TensorFlow version 2.16 has been officially released. However, as I understand it installation is not recommended for Ubuntu operating systems (?). My PC specs include an Ubuntu OS with RTX3060 Nvidia graphics card (CUDA-enabled). I have installed already TensorFlow version 2.15.0 and it works absolutely fine utilizing my GPU. Is there any news regarding when could users with Ubuntu OS and CUDA-enabled GPUs upgrade to TensorFlow 2.16 ?
2 Likes
I have Ubuntu 22.04.4 LTS. I also have a RTX3060. TF 2.15 works nice and well with the GPUs in the RTX3060 but TF 2.16 does not. I reported the issue to the Google team; as far as I can tell that issue still persists. What I find surprising is that the GPUs in my box work really well with Julia and with torch. Hopefully the TF team will fix this issue soon.
2 Likes
It turns out that by running pip install tensorflow[and-cuda] in a new conda environment with python version = 3.10 and use JAX as the Keras backend (edit as appropriate the local config file at ~/.keras/keras.json) works like a charm 
Welcome TensorFlow 2.16 !
2 Likes
Addressing cuDNN Loading Failure
The initial workaround focused on resolving the cuDNN library loading issue. It was discovered that TensorFlow 2.16 stopped loading cuDNN from the typical Python site-packages directory. The solution was to manually adjust the LD_LIBRARY_PATH environment variable to include the directory where cuDNN was located:
export LD_LIBRARY_PATH=~/anaconda3/lib/python3.10/site-packages/nvidia/cudnn/lib/:$LD_LIBRARY_PATH
After applying this change, TensorFlow was able to recognize and list the GPU devices, indicating that the cuDNN loading issue was resolved.
Encountering a New Issue with ptxas
Upon attempting to execute a deep learning model training script, a new error emerged related to the ptxas tool, which is part of the CUDA toolkit and is responsible for compiling PTX (Parallel Thread Execution) code to SASS (Streamlined Assembler) code. The error suggested a bug in the version of ptxas being used, affecting the XLA compilation process.
Resolving the ptxas Issue
The final solution to make TensorFlow 2.16.1 fully operational with GPU support involved locating a compatible version of ptxas that did not exhibit the reported bug. This compatible version was found within the site-packages directory of a Python installation, under a CUDA toolkit installation path:
...lib/python3.10/site-packages/nvidia/cuda_nvcc/bin
Manually add this specific path to the environment variables and limit this change to the Conda virtual environment specifically created for TensorFlow 2.16.1. This step will ensure that TensorFlow can correctly utilize the ptxas tool during the compilation process, essential for training deep learning models on the GPU.
Github issue link: TF 2.17.0 RC0 Fails to work with GPUs (and TF 2.16 too) · Issue #63362 · tensorflow/tensorflow · GitHub
1 Like
I am wondering if binary releases will get updated on this page? - TensorFlow for C 설치
I tried to see if 2.6.0 or 2.6.1 exist at the distribution link location but couldn’t find any. On macOS 2.6.1 is available via brew.
Out of curiosity, why didn’t you install keras + jax instead, in that case? Or do keras and also jax come with tensorflow?
@Mah_Neh I am a TensorFlow user and surely not familiar with JAX at all. Thus my primary aim was to use TensorFlow with my GPU. However, Keras 3 is backend-agnostic thus I just had to adjust appropriately my scripts from TensorFlow to pure Keras and by that way I could utilize JAX (which in many cases is actually faster).
1 Like
Interesting. Did you install jax backend then as well?
I have been in the same road, following their migration guide. In fact, I made tf work only because of your installation script, that adds the paths to nvidia libraries, so thank you.
1 Like
@Mah_Neh welcome. Indeed I did install JAX as well… Hopefully the next release of TensorFlow (perhaps in the next days) will be easier to install in order to utilize GPUs…
1 Like
Do you also ise pytorch? I like both of the frameworks
Yes indeed. I agree. Both frameworks are splendid. Perhaps PyTorch is less buggy but each framework has its pros and cons.
1 Like