Hi,
Could you simply add
with torch.no_grad():
in the example code? Otherwise it takes too much GPU memory as a normal user who simply need these embedings.
· Sign up or log in to comment
