Howto: installation on Windows

Worked for me! Thanks so much

Hi lollcat, Thanks. So, you were able to use fast.ai on a windows with no NVidia GPU ? Can you please confirm?

Hi rramjee. No - I followed WakeMeAtThree’s post on installation with GPU that originally gave the error “PyTorch no longer supports this GPU because it is too old.” and have now successfully run lesson1.ipynb.

Oh ok sorry. Anyways thanks for your response. Good luck.

I performed all the given steps but the it is pops up an import error on fastai.imports and fastai.structured but not fastai

https://forums.fast.ai/t/pytorch-1-0-is-officially-released/32212/15

Should help.

It says “you dont have access to this”

Copying my steps here as the link was to the v3 course and is a locked forum.

This is the steps I’m using for my Conda environment. Windows 10.
conda create -n fastai-3.7 python=3.7
conda activate fastai-3.7
conda install pytorch torchvision -c pytorch
conda install -c fastai fastai
conda install nb_conda_kernels
python -m ipykernel install --user --name fastai-3.7 --display-name “Python (fastai-3.7)”
conda install ipywidgets

This works on my 1080ti - but I have seen the ‘too old’ message on other machines. GTX 480 and possibly Quadra mobile for example.

1 Like

It worked on my 1050Ti machine, Thanks. But fastai.structured is not being imported. This is the problem with kaggle too. Though i can simply copy the functions, i want to ask if there is a way that can be fixed

Can’t clone the repository…on running the command “git clone https://www.github.com/fastai/fastai.git” I get an error namely:

fatal: unable to access ‘https://github.com/fastai/fastai.git/’: SSL certificate problem: self signed certificate in certificate chain

Specs:
CPU : Intel i7-7700HQ @2.8GHz
GPU : Nvidia 1050Ti (GP 107M) 4GB
RAM : 16GB

1 Like

Hi,

Setup fastai using instructions given by Jeremy. Excellent instructions and worked perfectly!

I set it up for ML using the commands below

cd courses\ml1
del fastai
mklink /d fastai …\old\fastai
cd …

I opened the courses\ml1\lesson1-rf.ipynb and imported the fastai libraries successfully.

Then I ran the following commands and set it up for DL

cd courses\dl1
del fastai
mklink /d fastai …\old\fastai
cd …

I opened courses\dl1\lesson1.ipynb. When I tried to import the fastai libraries it failed to find fastai.imports. When I tried to switch to courses\ml1\lesson1-rf.ipynb and tried to import the libraries, it failed too.

Can someone please explain what happened?

Thanks for your help!

Refer to this

Thanks naveed but I could not really understand. Perhaps if you mark the end of each line or format it better, it would more readable?

Click the upward arrow button

I followed these steps to setup my environment but a lot of libraries like seaborn,bcolz, graphviz were not present.
I individually installed them but graphviz still wasn’t getting imported!!

Using the way specified by jeremy, I’m getting this error:

Is your Conda up to date? I recall something like this a while back.

Yes, my conda was up to date.
But I found another way to install and everything was running great.
I’m listing my steps:

  1. Removed previous Anaconda installation along with all registry files using Revo Uninstaller.
  2. Installed miniconda
  3. git clone https://github.com/fastai/fastai
  4. cd fastai
  5. conda create -n fastai python=3.6 anaconda
  6. conda env update
  7. activate fastai

Hopefully this may help others.

2 Likes

Hey,
I didn’t find a post regarding the stable windows version of Pytorch 1.0.

So Since Pytorch 1.0 is stable now, the windows user can install Pytorch via conda. Is it also possible for the fastai 1.0 libary? Or do I still have to build it from source?