervana Cloud includes a number of additional features and fixes:
- Improved support for uploading large file volumes and fixed upload file count report
- Added support for SegNet demo
- Web UI branding updates
- To improve consistency, added “–details” flag for stream list and show commands as well as improvements in various ncloud options and response formatting.
- Added more verbose request printing for easier debugging in command line interface and webUI
- Included additional error info in failed job email notifications
- User and tenant admin improvements in the UI and CLI
Known Limitation: The datasets command line interface has been disabled at this time. This functionality has been replaced by volumes. Volumes supports all functionality provided by datasets, plus more. Users can still run training jobs against existing datasets.
For complete history see the ChangeLog
Thanks for using Nervana Cloud. If you believe you received this notification in error, please contact us at firstname.lastname@example.org.