TCAV for tiny-bert model report AssertionError: `inputs` must have type torch.Tensor but <class 'list'> found: error happened

[joshua-xia]

🐛 Bug

when I using TCAV on tiny-bert model, the AssertionError: inputs must have type torch.Tensor but <class 'list'> found: error happened

To Reproduce

Steps to reproduce the behavior:

1. The model need three parameters to do predict:
   def predict(model, dataloader):
   model.eval()
   for idxs, input_ids, input_mask, segment_ids in dataloader:
   with torch.no_grad():
   logits = model(input_ids, segment_ids, input_mask, labels=None)

2. so the dataloader feed three parameters

3. when invoke compute_cavs， the following error happened:
   File "/devtools/anaconda3/envs/xai-env/lib/python3.8/site-packages/captum/concept/_core/tcav.py", line 511, in compute_cavs
   self.generate_activations(concept_layers)
   File "/devtools/anaconda3/envs/xai-env/lib/python3.8/site-packages/captum/concept/_core/tcav.py", line 388, in generate_activations
   self.generate_activation(concept_layers[concept], concept)
   File "/devtools/anaconda3/envs/xai-env/lib/python3.8/site-packages/captum/concept/_core/tcav.py", line 360, in generate_activation
   activations = layer_act.attribute.wrapped( # type: ignore
   File "/devtools/anaconda3/envs/xai-env/lib/python3.8/site-packages/captum/attr/_core/layer/layer_activation.py", line 118, in attribute
   layer_eval = _forward_layer_eval(
   File "/devtools/anaconda3/envs/xai-env/lib/python3.8/site-packages/captum/_utils/gradient.py", line 182, in _forward_layer_eval
   return _forward_layer_eval_with_neuron_grads(
   File "/devtools/anaconda3/envs/xai-env/lib/python3.8/site-packages/captum/_utils/gradient.py", line 445, in _forward_layer_eval_with_neuron_grads
   saved_layer = _forward_layer_distributed_eval(
   File "/devtools/anaconda3/envs/xai-env/lib/python3.8/site-packages/captum/_utils/gradient.py", line 294, in _forward_layer_distributed_eval
   output = _run_forward(
   File "/devtools/anaconda3/envs/xai-env/lib/python3.8/site-packages/captum/_utils/common.py", line 455, in _run_forward
   inputs = _format_input(inputs)
   File "/devtools/anaconda3/envs/xai-env/lib/python3.8/site-packages/captum/_utils/common.py", line 167, in _format_input
   return _format_tensor_into_tuples(inputs)
   File "/devtools/anaconda3/envs/xai-env/lib/python3.8/site-packages/captum/_utils/common.py", line 159, in _format_tensor_into_tuples
   assert isinstance(
   AssertionError: inputs must have type torch.Tensor but <class 'list'> found:

Expected behavior

get the cavs score correctly

Environment

Describe the environment used for Captum

torch                         1.12.1
captum                        0.5.0

## Additional context

<!-- Add any other context about the problem here. -->

[joshua-xia]

fix:

captum/_utils/common.py

def _format_tensor_into_tuples(
inputs: Union[None, Tensor, Tuple[Tensor, ...]]
) -> Union[None, Tuple[Tensor, ...]]:
if inputs is None:
return None
if isinstance(inputs, List):
inputs = tuple(inputs)
if not isinstance(inputs, tuple):
assert isinstance(
inputs, torch.Tensor
), "inputs must have type " "torch.Tensor but {} found: ".format(type(inputs))
inputs = (inputs,)
return inputs

[aobo-y]

discussed in #1077