tensorflow is so buggy, you guys should just gave up and should migrate to TORCH this is so bad , i cant anymore.

[wicaksonoleksono]

Issue type

Build/Install

Have you reproduced the bug with TensorFlow Nightly?

Yes

Source

source

TensorFlow version

tf 2.16

Custom code

Yes

OS platform and distribution

ubuntu 24.04

Mobile device

wsl

Python version

3.10

Bazel version

No response

GCC/compiler version

No response

CUDA/cuDNN version

12.3

GPU model and memory

No response

Current behavior?

I dont wanna talk about it anymore , the fact that it's too high a tower (keras > tensorflow > graph > compiled code), it's just too complicated to troubleshoot and extend. JUST STOP IT you people are torturing your self, tell your boss to STOP developing tensorflow, and CLOSES IT already

Standalone code to reproduce the issue

Cuddn: 8.9 Cuda: 12.3

Installing wsl tensorflow

wsl --install

wsl --list --online
wsl --install Ubuntu-24.04

wsl -d Ubuntu-24.04


mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm -rf ~/miniconda3/miniconda.sh
~/miniconda3/bin/conda init bash




Second opt 
---------------------------------------------------------------------------------------
conda create -n variety_linux python=3.10
conda activate variety_linux
pip install tensorflow[and-cuda]


cd $CONDA_PREFIX
mkdir -p ./etc/conda/activate.d
mkdir -p ./etc/conda/deactivate.d
touch ./etc/conda/activate.d/env_vars.sh
touch ./etc/conda/deactivate.d/env_vars.sh



#!/bin/sh

# Store original LD_LIBRARY_PATH
export ORIGINAL_LD_LIBRARY_PATH="${LD_LIBRARY_PATH}" 

# Get the CUDNN directory
CUDNN_DIR=$(dirname $(dirname $(python -c "import nvidia.cudnn; print(nvidia.cudnn.__file__)")))

# Set LD_LIBRARY_PATH to include CUDNN directory
export LD_LIBRARY_PATH=$(find ${CUDNN_DIR}/*/lib/ -type d -printf "%p:")${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

# Get the ptxas directory
PTXAS_DIR=$(dirname $(dirname $(python -c "import nvidia.cuda_nvcc; print(nvidia.cuda_nvcc.__file__)")))

# Set PATH to include the directory containing ptxas
export PATH=$(find ${PTXAS_DIR}/*/bin/ -type d -printf "%p:")${PATH:+:${PATH}}




#!/bin/sh

# Unset environment variables
unset CUDNN_DIR
unset PTXAS_DIR

# Restore original LD_LIBRARY_PATH
export LD_LIBRARY_PATH="${ORIGINAL_LD_LIBRARY_PATH}"


python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"

Relevant log output

GPU is availabel , but it s not

[mihaimaruseac]

Canonical issue #63362. It also contains some work-arounds. TF 2.17 branch cut just happened, maybe this will get fixed now?

See also: #63362 (comment)

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No

[wicaksonoleksono]

Canonical issue #63362. It also contains some work-arounds. TF 2.17 branch cut just happened, maybe this will get fixed now?

See also: #63362 (comment)
still didnt work, i think the problem is tf 2.16 messes up with keras 3, thats why it doesnt work

[mihaimaruseac]

Yes. The comment there lists why this did not get caught.

Now there is a 2.17 release train, hopefully that fixes things