I tried you code in mnist dataset, but I got NaN error

[DeepBlurt]

No description provided.

[DeepBlurt]

Traceback (most recent call last):
File "train.py", line 43, in
model.train()
File "/home/ax/Anomaly/GANomaly/lib/model.py", line 280, in train
res = self.test()
File "/home/ax/Anomaly/GANomaly/lib/model.py", line 354, in test
auc = evaluate(self.gt_labels, self.an_scores, metric=self.opt.metric)
File "/home/ax/Anomaly/GANomaly/lib/evaluate.py", line 24, in evaluate
return roc(labels, scores)
File "/home/ax/Anomaly/GANomaly/lib/evaluate.py", line 49, in roc
fpr, tpr, _ = roc_curve(labels.numpy(), scores.numpy())
File "/home/ax/anaconda3/lib/python3.6/site-packages/sklearn/metrics/ranking.py", line 622, in roc_curve
y_true, y_score, pos_label=pos_label, sample_weight=sample_weight)
File "/home/ax/anaconda3/lib/python3.6/site-packages/sklearn/metrics/ranking.py", line 402, in _binary_clf_curve
assert_all_finite(y_score)
File "/home/ax/anaconda3/lib/python3.6/site-packages/sklearn/utils/validation.py", line 72, in assert_all_finite
_assert_all_finite(X.data if sp.issparse(X) else X, allow_nan)
File "/home/ax/anaconda3/lib/python3.6/site-packages/sklearn/utils/validation.py", line 56, in _assert_all_finite
raise ValueError(msg_err.format(type_err, X.dtype))
ValueError: Input contains NaN, infinity or a value too large for dtype('float32')

[DeepBlurt]

I guess the reason lies in the version of sklearn, or pytorch, But I don't know the exact version you used, waiting for your answer!
-------------------update-------------------
I down grade the version of sklearn, the error evaded, but the AUC result got nan:
Avg Run Time (ms/batch): 8.383 AUC: nan max AUC: 0.000

Could you please point out the reason?

[lzzlxxlsz]

have you run the code successufully? and how about the result?

[DeepBlurt]

have you run the code successufully? and how about the result?

I have run successfully, the result differs from the article, I got following AUC:

0.7897 , 0.4115 , 0.7857 , 0.6432 , 0.6759 , 0.7587, 0.745 , 0.5428 , 0.8484 , 0.5741
in MNIST, does this result make sense? Can you offer your result?

I modified the train step in code, because if not, it does not do any training, thus got NaN error. I make train step equal to ratio of number of samples and batch size per epoch.

[lzzlxxlsz]

sorry,I don't run this code.and does the result be similar with the test image?

[Funple]

have you run the code successufully? and how about the result?

I have run successfully, the result differs from the article, I got following AUC:

0.7897 , 0.4115 , 0.7857 , 0.6432 , 0.6759 , 0.7587, 0.745 , 0.5428 , 0.8484 , 0.5741
in MNIST, does this result make sense? Can you offer your result?

I modified the train step in code, because if not, it does not do any training, thus got NaN error. I make train step equal to ratio of number of samples and batch size per epoch.

Could you provide your changed code?

[DeepBlurt]

I find this code not stable, then I tried this code:
https://github.com/caiya55/ganomaly-tensorflow , you can have a try, this code is much stable and easy to understand.

[Funple]

Thank you for your effective reply.