Package dependency issue #29

I am not able to publish the pipeline into ML workspace now with the following package dependencies as specified in the code:

conda_dependencies = CondaDependencies.create(
conda_packages=["cudatoolkit=10.0"],
pip_packages=[
"azure-storage-blob==2.1.0",
"azureml-sdk",
"hickle==3.4.3",
"requests==2.21.0",
"sklearn",
"pandas",
"numpy",
"pillow==6.0.0",
"tensorflow-gpu==1.15",
"keras",
"matplotlib",
"seaborn",
]
)

The pip package dependencies is missing package versions, leads to failure while publish the pipeline into ML workspace.

A couple of obvious errors:

1. sklearn is no longer available now
2. Recurrent layer is not available with latest Tensorflow and Keras

Would be great if the code can be verified with library versions.