search

ContinuumIO / anaconda-issues

Anaconda issue tracking
648 stars 223 forks source link

Add latest versions of TensorFlow GPU to Conda #11967

Open johncaling40 opened 4 years ago

johncaling40 commented 4 years ago

Actual Behavior

The latest versions of TensorFlow GPU are not there.

Expected Behavior

The latest versions of TensorFlow GPU should be there

Steps to Reproduce

https://anaconda.org/anaconda/tensorflow-gpu Going there and seeing that the latest version(2.3) is not available for Windows and Linux

Anaconda or Miniconda version:

Latest version

Operating System:

Windows 10

conda info
``` PASTE OUTPUT HERE: active environment : None user config file : C:\Users\iamap\.condarc populated config files : C:\Users\iamap\.condarc conda version : 4.8.3 conda-build version : not installed python version : 3.8.3.final.0 virtual packages : __cuda=11.1 base environment : C:\ProgramData\Miniconda3 (writable) channel URLs : https://repo.anaconda.com/pkgs/main/win-64 https://repo.anaconda.com/pkgs/main/noarch https://repo.anaconda.com/pkgs/r/win-64 https://repo.anaconda.com/pkgs/r/noarch https://repo.anaconda.com/pkgs/msys2/win-64 https://repo.anaconda.com/pkgs/msys2/noarch package cache : C:\ProgramData\Miniconda3\pkgs C:\Users\iamap\.conda\pkgs C:\Users\iamap\AppData\Local\conda\conda\pkgs envs directories : C:\ProgramData\Miniconda3\envs C:\Users\iamap\.conda\envs C:\Users\iamap\AppData\Local\conda\conda\envs platform : win-64 user-agent : conda/4.8.3 requests/2.23.0 CPython/3.8.3 Windows/10 Windows/10.0.20180 administrator : True netrc file : None offline mode : False ```
conda list --show-channel-urls
``` PASTE OUTPUT HERE: # packages in environment at C:\ProgramData\Miniconda3: # # Name Version Build Channel anaconda-client 1.7.2 py38_0 defaults anaconda-navigator 1.9.12 py38_0 defaults attrs 19.3.0 py_0 defaults backcall 0.2.0 py_0 defaults bleach 3.1.5 py_0 defaults ca-certificates 2020.6.24 0 defaults certifi 2020.6.20 py38_0 defaults cffi 1.14.0 py38h7a1dbc1_0 defaults chardet 3.0.4 py38_1003 defaults clyent 1.2.2 py38_1 defaults colorama 0.4.3 py_0 defaults conda 4.8.3 py38_0 defaults conda-package-handling 1.6.1 py38h62dcd97_0 defaults console_shortcut 0.1.1 4 defaults cryptography 2.9.2 py38h7a1dbc1_0 defaults decorator 4.4.2 py_0 defaults defusedxml 0.6.0 py_0 defaults entrypoints 0.3 py38_0 defaults freetype 2.10.2 hd328e21_0 defaults icu 58.2 ha925a31_3 defaults idna 2.9 py_1 defaults importlib-metadata 1.7.0 py38_0 defaults importlib_metadata 1.7.0 0 defaults ipykernel 5.3.4 py38h5ca1d4c_0 defaults ipython 7.16.1 py38h5ca1d4c_0 defaults ipython_genutils 0.2.0 py38_0 defaults jedi 0.17.1 py38_0 defaults jinja2 2.11.2 py_0 defaults jpeg 9b hb83a4c4_2 defaults json5 0.9.5 py_0 defaults jsonschema 3.2.0 py38_0 defaults jupyter_client 6.1.6 py_0 defaults jupyter_core 4.6.3 py38_0 defaults jupyterlab 2.1.5 py_0 defaults jupyterlab_server 1.2.0 py_0 defaults libpng 1.6.37 h2a8f88b_0 defaults libsodium 1.0.18 h62dcd97_0 defaults libtiff 4.1.0 h56a325e_1 defaults lz4-c 1.9.2 h62dcd97_1 defaults m2w64-gcc-libgfortran 5.3.0 6 defaults m2w64-gcc-libs 5.3.0 7 defaults m2w64-gcc-libs-core 5.3.0 7 defaults m2w64-gmp 6.1.0 2 defaults m2w64-libwinpthread-git 5.0.0.4634.697f757 2 defaults markupsafe 1.1.1 py38he774522_0 defaults menuinst 1.4.16 py38he774522_0 defaults mistune 0.8.4 py38he774522_1000 defaults msys2-conda-epoch 20160418 1 defaults nbconvert 5.6.1 py38_0 defaults nbformat 5.0.7 py_0 defaults notebook 6.0.3 py38_0 defaults olefile 0.46 py_0 defaults openssl 1.1.1g he774522_0 defaults packaging 20.4 py_0 defaults pandoc 2.10 0 defaults pandocfilters 1.4.2 py38_1 defaults parso 0.7.0 py_0 defaults pickleshare 0.7.5 py38_1000 defaults pillow 7.2.0 py38hcc1f983_0 defaults pip 20.0.2 py38_3 defaults powershell_shortcut 0.0.1 3 defaults prometheus_client 0.8.0 py_0 defaults prompt-toolkit 3.0.5 py_0 defaults psutil 5.7.0 py38he774522_0 defaults pycosat 0.6.3 py38he774522_0 defaults pycparser 2.20 py_0 defaults pygments 2.6.1 py_0 defaults pyopenssl 19.1.0 py38_0 defaults pyparsing 2.4.7 py_0 defaults pyqt 5.9.2 py38ha925a31_4 defaults pyrsistent 0.16.0 py38he774522_0 defaults pysocks 1.7.1 py38_0 defaults python 3.8.3 he1778fa_0 defaults python-dateutil 2.8.1 py_0 defaults pytz 2020.1 py_0 defaults pywin32 227 py38he774522_1 defaults pywinpty 0.5.7 py38_0 defaults pyyaml 5.3.1 py38he774522_0 defaults pyzmq 19.0.1 py38ha925a31_1 defaults qt 5.9.7 vc14h73c81de_0 defaults qtpy 1.9.0 py_0 defaults requests 2.23.0 py38_0 defaults ruamel_yaml 0.15.87 py38he774522_0 defaults send2trash 1.5.0 py38_0 defaults setuptools 46.4.0 py38_0 defaults sip 4.19.13 py38ha925a31_0 defaults six 1.14.0 py38_0 defaults sqlite 3.31.1 h2a8f88b_1 defaults terminado 0.8.3 py38_0 defaults testpath 0.4.4 py_0 defaults tk 8.6.10 he774522_0 defaults tornado 6.0.4 py38he774522_1 defaults tqdm 4.46.0 py_0 defaults traitlets 4.3.3 py38_0 defaults urllib3 1.25.8 py38_0 defaults vc 14.1 h0510ff6_4 defaults vs2015_runtime 14.16.27012 hf0eaf9b_1 defaults wcwidth 0.2.5 py_0 defaults webencodings 0.5.1 py38_1 defaults wheel 0.34.2 py38_0 defaults win_inet_pton 1.1.0 py38_0 defaults wincertstore 0.2 py38_0 defaults winpty 0.4.3 4 defaults xmltodict 0.12.0 py_0 defaults xz 5.2.5 h62dcd97_0 defaults yaml 0.1.7 hc54c509_2 defaults zeromq 4.3.2 ha925a31_2 defaults zipp 3.1.0 py_0 defaults zlib 1.2.11 h62dcd97_4 defaults zstd 1.4.5 h04227a9_0 defaults ```
jayfurmanek commented 4 years ago

The delay may have to do with this: https://github.com/ContinuumIO/anaconda-issues/issues/11987

pzobel commented 4 years ago

Are there any updates yet? I would also like to use TF2.3 via Conda.

183amir commented 4 years ago

I would also like to see tensorflow 2.3 Linux version in the defaults channel even if the CPU version is available only for some time.

FarukEroglu2048 commented 4 years ago

Same for the CPU version of the TensorFlow, sadly Windows versions of both are still stuck at 2.1 - I would love to have 2.2+ through conda.

zehao-sean-huang commented 3 years ago

Same! Would love to see tensorflow gpu 2.3 working with conda!

jayfurmanek commented 3 years ago

I really don't mean to crash this issue, but a project I'm working on may help anyone (including Aanconda!) needing working tensorflow-gpu 2.3 conda recipes. It's called Open Cognitive Environment (https://github.com/open-ce) and we have conda recipes for many machine learning frameworks (including TensorFlow, which requires a few patches) that work with cpu and/or cuda. The recipes are Linux only and build on x86 and ppc64le and are easy to build using the included tools. Hope to see you there!

katietz commented 3 years ago

Thank you, @jayfurmanek. I will take a look how Open Cognitive Environment project has solved gpu variant. For me it has shown that I might need a more recent gcc toolchain for building this. But sure, I might overseen here soemthing. The gpu support on Ananconda's default channel is just turned of temporary. We intend to provide it ASAP.

183amir commented 3 years ago

Hi @jayfurmanek have you considered submitting your tensorflow recipe to conda-forge?

jayfurmanek commented 3 years ago

I could, but it may not help. TensorFlow is a bit difficult for conda-forge since it takes so long to build and forge mostly relies on free-tier CI build services which includes build time limits. That's also likely why the forge recipe is still on 1.14.

Note that in Open-CE we don't attempt any builds on CentOS6, only 7 (and now 8). In theory that shouldn't matter since the toolchain has its sysroot glibc which makes conda build fairly hermetic, but you never know - especially with TensorFlow.