This fixes Issue #27, where classification training crashes during compute_full_test_accuracy.
Masking out excess values of the last test batch from all_losses, all_y, and all_predictions would fail since those lists are empty and unused for classification. The fix was simply to wrap the relevant sections of code in if-statements that check if the problem type is something other than classification.
The test suite still runs without issue and running the CIFAR10 classification example with the data in test_data now finishes without crashing.
This fixes Issue #27, where classification training crashes during
compute_full_test_accuracy
.Masking out excess values of the last test batch from
all_losses
,all_y
, andall_predictions
would fail since those lists are empty and unused for classification. The fix was simply to wrap the relevant sections of code in if-statements that check if the problem type is something other than classification.The test suite still runs without issue and running the CIFAR10 classification example with the data in
test_data
now finishes without crashing.