change the int8 test to use unquantized bias (fp32)

Summary: change the test default to test the version we care about

Test Plan: ran the test

Reviewed By: amylittleyang

Differential Revision: D21725194

fbshipit-source-id: 243fcdf1dd5784768f6ceb2b46f9f1c9e64341eb

caffe2/contrib/fakelowp/test/test_int8_ops_nnpi.py

@@ -82,7 +82,7 @@ def _test_int8_quantize(self, n, m, rand_seed):
         m=st.integers(1, 1024),
         k=st.integers(1, 1024),
         rand_seed=st.integers(0, 65534),
-        quantize_bias=st.sampled_from([True]),
+        quantize_bias=st.sampled_from([False]),
     )
     def test_int8_fc(
         self, n, m, k, rand_seed, quantize_bias