Add TF_MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING (#18469)

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/__init__.py

@@ -2088,6 +2088,7 @@
             "TF_MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING",
             "TF_MODEL_FOR_PRETRAINING_MAPPING",
             "TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING",
+            "TF_MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING",
             "TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING",
             "TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING",
             "TF_MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING",
@@ -4582,6 +4583,7 @@
             TF_MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING,
             TF_MODEL_FOR_PRETRAINING_MAPPING,
             TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING,
+            TF_MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING,
             TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING,
             TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
             TF_MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING,

src/transformers/models/auto/__init__.py

@@ -111,6 +111,7 @@
         "TF_MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING",
         "TF_MODEL_FOR_PRETRAINING_MAPPING",
         "TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING",
+        "TF_MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING",
         "TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING",
         "TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING",
         "TF_MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING",
@@ -253,6 +254,7 @@
             TF_MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING,
             TF_MODEL_FOR_PRETRAINING_MAPPING,
             TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING,
+            TF_MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING,
             TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING,
             TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
             TF_MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING,

src/transformers/models/data2vec/modeling_tf_data2vec_vision.py

@@ -1352,8 +1352,8 @@ def masked_loss(real, pred):
             loss_ = loss_fct(real, pred)
             mask = tf.cast(mask, dtype=loss_.dtype)
             loss_ *= mask
-
-            return tf.reduce_sum(loss_) / tf.reduce_sum(mask)
+            reduced_masked_loss = tf.reduce_sum(loss_) / tf.reduce_sum(mask)
+            return tf.reshape(reduced_masked_loss, (1,))
 
         main_loss = masked_loss(labels, upsampled_logits)
         auxiliary_loss = masked_loss(labels, upsampled_auxiliary_logits)

src/transformers/models/segformer/modeling_tf_segformer.py

@@ -201,9 +201,9 @@ def __init__(self, config: SegformerConfig, hidden_size: int, **kwargs):
         self.dense = tf.keras.layers.Dense(hidden_size, name="dense")
         self.dropout = tf.keras.layers.Dropout(config.hidden_dropout_prob)
 
-    def call(self, hidden_states: tf.Tensor) -> tf.Tensor:
+    def call(self, hidden_states: tf.Tensor, training: bool = False) -> tf.Tensor:
         hidden_states = self.dense(hidden_states)
-        hidden_states = self.dropout(hidden_states)
+        hidden_states = self.dropout(hidden_states, training=training)
         return hidden_states
 
 
@@ -276,13 +276,13 @@ def __init__(
         self.dense2 = tf.keras.layers.Dense(out_features, name="dense2")
         self.dropout = tf.keras.layers.Dropout(config.hidden_dropout_prob)
 
-    def call(self, hidden_states: tf.Tensor, height: int, width: int) -> tf.Tensor:
+    def call(self, hidden_states: tf.Tensor, height: int, width: int, training: bool = False) -> tf.Tensor:
         hidden_states = self.dense1(hidden_states)
         hidden_states = self.depthwise_convolution(hidden_states, height, width)
         hidden_states = self.intermediate_act_fn(hidden_states)
-        hidden_states = self.dropout(hidden_states)
+        hidden_states = self.dropout(hidden_states, training=training)
         hidden_states = self.dense2(hidden_states)
-        hidden_states = self.dropout(hidden_states)
+        hidden_states = self.dropout(hidden_states, training=training)
         return hidden_states
 
 
@@ -749,7 +749,7 @@ def __init__(self, config: SegformerConfig, **kwargs):
 
         self.config = config
 
-    def call(self, encoder_hidden_states):
+    def call(self, encoder_hidden_states, training: bool = False):
         batch_size = shape_list(encoder_hidden_states[-1])[0]
 
         all_hidden_states = ()
@@ -773,9 +773,9 @@ def call(self, encoder_hidden_states):
             all_hidden_states += (encoder_hidden_state,)
 
         hidden_states = self.linear_fuse(tf.concat(all_hidden_states[::-1], axis=-1))
-        hidden_states = self.batch_norm(hidden_states)
+        hidden_states = self.batch_norm(hidden_states, training=training)
         hidden_states = self.activation(hidden_states)
-        hidden_states = self.dropout(hidden_states)
+        hidden_states = self.dropout(hidden_states, training=training)
 
         # logits of shape (batch_size, height/4, width/4, num_labels)
         logits = self.classifier(hidden_states)

src/transformers/utils/dummy_tf_objects.py

@@ -279,6 +279,9 @@ def __init__(self, *args, **kwargs):
 TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING = None
 
 
+TF_MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING = None
+
+
 TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING = None
 
 

tests/models/segformer/test_modeling_tf_segformer.py

@@ -27,6 +27,7 @@
 
 
 if is_tf_available():
+    import numpy as np
     import tensorflow as tf
 
     from transformers import TFSegformerForImageClassification, TFSegformerForSemanticSegmentation, TFSegformerModel
@@ -336,6 +337,9 @@ def recursive_check(tuple_object, dict_object):
     def test_dataset_conversion(self):
         super().test_dataset_conversion()
 
+    def check_keras_fit_results(self, val_loss1, val_loss2, atol=2e-1, rtol=2e-1):
+        self.assertTrue(np.allclose(val_loss1, val_loss2, atol=atol, rtol=rtol))
+
     @unittest.skipIf(
         not is_tf_available() or len(tf.config.list_physical_devices("GPU")) == 0,
         reason="TF (<=2.8) does not support backprop for grouped convolutions on CPU.",

tests/test_modeling_tf_common.py

@@ -62,11 +62,13 @@
     from transformers import (
         TF_MODEL_FOR_CAUSAL_LM_MAPPING,
         TF_MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING,
+        TF_MODEL_FOR_MASKED_IMAGE_MODELING_MAPPING,
         TF_MODEL_FOR_MASKED_LM_MAPPING,
         TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING,
         TF_MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING,
         TF_MODEL_FOR_PRETRAINING_MAPPING,
         TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING,
+        TF_MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING,
         TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING,
         TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
         TF_MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING,
@@ -170,6 +172,15 @@ def _prepare_for_class(self, inputs_dict, model_class, return_labels=False) -> d
                 inputs_dict["labels"] = tf.zeros(
                     (self.model_tester.batch_size, self.model_tester.seq_length), dtype=tf.int32
                 )
+            elif model_class in get_values(TF_MODEL_FOR_MASKED_IMAGE_MODELING_MAPPING):
+                num_patches = self.model_tester.image_size // self.model_tester.patch_size
+                inputs_dict["bool_masked_pos"] = tf.zeros(
+                    (self.model_tester.batch_size, num_patches**2), dtype=tf.int32
+                )
+            elif model_class in get_values(TF_MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING):
+                batch_size, num_channels, height, width = inputs_dict["pixel_values"].shape
+                inputs_dict["labels"] = tf.zeros((self.model_tester.batch_size, height, width), dtype=tf.int32)
+
         return inputs_dict
 
     def test_initialization(self):
@@ -1389,6 +1400,9 @@ def test_loss_computation(self):
 
                 self.assertTrue(loss.shape.as_list() == expected_loss_size or loss.shape.as_list() == [1])
 
+    def check_keras_fit_results(self, val_loss1, val_loss2, atol=1e-2, rtol=1e-3):
+        self.assertTrue(np.allclose(val_loss1, val_loss2, atol=atol, rtol=rtol))
+
     def test_keras_fit(self):
         config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
         for model_class in self.all_model_classes:
@@ -1468,7 +1482,7 @@ def test_keras_fit(self):
                 val_loss2 = history2.history["val_loss"][0]
                 self.assertTrue(not isnan(val_loss2))
                 accuracy2 = {key: val[0] for key, val in history2.history.items() if key.endswith("accuracy")}
-                self.assertTrue(np.allclose(val_loss1, val_loss2, atol=1e-2, rtol=1e-3))
+                self.check_keras_fit_results(val_loss1, val_loss2)
                 self.assertEqual(history1.history.keys(), history2.history.keys())
                 for key in history1.history.keys():
                     if not key.startswith("val_"):
@@ -1494,7 +1508,7 @@ def test_keras_fit(self):
                 val_loss3 = history3.history["val_loss"][0]
                 self.assertTrue(not isnan(val_loss3))
                 accuracy3 = {key: val[0] for key, val in history3.history.items() if key.endswith("accuracy")}
-                self.assertTrue(np.allclose(val_loss1, val_loss3, atol=1e-2, rtol=1e-3))
+                self.check_keras_fit_results(val_loss1, val_loss3)
                 self.assertEqual(history1.history.keys(), history3.history.keys())
                 if metrics:
                     self.assertTrue(len(accuracy1) == len(accuracy3) > 0, "Missing metrics!")