Internal change

PiperOrigin-RevId: 406906304

official/nlp/modeling/models/seq2seq_transformer.py

@@ -103,7 +103,7 @@ class Seq2SeqTransformer(tf.keras.Model):
         "beam_size": self._beam_size,
         "alpha": self._alpha,
         "encoder_layer": self.encoder_layer,
-        "decoder_layer": self.decoder_layer
+        "decoder_layer": self.decoder_layer,
     }
     base_config = super(Seq2SeqTransformer, self).get_config()
     return dict(list(base_config.items()) + list(config.items()))
@@ -122,14 +122,47 @@ class Seq2SeqTransformer(tf.keras.Model):
 
     return tf.reshape(logits, [batch_size, length, vocab_size])
 
+  def _parse_inputs(self, inputs):
+    """Parses the `call` inputs and returns an uniformed output."""
+    sources = inputs.get("inputs", None)
+    input_mask = inputs.get("input_masks", None)
+    embedded = inputs.get("embedded_inputs", None)
+
+    if sources is None and embedded is not None:
+      embedded_inputs = embedded
+      boolean_mask = input_mask
+      input_shape = tf_utils.get_shape_list(embedded, expected_rank=3)
+      source_dtype = embedded.dtype
+    elif sources is not None:
+      embedded_inputs = self.embedding_lookup(sources)
+      boolean_mask = tf.not_equal(sources, 0)
+      input_shape = tf_utils.get_shape_list(sources, expected_rank=2)
+      source_dtype = sources.dtype
+    else:
+      raise KeyError(
+          "The call method expects either `inputs` or `embedded_inputs` and "
+          "`input_masks` as input features.")
+
+    return embedded_inputs, boolean_mask, input_shape, source_dtype
+
   def call(self, inputs):
     """Calculate target logits or inferred target sequences.
 
     Args:
       inputs: a dictionary of tensors.
-        Feature `inputs`: int tensor with shape `[batch_size, input_length]`.
+        Feature `inputs` (optional): int tensor with shape
+          `[batch_size, input_length]`.
+        Feature `embedded_inputs` (optional): float tensor with shape
+          `[batch_size, input_length, embedding_width]`.
         Feature `targets` (optional): None or int tensor with shape
           `[batch_size, target_length]`.
+        Feature `input_masks` (optional): When providing the `embedded_inputs`,
+          the dictionary must provide a boolean mask marking the filled time
+          steps. The shape of the tensor is `[batch_size, input_length]`.
+        Either `inputs` or `embedded_inputs` and `input_masks` must be present
+        in the input dictionary. In the second case the projection of the
+        integer tokens to the transformer embedding space is skipped and
+        `input_masks` is expected to be present.
 
     Returns:
       If targets is defined, then return logits for each word in the target
@@ -144,21 +177,19 @@ class Seq2SeqTransformer(tf.keras.Model):
     Raises:
       NotImplementedError: If try to use padded decode method on CPU/GPUs.
     """
-    sources = inputs["inputs"]
-    targets = inputs.get("targets", None)
     # Prepare inputs to the layer stack by adding positional encodings and
     # applying dropout.
-    embedded_inputs = self.embedding_lookup(sources)
-    embedding_mask = tf.cast(tf.not_equal(sources, 0), embedded_inputs.dtype)
+    targets = inputs.get("targets", None)
+    (embedded_inputs, boolean_mask,
+     input_shape, source_dtype) = self._parse_inputs(inputs)
+    embedding_mask = tf.cast(boolean_mask, embedded_inputs.dtype)
     embedded_inputs *= tf.expand_dims(embedding_mask, -1)
     # Attention_mask generation.
-    input_shape = tf_utils.get_shape_list(sources, expected_rank=2)
     attention_mask = tf.cast(
-        tf.reshape(
-            tf.not_equal(sources, 0), [input_shape[0], 1, input_shape[1]]),
-        dtype=sources.dtype)
+        tf.reshape(boolean_mask, [input_shape[0], 1, input_shape[1]]),
+        dtype=source_dtype)
     broadcast_ones = tf.ones(
-        shape=[input_shape[0], input_shape[1], 1], dtype=sources.dtype)
+        shape=[input_shape[0], input_shape[1], 1], dtype=source_dtype)
     attention_mask = broadcast_ones * attention_mask
 
     pos_encoding = self.position_embedding(embedded_inputs)
@@ -206,8 +237,7 @@ class Seq2SeqTransformer(tf.keras.Model):
       # Add encoder output and attention bias to the cache.
       encoder_outputs = tf.cast(encoder_outputs, dtype=self.compute_dtype)
       attention_mask = tf.cast(
-          tf.reshape(
-              tf.not_equal(sources, 0), [input_shape[0], 1, input_shape[1]]),
+          tf.reshape(boolean_mask, [input_shape[0], 1, input_shape[1]]),
           dtype=self.compute_dtype)
       cache["encoder_outputs"] = encoder_outputs
       cache["encoder_decoder_attention_mask"] = attention_mask
@@ -252,7 +282,7 @@ class Seq2SeqTransformer(tf.keras.Model):
     self_attention_mask = tf.tile(self_attention_mask, [batch_size, 1, 1])
 
     attention_mask = tf.cast(
-        tf.expand_dims(tf.not_equal(sources, 0), axis=1), dtype=sources.dtype)
+        tf.expand_dims(boolean_mask, axis=1), dtype=source_dtype)
     attention_mask = tf.tile(attention_mask, [1, decoder_length, 1])
 
     outputs = self.decoder_layer(

official/nlp/modeling/models/seq2seq_transformer_test.py

@@ -26,12 +26,11 @@ from official.nlp.modeling.models import seq2seq_transformer
 
 class Seq2SeqTransformerTest(tf.test.TestCase, parameterized.TestCase):
 
-  def _build_model(self, padded_decode, decode_max_length):
+  def _build_model(self, padded_decode, decode_max_length, embedding_width):
     num_layers = 1
     num_attention_heads = 2
     intermediate_size = 32
     vocab_size = 100
-    embedding_width = 16
     encdec_kwargs = dict(
         num_layers=num_layers,
         num_attention_heads=num_attention_heads,
@@ -63,15 +62,19 @@ class Seq2SeqTransformerTest(tf.test.TestCase, parameterized.TestCase):
               strategy_combinations.default_strategy,
               strategy_combinations.cloud_tpu_strategy,
           ],
+          embed=[True, False],
+          is_training=[True, False],
           mode="eager"))
-  def test_create_model_with_ds(self, distribution):
+  def test_create_model_with_ds(self, distribution, embed, is_training):
     with distribution.scope():
       padded_decode = isinstance(
           distribution,
           (tf.distribute.TPUStrategy, tf.distribute.experimental.TPUStrategy))
       decode_max_length = 10
       batch_size = 4
-      model = self._build_model(padded_decode, decode_max_length)
+      embedding_width = 16
+      model = self._build_model(
+          padded_decode, decode_max_length, embedding_width)
 
       @tf.function
       def step(inputs):
@@ -83,23 +86,32 @@ class Seq2SeqTransformerTest(tf.test.TestCase, parameterized.TestCase):
         return tf.nest.map_structure(distribution.experimental_local_results,
                                      outputs)
 
+      if embed:
+        fake_inputs = dict(
+            embedded_inputs=np.zeros(
+                (batch_size, decode_max_length, embedding_width),
+                dtype=np.float32),
+            input_masks=np.ones((batch_size, decode_max_length), dtype=np.bool))
+      else:
         fake_inputs = dict(
             inputs=np.zeros((batch_size, decode_max_length), dtype=np.int32))
-      local_outputs = step(fake_inputs)
-      logging.info("local_outputs=%s", local_outputs)
-      self.assertEqual(local_outputs["outputs"][0].shape, (4, 10))
 
-      fake_inputs = dict(
-          inputs=np.zeros((batch_size, decode_max_length), dtype=np.int32),
-          targets=np.zeros((batch_size, 8), dtype=np.int32))
+      if is_training:
+        fake_inputs["targets"] = np.zeros((batch_size, 8), dtype=np.int32)
         local_outputs = step(fake_inputs)
         logging.info("local_outputs=%s", local_outputs)
         self.assertEqual(local_outputs[0].shape, (4, 8, 100))
+      else:
+        local_outputs = step(fake_inputs)
+        logging.info("local_outputs=%s", local_outputs)
+        self.assertEqual(local_outputs["outputs"][0].shape, (4, 10))
 
   @parameterized.parameters(True, False)
   def test_create_savedmodel(self, padded_decode):
     decode_max_length = 10
-    model = self._build_model(padded_decode, decode_max_length)
+    embedding_width = 16
+    model = self._build_model(
+        padded_decode, decode_max_length, embedding_width)
 
     class SaveModule(tf.Module):
 
@@ -111,14 +123,28 @@ class Seq2SeqTransformerTest(tf.test.TestCase, parameterized.TestCase):
       def serve(self, inputs):
         return self.model.call(dict(inputs=inputs))
 
+      @tf.function
+      def embedded_serve(self, embedded_inputs, input_masks):
+        return self.model.call(
+            dict(embedded_inputs=embedded_inputs, input_masks=input_masks))
+
     save_module = SaveModule(model)
     if padded_decode:
-      tensor_shape = (4, 10)
+      tensor_shape = (4, decode_max_length)
+      embedded_tensor_shape = (4, decode_max_length, embedding_width)
     else:
       tensor_shape = (None, None)
+      embedded_tensor_shape = (None, None, embedding_width)
     signatures = dict(
         serving_default=save_module.serve.get_concrete_function(
-            tf.TensorSpec(shape=tensor_shape, dtype=tf.int32, name="inputs")))
+            tf.TensorSpec(shape=tensor_shape, dtype=tf.int32, name="inputs")),
+        embedded_serving=save_module.embedded_serve.get_concrete_function(
+            tf.TensorSpec(
+                shape=embedded_tensor_shape, dtype=tf.float32,
+                name="embedded_inputs"),
+            tf.TensorSpec(
+                shape=tensor_shape, dtype=tf.bool, name="input_masks"),
+            ))
     tf.saved_model.save(save_module, self.get_temp_dir(), signatures=signatures)