Logloss metric in Fastai

user_10697488 0
Viewed 431

i'm doing a competition in zindi plateform which they are using The evaluation metric for this challenge as Log Loss.

so i'm working with fastai library and i want the metric log loss .. i didn't find LogLoss as metric in this library ! i tried some codes like the function provided by sklearn from sklearn.metrics import log_loss but i didn't work
the link of the competition : https://zindi.africa/competitions/basic-needs-basic-rights-kenya-tech4mentalhealth

fast-ai pytorch metrics
1 Answers

if needed as a metric (typically mostly used as a loss) you should be able to use cross_entropy function from pytorch:

import torch.nn.functional as F

metrics=[F.cross_entropy,(plus other metrics if needed)]

model= cnn_learner(data, model, metrics=metrics,...)

user_33402990
Related