Thank you Jeremy and Sylvain for sharing your book for free. This is a nice act of generosity of both of you in sharing your knowledge with the risk of lowering a potential revenue. However, I think both your gesture and your daily contribution in democratizing deep learning inspires gratitude. I am confident that a lot of people (especially in this community), who can afford, will not hesitate to buy your book as an expression of gratitude towards you for making our deep learning journey more meaningful and more fun: I am one of them, and I am looking forward to have the paper version of your book in my hands.
Thank you again
Link to the book
Looks like the fastbook notebooks require fastai version2. Is there any post which gives details of installing fastai v2 and its dependencies on a virtual environment??
I suggest to view our most friendliest person muellerzr’s walkthru because he wrote few notebooks on fastaiv2 that can run on colab. He said the stable version of fastaiv2 comes out one June and the naming in fastaiv2 may be changed before it.
So far I got my 01 intro notebook for V2 code working that is for installing fastai v2 and its dependencies on a virtual environment:
Anyways, thx for Jeremy and Sylvain for sharing your book for free.
Hi, have you figured this out ? Can you please share.
They have added an requirement.txt.
You can just run pip install for it.
reading the book notebooks on windows will also need
I am using Python 3.6 within a virtual environment. On pip install requirements.txt I am getting an error
ERROR: No matching distribution found for requirements.txt
Should we give specific versions to overcome this error??
Maybe your problem is you did not direct into the folder, so I suggest this code:
!pip install -r /content/fastbook/requirements.txt
if it fails, you can try my installation:
check out my colab notebook: https://github.com/JonathanSum/Fastbook_colab
You can just try to run all the install I did on notebook1 to see it can solve the problem or not.
That was indeed helpful. However still I am encountering the following error:
ERROR: No matching distribution found for torch>=1.3.0 (from fastai2>=0.0.11->-r F:\fastbook\requirements.txt (line 1))
I am using python 3.6. Is it an issue with the python version?
It says the Pytorch that you are using is missing or not in the version that is 1.30 or above.
You need to install Pytorch following this website:
Run this code if you are in window with conda: conda install pytorch torchvision cudatoolkit=10.1 -c pytorch
If that doesn’t work, try these codes:
!pip install -q feather-format kornia pyarrow wandb nbdev fastprogress fastai2 fastcore --upgrade
!pip install torch==1.3.1
!pip install torchvision==0.4.2
!pip install Pillow==6.2.1 --upgrade
It is from muellerzr’s walk thru, and I also put them in the notebook that I showed you. Didn’t you try it?
I added this single cell on the first notebook to get it running on colab:
Hi, I started reading the fastbook and there was reference to https://book.fast.ai/ for nstalling the GPUs. This link seems to be broken. Can someone please help me here?
It will be up once the book is “officially” release most likely in July
in google colab use this
!pip install fastai2
and you are good to go
Using Google Colab with the
!pip install fastai2 snippet you provided seems like a great way to get going until “the instructions to get connected to a GPU deep learning server” are up on https://book.fast.ai
My copy just arrived today (), so I assume others will be running into this soon. If the above isn’t enough detail, I believe there are more detailed instructions for working with Colab available from @muellerzr floating around on the forums as well - at least until https://book.fast.ai has official instructions.
We’re working on that now, and should be pretty easy to install and get it going
For the time being, a (ish clunky) way to get it going is by doing:
!wget -O https://raw.githubusercontent.com/fastai/course-v4/master/nbs/images/grizzly.jpg 'images/grizzly.jpg'
!pip install -r requirements.txt
Eventually this will become easier (like a
sh script, similar to course-v3)