Add support for module input of type list #56
Description
Wonderful project! I tried some baseline models and it worked well.
However, it seems that it does not support modules which take only one input argument of type list. For instance:
def forward(self, feats: list[Tensor]):
assert len(feats) == len(self.in_channels)
proj_feats = [self.input_proj[i](feat) for i, feat in enumerate(feats)]
....And if I put this specific model to torchexplorer, I get the following error:
File "/root/miniconda3/envs/rtdetr/lib/python3.10/site-packages/torchexplorer/hook/hook.py", line 329, in process_tensor
return tensor + dummy_tensor if torch.is_floating_point(tensor) else tensor
TypeError: is_floating_point(): argument 'input' (position 1) must be Tensor, not listI tried to modify the hooks but it did not work out.
Any suggestions?
@spfrommer