I'm trying to install Tensorflow on a computer with a J1800 CPU. I know this CPU does not have the AVX extension. I'm trying to find a wheel package of tensorflow without an AVX extension. Every wheel I try gets the same error. Here's one example:
pip3 install https://raw.githubusercontent.com/fo40225/tensorflow-windows-wheel/master/1.14.0/py37/CPU/sse2/tensorflow-1.14.0-cp37-cp37m-win_amd64.whl
: ERROR: tensorflow-1.14.0-cp37-cp37m-win_amd64.whl is not a supported wheel on this platform.
I'm running Python 3.8 64Bit
Does anyone know a wheel that will work on this CPU? Or am I making another mistake?
Only tensorflow >= 2.2 is available on python 3.8.
In order to work with tensorflow 1.14 you need to use another version of python (3.6 or 3.7)
Assuming you are installing tensorflow 1.14 on python 3.8. python 3.8 works with TF 2.0 only. You can either change the python version.
If you have any version like 3.x you can current python version to those by alias
eg :
alias python3 = python3.x
where x > 8
other wise checkout 2.0 and reinstall for the source.
Related
I have Python3.11.0 on a Windows10 PC.
Trying to install tensorflow using:
pip install tensorflow
gives error. Upon visiting the tensorflow site I realised that it supports only 3.7 - 3.10
Should I downgrade the python version or is any workaround available?
Yes, you should downgrade python to < 3.11 until the wheels are updated to support python 3.11
Yes, you could create a conda environment using python 3.10 to use tensorflow. Here's a link with instructions to follow : https://www.tensorflow.org/install/pip#macos
tensorflow currently supports from python 3.7 - 3.10 4
updated code to work with tensorflow 2.4 to get around issue
I'm trying to install Tensorflow 1.9 on a Raspberry Pi as it is a requirement of the code I want to run. It installs fine on my Macbook using pip install tensorflow==1.9.0, but on the Pi I get the error:
Could not find a version that satisfies the requirement tensorflow==1.9.0 (from versions: 0.11.0, 1.11.0, 1.12.0, 1.13.1, 1.14.0)
No matching distribution found for tensorflow==1.9.0
I'm using a Conda (miniconda3) environment with Python 3.6.
Would using Ubuntu Desktop instead of Raspberry Pi OS work maybe? Or is there perhaps a way to build it from https://github.com/tensorflow/tensorflow/tree/r1.9?
You need to install python version 3.6 or other compatible versions in conda. Tensorflow does not work on versions higher than 3.8 and has identified issues with some other versions. I recommend using python 3.6 through conda. You will also probably need to move the
libcrypto-1_1-x64.dll
libssl-1_1-x64.dll
files from anaconda3\Library\bin to anaconda3/DLLs. You can find more details about this issue if you run into it here.
Hope this answers your question!
To install Tensorflow on Raspberry Pi run:
sudo apt install libatlas-base-dev
and then
pip3 install tensorflow
You can also compilte TensorflowLite and use if you wish. The compile TensorflowLite on RPi refer this link
I have installed both tensorflow 2.2.0 and tensorflow 1.15.0(by pip install tensorflow-gpu==1.15.0). The tensorflow 2 is installed in the base environment of Anaconda 3, while the tensorflow 1 is installed in a separate environment.
The tensorflow 2.2.0 can recognize gpu based on a simple test:
if tf.test.gpu_device_name():
print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))
//output: Default GPU Device: /device:GPU:0
But the tensorflow 1.15.0 can not detect gpu.
For your information, my system environment is python + cuda 10.1 + vs 2015.
The tensosflow versions 1.15.0 to 1.15.3 (the latest version) are all compiled against Cuda 10.0. Downgrading the cuda 10.1 to cuda 10.0 solved the problem.
Also be aware of the python version. It is recommended to install the tensorflow .whl file (as listed at https://nero-mirror.stanford.edu/pypi/simple/tensorflow-gpu/) for the specific python version. As for installation, see How do I install a Python package with a .whl file?
Tensorflow 1.15 expects cuda 10.0 , but I managed to make it work with cuda 10.1 by installing the following packages with Anaconda: cudatoolkit (10.0) and cudnn (7.6.5). So, after running
conda install cudatoolkit=10.0
conda install cudnn=7.6.5
tensorflow 1.15 was able to find and use GPU (which is using cuda 10.1).
PS: I understand your environment is Windows based, but this question pops on Google for the same problem happening on Linux (where I tested this solution). Might be useful also on Windows.
It might have to do with the version compatibility of TF, Cuda and CuDNN. This post has it discussed thoroughly.
Have you tried installing Anaconda? it downloads all the requirements and make it easy for you with just a few clicks.
I was using tensorflow 2.0 with mkl (avx/avx2 optimization) and has zero problem. I found yesterday that tf 2.1 is available on anaconda, so I upgrade it to the latest. however, after upgrade, tf complains that "Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2".
To make sure I installed the correct version. I uninstalled all tensorflow components, and then do a fresh install using command conda install tensorflow-mkl. The installation succeed but the problem persists.
Could anyone advise how to solve this? Very appreciate the help.
Best
I have been trying to install tensorflow-gpu on windows 10, via
pip3 install --upgrade tensorflow-gpu
When I do this I break the current installation of ordinary tensorflow!, and get this error: On Windows, running "import tensorflow" generates No module named "_pywrap_tensorflow" error.
Somehow I manage to fix this by re-installing ordinary tensorflow, but then when I import tensorflow in python 3.5.2 and try to identify my GPU, No device is found!
I have a Cuda 9.0 installed alongside cudnn64_6 defined as a DLL in CUDA/v9.0/bin, and I can run the nbody test program without problems and I can see the GPU being used for that demo application.
Is there any known issue with tensorflow-gpu 1.3.0?
Really struggling on this. Why does it have to be so problematic installing this library!
Please help
mg
TensorFlow 1.3 (and 1.4) require CUDA 8.0 and do not support later versions. You will either need to downgrade CUDA to 8.0 or make a custom build from source.