How to install objax with GPU support? - python

I have followed the objax documentation to install the library with GPU support: https://objax.readthedocs.io/en/stable/installation_setup.html
i.e.
pip install --upgrade objax
CUDA_VERSION=11.6
pip install -f https://storage.googleapis.com/jax-releases/jax_releases.html jaxlib==`python3 -c 'import jaxlib; print(jaxlib.__version__)'`+cuda`echo $CUDA_VERSION | sed s:\\\.::g`
However the last step doesn't work. I get the following error message:
ERROR: Could not find a version that satisfies the requirement jaxlib==0.3.15+cuda116 (from versions: 0.1.32, 0.1.40, 0.1.41, 0.1.42, 0.1.43, 0.1.44, 0.1.46, 0.1.50, 0.1.51, 0.1.52, 0.1.55, 0.1.56, 0.1.57, 0.1.58, 0.1.59, 0.1.60, 0.1.61, 0.1.62, 0.1.63, 0.1.64, 0.1.65, 0.1.66, 0.1.67, 0.1.68, 0.1.69, 0.1.70, 0.1.71, 0.1.72, 0.1.73, 0.1.74, 0.1.75, 0.1.76, 0.3.0, 0.3.2, 0.3.5, 0.3.7, 0.3.8, 0.3.10, 0.3.14, 0.3.15)
ERROR: No matching distribution found for jaxlib==0.3.15+cuda116
I have tried with multiple versions of python/CUDA, but I always get this error.
Executing pip install --upgrade pip at the begining does not help.
System description:
Operating system: Ubuntu 20.04.4 LTS
CUDA Version: 11.6
Python version: 3.8.13

JAX recently updated its GPU installation instructions, which you can find here: https://github.com/google/jax#pip-installation-gpu-cuda
In particular, the CUDA wheels are now located at https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
So, for example, you can install JAX with
$ pip install "jax[cuda11_cudnn805]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
and replace cuda11 and cudnn805 respectively with the appropriate CUDA and CUDNN version for your system, ensuring that they match the versions listed in the index at the above URL.
I've sent a pull request to the objax repository to update the instructions you were following: https://github.com/google/objax/pull/246

The doc page was last updated 2 year ago. It's outdated and due to this it currently lies. Take a close look at https://storage.googleapis.com/jax-releases/jax_releases.html — there're currently no releases with CUDA.
As for no-CUDA versions you can install them directly from PyPI:
pip install jaxlib
For CUDA version — try to compile from sources. Or use conda: https://anaconda.org/search?q=jaxlib

Related

Can't install tensorflow on ubuntu 20.04.3 Python 3.6.15 pip 21.2.4

I have been around the mulberry bush for a few days trying to get tensorflow installed on an Ubuntu system. My specifications are;
Ubuntu: 20.04.3
Python (virtual environment): 3.6.15
Pip: 21.2.4
Tensorflow version wanted: 1.13.1
Tensorflow version will take: anything at this point
I followed a number of answers on the web. Got a compatible version of Python. Got the latest pip....
When I use the command;
pip3 install tensorflow==1.13.1
I get;
ERROR: Could not find a version that satisfies the requirement tensorflow==1.13.1 (from versions: none)
ERROR: No matching distribution found for tensorflow==1.13.1
when I do a pip_search on tensorflow, it shows tensorflow 2.6.0.
But when I do;
pip3 install tensorflow==2.6.0
I get the same error.
when I use the --upgrade switch to pip, I get the same error. I tried to install from a wheel file, but the file was incompatible on my system.
Any help is appreciated.
As per documentation of Tensorflow, version 1's final release that can be downloaded/installed is 1.15. Below this version are deprecated and no longer available to use in development.
Read it about here

Error while installing opencv-contrib-python 3.3.0.9

I am trying to install opencv-contrib-python 3.3.0.9 using command python -m pip install opencv-contrib-python==3.3.0.9.
However it is throwing the following error:
Could not find a version that satisfies the requirement opencv-contrib-python==3.3.0.9 (from versions: 3.4.2.16, 3.4.2.17, 3.4.3.18, 3.4.4.19, 3.4.5.20, 3.4.6.27, 3.4.7.28, 3.4.8.29, 4.0.0.21, 4.0.1.23, 4.0.1.24, 4.
1.0.25, 4.1.1.26, 4.1.2.30)
No matching distribution found for opencv-contrib-python==3.3.0.9
I am using Python 3.7, opencv 4.1.2.30 and PyCharm.
Can you install a newer version of opencv-contrib-python by using
pip install opencv-contrib-python?
According to https://pypi.org/project/opencv-contrib-python/3.3.0.9/, that version does not support Python 3.7 and it also provides a link to the latest version which mention support Python 3.7 and installation guide (i.e. pip install opencv-contrib-python)

How to install PyTorch on Python 3.7 / Windows 10 with pip

I am trying to install PyTorch 1.3 using pip :
pip3 install torch==1.3.1+cpu torchvision==0.4.2+cpu -f https://download.pytorch.org/whl/torch_stable.html
I got the command line from this page :
https://pytorch.org/get-started/locally/#start-locally
It fails, with the following error message :
ERROR: Could not find a version that satisfies the requirement
torch==1.3.1+cpu (from versions: 0.1.2, 0.1.2.post1, 0.1.2.post2)
ERROR: No matching distribution found for torch==1.3.1+cpu
Using verbose mode (-v) give a little more info :
Skipping link: unsupported archive format: .html:
https://download.pytorch.org/whl/torch_stable.html (from -f)
Analyzing links from page https://pypi.org/simple/torch/
It seems pip cannot use the url I provide and will get packages from another URL.
That one support at best Python 3.6. Since I have installed 3.7.5 (which is support by URL initially provided) none of the packages match and it fails.
I could downgrade to 3.6 but since 3.7 is officially supported I would like to try to fix that issue first.
EDIT : pip is up to date. I got exact same error on both Windows 10 and 7 64 bit machines.
Also: the PyTorch page says :
Please ensure that you have met the prerequisites below (e.g., numpy)
But AFAIK does not list any library in that section. During my tests, numpy was installed and up to date.
As requested, here is the full output of pip :
https://pastebin.com/8Wpqp6du
You error is being triggered because the version 1.3.1+cpu doesn't exist.
If you go to pytorch.org you will be able to select a version of pytorch and your OS, where it will give you a command to install pyTorch correctly, for Python 3.7 and PIP use the following:
pip3 install --find-links https://download.pytorch.org/whl/torch_stable.html torch==1.3.1 torchvision==0.4.2
Remember to check that the pip3 binary you are using is the one installed over Python 3.7.
Also, mind that the documentation you are pointing at is for MacOS.

How to install yolo3 library in python

I am trying to install yolo3 library in python. But, it does not. I typed pip install yolo3
ERROR: Could not find a version that satisfies the requirement yolo3 (from versions: none)'
ERROR: No matching distribution found for yolo3
yolo34py comes in 2 variants, CPU Only Version and GPU Version. Installation may take a while since it involves downloading and compiling of darknet.
CPU Only Version
pip3 install yolo34py
GPU Version:
pip3 install yolo34py-gpu
pypi Reference : https://pypi.org/project/yolo34py/

Anaconda3, Python 3.6 - pip install imagenet_utils - error message: Could not find a version that satisfies the requirement imagenet_utils

Anaconda3, Python 3.6, pip version 9.0.1
pip install imagenet_utils
Error message:
Could not find a version that satisfies the requirement imagenet_utils (from versions: ) No matching distribution found for imagenet_utils
Please, help.
Read me docs on github indicate compatibility up to 3.5. Take a look at some comparable libraries to avoid this issue here:
5 Genius Python Deep Learning Libraries.
Anaconda also offers a conda install version of Keras where this specific package can be accessed:
conda install -c conda-forge keras
then for python script would read:
from keras import imagenet_utils
Assuming compatibility is not the issue, you will need to verify the anaconda package location and make sure that you are installing to that location.

Categories

Resources