Unable to install fastai on GCloud


I just finished the first part of the 2019 course. I looked around for a few projects and text summarization seemed pretty cool. I’m working on this model to summarize twitter conversations related to anime.

However, I’m facing some problems setting up fastai on my GCloud server. I used a conda env to install jupyter and the IPython kernel. When I try to install fastai, my PyTorch v1.x installation is removed and 0.4 is installed. So, I’m currently writing in PyTorch.

I have fastai 0.7. When I run

from fastai.vision import *

I get the following error

ImportError: No module named ‘fastai.vision’

I get this error when I try to import fastai.text

re_tok = re.compile(f’([{string.punctuation}“”¨«»®´·º½¾¿¡§£₤‘’])’

What could be the reason? And, how do I fix this?

Thank you :slight_smile:

The v3 course uses fastai v1. Have a look here for a full guide for GCP: https://course.fast.ai/start_gcp.html

I have gone through that page. But, the problem is that the instance I’m using right now was already setup. It’s more like a hand-down. So, I don’t know which image (PyTorch/Tensorflow) it has built into it.

Hi Pranesh,

I am facing a similar problem when installed fastai library on AWS Sagemaker at my work place.

Just enter in the CLI:
pip uninstall fastai
conda install -c pytorch -c fastai fastai