I need to have probability score for each testing sample. What changes I should make in following code to get those probability score for each sample
eta = nn.functional.softmax(logits,-1)
this gives you probability of each class given sample x
(P(i|x)).
this gives you probability of each class for each sample.
can you give me an example of what you want
it should work.
replace
for imageName,realLabel,actualLabel,logitValue in zip(tested_file_name,real_label,actual_pred,logits):
with
for imageName,realLabel,actualLabel,logitValue in zip(tested_file_name,real_label,actual_pred,eta):
some where in your for loop print it.
something is off
how many class do you have ?
what is you batch size??
eta should be NxC
and eta.sum(-1) = 1
can you give me eta shape.
and what you’re printing in for loop
in for loop you should print logitValue
not eta
itself.
but still something is off,
you should have eta shape as 16x9??
also all of tested_file_name,real_label,actual_pred,logits
should have 16 element or have 16 row (if they’re matrices).
okay, i have printed logitvalue
it’s better now.
i think you can report predicted label probability as confidence.
in your example >>>> 0.1517