[quant] Fix flaky test test_histogram_observer_against_reference (pyt…

…orch#46957)

Summary:
Pull Request resolved: pytorch#46957

Possibly due to use of large tensor in hypothesis. Reducing the size to see if it helps

Test Plan:
python test/test_quantization.py TestRecordHistogramObserver.test_histogram_observer_against_reference

Imported from OSS

Reviewed By: jerryzh168

Differential Revision: D24580137

fbshipit-source-id: f44ab059796fba97cccb12353c13803bf49214a1

test/quantization/test_workflow_module.py

@@ -736,7 +736,7 @@ def test_histogram_observer_same_inputs(self):
         self.assertEqual(myobs.max_val, 8.0)
         self.assertEqual(myobs.histogram, [2., 3., 3.])
 
-    @given(N=st.sampled_from([10, 1000, 10**6]),
+    @given(N=st.sampled_from([10, 1000]),
            bins=st.sampled_from([256, 512, 1024, 2048]),
            dtype=st.sampled_from([torch.qint8, torch.quint8]),
            qscheme=st.sampled_from([torch.per_tensor_affine, torch.per_tensor_symmetric]),