So i’m a completely noob and i just finished lesson 1.
For those who don’t want or can’t afford to rent AWS server, i suggest you do it on your own PC.
i’m running on windows 10 64 bit, the tutorial run on ubuntu server and you know what ?
I’m a gamer, gamer don’t use ubuntu or linux.
i don’t know anything about ubuntu or linux. so i will explain everything i can to help you go through with windows
First you need to download anaconda for python 2.7, CUDA toolkit and CUDNN. for those who are using AMD GPU, forget it.
Here is the link for anaconda
https://www.continuum.io/downloads
Nvidia toolkit 8.0
http://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/
CUDNN v5.1
https://developer.nvidia.com/cudnn
after you install anaconda, open the windows command prompt or type CMD.
type
conda create -n nooblearnAI python=2.7
after that active the environment by typing
activate nooblearnAI
congrat you have your first environment.
next step to install the prerequisite library.
http://wiki.fast.ai/index.php/Python_libraries
you install those lib by typing pip install theano or conda install theano.
Remember to install jupyter notebook too by typing pip install jupyter notebook.
For some reason the latest version of keras does not work. So you have to install keras1.1.
pip install 'keras<2'
thanks to @ville .
ok so we good to go. now in our env type jupyter notebook to open the ipython notebook.
For lesson 1 dog and cat, i suggest you to download the data from this site
http://files.fast.ai/files/dogscats.zip
and the course material too
https://github.com/fastai/courses
after you download the dog and cat files, create the data folder inside the course material folder
and unzip the files.
now open the jupyter notebook, try to run every cell from top to bottom, if you got error like no module named… don’t worry, no big deal, i have that error like 100 times and i don’t give up. go back to command prompt, active the environment and type pip install the module name again.
i will try to write the next tutorial when i have time.
but some of my word. Deep learning is hard, I make a lot of mistake just to run the lesson 1. it take me half month to figured out how to run it. So don’t give up. i quit school when i was in high school, so no bachelor degree ? don’t worry, no PhD ? who care ? you still can learn, it just take longer.