What's new in WML CE 1.6.1
Check here for the latest updates to IBM Watson Machine Learning Community Edition information for the 1.6.1 release.
What's new
TensorFlow Serving and TensorFlow Serving API 1.14 are now included.
Keras-team Keras 2.2.4 is now included.
NVIDIA DALI 0.9.0 is now included.
CPU-only packages for TensorFlow and IBM-enhanced Caffe.
Support for WML CE on x86.
Technology previews
- NVIDIA TensorRT
- PyTorch: torchtext and pytext natural language support
- PyTorch: NVIDIA APEX Automatic Mixed Precision and optimizer support
- TensorFlow 2.0
- Snap ML: snap-ml-spark library
Behavioral changes
TensorFlow changed the environment variable name TF_CUDA_HOST_MEM_LIMIT_IN_MB
to
TF_GPU_HOST_MEM_LIMIT_IN_MB
. This may affect scripts written to use models with
TensorFlow Large Model Support. See Getting started with TensorFlow large model support for
more information.
Deprecations
Python 2.7 support: The Python core team plans to stop supporting Python 2 on January 1st 2020. Ahead of this, many deep learning frameworks and dependent libraries are or are planning to drop support for Python 2.7.
TensorFlow Large Model Support (TFLMS) V1 is no longer included with WML CE. Please refer to Getting started with TensorFlow large model support
Post WML CE 1.6.1 release, Conda packages for
pai4sk
, py-xgboost-cpu
, py-xgboost-gpu
, and
snapml-spark
will not have support for Python 2.7