Lesson 1: name 'accuracy_np' is not defined

dhiman63 · February 16, 2018, 6:05am

I am getting an error for the execution of the following statement:

accuracy_np(probs, y)

The Error is:

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-75-98448a3af0e1> in <module>()
----> 1 accuracy_np(probs, y)

NameError: name 'accuracy_np' is not defined

Any help is appreciated!

Brad_S · February 16, 2018, 6:33am

have you run all the imports at the top? is everything else running?

go into the arguments and do shift-tab. Does it find anything?
it should - a file PATH\fastai\courses\dl1\fastai\metrics.py

is that on your machine?

laphi · February 16, 2018, 6:33am

That function should be defined in fastai/metrics.py. Maybe your local git repository is out of date? Try running git pull in your fastai directory to pull the latest changes.

github.com

fastai/fastai/blob/master/fastai/metrics.py

from .imports import *
from .torch_imports import *

def accuracy_np(preds, targs):
    preds = np.argmax(preds, 1)
    return (preds==targs).mean()

def accuracy(preds, targs):
    preds = torch.max(preds, dim=1)[1]
    return (preds==targs).float().mean()

def accuracy_thresh(thresh):
    return lambda preds,targs: accuracy_multi(preds, targs, thresh)

def accuracy_multi(preds, targs, thresh):
    return ((preds>thresh)==targs).float().mean()

You could also copy/paste this function from github into your notebook, althought this is a one time solution

dhiman63 · February 16, 2018, 6:42am

I am running in Google Collab!

Everything else above is running perfectly!

Brad_S · February 16, 2018, 6:52am

and this part?

dhiman63 · February 16, 2018, 7:03am

It gives: no docstring availble

I have just done !pip install --upgrade fastai

the file is not on the machine

navneetkrch · February 19, 2018, 7:59am

Were you able to solve the error?
Can you share details/ notebook?

dhiman63 · February 20, 2018, 4:16am

No not yet! not even sure what is the issue!

nikhil_no_1 · February 26, 2018, 1:59pm

I am having a related issue.
Check this out. Wiki: Lesson 1 - #183 by nikhil_no_1
I am using accuracy function.
This is the source for it.

def accuracy(preds, targs):
preds = np.argmax(preds, axis=1)
return (preds==targs).mean()

Moreover I replaced probs with log_preds where ever it gave axis error. Just wanted to run through it once. Will now go through each line step by step to understand if that is the right thing to do.

-Nikhil

dhiman63 · March 2, 2018, 5:08am

This post here seems to work for me! Thanks to @paultimothymooney!

Get the Collab Notebook here!