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

Merge branch 'master' into t5

Browse Source
This commit is contained in:
thomwolf
2019-12-10 12:58:48 +01:00
parent 608a8f5b56 e57d00ee10
commit 0558c9cb9b
169 changed files with 13409 additions and 3288 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
templates/adding_a_new_example_script/run_xxx.py
View File

@@ -43,7 +43,7 @@ from transformers import (WEIGHTS_NAME, BertConfig,
XLNetTokenizer,
DistilBertConfig, DistilBertForQuestionAnswering, DistilBertTokenizer)
from transformers import AdamW, WarmupLinearSchedule
from transformers import AdamW, get_linear_schedule_with_warmup
from utils_squad import (read_squad_examples, convert_examples_to_features,
RawResult, write_predictions,
@@ -98,7 +98,7 @@ def train(args, train_dataset, model, tokenizer):
{'params': [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], 'weight_decay': 0.0}
]
optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate, eps=args.adam_epsilon)
scheduler = WarmupLinearSchedule(optimizer, warmup_steps=args.warmup_steps, t_total=t_total)
scheduler = get_linear_schedule_with_warmup(optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total)
if args.fp16:
try:
from apex import amp

14
templates/adding_a_new_model/README.md
View File

@@ -7,7 +7,7 @@ The library is designed to incorporate a variety of models and code bases. As su
One important point though is that the library has the following goals impacting the way models are incorporated:
- one specific feature of the API is the capability to run the model and tokenizer inline. The tokenization code thus often have to be slightly adapted to allow for running in the python interpreter.
- the package is also designed to be as self-consistent and with a small and reliable set of packages dependencies. In consequence, additional dependencies are usually not allowed when adding a model but can be allowed for the inclusion of a new tokenizer (recent examples of dependencies added for tokenizer specificites includes `sentencepiece` and `sacremoses`). Please make sure to check the existing dependencies when possible before adding a new one.
- the package is also designed to be as self-consistent and with a small and reliable set of packages dependencies. In consequence, additional dependencies are usually not allowed when adding a model but can be allowed for the inclusion of a new tokenizer (recent examples of dependencies added for tokenizer specificities include `sentencepiece` and `sacremoses`). Please make sure to check the existing dependencies when possible before adding a new one.
For a quick overview of the library organization, please check the [QuickStart section of the documentation](https://huggingface.co/transformers/quickstart.html).
@@ -20,7 +20,7 @@ Here an overview of the general workflow:
- [ ] add tests
- [ ] finalize
Let's details what should be done at each step
Let's detail what should be done at each step
## Adding model/configuration/tokenization classes
@@ -28,16 +28,16 @@ Here is the workflow for adding model/configuration/tokenization classes:
- [ ] copy the python files from the present folder to the main folder and rename them, replacing `xxx` with your model name,
- [ ] edit the files to replace `XXX` (with various casing) with your model name
- [ ] copy-past or create a simple configuration class for your model in the `configuration_...` file
- [ ] copy-past or create the code for your model in the `modeling_...` files (PyTorch and TF 2.0)
- [ ] copy-past or create a tokenizer class for your model in the `tokenization_...` file
- [ ] copy-paste or create a simple configuration class for your model in the `configuration_...` file
- [ ] copy-paste or create the code for your model in the `modeling_...` files (PyTorch and TF 2.0)
- [ ] copy-paste or create a tokenizer class for your model in the `tokenization_...` file
# Adding conversion scripts
Here is the workflow for the conversion scripts:
- [ ] copy the conversion script (`convert_...`) from the present folder to the main folder.
- [ ] edit this scipt to convert your original checkpoint weights to the current pytorch ones.
- [ ] edit this script to convert your original checkpoint weights to the current pytorch ones.
# Adding tests:
@@ -58,5 +58,5 @@ You can then finish the addition step by adding imports for your classes in the
- [ ] add your models and tokenizer to `pipeline.py`
- [ ] add a link to your conversion script in the main conversion utility (currently in `__main__` but will be moved to the `commands` subfolder in the near future)
- [ ] edit the PyTorch to TF 2.0 conversion script to add your model in the `convert_pytorch_checkpoint_to_tf2.py` file
- [ ] add a mention of your model in the doc: `README.md` and the documentation it-self at `docs/source/pretrained_models.rst`.
- [ ] add a mention of your model in the doc: `README.md` and the documentation itself at `docs/source/pretrained_models.rst`.
- [ ] upload the pretrained weigths, configurations and vocabulary files.

12
templates/adding_a_new_model/modeling_tf_xxx.py
View File

@@ -34,7 +34,7 @@ import numpy as np
import tensorflow as tf
from .configuration_xxx import XxxConfig
from .modeling_tf_utils import TFPreTrainedModel, get_initializer
from .modeling_tf_utils import TFPreTrainedModel, get_initializer, shape_list
from .file_utils import add_start_docstrings
logger = logging.getLogger(__name__)
@@ -51,7 +51,7 @@ TF_XXX_PRETRAINED_MODEL_ARCHIVE_MAP = {
####################################################
# TF 2.0 Models are constructed using Keras imperative API by sub-classing
# - tf.keras.layers.Layer for the layers and
# - TFPreTrainedModel for the models (it-self a sub-class of tf.keras.Model)
# - TFPreTrainedModel for the models (itself a sub-class of tf.keras.Model)
####################################################
####################################################
@@ -123,9 +123,9 @@ class TFXxxMainLayer(tf.keras.layers.Layer):
input_ids = inputs
if attention_mask is None:
attention_mask = tf.fill(tf.shape(input_ids), 1)
attention_mask = tf.fill(shape_list(input_ids), 1)
if token_type_ids is None:
token_type_ids = tf.fill(tf.shape(input_ids), 0)
token_type_ids = tf.fill(shape_list(input_ids), 0)
# We create a 3D attention mask from a 2D tensor mask.
# Sizes are [batch_size, 1, 1, to_seq_length]
@@ -257,6 +257,10 @@ XXX_INPUTS_DOCSTRING = r"""
Mask to nullify selected heads of the self-attention modules.
Mask values selected in ``[0, 1]``:
``1`` indicates the head is **not masked**, ``0`` indicates the head is **masked**.
**inputs_embeds**: (`optional`) ``Numpy array`` or ``tf.Tensor`` of shape ``(batch_size, sequence_length, embedding_dim)``:
Optionally, instead of passing ``input_ids`` you can choose to directly pass an embedded representation.
This is useful if you want more control over how to convert `input_ids` indices into associated vectors
than the model's internal embedding lookup matrix.
"""
@add_start_docstrings("The bare Xxx Model transformer outputing raw hidden-states without any specific head on top.",

49
templates/adding_a_new_model/modeling_xxx.py
View File

@@ -122,7 +122,7 @@ def load_tf_weights_in_xxx(model, config, tf_checkpoint_path):
####################################################
# PyTorch Models are constructed by sub-classing
# - torch.nn.Module for the layers and
# - PreTrainedModel for the models (it-self a sub-class of torch.nn.Module)
# - PreTrainedModel for the models (itself a sub-class of torch.nn.Module)
####################################################
####################################################
@@ -240,6 +240,10 @@ XXX_INPUTS_DOCSTRING = r"""
Mask to nullify selected heads of the self-attention modules.
Mask values selected in ``[0, 1]``:
``1`` indicates the head is **not masked**, ``0`` indicates the head is **masked**.
**inputs_embeds**: (`optional`) ``torch.FloatTensor`` of shape ``(batch_size, sequence_length, embedding_dim)``:
Optionally, instead of passing ``input_ids`` you can choose to directly pass an embedded representation.
This is useful if you want more control over how to convert `input_ids` indices into associated vectors
than the model's internal embedding lookup matrix.
"""
@add_start_docstrings("The bare Xxx Model transformer outputting raw hidden-states without any specific head on top.",
@@ -296,11 +300,22 @@ class XxxModel(XxxPreTrainedModel):
for layer, heads in heads_to_prune.items():
self.encoder.layer[layer].attention.prune_heads(heads)
def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None):
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None):
if input_ids is not None and inputs_embeds is not None:
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
elif input_ids is not None:
input_shape = input_ids.size()
elif inputs_embeds is not None:
input_shape = inputs_embeds.size()[:-1]
else:
raise ValueError("You have to specify either input_ids or inputs_embeds")
device = input_ids.device if input_ids is not None else inputs_embeds.device
if attention_mask is None:
attention_mask = torch.ones_like(input_ids)
attention_mask = torch.ones(input_shape, device=device)
if token_type_ids is None:
token_type_ids = torch.zeros_like(input_ids)
token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)
# We create a 3D attention mask from a 2D tensor mask.
# Sizes are [batch_size, 1, 1, to_seq_length]
@@ -334,7 +349,7 @@ class XxxModel(XxxPreTrainedModel):
##################################
# Replace this with your model code
embedding_output = self.embeddings(input_ids, position_ids=position_ids, token_type_ids=token_type_ids)
embedding_output = self.embeddings(input_ids=input_ids, position_ids=position_ids, token_type_ids=token_type_ids, inputs_embeds=inputs_embeds)
encoder_outputs = self.encoder(embedding_output, extended_attention_mask, head_mask=head_mask)
sequence_output = encoder_outputs[0]
outputs = (sequence_output,) + encoder_outputs[1:] # add hidden_states and attentions if they are here
@@ -385,14 +400,15 @@ class XxxForMaskedLM(XxxPreTrainedModel):
def get_output_embeddings(self):
return self.lm_head
def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None,
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None,
masked_lm_labels=None):
outputs = self.transformer(input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask)
head_mask=head_mask,
inputs_embeds=inputs_embeds)
sequence_output = outputs[0]
prediction_scores = self.cls(sequence_output)
@@ -450,14 +466,15 @@ class XxxForSequenceClassification(XxxPreTrainedModel):
self.init_weights()
def forward(self, input_ids, attention_mask=None, token_type_ids=None,
position_ids=None, head_mask=None, labels=None):
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None,
position_ids=None, head_mask=None, inputs_embeds=None, labels=None):
outputs = self.transformer(input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask)
head_mask=head_mask,
inputs_embeds=inputs_embeds)
pooled_output = outputs[1]
@@ -521,14 +538,15 @@ class XxxForTokenClassification(XxxPreTrainedModel):
self.init_weights()
def forward(self, input_ids, attention_mask=None, token_type_ids=None,
position_ids=None, head_mask=None, labels=None):
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None,
position_ids=None, head_mask=None, inputs_embeds=None, labels=None):
outputs = self.transformer(input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask)
head_mask=head_mask,
inputs_embeds=inputs_embeds)
sequence_output = outputs[0]
@@ -604,14 +622,15 @@ class XxxForQuestionAnswering(XxxPreTrainedModel):
self.init_weights()
def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None,
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None,
start_positions=None, end_positions=None):
outputs = self.transformer(input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask)
head_mask=head_mask,
inputs_embeds=inputs_embeds)
sequence_output = outputs[0]

7
templates/adding_a_new_model/tests/modeling_tf_xxx_test.py
View File

@@ -18,11 +18,11 @@ from __future__ import print_function
import unittest
import shutil
import pytest
import sys
from .modeling_tf_common_test import (TFCommonTestCases, ids_tensor)
from .configuration_common_test import ConfigTester
from .utils import require_tf, slow
from transformers import XxxConfig, is_tf_available
@@ -33,10 +33,9 @@ if is_tf_available():
TFXxxForTokenClassification,
TFXxxForQuestionAnswering,
TF_XXX_PRETRAINED_MODEL_ARCHIVE_MAP)
else:
pytestmark = pytest.mark.skip("Require TensorFlow")
@require_tf
class TFXxxModelTest(TFCommonTestCases.TFCommonModelTester):
all_model_classes = (TFXxxModel, TFXxxForMaskedLM, TFXxxForQuestionAnswering,
@@ -244,7 +243,7 @@ class TFXxxModelTest(TFCommonTestCases.TFCommonModelTester):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xxx_for_token_classification(*config_and_inputs)
@pytest.mark.slow
@slow
def test_model_from_pretrained(self):
cache_dir = "/tmp/transformers_test/"
for model_name in ['xxx-base-uncased']:

12
templates/adding_a_new_model/tests/modeling_xxx_test.py
View File

@@ -18,12 +18,12 @@ from __future__ import print_function
import unittest
import shutil
import pytest
from transformers import is_torch_available
from .modeling_common_test import (CommonTestCases, ids_tensor)
from .configuration_common_test import ConfigTester
from .utils import require_torch, slow, torch_device
if is_torch_available():
from transformers import (XxxConfig, XxxModel, XxxForMaskedLM,
@@ -31,10 +31,9 @@ if is_torch_available():
XxxForQuestionAnswering, XxxForSequenceClassification,
XxxForTokenClassification, XxxForMultipleChoice)
from transformers.modeling_xxx import XXX_PRETRAINED_MODEL_ARCHIVE_MAP
else:
pytestmark = pytest.mark.skip("Require Torch")
@require_torch
class XxxModelTest(CommonTestCases.CommonModelTester):
all_model_classes = (XxxModel, XxxForMaskedLM, XxxForQuestionAnswering,
@@ -131,6 +130,7 @@ class XxxModelTest(CommonTestCases.CommonModelTester):
def create_and_check_xxx_model(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
model = XxxModel(config=config)
model.to(torch_device)
model.eval()
sequence_output, pooled_output = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids)
sequence_output, pooled_output = model(input_ids, token_type_ids=token_type_ids)
@@ -148,6 +148,7 @@ class XxxModelTest(CommonTestCases.CommonModelTester):
def create_and_check_xxx_for_masked_lm(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
model = XxxForMaskedLM(config=config)
model.to(torch_device)
model.eval()
loss, prediction_scores = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, masked_lm_labels=token_labels)
result = {
@@ -162,6 +163,7 @@ class XxxModelTest(CommonTestCases.CommonModelTester):
def create_and_check_xxx_for_question_answering(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
model = XxxForQuestionAnswering(config=config)
model.to(torch_device)
model.eval()
loss, start_logits, end_logits = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids,
start_positions=sequence_labels, end_positions=sequence_labels)
@@ -182,6 +184,7 @@ class XxxModelTest(CommonTestCases.CommonModelTester):
def create_and_check_xxx_for_sequence_classification(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
config.num_labels = self.num_labels
model = XxxForSequenceClassification(config)
model.to(torch_device)
model.eval()
loss, logits = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=sequence_labels)
result = {
@@ -197,6 +200,7 @@ class XxxModelTest(CommonTestCases.CommonModelTester):
def create_and_check_xxx_for_token_classification(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
config.num_labels = self.num_labels
model = XxxForTokenClassification(config=config)
model.to(torch_device)
model.eval()
loss, logits = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=token_labels)
result = {
@@ -243,7 +247,7 @@ class XxxModelTest(CommonTestCases.CommonModelTester):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_xxx_for_token_classification(*config_and_inputs)
@pytest.mark.slow
@slow
def test_model_from_pretrained(self):
cache_dir = "/tmp/transformers_test/"
for model_name in list(XXX_PRETRAINED_MODEL_ARCHIVE_MAP.keys())[:1]:

2
templates/adding_a_new_model/tokenization_xxx.py
View File

@@ -172,7 +172,7 @@ class XxxTokenizer(PreTrainedTokenizer):
special tokens for the model
Returns:
A list of integers in the range [0, 1]: 0 for a special token, 1 for a sequence token.
A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
"""
if already_has_special_tokens: