Some inconsistencies in code vs. paper

[lisa-lthorrold]

Hi,

I node some inconsistencies, particular with the Hyperparameter selection in B5: chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://openreview.net/attachment?id=aExAsh1UHZo&name=supplementary_material

which mentions specific values for lambda zero and lambda one. According to loss.py (line 70,71). it seems a constant lambda is used for both, and also the lambda's are not divided by 2. Could you please clarify the discrepancies?