Get up-to-date

[khatchad]

Sorry for the large update.

• Investigate decorator issue on tutorial code (Investigate decorator issue on tutorial code ponder-lab/ML#68)
• Fix comment.
• Add interprocedural dataset support (Add interprocedural dataset support ponder-lab/ML#69)
• Change log from warning to info.
• Add support for non-scalar datasets (Add support for non-scalar datasets ponder-lab/ML#70)
• Use interprocedural analysis for property read case.
• Add logging.
• Fix non-scalar determination.
• Add logger.
• Fix logger.
• Add logging around callable workaround.
• Handle callables with lexical scoping.
• Formatting.
• Handle possible null receivers in callable lookup.
• Change log message.
• Add log.
• Add the ability to process datasets coming out of enumerate() (Add the ability to process datasets coming out of enumerate() ponder-lab/ML#71)
• Fix tests for input files with multiple functions.
• Enhance TF2 tests (Enhance TF2 tests ponder-lab/ML#73)
• Separate tests by calling context (Separate tests by calling context ponder-lab/ML#74)
• Add tf.keras.layers.Dense.
• Move comment.
• Additional check.
• New API.
• Add autoencoder test.
• Move the value number check outside the context loop.
• More the sigmoid function to math.
• Format.
• Add API tf.add().
• Check the context/funciton PK mappings. (Check the context/funciton PK mappings. ponder-lab/ML#77)
• Add URL.
• Add dataset generator and test.
• tf.nn.softmax isn't a tensor generator.
• Add URL.
• Apply spotless.
• Let's make the two tests only diff in the invocation.
• Prevent infinite recursion when processing instructions when finding tensor data flow sources.
  • When we look for dataset tensor sources, we recurse up to the original defining instruction. However, there's a possibility that instructions have mutually recursive definitions due to the phi instruction. Let's prevent against that.

[msridhar]

Did not do a deep review, but LGTM :-)