Object detection API installation does not recognize tensorflow-gpu==2.5.0 already installed in the Conda virtual environment

[rsun-bdti]

Prerequisites

Please answer the following questions for yourself before submitting an issue.

• I am using the latest TensorFlow Model Garden release and TensorFlow 2.
• I am reporting the issue to the correct repository. (Model Garden official or research directory)
• I checked to make sure that this issue has not already been filed.

1. The entire URL of the file you are using

https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2.md#python-package-installation

https://github.com/tensorflow/models/blob/master/research/object_detection/packages/tf2/setup.py

2. Describe the bug

Running setup.py to install object detection API installs tensorflow==2.6.0 official in my Conda virtual environment. tensorflow-gpu==2.5.0 was already installed in the environment, and it should be used instead.

The behavior is similar to the complaint in issue #9267, except that was in a docker container. In my case, it's the Conda virtual environment.

Is this a bug? Or is it something wrong with my Conda configuration?

3. Steps to reproduce

Follow the steps below to reproduce the bug.

nano conda-env-tf25-cuda11-p36.yaml

name: env-test
channels:
  - conda-forge
  - defaults
dependencies:
  - pip=21.2.4=pyhd8ed1ab_0
  - python=3.6.13=hb7a2778_1_cpython
  - flatbuffers==1.12
  - matplotlib==3.3.4
  - numpy==1.19.5
  - pandas==1.1.5
  - pillow==8.3.2
  - tensorboard==2.5.0
  - pip:
    - opencv-python==4.5.3.56
    - opencv-python-headless==4.5.3.56
    - protobuf==3.17.3
    - scipy==1.5.4
    - tensorflow-gpu==2.5.0

conda -V

conda 4.10.3

conda env create --file conda-env-tf25-cuda11-p36.yaml --force

conda activate env-test

python -c 'import tensorflow as tf; print(tf.__version__)'

2021-09-17 17:33:17.425787: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2.5.0

git clone https://github.com/tensorflow/models.git

cd models/research

protoc object_detection/protos/*.proto --python_out=.

cp object_detection/packages/tf2/setup.py .

python -m pip install --use-feature=2020-resolver .

(Terminal displayed information about many packages being installed)

python -c 'import tensorflow as tf; print(tf.__version__)'

2.6.0

Not only was the Tensorflow version changed from 2.5.0 to 2.6.0, but also no CUDA information was displayed.

4. Expected behavior

tensorflow-gpu==2.5.0 in the Conda virtual environment should not be replaced by Tensorflow 2.6.0 CPU.

5. Additional context

Include any logs that would be helpful to diagnose the problem.

6. System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 18.04
• Mobile device name if the issue happens on a mobile device:
• TensorFlow installed from (source or binary): Binary
• TensorFlow version (use command below): tensorflow-gpu==2.5.0
• Python version: 3.6.13
• Bazel version (if compiling from source):
• GCC/Compiler version (if compiling from source):
• CUDA/cuDNN version: CUDA 11.0/cuDNN 8.2.4
• GPU model and memory: GeForce RTS 2080 TI

[MrinalJain17]

Faced a similar issue.

Internally, the object detection package installs tf-models-official, which has a dependency on tensorflow-text. Pip ends up fetching the latest version of text (2.6.0), which in turn requires TF 2.6. That is probably why your TF installation is overridden.

A potential workaround could be to manually install tensorflow-text (2.4) before installing the object detection package.

models/official/pip_package/setup.py

Line 65 in 87769dc

install_requires.append('tensorflow-text>=2.4.0')

[LUYIJUN1206074121]

I have the same question, may I know you how to solove it?

[thmsgntz]

Hello, have a look a this issue #9911, might help you.