You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When using the NNsight class to wrap pre-existing pytorch modules (e.g. CLIP-ViT) where positional arguments are optional, users are currently forced to provide a positional argument (e.g. inputs).
Positional arguments to the NNsight forward should be made optional when the underlying model allows for it.
See discord discussion about this issue for a CLIP-VIT model: Discussion link
The text was updated successfully, but these errors were encountered:
When using the
NNsight
class to wrap pre-existing pytorch modules (e.g. CLIP-ViT) where positional arguments are optional, users are currently forced to provide a positional argument (e.g. inputs).Positional arguments to the
NNsight
forward should be made optional when the underlying model allows for it.See discord discussion about this issue for a CLIP-VIT model: Discussion link
The text was updated successfully, but these errors were encountered: