albert flax (#13294)

* albert flax

* year -> 2021

* docstring updated for flax

* removed head_mask

* removed from_pt

* removed passing attention_mask to embedding layer

tests/test_modeling_flax_albert.py

@@ -0,0 +1,161 @@
+# 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 AlbertConfig, 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.albert.modeling_flax_albert import (
+        FlaxAlbertForMaskedLM,
+        FlaxAlbertForMultipleChoice,
+        FlaxAlbertForPreTraining,
+        FlaxAlbertForQuestionAnswering,
+        FlaxAlbertForSequenceClassification,
+        FlaxAlbertForTokenClassification,
+        FlaxAlbertModel,
+    )
+
+
+class FlaxAlbertModelTester(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])
+
+        token_type_ids = None
+        if self.use_token_type_ids:
+            token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.type_vocab_size)
+
+        config = AlbertConfig(
+            vocab_size=self.vocab_size,
+            hidden_size=self.hidden_size,
+            num_hidden_layers=self.num_hidden_layers,
+            num_attention_heads=self.num_attention_heads,
+            intermediate_size=self.intermediate_size,
+            hidden_act=self.hidden_act,
+            hidden_dropout_prob=self.hidden_dropout_prob,
+            attention_probs_dropout_prob=self.attention_probs_dropout_prob,
+            max_position_embeddings=self.max_position_embeddings,
+            type_vocab_size=self.type_vocab_size,
+            is_decoder=False,
+            initializer_range=self.initializer_range,
+        )
+
+        return config, input_ids, token_type_ids, attention_mask
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        config, input_ids, token_type_ids, attention_mask = config_and_inputs
+        inputs_dict = {"input_ids": input_ids, "token_type_ids": token_type_ids, "attention_mask": attention_mask}
+        return config, inputs_dict
+
+
+@require_flax
+class FlaxAlbertModelTest(FlaxModelTesterMixin, unittest.TestCase):
+
+    all_model_classes = (
+        (
+            FlaxAlbertModel,
+            FlaxAlbertForPreTraining,
+            FlaxAlbertForMaskedLM,
+            FlaxAlbertForMultipleChoice,
+            FlaxAlbertForQuestionAnswering,
+            FlaxAlbertForSequenceClassification,
+            FlaxAlbertForTokenClassification,
+            FlaxAlbertForQuestionAnswering,
+        )
+        if is_flax_available()
+        else ()
+    )
+
+    def setUp(self):
+        self.model_tester = FlaxAlbertModelTester(self)
+
+    @slow
+    def test_model_from_pretrained(self):
+        for model_class_name in self.all_model_classes:
+            model = model_class_name.from_pretrained("albert-base-v2")
+            outputs = model(np.ones((1, 1)))
+            self.assertIsNotNone(outputs)
+
+
+@require_flax
+class FlaxAlbertModelIntegrationTest(unittest.TestCase):
+    @slow
+    def test_inference_no_head_absolute_embedding(self):
+        model = FlaxAlbertModel.from_pretrained("albert-base-v2")
+        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(
+            [[[-0.6513, 1.5035, -0.2766], [-0.6515, 1.5046, -0.2780], [-0.6512, 1.5049, -0.2784]]]
+        )
+
+        self.assertTrue(jnp.allclose(output[:, 1:4, 1:4], expected_slice, atol=1e-4))