Protobuf update breaks TF37 base image

[Gholtes]

The May 28 update of Protobuf contains changes that cause TensorFlow to install but fail to import into Python at runtime when installed as per the TF3y sample. I suggest changing the base docker file in the TF37 example to specify a version of Protobuf (such as 3.20.1) where this is not an issue.

Suggested change to dockerfile for the TF37 base image, which is downloaded by panorama_test_utility.download_artifacts_gpu_sample('tensorflow', account_id)

Current:
RUN python3.7 -m pip install --no-cache-dir --verbose future==0.18.2 mock==3.0.5 h5py==2.10.0 keras_preprocessing==1.1.1 keras_applications==1.0.8 gast==0.2.2 futures protobuf pybind11

Suggestion:
RUN python3.7 -m pip install --no-cache-dir --verbose future==0.18.2 mock==3.0.5 h5py==2.10.0 keras_preprocessing==1.1.1 keras_applications==1.0.8 gast==0.2.2 futures protobuf==3.20.1 pybind11

More info:
https://pypi.org/project/protobuf/3.20.1/#history
https://developers.google.com/protocol-buffers/docs/news/2022-05-06#python-updates

[shimomut]

@Gholtes thank you for reporting this problem. We also noticed this issue, and are working on it. We will use this issue item to track.