DEEPaaS Release Notes

Current Release Notes

v2.4-45

Upgrade Notes

Removed deprecated loading of “deepaas-test” model. If you wish to develop or debug your model, use the “demo-app” instead: https://github.com/ai4os-hub/ai4os-demo-app

Removed the loading of various models, now only a single model can be served. If you have more than one model available in the system please specify a single one via the command line option --model-name.

v2.4

Upgrade Notes

The command deepaas-predict has been removed, please use deepaas-cli instead.

v2.3.0

Prelude

Model loading is being refactored, so that only a single model will be available for serving.

Deprecation Notes

Loading the default deepaas-test model if no models are available is now marked as deprecated and will be removed in the next major version.

Loading several models is now marked as deprecated. If you have several models installed, please use the model-name configuration option or the –model-name CLI option.

2.2.0

Upgrade Notes

OpenWhisk integration is now fully removed.

2.1.0

Deprecation Notes

OpenWhisk integration is marked as deprecated.

2.0.0

Prelude

Fallback loading of arguments (from V1 style) is now removed.

This new release removes support for the V1 API (marked as deprecated in version 1.0.0).

Upgrade Notes

The old way of returning the arguments, to load the arguments using the old get_*_args functions is now removed. If you were still relying on them (check the logs) you must migrate to using the argument parser as soon as possible.

V1 API is now removed, therefore all modules using the deepaas.model entrypoint will stop working. The --enable-v1 configuration option is now removed.

Other Notes

V1 support is now removed.

1.3.0

New Features

Include deepaas-cli a new command line tool that provides the same functionality as the API through the commandline. This is useful for execution of batch tasks in a HPC/HTC or batch system cluster.

1.2.1

Bug Fixes

Fix [#83](https://github.com/indigo-dc/DEEPaaS/issues/87) out out memory errors due to the usage of two different executor pools.

1.2.0

New Features

Include a new command line tool to execute inference and prediction calls from the shell, whitout spawning a server and making cURL requests to it. This is useful for execution of batch tasks in a HPC/HTC or batch system cluster.

Bug Fixes

This new release fixes an error introduced by webargs in its 6.0.0 version. Since we cannot yet fix it (aiohttp-webargs needs to be updated as well) we pin webargs to the 5.X.X versions https://github.com/indigo-dc/DEEPaaS/issues/82

Set debug in OpenWhisk mode as configured by the user.