Fix unit test listing (#1574)

Summary:
Pull Request resolved: #1574

On a clean, stable fbcode checkout, running `buck test fbcode//pytorch/captum/tests/utils:test_progress` yields `NO TESTS RAN` consistently.

Looking at the TestX dashboard, some Captum tests had not been run since 5/22, which coincides with D75113906, e.g. [pytorch/captum/tests/utils:test_progress - test_nested_simple_progress](https://www.internalfb.com/intern/test/562950126652826?ref_report_id=0):
https://www.internalfb.com/intern/px/p/7nGLJ

My hypothesis is that the current `BaseTest` assignment in init confuses unit test listing/discovery, however not all the Captum unit tests that extend `BaseTest` are impacted... frankly not sure why. However the repro on the affected tests are consistent
Some example tests that are not affected: test_feature_permutation, test_common
Affected examples: test_linear_model, test_jacobian, test_infidelity

Running from fbcode stable, ~1500 tests run:
```
[sarahtran@8176.od /data/sandcastle/boxes/fbsource (e78e7566fd)]$ buck test fbcode//pytorch/captum/tests/...
File changed: fbcode//pytorch/captum/captum/testing/helpers/__init__.py
⃠ Omitted: pytorch/captum/tests/influence/fb/examples/simple:captum.influence.simple_example - main
Test has been omitted from this run because of the reason: Fails label filtering because: Negative filters present: slow
No missing positive labels.
Buck UI: https://www.internalfb.com/buck2/110b3a2a-43f5-4960-9025-108d1a854baf
Test UI: https://www.internalfb.com/intern/testinfra/testrun/15199648830072998
Network: Up: 728KiB  Down: 621KiB  (reSessionID-5fea945a-5dc6-43ad-b7f0-cd3d53656a22)
Executing actions. Remaining     0/243
Command: test.     Finished 127 cache (100% hit)
Time elapsed: 4:02.8s
Tests finished: Pass 1561. Fail 0. Fatal 0. Skip 1. Build failure 0
```

Reviewed By: jjuncho

Differential Revision: D75927248

fbshipit-source-id: aa95489178fadcb809f4d5884b4cbaedb3109081

Author: sarahtranfb

captum/testing/attr/helpers/attribution_delta_util.py

@@ -19,6 +19,8 @@ def assert_attribution_delta(
 ) -> None:
     if not is_layer:
         for input, attribution in zip(inputs, attributions):
+            # Pyre error in OSS environment only
+            # pyre-ignore[16]: `BaseTest` has no attribute `assertEqual`
             test.assertEqual(attribution.shape, input.shape)
     if isinstance(inputs, tuple):
         bsz = inputs[0].shape[0]
@@ -34,6 +36,8 @@ def assert_delta(
     test: BaseTest, delta: Tensor, delta_thresh: Union[Tensor, float] = 0.0006
 ) -> None:
     delta_condition = (delta.abs() < delta_thresh).all()
+    # Pyre error in OSS environment only
+    # pyre-ignore[16]: `BaseTest` has no attribute `assertTrue`
     test.assertTrue(
         delta_condition,
         "Sum of SHAP values {} does"

captum/testing/helpers/__init__.py

@@ -4,16 +4,14 @@
 
 try:
     from captum.testing.helpers.fb.internal_base import (  # type: ignore
-        FbBaseTest as _BaseTest,
+        FbBaseTest as BaseTest,
     )
 
 except ImportError:
     # tests/helpers/__init__.py:13: error: Incompatible import of "_BaseTest"
     # (imported name has type "type[BaseTest]", local name has type
     # "type[FbBaseTest]")  [assignment]
-    from captum.testing.helpers.basic import BaseTest as _BaseTest  # type: ignore
-
-BaseTest = _BaseTest
+    from captum.testing.helpers.basic import BaseTest  # type: ignore
 
 __all__ = [
     "BaseTest",