[cleanup] Hoist ModelTester objects to top level (#4939)
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>
This commit is contained in:
Amil Khare
@@ -35,6 +35,211 @@ if is_tf_available():
|
||||
)
|
||||
|
||||
|
||||
class TFXLMModelTester:
|
||||
def __init__(
|
||||
self, parent,
|
||||
):
|
||||
self.parent = parent
|
||||
self.batch_size = 13
|
||||
self.seq_length = 7
|
||||
self.is_training = True
|
||||
self.use_input_lengths = True
|
||||
self.use_token_type_ids = True
|
||||
self.use_labels = True
|
||||
self.gelu_activation = True
|
||||
self.sinusoidal_embeddings = False
|
||||
self.causal = False
|
||||
self.asm = False
|
||||
self.n_langs = 2
|
||||
self.vocab_size = 99
|
||||
self.n_special = 0
|
||||
self.hidden_size = 32
|
||||
self.num_hidden_layers = 5
|
||||
self.num_attention_heads = 4
|
||||
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.summary_type = "last"
|
||||
self.use_proj = True
|
||||
self.scope = None
|
||||
self.bos_token_id = 0
|
||||
|
||||
def prepare_config_and_inputs(self):
|
||||
input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
|
||||
input_mask = ids_tensor([self.batch_size, self.seq_length], 2, dtype=tf.float32)
|
||||
|
||||
input_lengths = None
|
||||
if self.use_input_lengths:
|
||||
input_lengths = (
|
||||
ids_tensor([self.batch_size], vocab_size=2) + self.seq_length - 2
|
||||
) # small variation of seq_length
|
||||
|
||||
token_type_ids = None
|
||||
if self.use_token_type_ids:
|
||||
token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.n_langs)
|
||||
|
||||
sequence_labels = None
|
||||
token_labels = None
|
||||
is_impossible_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)
|
||||
is_impossible_labels = ids_tensor([self.batch_size], 2, dtype=tf.float32)
|
||||
|
||||
config = XLMConfig(
|
||||
vocab_size=self.vocab_size,
|
||||
n_special=self.n_special,
|
||||
emb_dim=self.hidden_size,
|
||||
n_layers=self.num_hidden_layers,
|
||||
n_heads=self.num_attention_heads,
|
||||
dropout=self.hidden_dropout_prob,
|
||||
attention_dropout=self.attention_probs_dropout_prob,
|
||||
gelu_activation=self.gelu_activation,
|
||||
sinusoidal_embeddings=self.sinusoidal_embeddings,
|
||||
asm=self.asm,
|
||||
causal=self.causal,
|
||||
n_langs=self.n_langs,
|
||||
max_position_embeddings=self.max_position_embeddings,
|
||||
initializer_range=self.initializer_range,
|
||||
summary_type=self.summary_type,
|
||||
use_proj=self.use_proj,
|
||||
bos_token_id=self.bos_token_id,
|
||||
)
|
||||
|
||||
return (
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
)
|
||||
|
||||
def create_and_check_xlm_model(
|
||||
self,
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
):
|
||||
model = TFXLMModel(config=config)
|
||||
inputs = {"input_ids": input_ids, "lengths": input_lengths, "langs": token_type_ids}
|
||||
outputs = model(inputs)
|
||||
|
||||
inputs = [input_ids, input_mask]
|
||||
outputs = model(inputs)
|
||||
sequence_output = outputs[0]
|
||||
result = {
|
||||
"sequence_output": sequence_output.numpy(),
|
||||
}
|
||||
self.parent.assertListEqual(
|
||||
list(result["sequence_output"].shape), [self.batch_size, self.seq_length, self.hidden_size]
|
||||
)
|
||||
|
||||
def create_and_check_xlm_lm_head(
|
||||
self,
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
):
|
||||
model = TFXLMWithLMHeadModel(config)
|
||||
|
||||
inputs = {"input_ids": input_ids, "lengths": input_lengths, "langs": token_type_ids}
|
||||
outputs = model(inputs)
|
||||
|
||||
logits = outputs[0]
|
||||
|
||||
result = {
|
||||
"logits": logits.numpy(),
|
||||
}
|
||||
|
||||
self.parent.assertListEqual(list(result["logits"].shape), [self.batch_size, self.seq_length, self.vocab_size])
|
||||
|
||||
def create_and_check_xlm_qa(
|
||||
self,
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
):
|
||||
model = TFXLMForQuestionAnsweringSimple(config)
|
||||
|
||||
inputs = {"input_ids": input_ids, "lengths": input_lengths}
|
||||
|
||||
start_logits, end_logits = model(inputs)
|
||||
|
||||
result = {
|
||||
"start_logits": start_logits.numpy(),
|
||||
"end_logits": end_logits.numpy(),
|
||||
}
|
||||
|
||||
self.parent.assertListEqual(list(result["start_logits"].shape), [self.batch_size, self.seq_length])
|
||||
self.parent.assertListEqual(list(result["end_logits"].shape), [self.batch_size, self.seq_length])
|
||||
|
||||
def create_and_check_xlm_sequence_classif(
|
||||
self,
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
):
|
||||
model = TFXLMForSequenceClassification(config)
|
||||
|
||||
inputs = {"input_ids": input_ids, "lengths": input_lengths}
|
||||
|
||||
(logits,) = model(inputs)
|
||||
|
||||
result = {
|
||||
"logits": logits.numpy(),
|
||||
}
|
||||
|
||||
self.parent.assertListEqual(list(result["logits"].shape), [self.batch_size, self.type_sequence_label_size])
|
||||
|
||||
def prepare_config_and_inputs_for_common(self):
|
||||
config_and_inputs = self.prepare_config_and_inputs()
|
||||
(
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
) = config_and_inputs
|
||||
inputs_dict = {
|
||||
"input_ids": input_ids,
|
||||
"token_type_ids": token_type_ids,
|
||||
"langs": token_type_ids,
|
||||
"lengths": input_lengths,
|
||||
}
|
||||
return config, inputs_dict
|
||||
|
||||
|
||||
@require_tf
|
||||
class TFXLMModelTest(TFModelTesterMixin, unittest.TestCase):
|
||||
|
||||
@@ -47,244 +252,8 @@ class TFXLMModelTest(TFModelTesterMixin, unittest.TestCase):
|
||||
(TFXLMWithLMHeadModel,) if is_tf_available() else ()
|
||||
) # TODO (PVP): Check other models whether language generation is also applicable
|
||||
|
||||
class TFXLMModelTester(object):
|
||||
def __init__(
|
||||
self,
|
||||
parent,
|
||||
batch_size=13,
|
||||
seq_length=7,
|
||||
is_training=True,
|
||||
use_input_lengths=True,
|
||||
use_token_type_ids=True,
|
||||
use_labels=True,
|
||||
gelu_activation=True,
|
||||
sinusoidal_embeddings=False,
|
||||
causal=False,
|
||||
asm=False,
|
||||
n_langs=2,
|
||||
vocab_size=99,
|
||||
n_special=0,
|
||||
hidden_size=32,
|
||||
num_hidden_layers=5,
|
||||
num_attention_heads=4,
|
||||
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,
|
||||
summary_type="last",
|
||||
use_proj=True,
|
||||
scope=None,
|
||||
bos_token_id=0,
|
||||
):
|
||||
self.parent = parent
|
||||
self.batch_size = batch_size
|
||||
self.seq_length = seq_length
|
||||
self.is_training = is_training
|
||||
self.use_input_lengths = use_input_lengths
|
||||
self.use_token_type_ids = use_token_type_ids
|
||||
self.use_labels = use_labels
|
||||
self.gelu_activation = gelu_activation
|
||||
self.sinusoidal_embeddings = sinusoidal_embeddings
|
||||
self.asm = asm
|
||||
self.n_langs = n_langs
|
||||
self.vocab_size = vocab_size
|
||||
self.n_special = n_special
|
||||
self.summary_type = summary_type
|
||||
self.causal = causal
|
||||
self.use_proj = use_proj
|
||||
self.hidden_size = hidden_size
|
||||
self.num_hidden_layers = num_hidden_layers
|
||||
self.num_attention_heads = num_attention_heads
|
||||
self.hidden_dropout_prob = hidden_dropout_prob
|
||||
self.attention_probs_dropout_prob = attention_probs_dropout_prob
|
||||
self.max_position_embeddings = max_position_embeddings
|
||||
self.n_langs = n_langs
|
||||
self.type_sequence_label_size = type_sequence_label_size
|
||||
self.initializer_range = initializer_range
|
||||
self.summary_type = summary_type
|
||||
self.num_labels = num_labels
|
||||
self.num_choices = num_choices
|
||||
self.scope = scope
|
||||
self.bos_token_id = bos_token_id
|
||||
|
||||
def prepare_config_and_inputs(self):
|
||||
input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
|
||||
input_mask = ids_tensor([self.batch_size, self.seq_length], 2, dtype=tf.float32)
|
||||
|
||||
input_lengths = None
|
||||
if self.use_input_lengths:
|
||||
input_lengths = (
|
||||
ids_tensor([self.batch_size], vocab_size=2) + self.seq_length - 2
|
||||
) # small variation of seq_length
|
||||
|
||||
token_type_ids = None
|
||||
if self.use_token_type_ids:
|
||||
token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.n_langs)
|
||||
|
||||
sequence_labels = None
|
||||
token_labels = None
|
||||
is_impossible_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)
|
||||
is_impossible_labels = ids_tensor([self.batch_size], 2, dtype=tf.float32)
|
||||
|
||||
config = XLMConfig(
|
||||
vocab_size=self.vocab_size,
|
||||
n_special=self.n_special,
|
||||
emb_dim=self.hidden_size,
|
||||
n_layers=self.num_hidden_layers,
|
||||
n_heads=self.num_attention_heads,
|
||||
dropout=self.hidden_dropout_prob,
|
||||
attention_dropout=self.attention_probs_dropout_prob,
|
||||
gelu_activation=self.gelu_activation,
|
||||
sinusoidal_embeddings=self.sinusoidal_embeddings,
|
||||
asm=self.asm,
|
||||
causal=self.causal,
|
||||
n_langs=self.n_langs,
|
||||
max_position_embeddings=self.max_position_embeddings,
|
||||
initializer_range=self.initializer_range,
|
||||
summary_type=self.summary_type,
|
||||
use_proj=self.use_proj,
|
||||
bos_token_id=self.bos_token_id,
|
||||
)
|
||||
|
||||
return (
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
)
|
||||
|
||||
def create_and_check_xlm_model(
|
||||
self,
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
):
|
||||
model = TFXLMModel(config=config)
|
||||
inputs = {"input_ids": input_ids, "lengths": input_lengths, "langs": token_type_ids}
|
||||
outputs = model(inputs)
|
||||
|
||||
inputs = [input_ids, input_mask]
|
||||
outputs = model(inputs)
|
||||
sequence_output = outputs[0]
|
||||
result = {
|
||||
"sequence_output": sequence_output.numpy(),
|
||||
}
|
||||
self.parent.assertListEqual(
|
||||
list(result["sequence_output"].shape), [self.batch_size, self.seq_length, self.hidden_size]
|
||||
)
|
||||
|
||||
def create_and_check_xlm_lm_head(
|
||||
self,
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
):
|
||||
model = TFXLMWithLMHeadModel(config)
|
||||
|
||||
inputs = {"input_ids": input_ids, "lengths": input_lengths, "langs": token_type_ids}
|
||||
outputs = model(inputs)
|
||||
|
||||
logits = outputs[0]
|
||||
|
||||
result = {
|
||||
"logits": logits.numpy(),
|
||||
}
|
||||
|
||||
self.parent.assertListEqual(
|
||||
list(result["logits"].shape), [self.batch_size, self.seq_length, self.vocab_size]
|
||||
)
|
||||
|
||||
def create_and_check_xlm_qa(
|
||||
self,
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
):
|
||||
model = TFXLMForQuestionAnsweringSimple(config)
|
||||
|
||||
inputs = {"input_ids": input_ids, "lengths": input_lengths}
|
||||
|
||||
start_logits, end_logits = model(inputs)
|
||||
|
||||
result = {
|
||||
"start_logits": start_logits.numpy(),
|
||||
"end_logits": end_logits.numpy(),
|
||||
}
|
||||
|
||||
self.parent.assertListEqual(list(result["start_logits"].shape), [self.batch_size, self.seq_length])
|
||||
self.parent.assertListEqual(list(result["end_logits"].shape), [self.batch_size, self.seq_length])
|
||||
|
||||
def create_and_check_xlm_sequence_classif(
|
||||
self,
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
):
|
||||
model = TFXLMForSequenceClassification(config)
|
||||
|
||||
inputs = {"input_ids": input_ids, "lengths": input_lengths}
|
||||
|
||||
(logits,) = model(inputs)
|
||||
|
||||
result = {
|
||||
"logits": logits.numpy(),
|
||||
}
|
||||
|
||||
self.parent.assertListEqual(list(result["logits"].shape), [self.batch_size, self.type_sequence_label_size])
|
||||
|
||||
def prepare_config_and_inputs_for_common(self):
|
||||
config_and_inputs = self.prepare_config_and_inputs()
|
||||
(
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_lengths,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
is_impossible_labels,
|
||||
input_mask,
|
||||
) = config_and_inputs
|
||||
inputs_dict = {
|
||||
"input_ids": input_ids,
|
||||
"token_type_ids": token_type_ids,
|
||||
"langs": token_type_ids,
|
||||
"lengths": input_lengths,
|
||||
}
|
||||
return config, inputs_dict
|
||||
|
||||
def setUp(self):
|
||||
self.model_tester = TFXLMModelTest.TFXLMModelTester(self)
|
||||
self.model_tester = TFXLMModelTester(self)
|
||||
self.config_tester = ConfigTester(self, config_class=XLMConfig, emb_dim=37)
|
||||
|
||||
def test_config(self):
|
||||
|
||||
Reference in New Issue
Block a user