Issue with load_learner

Enjoying the course. I’ve been using to do most of my work. I’ve noticed that when I run the notebook, I persistently cannot get load_learner to work on there, regardless of the notebook. Has anyone else had this issue? I had first cloned the fastbook repo and begun working. Anyone else had this issue? Attached is a screen shot…

Is fastai installed? And specifically fastai v2?

What does pip show fastai give.

1 Like

Yep! I’m using their fastai instance. This is what pip show fastai gave:

Name: fastai
Version: 2.0.13
Summary: fastai simplifies training fast and accurate neural nets using modern best practices
Author: Jeremy Howard, Sylvain Gugger, and contributors
License: Apache Software License 2.0
Location: /home/user/anaconda3/envs/fastai/lib/python3.8/site-packages
Requires: requests, pyyaml, spacy, pandas, torchvision, pillow, torch, matplotlib, packaging, fastprogress, fastcore, pip, scikit-learn, scipy
Required-by: fastbook

Hmmm… What version of fastcore?

pip show fastcore

Also you should try to match the exact versions of what you trained in for when you deploy (load it in). When dealing with fastai this means keeping track of three packages:

  • fastcore
  • fastai
  • torch

version 1.0.13 for fast core and 1.6 for torch. Your help is very appreciated!

Wow very old versions :slight_smile: (That’s September of last year!)

For some particular reason it’s not finding that you have fastcore installed, and I don’t know why. Did you always use this environment with those specific versions? (when training)

yeah I did. I’m wondering if I should just their stock ubuntu instance and install everything myself when I get in there each time.

Personally that’s what I do (I use colab so I do that each time). This way you get the most updated features. Again not particularly sure why it can’t find fastcore.basics when pip show fastcore returns what we want

can you share an example of what you usually do whenever you start your notebook? I’d just been cloning fastbook in to my home directory to work from there,

so on a fresh ubuntu machine we should just have to do the following if we’re using Anaconda, correct?

conda install -c fastai -c pytorch -c anaconda fastai gh anaconda