justforfunc live: getting started with Tensorflow Go

preview_player
Показать описание

How hard can it be? Well, basically ended up writing a hello world of Tensorflow in over an hour.

Most of the time I spent was on Dockerfiles and setting things up ... but at least next time everything will be ready!
Рекомендации по теме
Комментарии
Автор

This is extremely painful, I went through this same nightmare. It shouldn't be this hard. 😢

vdemario
Автор

The python dependency hell. Thats one reason why we love go :-)

PapaNoah
Автор

This is longer and complicated than usual episodes but to be honest that make it even more intersting. Hope to see more like this ;)

kianostad
Автор

Hi Francesc, You can easily install pip packages in macOS with $ pip install --user tensorflow (with no sudo). It installs the packages in your home directory which you have permisions ('--user' it solves lots of the 'OperationNotPermited' issues) If you want to install system-wide pip packages they may interfere with the existing pip system packages and probably break everything, which is bad :(
Great videos BTW!! :)

vtomasr
Автор

I really dig your reaction at 18:40 "sudo not found". I think I saw a silent punch to the computer.

dirkolbrich
Автор

Do more live sessions. It's really educational to see how you solve the problems, not just follow the correct path from the beginning.

lNLl
Автор

I think the problem is the python version that comes with osx. It is easier if you install a python version with homebrew and then install tensorflow with pip. Also to install python packages with user rights, you want to use the --user flag and not sudo change parts of the global python installation.

JohannesHeuel
Автор

many people get suck before they even can do the cool stuffs, Thanks and hope to see the next video soon

andreaswestberg
Автор

Yet another great video. Thank you.

A question: Is the source code for this accessible ?

zveinn
Автор

Francesc you should of gone the virtualenv route. Python is too coupled to your system, and developing anything with dependencies on your machine is just a recipe to mess up your system's python. The virtualenv lets you build a completely decoupled python environment from your machine and you could install the numpy you needed in that virtualenv. #pythonproblems#archaic

daviddesmarais-michaud
Автор

Try writing pip install tensorflow please at 28:20

chaddibox
Автор

The thing is tf Variable is a python thing. Which is not available in C/C++, so no bindings for Go

iamravone
Автор

Oh the joy of yak shaving: homebrew is owned by another user, and that's why you faced so many errors there, and they might not ended even if you chown /usr/local/Homebrew. Switching to docker was the right choice. I would have set up a VM instead, but it's just a matter of taste (and "de gustibus non est disputandum", ain't it?). Thanks for the video.

julian
Автор

I want more live/video like this, want to see more troubling with tensorflow-go (I am sure! you will find the solution). since most of live streamers/uploaders want to hide even though some people might have same problem....

applePrincess
Автор

You should post on twitter and reddit as well.

kshitijkumar
Автор

It's maybe time to try a real package manager, with a real OS. I had only to do `sudo pacman -Sy python python-pip && pip install --user tenserflow`. And all was fine. MacOS is a fucking garbage OS useless.

But I appreciated to see you try hard to setup your env, that was fun xD.
Thanks for all your job.

pmartin
Автор

for work with python must have iron egg skill :)

rootruslan