[FlaxWav2Vec2Model] Fix bug in attention mask (#16725)

* [FlaxWav2Vec2Model] Fix bug in attention mask

* more fixes

* add (Flax)SpeechEncoderDecoderModel PT-FX cross-test

src/transformers/models/wav2vec2/modeling_flax_wav2vec2.py

@@ -920,7 +920,7 @@ class FlaxWav2Vec2PreTrainedModel(FlaxPreTrainedModel):
     def _get_feat_extract_output_lengths(
         self, input_lengths: Union[jnp.ndarray, int], add_adapter: Optional[bool] = None
     ):
-        return self.module._get_feat_extract_output_lengths(input_lengths)
+        return self.module._get_feat_extract_output_lengths(input_lengths, add_adapter=add_adapter)
 
 
 class FlaxWav2Vec2Module(nn.Module):
@@ -956,15 +956,10 @@ class FlaxWav2Vec2Module(nn.Module):
 
         # make sure that no loss is computed on padded inputs
         if attention_mask is not None:
-            # compute real output lengths according to convolution formula
-            output_lengths = self._get_feat_extract_output_lengths(attention_mask.sum(-1).astype("i4"))
-
-            attention_mask = jnp.zeros(extract_features.shape[:2], dtype=self.dtype)
-
-            # these two operations makes sure that all values
-            # before the output lengths indices are attended to
-            attention_mask = attention_mask.at[jnp.arange(attention_mask.shape[0]), output_lengths - 1].set(1)
-            attention_mask = jnp.flip(jnp.flip(attention_mask, -1).cumsum(-1), -1).astype("bool")
+            # compute reduced attention_mask corresponding to feature vectors
+            attention_mask = self._get_feature_vector_attention_mask(
+                extract_features.shape[1], attention_mask, add_adapter=False
+            )
 
         hidden_states, extract_features = self.feature_projection(extract_features, deterministic=deterministic)
         if mask_time_indices is not None:  # apply SpecAugment along time axis with given indices
@@ -1034,12 +1029,10 @@ class FlaxWav2Vec2Module(nn.Module):
         batch_size = attention_mask.shape[0]
 
         attention_mask = jnp.zeros((batch_size, feature_vector_length), dtype=attention_mask.dtype)
-        # these two operations makes sure that all values before the output lengths idxs are attended to
-        idx = (jnp.arange(attention_mask.shape[0]), output_lengths - 1)
-        attention_mask = attention_mask.at[idx].set(1)
-        attention_mask = jnp.flip(jnp.flip(attention_mask, axis=-1).cumsum(axis=-1), axis=-1)
-
-        attention_mask = jnp.array(attention_mask, dtype=bool)
+        # these two operations makes sure that all values
+        # before the output lengths indices are attended to
+        attention_mask = attention_mask.at[jnp.arange(attention_mask.shape[0]), output_lengths - 1].set(1)
+        attention_mask = jnp.flip(jnp.flip(attention_mask, -1).cumsum(-1), -1).astype("bool")
         return attention_mask
 
 
@@ -1286,11 +1279,15 @@ class FlaxWav2Vec2ForPreTrainingModule(nn.Module):
             attentions=outputs.attentions,
         )
 
-    def _get_feat_extract_output_lengths(self, input_lengths: Union[jnp.ndarray, int]):
+    def _get_feat_extract_output_lengths(
+        self, input_lengths: Union[jnp.ndarray, int], add_adapter: Optional[bool] = None
+    ):
         """
         Computes the output length of the convolutional layers
         """
 
+        add_adapter = self.config.add_adapter if add_adapter is None else add_adapter
+
         def _conv_out_length(input_length, kernel_size, stride):
             # 1D convolutional layer output length formula taken
             # from https://pytorch.org/docs/stable/generated/torch.nn.Conv1d.html
@@ -1299,6 +1296,10 @@ class FlaxWav2Vec2ForPreTrainingModule(nn.Module):
         for kernel_size, stride in zip(self.config.conv_kernel, self.config.conv_stride):
             input_lengths = _conv_out_length(input_lengths, kernel_size, stride)
 
+        if add_adapter:
+            for _ in range(self.config.num_adapter_layers):
+                input_lengths = _conv_out_length(input_lengths, 1, self.config.adapter_stride)
+
         return input_lengths
 
 

tests/speech_encoder_decoder/test_modeling_flax_speech_encoder_decoder.py

@@ -539,6 +539,12 @@ class FlaxEncoderDecoderMixin:
         self.check_equivalence_pt_to_flax(config, decoder_config, inputs_dict)
         self.check_equivalence_flax_to_pt(config, decoder_config, inputs_dict)
 
+        # check `add_adapter` works as expected
+        config.add_adapter = True
+        self.assertTrue(config.add_adapter)
+        self.check_equivalence_pt_to_flax(config, decoder_config, inputs_dict)
+        self.check_equivalence_flax_to_pt(config, decoder_config, inputs_dict)
+
     @slow
     def test_real_model_save_load_from_pretrained(self):
         model_2 = self.get_pretrained_model()