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
Is your feature request related to a problem? Please describe.
Right now compilation will fail with torchtrtc if you use custom operators, i.e. ops defined by libraries other than torch. This is because the PyTorch operator registry is missing information about these operators and so when trying to deserialize, TorchScript does not know what to do
This includes torchvision and torch_squeeze (#876 )
Describe the solution you'd like
Be able to add paths to .so files that need to be DL_OPEN before compilation so that custom ops are available for deserialization.
Describe alternatives you've considered
Right now the work around is to use C++ or Python with the libraries imported by the user
Additional context
The text was updated successfully, but these errors were encountered:
Is your feature request related to a problem? Please describe.
Right now compilation will fail with
torchtrtc
if you use custom operators, i.e. ops defined by libraries other thantorch
. This is because the PyTorch operator registry is missing information about these operators and so when trying to deserialize, TorchScript does not know what to doThis includes
torchvision
andtorch_squeeze
(#876 )Describe the solution you'd like
Be able to add paths to .so files that need to be
DL_OPEN
before compilation so that custom ops are available for deserialization.Describe alternatives you've considered
Right now the work around is to use C++ or Python with the libraries imported by the user
Additional context
The text was updated successfully, but these errors were encountered: