Tensorflow 2.10.0 not detecting GPU

Question

I've created a conda environment and installed tensorflow as such:

conda create -n foo python=3.10
conda activate foo
conda install mamba
mamba install tensorflow -c conda-forge
mamba install cudnn cudatoolkit

This installed TensorFlow 2.10.0. I've installed CUDA 11.2 and cuDNN 8.1, and then try to run the following:

import tensorflow as tf

print(f"GPUs available: {tf.config.list_physical_devices('GPU')}")

but it just returns an empty list. I have a 3060ti that I want to use for my ML projects but TensorFlow is not detecting it. I found similar questions to mine, like this, this and this but they use the old version of TensorFlow, which would install tensorflow-gpu and is no longer supported. How can I fix this, or even attempt to troubleshoot it.

I'm using a Windows 10 machine

Output of nvidia-smi:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 528.24       Driver Version: 528.24       CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ... WDDM  | 00000000:09:00.0  On |                  N/A |
| 30%   43C    P8    16W / 200W |    809MiB /  8192MiB |      3%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      7176    C+G   ...perience\NVIDIA Share.exe    N/A      |
|    0   N/A  N/A      9240    C+G   C:\Windows\explorer.exe         N/A      |
|    0   N/A  N/A     12936    C+G   ...cw5n1h2txyewy\LockApp.exe    N/A      |
|    0   N/A  N/A     13652    C+G   ...e\PhoneExperienceHost.exe    N/A      |
|    0   N/A  N/A     14020    C+G   ...2txyewy\TextInputHost.exe    N/A      |
|    0   N/A  N/A     14888    C+G   ...ser\Application\brave.exe    N/A      |
|    0   N/A  N/A     15112    C+G   ...5n1h2txyewy\SearchApp.exe    N/A      |
|    0   N/A  N/A     16516    C+G   ...oft OneDrive\OneDrive.exe    N/A      |
|    0   N/A  N/A     18296    C+G   ...aming\Spotify\Spotify.exe    N/A      |
|    0   N/A  N/A     18624    C+G   ...in7x64\steamwebhelper.exe    N/A      |
|    0   N/A  N/A     18672    C+G   ...\app-1.0.9010\Discord.exe    N/A      |
|    0   N/A  N/A     18828    C+G   ...lPanel\SystemSettings.exe    N/A      |
|    0   N/A  N/A     19284    C+G   ...Central\Razer Central.exe    N/A      |
|    0   N/A  N/A     20020    C+G   ...arp.BrowserSubprocess.exe    N/A      |
|    0   N/A  N/A     22912    C+G   ...8wekyb3d8bbwe\Cortana.exe    N/A      |
|    0   N/A  N/A     24848    C+G   ...ontend\Docker Desktop.exe    N/A      |
|    0   N/A  N/A     25804    C+G   ...y\ShellExperienceHost.exe    N/A      |
|    0   N/A  N/A     27064    C+G   ...8bbwe\WindowsTerminal.exe    N/A      |
+-----------------------------------------------------------------------------+

Output of nvcc -V:

Copyright (c) 2005-2021 NVIDIA Corporation
Built on Sun_Feb_14_22:08:44_Pacific_Standard_Time_2021
Cuda compilation tools, release 11.2, V11.2.152
Build cuda_11.2.r11.2/compiler.29618528_0

I ran a dummy code as such:

import tensorflow as tf
import numpy as np


def make_nn():
    model = tf.keras.models.Sequential()
    model.add(tf.keras.layers.Dense(1, input_shape=(1,)))
    model.compile(loss='mean_squared_error', optimizer='sgd')
    return model

def dataset():
    x = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9])
    y = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9])
    return tf.data.Dataset.from_tensor_slices((x, y)).batch(1)



def main():
    model = make_nn()
    model.fit(dataset(), epochs=1, steps_per_epoch=9)

if __name__ == '__main__':
    print(f"GPUs available: {tf.config.list_physical_devices('GPU')}")
    print(f"Built with cuda: {tf.test.is_built_with_cuda()}")

    main()

and it gave me the following log:

GPUs available: []
Built with cuda: False
2023-02-06 09:47:32.744450: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2023-02-06 09:47:32.779280: I tensorflow/core/common_runtime/process_util.cc:146] Creating new thread pool with default inter op setting: 2. Tune using inter_op_parallelism_threads for best performance.

Looks like it's using a CPU build

At first I'd check if GPU driver is installed and up to date. — Gameplay, Feb 06 '23 at 09:28
Did you create an account to NVidia then download and install cuDNN? — Corralien, Feb 06 '23 at 09:30
Oh @Corralien, looks like it's taking CUDA version 12 that I had installed a while back — João Areias, Feb 06 '23 at 09:41
Standard troubleshooting is to look at the tensorflow logs printed to terminal, they have all the information you need about finding GPU libraries. — Dr. Snoopy, Feb 06 '23 at 11:44
@Dr.Snoopy it doesn't print anything, it just starts using the CPU. I noticed it wasn't detecting it just because the training of my nn was way too slow. Is there a logfile that TensorFlow creates, or something like that that I can use to check? — João Areias, Feb 06 '23 at 12:08
There is no log file, its all printed to the terminal, so it depends on how you are running python exactly, in jupyter this is sometimes printed to the terminal running jupyter. — Dr. Snoopy, Feb 06 '23 at 12:11
OK, so you are using a non-GPU enabled version of Tensorflow. Probably not the one installed by conda. — talonmies, Feb 06 '23 at 12:45
@Dr.Snoopy well, I printed `tf.__version__` and showed 2.10, why? — João Areias, Feb 06 '23 at 13:17
Because TensorFlow 2.11 does not support GPUs in Windows anymore. Also anaconda might have tensorflow-gpu packages that you need to install. — Dr. Snoopy, Feb 06 '23 at 13:18
@Dr.Snoopy Really? I didn't know that. If I run `mamba install tensorflow-gpu` it looks for TensorFlow 1.15 — João Areias, Feb 06 '23 at 13:21

score 0 · Answer 1 · answered Feb 06 '23 at 14:03

Probably not the best solution, but I downgraded TensorFlow back to version 2.6.0 which was previously installed and it worked, which is a bummer, I wanted to try some more recent features, but for the time being looks like this will suffice. If anyone is facing the same issues, this is the current conda environment that I'm using

score 0 · Answer 2 · edited Mar 22 '23 at 15:31

0

I solved it by creating another environment with python = 3.9 and tensorflow = 2.8.1. Also, I have cuda =11.4

edited Mar 22 '23 at 15:31

Ajeet Verma

2,938
3
13
24

answered Mar 16 '23 at 00:14

Nour Alden

29
4

and thanks for dislike before 3 minute the fastest response – Nour Alden Mar 16 '23 at 00:15
i add picture to believe me – Nour Alden Mar 16 '23 at 00:16

score -1 · Answer 3 · answered Feb 06 '23 at 20:19

If you use conda-forge you may need to set the environment variable CONDA_OVERRIDE_CUDA to force installing the gpu enabled version of tensorflow as explained here https://conda-forge.org/docs/user/tipsandtricks.html#installing-cuda-enabled-packages-like-tensorflow-and-pytorch. Under bash that would be something like

CONDA_OVERRIDE_CUDA="11.2" conda install "tensorflow==2.8" -c conda-forge

Tensorflow 2.10.0 not detecting GPU

3 Answers3

Linked