Fix missing output_attentions in PT/Flax equivalence test (#16271)

* fix - set output_attentions to True

* Update tests/test_modeling_flax_common.py

* update for has_attentions

* overwrite check_outputs in FlaxBigBirdModelTest
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
Co-authored-by: Suraj Patil <surajp815@gmail.com>

tests/big_bird/test_modeling_flax_big_bird.py

@@ -190,3 +190,12 @@ class FlaxBigBirdModelTest(FlaxModelTesterMixin, unittest.TestCase):
                 for jitted_output, output in zip(jitted_outputs, outputs):
 
                     self.assertEqual(jitted_output.shape, output.shape)
+
+    # overwrite from common in order to skip the check on `attentions`
+    def check_outputs(self, fx_outputs, pt_outputs, model_class, names):
+        # `bigbird_block_sparse_attention` in `FlaxBigBird` returns `attention_probs = None`, while in PyTorch version,
+        # an effort was done to return `attention_probs` (yet to be verified).
+        if type(names) == str and names.startswith("attentions"):
+            return
+        else:
+            super().check_outputs(fx_outputs, pt_outputs, model_class, names)

tests/test_modeling_flax_common.py

@@ -120,6 +120,7 @@ class FlaxModelTesterMixin:
     test_mismatched_shapes = True
     is_encoder_decoder = False
     test_head_masking = False
+    has_attentions = True
 
     def _prepare_for_class(self, inputs_dict, model_class):
         inputs_dict = copy.deepcopy(inputs_dict)
@@ -168,6 +169,7 @@ class FlaxModelTesterMixin:
             dict_inputs = self._prepare_for_class(inputs_dict, model_class)
             check_equivalence(model, tuple_inputs, dict_inputs, {"output_hidden_states": True})
 
+    # (Copied from tests.test_modeling_common.ModelTesterMixin.check_outputs)
     def check_outputs(self, fx_outputs, pt_outputs, model_class, names):
         """
         Args:
@@ -204,8 +206,7 @@ class FlaxModelTesterMixin:
             pt_outputs[pt_nans] = 0
             fx_outputs[pt_nans] = 0
 
-            max_diff = np.amax(np.abs(fx_outputs - pt_outputs))
-            self.assertLessEqual(max_diff, 1e-5)
+            self.assert_almost_equals(fx_outputs, pt_outputs, 1e-5)
         else:
             raise ValueError(
                 f"`fx_outputs` should be a `tuple` or an instance of `jnp.ndarray`. Got {type(fx_outputs)} instead."
@@ -222,6 +223,7 @@ class FlaxModelTesterMixin:
 
                 # Output all for aggressive testing
                 config.output_hidden_states = True
+                config.output_attentions = self.has_attentions
 
                 # prepare inputs
                 prepared_inputs_dict = self._prepare_for_class(inputs_dict, model_class)
@@ -274,7 +276,7 @@ class FlaxModelTesterMixin:
 
                 # Output all for aggressive testing
                 config.output_hidden_states = True
-                # Pure convolutional models have no attention
+                config.output_attentions = self.has_attentions
 
                 # prepare inputs
                 prepared_inputs_dict = self._prepare_for_class(inputs_dict, model_class)
@@ -314,6 +316,7 @@ class FlaxModelTesterMixin:
 
                 # send pytorch model to the correct device
                 pt_model_loaded.to(torch_device)
+                pt_model_loaded.eval()
 
                 with torch.no_grad():
                     pt_outputs_loaded = pt_model_loaded(**pt_inputs)