Suppose we have a torch model like this
from script_class import CustomClass
class Model(nn.Module):
def __init__(self):
super(Model, self).__init___()
self.object = CustomClass()
def forward(self):
print('hello')
Then i script the model and save it as ‘model.pt’
Next when I load it without import CustomClass and run it, it will crash with unreadable c++ errors
How do I bake the class into the scripted model, something like torch.jit.export but for data structures?