Getting load_learner not defined

fastai
1

Kindly find the response inline:
from fastai import *
from fastai.vision import *
from fastai.vision import image
model = load_learner("./weights/",‘mobilenetv2_8epochs_best.pkl’)

Was getting NameError: name ‘load_learner’ is not defined in fast ai version 2.0
for fast ai version 1.6 it is working fine

Kindly help

2

Hi

Since release fastai2 (few days ago) , name pip package “fastai” becomes fastai in version 2
and your code applies to version 1

You may install fastai version 1 directly from repo https://github.com/fastai/fastai1 if you have to.

pip install https://github.com/fastai/fastai1/archive/master.zip

But it is preferable to switch to version 2 (but you will have to migrate your network)

3

It’s not that it’s not defined, importing in fastai2 is different. Please read the documentation for examples. Also fastai2 is not backwards compatible with v1, so you should do as @Kornel, otherwise you’ll fail.

4

Hi, I am facing a similar issue.
I have a fastai text model that was trained from fastai v1. It is stored in Microsoft Azure Storage as a block blob and I have been using load_learner to load it from a file path and run it on a model in production.

from fastai import load_learner
path = ‘/dbfs/mnt/lake/users/helena_shi’
learn = load_learner(path, ‘final_whole’)
learn.predict(text)

I have installed fastai v1 with pip install https://github.com/fastai/fastai1/archive/master.zip but that only installs fastai v1.0.34 when I need 1.4.0 to use load_learner.
Alternatively, does fastai v2 have a similiar function to directly load a model from a path and predict with or do I need to retrain my existing model with the new syntax?
Please advise - thank you!

2 Likes
5

Did you ever get this figured out? I am using fastai v2.4. My model was trained on Google on 2-17-21 after the release of fastai2. I assume that means that my model was created using v2 but I am not positive.
Not sure if that makes a difference seeing as how the error is that load_learner is not defined which in my mind has nothing to do with creating the model as others have suggested.

6

I am facing same issue. but nobody has given any solution to this issue yet ?

7

I ended up reinstalling previous versions of fastai and torch to get around the upgrade, as I had models already trained with fastai v1.

!pip install torch==1.7.1
!pip install torchvision==0.1.8
!pip install fastai==1.0.60