SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Flax Big Bird (#11967)

Browse Source 
* add flax bert

* bert -> bigbird

* original_full ported

* add debugger

* init block sparse

* fix copies ; gelu_fast -> gelu_new

* block sparse port

* fix block sparse

* block sparse working

* all ckpts working

* fix-copies

* make quality

* init tests

* temporary fix for FlaxBigBirdForMultipleChoice

* skip test_attention_outputs

* fix

* gelu_fast -> gelu_new ; fix multiple choice model

* remove nsp

* fix sequence classifier

* fix

* make quality

* make fix-copies

* finish

* Delete debugger.ipynb

* Update src/transformers/models/big_bird/modeling_flax_big_bird.py

* make style

* finish

* bye bye jit flax tests

Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
This commit is contained in:
Vasudev Gupta
2021-06-15 00:31:03 +05:30
committed by GitHub
parent a156da9a23
commit d9c0d08f9a
17 changed files with 2434 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

164
tests/test_modeling_flax_big_bird.py Normal file
View File

@@ -0,0 +1,164 @@
# 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
import numpy as np
from transformers import BigBirdConfig, 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():
from transformers.models.big_bird.modeling_flax_big_bird import (
FlaxBigBirdForMaskedLM,
FlaxBigBirdForMultipleChoice,
FlaxBigBirdForPreTraining,
FlaxBigBirdForQuestionAnswering,
FlaxBigBirdForSequenceClassification,
FlaxBigBirdForTokenClassification,
FlaxBigBirdModel,
)
class FlaxBigBirdModelTester(unittest.TestCase):
def __init__(
self,
parent,
batch_size=13,
seq_length=56,
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_new",
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,
attention_type="block_sparse",
use_bias=True,
rescale_embeddings=False,
block_size=4,
num_random_blocks=3,
):
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
self.rescale_embeddings = rescale_embeddings
self.attention_type = attention_type
self.use_bias = use_bias
self.block_size = block_size
self.num_random_blocks = num_random_blocks
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 = BigBirdConfig(
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,
attention_type=self.attention_type,
block_size=self.block_size,
num_random_blocks=self.num_random_blocks,
use_bias=self.use_bias,
rescale_embeddings=self.rescale_embeddings,
)
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 FlaxBigBirdModelTest(FlaxModelTesterMixin, unittest.TestCase):
all_model_classes = (
(
FlaxBigBirdModel,
FlaxBigBirdForPreTraining,
FlaxBigBirdForMaskedLM,
FlaxBigBirdForMultipleChoice,
FlaxBigBirdForQuestionAnswering,
FlaxBigBirdForSequenceClassification,
FlaxBigBirdForTokenClassification,
)
if is_flax_available()
else ()
)
test_attn_probs = False
def setUp(self):
self.model_tester = FlaxBigBirdModelTester(self)
@slow
def test_model_from_pretrained(self):
for model_class_name in self.all_model_classes:
model = model_class_name.from_pretrained("google/bigbird-roberta-base", from_pt=True)
outputs = model(np.ones((1, 1)))
self.assertIsNotNone(outputs)
def test_attention_outputs(self):
if self.test_attn_probs:
super().test_attention_outputs()

1
tests/test_modeling_flax_clip.py
View File

@@ -342,6 +342,7 @@ class FlaxCLIPModelTest(FlaxModelTesterMixin, unittest.TestCase):
def test_hidden_states_output(self):
pass
@slow
def test_jit_compilation(self):
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()

3
tests/test_modeling_flax_common.py
View File

@@ -23,7 +23,7 @@ import numpy as np
import transformers
from transformers import is_flax_available, is_torch_available
from transformers.models.auto import get_values
from transformers.testing_utils import is_pt_flax_cross_test, require_flax
from transformers.testing_utils import is_pt_flax_cross_test, require_flax, slow
if is_flax_available():
@@ -273,6 +273,7 @@ class FlaxModelTesterMixin:
for output_loaded, output in zip(outputs_loaded, outputs):
self.assert_almost_equals(output_loaded, output, 1e-3)
@slow
def test_jit_compilation(self):
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()

1
tests/test_modeling_flax_vit.py
View File

@@ -179,6 +179,7 @@ class FlaxViTModelTest(FlaxModelTesterMixin, unittest.TestCase):
self.assertListEqual(arg_names[:1], expected_arg_names)
# We neeed to override this test because ViT expects pixel_values instead of input_ids
@slow
def test_jit_compilation(self):
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()