I use AOTAutograd to produce FX graphs consisting of canonical Aten ops. It likes this.
%arange : [num_users=1] = call_function[target=torch.ops.aten.arange.start](args = (0, 13), kwargs = {dtype: torch.int64, device: cpu, pin_memory: False})
%unsqueeze : [num_users=1] = call_function[target=torch.ops.aten.unsqueeze.default](args = (%arange, 0), kwargs = {})
%view : [num_users=64] = call_function[target=torch.ops.aten.view.default](args = (%unsqueeze, [-1, 13]), kwargs = {})
...
For some reasons, I want to get intermediate results in these ops, such as the value of %unsqueeze. How can I get it? Thank you for your help.