[Test refactor 1/5] Per-folder tests reorganization (#15725)

* Per-folder tests reorganization

Co-authored-by: sgugger <sylvain.gugger@gmail.com>
Co-authored-by: Stas Bekman <stas@stason.org>

tests/distilbert/test_modeling_distilbert.py

@@ -0,0 +1,294 @@
+# coding=utf-8
+# Copyright 2020 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import os
+import tempfile
+import unittest
+
+from transformers import DistilBertConfig, is_torch_available
+from transformers.testing_utils import require_torch, require_torch_gpu, slow, torch_device
+
+from ..test_configuration_common import ConfigTester
+from ..test_modeling_common import ModelTesterMixin, ids_tensor, random_attention_mask
+
+
+if is_torch_available():
+    import torch
+
+    from transformers import (
+        DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
+        DistilBertForMaskedLM,
+        DistilBertForMultipleChoice,
+        DistilBertForQuestionAnswering,
+        DistilBertForSequenceClassification,
+        DistilBertForTokenClassification,
+        DistilBertModel,
+    )
+
+
+class DistilBertModelTester(object):
+    def __init__(
+        self,
+        parent,
+        batch_size=13,
+        seq_length=7,
+        is_training=True,
+        use_input_mask=True,
+        use_token_type_ids=False,
+        use_labels=True,
+        vocab_size=99,
+        hidden_size=32,
+        num_hidden_layers=5,
+        num_attention_heads=4,
+        intermediate_size=37,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=512,
+        type_vocab_size=16,
+        type_sequence_label_size=2,
+        initializer_range=0.02,
+        num_labels=3,
+        num_choices=4,
+        scope=None,
+    ):
+        self.parent = parent
+        self.batch_size = batch_size
+        self.seq_length = seq_length
+        self.is_training = is_training
+        self.use_input_mask = use_input_mask
+        self.use_token_type_ids = use_token_type_ids
+        self.use_labels = use_labels
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.hidden_act = hidden_act
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.type_vocab_size = type_vocab_size
+        self.type_sequence_label_size = type_sequence_label_size
+        self.initializer_range = initializer_range
+        self.num_labels = num_labels
+        self.num_choices = num_choices
+        self.scope = scope
+
+    def prepare_config_and_inputs(self):
+        input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
+
+        input_mask = None
+        if self.use_input_mask:
+            input_mask = random_attention_mask([self.batch_size, self.seq_length])
+
+        sequence_labels = None
+        token_labels = None
+        choice_labels = None
+        if self.use_labels:
+            sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
+            token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
+            choice_labels = ids_tensor([self.batch_size], self.num_choices)
+
+        config = self.get_config()
+
+        return config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+
+    def get_config(self):
+        return DistilBertConfig(
+            vocab_size=self.vocab_size,
+            dim=self.hidden_size,
+            n_layers=self.num_hidden_layers,
+            n_heads=self.num_attention_heads,
+            hidden_dim=self.intermediate_size,
+            hidden_act=self.hidden_act,
+            dropout=self.hidden_dropout_prob,
+            attention_dropout=self.attention_probs_dropout_prob,
+            max_position_embeddings=self.max_position_embeddings,
+            initializer_range=self.initializer_range,
+        )
+
+    def create_and_check_distilbert_model(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = DistilBertModel(config=config)
+        model.to(torch_device)
+        model.eval()
+        result = model(input_ids, input_mask)
+        result = model(input_ids)
+        self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
+
+    def create_and_check_distilbert_for_masked_lm(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = DistilBertForMaskedLM(config=config)
+        model.to(torch_device)
+        model.eval()
+        result = model(input_ids, attention_mask=input_mask, labels=token_labels)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size))
+
+    def create_and_check_distilbert_for_question_answering(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = DistilBertForQuestionAnswering(config=config)
+        model.to(torch_device)
+        model.eval()
+        result = model(
+            input_ids, attention_mask=input_mask, start_positions=sequence_labels, end_positions=sequence_labels
+        )
+        self.parent.assertEqual(result.start_logits.shape, (self.batch_size, self.seq_length))
+        self.parent.assertEqual(result.end_logits.shape, (self.batch_size, self.seq_length))
+
+    def create_and_check_distilbert_for_sequence_classification(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_labels = self.num_labels
+        model = DistilBertForSequenceClassification(config)
+        model.to(torch_device)
+        model.eval()
+        result = model(input_ids, attention_mask=input_mask, labels=sequence_labels)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_labels))
+
+    def create_and_check_distilbert_for_token_classification(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_labels = self.num_labels
+        model = DistilBertForTokenClassification(config=config)
+        model.to(torch_device)
+        model.eval()
+
+        result = model(input_ids, attention_mask=input_mask, labels=token_labels)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.num_labels))
+
+    def create_and_check_distilbert_for_multiple_choice(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_choices = self.num_choices
+        model = DistilBertForMultipleChoice(config=config)
+        model.to(torch_device)
+        model.eval()
+        multiple_choice_inputs_ids = input_ids.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous()
+        multiple_choice_input_mask = input_mask.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous()
+        result = model(
+            multiple_choice_inputs_ids,
+            attention_mask=multiple_choice_input_mask,
+            labels=choice_labels,
+        )
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_choices))
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        (config, input_ids, input_mask, sequence_labels, token_labels, choice_labels) = config_and_inputs
+        inputs_dict = {"input_ids": input_ids, "attention_mask": input_mask}
+        return config, inputs_dict
+
+
+@require_torch
+class DistilBertModelTest(ModelTesterMixin, unittest.TestCase):
+
+    all_model_classes = (
+        (
+            DistilBertModel,
+            DistilBertForMaskedLM,
+            DistilBertForMultipleChoice,
+            DistilBertForQuestionAnswering,
+            DistilBertForSequenceClassification,
+            DistilBertForTokenClassification,
+        )
+        if is_torch_available()
+        else None
+    )
+    fx_compatible = True
+    test_pruning = True
+    test_torchscript = True
+    test_resize_embeddings = True
+    test_resize_position_embeddings = True
+
+    def setUp(self):
+        self.model_tester = DistilBertModelTester(self)
+        self.config_tester = ConfigTester(self, config_class=DistilBertConfig, dim=37)
+
+    def test_config(self):
+        self.config_tester.run_common_tests()
+
+    def test_distilbert_model(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_model(*config_and_inputs)
+
+    def test_for_masked_lm(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_for_masked_lm(*config_and_inputs)
+
+    def test_for_question_answering(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_for_question_answering(*config_and_inputs)
+
+    def test_for_sequence_classification(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_for_sequence_classification(*config_and_inputs)
+
+    def test_for_token_classification(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_for_token_classification(*config_and_inputs)
+
+    def test_for_multiple_choice(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_for_multiple_choice(*config_and_inputs)
+
+    @slow
+    def test_model_from_pretrained(self):
+        for model_name in DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
+            model = DistilBertModel.from_pretrained(model_name)
+            self.assertIsNotNone(model)
+
+    @slow
+    @require_torch_gpu
+    def test_torchscript_device_change(self):
+        config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
+        for model_class in self.all_model_classes:
+
+            # BertForMultipleChoice behaves incorrectly in JIT environments.
+            if model_class == DistilBertForMultipleChoice:
+                return
+
+            config.torchscript = True
+            model = model_class(config=config)
+
+            inputs_dict = self._prepare_for_class(inputs_dict, model_class)
+            traced_model = torch.jit.trace(
+                model, (inputs_dict["input_ids"].to("cpu"), inputs_dict["attention_mask"].to("cpu"))
+            )
+
+            with tempfile.TemporaryDirectory() as tmp:
+                torch.jit.save(traced_model, os.path.join(tmp, "traced_model.pt"))
+                loaded = torch.jit.load(os.path.join(tmp, "traced_model.pt"), map_location=torch_device)
+                loaded(inputs_dict["input_ids"].to(torch_device), inputs_dict["attention_mask"].to(torch_device))
+
+
+@require_torch
+class DistilBertModelIntergrationTest(unittest.TestCase):
+    @slow
+    def test_inference_no_head_absolute_embedding(self):
+        model = DistilBertModel.from_pretrained("distilbert-base-uncased")
+        input_ids = torch.tensor([[0, 345, 232, 328, 740, 140, 1695, 69, 6078, 1588, 2]])
+        attention_mask = torch.tensor([[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]])
+        with torch.no_grad():
+            output = model(input_ids, attention_mask=attention_mask)[0]
+        expected_shape = torch.Size((1, 11, 768))
+        self.assertEqual(output.shape, expected_shape)
+        expected_slice = torch.tensor(
+            [[[-0.1639, 0.3299, 0.1648], [-0.1746, 0.3289, 0.1710], [-0.1884, 0.3357, 0.1810]]]
+        )
+
+        self.assertTrue(torch.allclose(output[:, 1:4, 1:4], expected_slice, atol=1e-4))

tests/distilbert/test_modeling_flax_distilbert.py

@@ -0,0 +1,152 @@
+# Copyright 2021 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+
+import numpy as np
+
+from transformers import DistilBertConfig, is_flax_available
+from transformers.testing_utils import require_flax, slow
+
+from ..test_modeling_flax_common import FlaxModelTesterMixin, ids_tensor, random_attention_mask
+
+
+if is_flax_available():
+    import jax.numpy as jnp
+    from transformers.models.distilbert.modeling_flax_distilbert import (
+        FlaxDistilBertForMaskedLM,
+        FlaxDistilBertForMultipleChoice,
+        FlaxDistilBertForQuestionAnswering,
+        FlaxDistilBertForSequenceClassification,
+        FlaxDistilBertForTokenClassification,
+        FlaxDistilBertModel,
+    )
+
+
+class FlaxDistilBertModelTester(unittest.TestCase):
+    def __init__(
+        self,
+        parent,
+        batch_size=13,
+        seq_length=7,
+        is_training=True,
+        use_attention_mask=True,
+        use_token_type_ids=True,
+        use_labels=True,
+        vocab_size=99,
+        hidden_size=32,
+        num_hidden_layers=5,
+        num_attention_heads=4,
+        intermediate_size=37,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=512,
+        type_vocab_size=16,
+        type_sequence_label_size=2,
+        initializer_range=0.02,
+        num_choices=4,
+    ):
+        self.parent = parent
+        self.batch_size = batch_size
+        self.seq_length = seq_length
+        self.is_training = is_training
+        self.use_attention_mask = use_attention_mask
+        self.use_token_type_ids = use_token_type_ids
+        self.use_labels = use_labels
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.hidden_act = hidden_act
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.type_vocab_size = type_vocab_size
+        self.type_sequence_label_size = type_sequence_label_size
+        self.initializer_range = initializer_range
+        self.num_choices = num_choices
+
+    def prepare_config_and_inputs(self):
+        input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
+
+        attention_mask = None
+        if self.use_attention_mask:
+            attention_mask = random_attention_mask([self.batch_size, self.seq_length])
+
+        config = DistilBertConfig(
+            vocab_size=self.vocab_size,
+            dim=self.hidden_size,
+            n_layers=self.num_hidden_layers,
+            n_heads=self.num_attention_heads,
+            hidden_dim=self.intermediate_size,
+            hidden_act=self.hidden_act,
+            dropout=self.hidden_dropout_prob,
+            attention_dropout=self.attention_probs_dropout_prob,
+            max_position_embeddings=self.max_position_embeddings,
+            initializer_range=self.initializer_range,
+            tie_weights_=True,
+        )
+
+        return config, input_ids, attention_mask
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        config, input_ids, attention_mask = config_and_inputs
+        inputs_dict = {"input_ids": input_ids, "attention_mask": attention_mask}
+        return config, inputs_dict
+
+
+@require_flax
+class FlaxDistilBertModelTest(FlaxModelTesterMixin, unittest.TestCase):
+
+    all_model_classes = (
+        (
+            FlaxDistilBertModel,
+            FlaxDistilBertForMaskedLM,
+            FlaxDistilBertForMultipleChoice,
+            FlaxDistilBertForQuestionAnswering,
+            FlaxDistilBertForSequenceClassification,
+            FlaxDistilBertForTokenClassification,
+            FlaxDistilBertForQuestionAnswering,
+        )
+        if is_flax_available()
+        else ()
+    )
+
+    def setUp(self):
+        self.model_tester = FlaxDistilBertModelTester(self)
+
+    @slow
+    def test_model_from_pretrained(self):
+        for model_class_name in self.all_model_classes:
+            model = model_class_name.from_pretrained("distilbert-base-uncased")
+            outputs = model(np.ones((1, 1)))
+            self.assertIsNotNone(outputs)
+
+
+@require_flax
+class FlaxDistilBertModelIntegrationTest(unittest.TestCase):
+    @slow
+    def test_inference_no_head_absolute_embedding(self):
+        model = FlaxDistilBertModel.from_pretrained("distilbert-base-uncased")
+        input_ids = np.array([[0, 345, 232, 328, 740, 140, 1695, 69, 6078, 1588, 2]])
+        attention_mask = np.array([[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]])
+        output = model(input_ids, attention_mask=attention_mask)[0]
+        expected_shape = (1, 11, 768)
+        self.assertEqual(output.shape, expected_shape)
+        expected_slice = np.array([[[-0.1639, 0.3299, 0.1648], [-0.1746, 0.3289, 0.1710], [-0.1884, 0.3357, 0.1810]]])
+
+        self.assertTrue(jnp.allclose(output[:, 1:4, 1:4], expected_slice, atol=1e-4))

tests/distilbert/test_modeling_tf_distilbert.py

@@ -0,0 +1,247 @@
+# coding=utf-8
+# Copyright 2020 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import unittest
+
+from transformers import DistilBertConfig, is_tf_available
+from transformers.testing_utils import require_tf, slow
+
+from ..test_configuration_common import ConfigTester
+from ..test_modeling_tf_common import TFModelTesterMixin, ids_tensor
+
+
+if is_tf_available():
+    import tensorflow as tf
+
+    from transformers.models.distilbert.modeling_tf_distilbert import (
+        TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
+        TFDistilBertForMaskedLM,
+        TFDistilBertForMultipleChoice,
+        TFDistilBertForQuestionAnswering,
+        TFDistilBertForSequenceClassification,
+        TFDistilBertForTokenClassification,
+        TFDistilBertModel,
+    )
+
+
+class TFDistilBertModelTester:
+    def __init__(
+        self,
+        parent,
+    ):
+        self.parent = parent
+        self.batch_size = 13
+        self.seq_length = 7
+        self.is_training = True
+        self.use_input_mask = True
+        self.use_token_type_ids = False
+        self.use_labels = True
+        self.vocab_size = 99
+        self.hidden_size = 32
+        self.num_hidden_layers = 5
+        self.num_attention_heads = 4
+        self.intermediate_size = 37
+        self.hidden_act = "gelu"
+        self.hidden_dropout_prob = 0.1
+        self.attention_probs_dropout_prob = 0.1
+        self.max_position_embeddings = 512
+        self.type_vocab_size = 16
+        self.type_sequence_label_size = 2
+        self.initializer_range = 0.02
+        self.num_labels = 3
+        self.num_choices = 4
+        self.scope = None
+
+    def prepare_config_and_inputs(self):
+        input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
+
+        input_mask = None
+        if self.use_input_mask:
+            input_mask = ids_tensor([self.batch_size, self.seq_length], vocab_size=2)
+
+        sequence_labels = None
+        token_labels = None
+        choice_labels = None
+        if self.use_labels:
+            sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
+            token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
+            choice_labels = ids_tensor([self.batch_size], self.num_choices)
+
+        config = DistilBertConfig(
+            vocab_size=self.vocab_size,
+            dim=self.hidden_size,
+            n_layers=self.num_hidden_layers,
+            n_heads=self.num_attention_heads,
+            hidden_dim=self.intermediate_size,
+            hidden_act=self.hidden_act,
+            dropout=self.hidden_dropout_prob,
+            attention_dropout=self.attention_probs_dropout_prob,
+            max_position_embeddings=self.max_position_embeddings,
+            initializer_range=self.initializer_range,
+        )
+
+        return config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+
+    def create_and_check_distilbert_model(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = TFDistilBertModel(config=config)
+        inputs = {"input_ids": input_ids, "attention_mask": input_mask}
+
+        result = model(inputs)
+
+        inputs = [input_ids, input_mask]
+
+        result = model(inputs)
+
+        self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
+
+    def create_and_check_distilbert_for_masked_lm(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = TFDistilBertForMaskedLM(config=config)
+        inputs = {"input_ids": input_ids, "attention_mask": input_mask}
+        result = model(inputs)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size))
+
+    def create_and_check_distilbert_for_question_answering(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = TFDistilBertForQuestionAnswering(config=config)
+        inputs = {
+            "input_ids": input_ids,
+            "attention_mask": input_mask,
+        }
+        result = model(inputs)
+        self.parent.assertEqual(result.start_logits.shape, (self.batch_size, self.seq_length))
+        self.parent.assertEqual(result.end_logits.shape, (self.batch_size, self.seq_length))
+
+    def create_and_check_distilbert_for_sequence_classification(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_labels = self.num_labels
+        model = TFDistilBertForSequenceClassification(config)
+        inputs = {"input_ids": input_ids, "attention_mask": input_mask}
+        result = model(inputs)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_labels))
+
+    def create_and_check_distilbert_for_multiple_choice(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_choices = self.num_choices
+        model = TFDistilBertForMultipleChoice(config)
+        multiple_choice_inputs_ids = tf.tile(tf.expand_dims(input_ids, 1), (1, self.num_choices, 1))
+        multiple_choice_input_mask = tf.tile(tf.expand_dims(input_mask, 1), (1, self.num_choices, 1))
+        inputs = {
+            "input_ids": multiple_choice_inputs_ids,
+            "attention_mask": multiple_choice_input_mask,
+        }
+        result = model(inputs)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_choices))
+
+    def create_and_check_distilbert_for_token_classification(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_labels = self.num_labels
+        model = TFDistilBertForTokenClassification(config)
+        inputs = {"input_ids": input_ids, "attention_mask": input_mask}
+        result = model(inputs)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.num_labels))
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        (config, input_ids, input_mask, sequence_labels, token_labels, choice_labels) = config_and_inputs
+        inputs_dict = {"input_ids": input_ids, "attention_mask": input_mask}
+        return config, inputs_dict
+
+
+@require_tf
+class TFDistilBertModelTest(TFModelTesterMixin, unittest.TestCase):
+
+    all_model_classes = (
+        (
+            TFDistilBertModel,
+            TFDistilBertForMaskedLM,
+            TFDistilBertForQuestionAnswering,
+            TFDistilBertForSequenceClassification,
+            TFDistilBertForTokenClassification,
+            TFDistilBertForMultipleChoice,
+        )
+        if is_tf_available()
+        else None
+    )
+    test_head_masking = False
+    test_onnx = False
+
+    def setUp(self):
+        self.model_tester = TFDistilBertModelTester(self)
+        self.config_tester = ConfigTester(self, config_class=DistilBertConfig, dim=37)
+
+    def test_config(self):
+        self.config_tester.run_common_tests()
+
+    def test_distilbert_model(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_model(*config_and_inputs)
+
+    def test_for_masked_lm(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_for_masked_lm(*config_and_inputs)
+
+    def test_for_question_answering(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_for_question_answering(*config_and_inputs)
+
+    def test_for_sequence_classification(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_for_sequence_classification(*config_and_inputs)
+
+    def test_for_multiple_choice(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_for_multiple_choice(*config_and_inputs)
+
+    def test_for_token_classification(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_distilbert_for_token_classification(*config_and_inputs)
+
+    @slow
+    def test_model_from_pretrained(self):
+        for model_name in list(TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]):
+            model = TFDistilBertModel.from_pretrained(model_name)
+            self.assertIsNotNone(model)
+
+
+@require_tf
+class TFDistilBertModelIntegrationTest(unittest.TestCase):
+    @slow
+    def test_inference_masked_lm(self):
+        model = TFDistilBertModel.from_pretrained("distilbert-base-uncased")
+        input_ids = tf.constant([[0, 1, 2, 3, 4, 5]])
+        output = model(input_ids)[0]
+
+        expected_shape = [1, 6, 768]
+        self.assertEqual(output.shape, expected_shape)
+
+        expected_slice = tf.constant(
+            [
+                [
+                    [0.19261885, -0.13732955, 0.4119799],
+                    [0.22150156, -0.07422661, 0.39037204],
+                    [0.22756018, -0.0896414, 0.3701467],
+                ]
+            ]
+        )
+        tf.debugging.assert_near(output[:, :3, :3], expected_slice, atol=1e-4)

tests/distilbert/test_tokenization_distilbert.py

@@ -0,0 +1,43 @@
+# coding=utf-8
+# Copyright 2020 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+from transformers import DistilBertTokenizer, DistilBertTokenizerFast
+from transformers.testing_utils import require_tokenizers, slow
+
+from ..bert.test_tokenization_bert import BertTokenizationTest
+
+
+@require_tokenizers
+class DistilBertTokenizationTest(BertTokenizationTest):
+
+    tokenizer_class = DistilBertTokenizer
+    rust_tokenizer_class = DistilBertTokenizerFast
+    test_rust_tokenizer = True
+
+    @slow
+    def test_sequence_builders(self):
+        tokenizer = DistilBertTokenizer.from_pretrained("distilbert-base-uncased")
+
+        text = tokenizer.encode("sequence builders", add_special_tokens=False)
+        text_2 = tokenizer.encode("multi-sequence build", add_special_tokens=False)
+
+        encoded_sentence = tokenizer.build_inputs_with_special_tokens(text)
+        encoded_pair = tokenizer.build_inputs_with_special_tokens(text, text_2)
+
+        assert encoded_sentence == [tokenizer.cls_token_id] + text + [tokenizer.sep_token_id]
+        assert encoded_pair == [tokenizer.cls_token_id] + text + [tokenizer.sep_token_id] + text_2 + [
+            tokenizer.sep_token_id
+        ]