具有虚拟环境的 vscode 中的 Jupyter 笔记本无法导入 tensorflow

Question

I'm attempting to create an isolated virtual environment running tensorflow & tf2onnx using a jupyter notebook in vscode.

The tf2onnx packge recommends python 3.7, and my local 3.7.9 version usually works well with tensorflow projects, so I have local and global versions set to 3.7.9 using pyenv.

The following is my setup procedure:

python -m venv .venv

Then after starting a new terminal in vscode:

pip install tensorflow==2.7.0

pip freeze > requirements.txt

After this, in a cell in my jupyter notebook, the following line fails

import tensorflow.keras as keras

Exception:

 TypeError: Descriptors cannot not be created directly. If this call came from a _pb2.py file, your generated code is out of date and must be regenerated with protoc >= 3.19.0. If you cannot immediately regenerate your protos, some other possible workarounds are: 1. Downgrade the protobuf package to 3.20.x or lower. 2. Set PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python (but this will use pure-Python parsing and will be much slower).

At this point, the protobuf package version is showing as v4.21.0 in my requirements file. I've attempted to pre-install the 3.20.1 version into the virtual environment before installing tensorflow but this yields no effect.

Here is the full requirements file after installing tensorflow:

absl-py==1.0.0
astunparse==1.6.3
cachetools==5.1.0
certifi==2022.5.18.1
charset-normalizer==2.0.12
flatbuffers==2.0
gast==0.4.0
google-auth==2.6.6
google-auth-oauthlib==0.4.6
google-pasta==0.2.0
grpcio==1.46.3
h5py==3.7.0
idna==3.3
importlib-metadata==4.11.4
keras==2.7.0
Keras-Preprocessing==1.1.2
libclang==14.0.1
Markdown==3.3.7
numpy==1.21.6
oauthlib==3.2.0
opt-einsum==3.3.0
protobuf==4.21.0
pyasn1==0.4.8
pyasn1-modules==0.2.8
requests==2.27.1
requests-oauthlib==1.3.1
rsa==4.8
six==1.16.0
tensorboard==2.9.0
tensorboard-data-server==0.6.1
tensorboard-plugin-wit==1.8.1
tensorflow==2.7.0
tensorflow-estimator==2.7.0
tensorflow-io-gcs-filesystem==0.26.0
termcolor==1.1.0
typing-extensions==4.2.0
urllib3==1.26.9
Werkzeug==2.1.2
wrapt==1.14.1
zipp==3.8.0

Answer 1

A recent change in protobuf is causing TensorFlow to break . Downgrading before installing TensorFlow might not work because TensorFlow might be bumping up the version itself. Check if that is what happens during the installation.

You might want to either:

Downgrade with

pip install --upgrade "protobuf<=3.20.1"

after installing TensorFlow, or

Upgrade TensorFlow to the latest version, as TensorFlow has updated their setup file in their 2.9.1 release.