search

DeepLabCut / DeepLabCut-live

SDK for running DeepLabCut on a live video stream
https://elifesciences.org/articles/61909
Other
190 stars 50 forks source link

Do I still need to specify `python=3.7` and `tensorflow-gpu==1.13.1` when creating the conda environment? #45

Closed jonmatthis closed 3 years ago

jonmatthis commented 3 years ago

The conda environment creation instructions at https://github.com/DeepLabCut/DeepLabCut-live/blob/master/docs/install_desktop.md say that I should use this command to create the environemnt -

conda create -n dlc-live python=3.7 tensorflow-gpu==1.13.1

...but I recall seeing elsewhere in the documentation that DLC now supports Tensorflow 2.x . I see that this page was last updated in January, so I figure it might have been written before that support was added

Do I still need to specify tensorflow-gpu==1.13.1 when creating this conda environment?

And while we're at it - Is python=3.7 still correct as well?

Thank you!!

MMathisLab commented 3 years ago

Python 3.7 yes, tensorflow, no - whatever version you exported your model(s) with is what you should use :)

jonmatthis commented 3 years ago

Great, thanks!

FYI, though - When I removed the tensorflow-gpu==1.13.1 bit from the deeplabcut installation, it produced an error related to SciPy: 

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
tensorflow 2.3.0 requires scipy==1.4.1, but you have scipy 1.5.4 which is incompatible.
AlexEMG commented 3 years ago

Then just downgrade the scipy (although likely causes no problems)