Fix or disabled pyre issues that show up after D65753120

Summary: Fix or disabled pyre issues that show up after update to Pyre

Differential Revision: D65826206

captum/attr/_core/dataloader_attr.py

@@ -87,6 +87,8 @@ def _perturb_inputs(
         else:
             baseline = baselines[attr_inp_count]
 
+            # pyre-fixme[58]: `*` is not supported for operand types `object` and
+            #  `Tensor`.
             perturbed_inp = inp * pert_mask + baseline * (1 - pert_mask)
             perturbed_inputs.append(perturbed_inp)
 

captum/attr/_core/feature_ablation.py

@@ -806,6 +806,7 @@ def _construct_ablated_input(
             dim=0,
         ).long()
         current_mask = current_mask.to(expanded_input.device)
+        assert baseline is not None, "baseline must be provided"
         ablated_tensor = (
             expanded_input * (1 - current_mask).to(expanded_input.dtype)
         ) + (baseline * current_mask.to(expanded_input.dtype))

captum/attr/_core/occlusion.py

@@ -316,6 +316,7 @@ def _construct_ablated_input(
             ],
             dim=0,
         ).long()
+        assert baseline is not None, "baseline should not be None"
         ablated_tensor = (
             expanded_input
             * (

captum/metrics/_core/infidelity.py

@@ -619,6 +619,8 @@ def _next_infidelity_tensors(
         inputs_fwd = torch.repeat_interleave(
             inputs_fwd, current_n_perturb_samples, dim=0
         )
+        # pyre-fixme[58]: `-` is not supported for operand types `Tensor` and
+        #  `Union[Future[Tensor], Tensor]`.
         perturbed_fwd_diffs = inputs_fwd - inputs_perturbed_fwd
         attributions_expanded = tuple(
             torch.repeat_interleave(attribution, current_n_perturb_samples, dim=0)

captum/module/gaussian_stochastic_gates.py

@@ -133,7 +133,9 @@ def _get_gate_active_probs(self) -> Tensor:
             probs (Tensor): probabilities tensor of the gates are active
                 in shape(n_gates)
         """
-        x = self.mu / self.std
+        std = self.std
+        assert std is not None, "std should not be None"
+        x = self.mu / std
         return 0.5 * (1 + torch.erf(x / math.sqrt(2)))
 
     @classmethod

tests/attr/test_input_x_gradient.py

@@ -117,7 +117,9 @@ def _input_x_gradient_classification_assert(self, nt_type: str = "vanilla") -> N
             attributions = input_x_grad.attribute(input, target)
             output = model(input)[:, target]
             output.backward()
-            expected = input.grad * input
+            input_grad = input.grad
+            assert input_grad is not None
+            expected = input_grad * input
             assertTensorAlmostEqual(self, attributions, expected, 0.00001, "max")
         else:
             nt = NoiseTunnel(input_x_grad)