Background

• Python2 is eol
• We still use some Python2 UDF in https://github.com/exasol/data-science-examples/blob/master/examples/tensorflow-with-gpu-preview/tensorflow-gpu-preview.ipynb

Acceptance Criteria

• Replace all Python2 UDFs with Python3

Add graphics/screenshot, proper layout and full formulations to tutorial made in #39, so it is ready for use.

Add a Jupyter Notebook example showcasing how to connect AzureML to an Exasol Database and access the data from ExasolDB in AzureML.
This should include all functionality and key-words.

Might be similar to ConnectSagemakerToExasol.ipynb and LoadExampleDataIntoExasol.ipynb

Add a Jupyter Notebook example showcasing how to apply a simple ML algorithm from AzureML on data from an ExasolDB by training and testing the model.
This should include all functionality and key-words.

Builds on #39

Might be similar to TrainSagemakerModelWithExasolData.ipynb .

Background

1. Python3.6 is deprecated and we should use a new version
2. There log location has been changed and needs to updated in the tutorial

Acceptance Criteria

1. Switch from Python 3.6 to Python3.8 in the tutorial (script-languages.ipynb)
2. Adjust the log location: exaslct.log => main.log

Background

• We develop the sagemaker extensions, if the following issue is implemented create an example for it exasol/sagemaker-extension#3

Acceptance Criteria

• To show the user how to use it, create a notebook which shows deployment and usage on an example dataset

Currently, the BucketFS connection information get embedded into the UDF code (classification.ipynb). A much cleaner method is, to provide this information via a connection. Either, by providing the BucketFS URL directly, or by providing a path to a configuration file in the BucketFS.

• In order to upload an object into BucketFS, the notebook creates and uses an UDF called EXABUCKET_HELPER.
• Then this UDF is called when an object is going to be uploaded. Instead of creating an UDF to upload an object into BucketFS, we can use bucketfs connection object and bucketfs_python libraries.

Add graphics/screenshot, proper layout and full formulations to tutorial made in #41 , so it is ready for use.

• Jupyter with pyexasol and r-exasol
• Exasol docker-db
• Loading the data-science-examples repos
• Containing required language container

We want to reuse the LoadExampleDataIntoExasol.ipynb for the AzureML tutorial.

Add some explanation to https://github.com/exasol/data-science-examples/blob/main/tutorials/machine-learning/python/sagemaker/LoadExampleDataIntoExasol.ipynb to make it more reader friendly.

Add a Jupyter Notebook example showcasing invoke a trained AzureML model from Exasol Database via UDF.
This should include all functionality and key-words.

Builds on #40

Might be similar to UseSagemakerModelFromExasol.ipynb .

Test Procedure:

• Start Exasol Database Docker Container
• Start Jupyter Notebook Docker Container
• Tranform Notebook to Script with jupyter nbconvert
• run script

Add an example for connecting from AWS Sagemaker to Exasol.

Add example for training Sagemaker model with data exasol.

Add an example for the SageMaker guide on how to use the SageMaker model from inside of the Exasol database.

As mentioned here, we directly import of the header of the scania trucks data set into exasol at the moment, and as a result use
all_columns = exasol.export_to_pandas("SELECT * FROM IDA.TRAIN LIMIT 1;")
for the reading of the header, which is not great.

please update in the three files mentioned below by using this or this:

we import the data using this notebook from sagemaker tutorial for the data import which needs to be changed, and then the data is read in the sagemaker tutorial and in the azureml tutorial

Background

• We plan to rename the flavor names of the script-languages exasol/script-languages-release#566
• We need to adapt this here as well

Acceptance Criteria

• Use new flavor names

Add a example for the SageMaker guide on how to load example data.

File error_code_config.yml contains invalid yaml syntax.

• tab characters are not allowed
• indention should be correct

Please compare contents of the corresponding file for project-keeper for reference.

Pip installs tensorflow 2 if the version is not pinned. However, the UDF still uses version 1.3.

Requirements:

• Show how to build a language container
• Show where logs can be found
• Show how a flavor looks like
• Show how to customize a container
• Show how to upload a container
• Show how to activate a container
• Explain caching and hashes
• Explain content of a container
• Show utility scripts to investigate container

Add graphics/screenshot, proper layout and full formulations to tutorial made in #40 , so it is ready for use.

Currently there is a bug in the saas online interface where the path of bucketfs files is shown incorrectly. this is mentioned in the azureml tutorial. once this bug is fixed, remove the two mentions of it.