Q7 - Taking too long to run validation_curve on data

M2. Selecting the best model M2. Wrap-up quiz 2
#1

Hello,

Is there a way to speed up the cross-validation that uses parameter range? I’m using n_jobs=-1 to speed up but the cell is still running and it’s been more than 30 minutes.

#2

Hmmm weird maybe it is related to Very slow cross validation and accuracy reproducibility - #7 by lesteve ?

#3

I’ll try the proposed solution. It might be related.

#4

For the record the full correction of this wrap-up quiz takes less than two seconds for me inside the FUN Jupyter.

#5

Hmm, I stopped and restarted the Jupyter Server. It looks like it did not help.

#6

It worked after restarting again. :slight_smile:

#7

Great the good old “turn off and on again” but multiple times :wink:

1 Like
#8

Hi! I hope you manage to solve it. In my case, I solved the same issue by naming the parameter to tune correctly (get_params())
Oddly enough, in such case, the notebook keeps running rather than returning an error

3 Likes
#9

same with me, this was the solution param_name=“kneighborsclassifier__n_neighbors”

For some reason when it is wrong instead of raising an error it keeps running forever

1 Like
#10

Thanks for the report, I can reproduce indeed. Looking a bit in more details, this is due to an IPython bug https://github.com/ipython/ipython/issues/12467.

We should probably use IPython < 8 until for the MOOC until this IPython bug is fixed.

#11

This has now been done on the FUN Jupyter notebooks, you may need to restart your server by doing something along these lines:

2 Likes
#12

Can you share the block of code for using validation_curve

I am doing it like this but get an error that it takes 3 arguments and am passing 5.

neighbors_range = [1, 2, 5, 10, 20, 50, 100, 200, 500]
model = make_pipeline(numerical_processor,KNeighborsClassifier())
from sklearn.model_selection import validation_curve
train_scores,test_scores = validation_curve(
KNeighborsClassifier(),
data,
target,
“n_neighbors”,
neighbors_range)
train_scores_mean = train_scores.mean()
test_scores_mean = test_scores.mean()