The directory home/hl/miniconda3/envs/StyleFlow/lib/python3.7/site-packages/ contains two subdirectories related to tensorflow:
The tensorflow directory and the tensorflow_core directory contain different TensorFlow library files: 1. The tensorflow directory contains the Python language package of TensorFlow, including:- tensorflow.py: the Python interface entry file of TensorFlow.
- tools: Some TensorFlow utility scripts, such as freeze_graph conversion pb model, etc.
-keras: TensorFlow Keras high-level API.
- estimator: TensorFlow Estimator high-level API.
- examples: TensorFlow official models and tutorial examples. So if you want to call TensorFlow in Python, you need to add this directory to the PYTHONPATH environment variable
The tensorflow_core directory contains TensorFlow's C++ underlying library and header files, including: - frameworks/: TensorFlow framework core implementation, including computing, memory management, compilation, etc.
- kernels/: Various op core implementations of TensorFlow, including numerical calculation, matrix operation, NN and other ops.
- graph/: TensorFlow computation graph and session implementation.
- common/: Some TensorFlow common utility function implementations.
- lib: Compiled and generated TensorFlow library files, such as libtensorflow_cc.so and libtensorflow_framework.so, etc. So if you want to develop TensorFlow C/C++ extensions, or compile third-party libraries that rely on TensorFlow, you need to include the header files in this directory and link the library files.