AntreasAntoniou/Learning_to_Learn_via_Self-Critique

Output size is too small ...

Opened this issue · 5 comments

I followed the instructions to install and run the provided scripts for miniimagenet. The standard ones go through but the intrinsic ones prompt:

/few_shot_learning_system.py", line 571, in forward
support_set_features.shape[-1]).squeeze()
RuntimeError: Given input size: (48x10x10). Calculated output size: (48x0x0). Output size is too small

Can you confirm that you are using the default experiment config files? If so, have you made any changes to the code? If not, I'll need to investigate this. I triple tested this, but sometimes bugs can still make it through.

I am using the default ones, e.g.:
sh mini-imagenet_embedding_variant_intrinsic_5_way_5_shot_48_preds_20_LSLR_conditioned_0_few_shot.sh 0

I followed the instructions step by step and never changed any file.

Ok, thank you for bringing this up. I shall investigate at once.

Ok, I have solved this issue. It was an error stemming from a last-minute variable rename that I did. Apologies.

Let me know if it works now.