Hi, sorry I forgot to put the link to my medium post! All my code is in screenshot if you read the medium post. Thanks!
1 Like
Hello, I’m new around here and have watched a few of the videos, am hooked, and I think what you’re up to here is amazing, so thank you!
I’ve got a question about the fastai library version. I’ve set up my own ubuntu dual-boot workspace, putting my 1080ti to more productive use than what I usually use it for… and have the old/0.7 version of the notebooks running smoothly. I’m sure that the general deep learning concepts from the now old version of the course are still just as valuable and I will go through them regardless, however I’m wondering about if at some point we can look forward to a new or updated version using the fastai 1.0 libray, considering it’s a from scratch rewrite? Do you think there is any harm at all in getting deep into the 0.7 course at this point? (just to be clear when I say old, I still mean the 2018 version, I say old because of the “old” wrapper version used in the course)
Yes, early 2019.
Still very useful - the concepts don’t change, and that’s what takes time to learn. Library changes are easy to learn.
1 Like
Thank you for your work!!
HI - thank you for all the work you have done. I am just starting and have seen the intro to Random Forests lesson 1. I am not sure how to proceed with library/version changes. I use the following code that I found in the forum to install fasai. Some of the libraries work, but I cannot get “from fastai.structured import *” (interested in the add_datepart function) to install. I see in the forum that the library has moved, but I am not sure what I need to do to access this “old” library.
#installs FastAI
#https://github.com/fastai/course-v3/blob/7dfe63de54daf2210f5b23f237e70ee246be5d0c/docs/google_colab_tutorial.md
from os.path import exists
from wheel.pep425tags import get_abbr_impl, get_impl_ver, get_abi_tag
platform = ‘{}{}-{}’.format(get_abbr_impl(), get_impl_ver(), get_abi_tag())
cuda_output = !ldconfig -p|grep cudart.so|sed -e ‘s/..([0-9]).([0-9]*)$/cu\1\2/’
accelerator = cuda_output[0] if exists(’/dev/nvidia0’) else ‘cpu’
!pip install torch_nightly -f https://download.pytorch.org/whl/nightly/{accelerator}/torch_nightly.html
!pip install fastai==1.0.28
import torch
print(torch.version)
print(torch.cuda.is_available())
print(torch.backends.cudnn.enabled)
I want to install fastai v0.7 using conda because I want to try out the Introduction to machine learning course.
Currently, on installing fast ai, I get an error that module structured not found.
Is there any way around ?
I cloned the fastai.git directory. I try to do conda env update on environment-cpu.yml. It is giving me a bunch of conda verification errors like CondaVerificationError: The package for numpy-base located at C:\Users\ssengupta
\AppData\Local\Continuum\anaconda2\pkgs\numpy-base-1.16.2-py36hc3f5095_0
appears to be corrupted. The path ‘Scripts/f2py.exe’
specified in the package manifest cannot be found.
Hey, for anyone who had difficulty running the notebooks on a windows 10 machine, this is what worked for me:
…Documents\Repo\fastai\courses\ml1>mklink /d fastai …\old\fastai
The existing symlink worked through cmd but not from jupyter.
The symlink created above did the trick. Hope that helps.