Viveka Kulharia

Implementation of 2nd order gradient in maml

1

In line 126 of few_shot/maml.py: `meta_batch_loss.backward()` however, the graph for meta_batch_loss does not use model parameters. Each loss used to get meta_batch_loss was obtained using fast_weights. So while doing backwards...

model.eval for temperature scaling?

1

Hi, I was wondering that as dropout is used, shouldn't we be using original_model.eval() for temperature scaling? As the logits obtained using one forward pass is used to get temperature...

Front end should also be updated while updating Q net