# coding=utf-8
# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
# Copyright (c) 2018, NVIDIA CORPORATION. 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.
""" Configuration base class and utilities."""
import copy
import json
import logging
import os
from typing import Dict, Tuple
from .file_utils import CONFIG_NAME, cached_path, hf_bucket_url, is_remote_url
logger = logging.getLogger(__name__)
class PretrainedConfig(object):
r""" Base class for all configuration classes.
Handles a few parameters common to all models' configurations as well as methods for loading/downloading/saving configurations.
Note:
A configuration file can be loaded and saved to disk. Loading the configuration file and using this file to initialize a model does **not** load the model weights.
It only affects the model's configuration.
Class attributes (overridden by derived classes):
- ``model_type``: a string that identifies the model type, that we serialize into the JSON file, and that we use to recreate the correct object in :class:`~transformers.AutoConfig`.
Args:
finetuning_task (:obj:`string` or :obj:`None`, `optional`, defaults to :obj:`None`):
Name of the task used to fine-tune the model. This can be used when converting from an original (TensorFlow or PyTorch) checkpoint.
num_labels (:obj:`int`, `optional`, defaults to `2`):
Number of classes to use when the model is a classification model (sequences/tokens)
output_hidden_states (:obj:`bool`, `optional`, defaults to :obj:`False`):
Should the model returns all hidden-states.
output_attentions (:obj:`bool`, `optional`, defaults to :obj:`False`):
Should the model returns all attentions.
torchscript (:obj:`bool`, `optional`, defaults to :obj:`False`):
Is the model used with Torchscript (for PyTorch models).
"""
model_type: str = ""
def __init__(self, **kwargs):
# Attributes with defaults
self.output_hidden_states = kwargs.pop("output_hidden_states", False)
self.output_attentions = kwargs.pop("output_attentions", False)
self.use_cache = kwargs.pop("use_cache", True) # Not used by all models
self.torchscript = kwargs.pop("torchscript", False) # Only used by PyTorch models
self.use_bfloat16 = kwargs.pop("use_bfloat16", False)
self.pruned_heads = kwargs.pop("pruned_heads", {})
# Is decoder is used in encoder-decoder models to differentiate encoder from decoder
self.is_encoder_decoder = kwargs.pop("is_encoder_decoder", False)
self.is_decoder = kwargs.pop("is_decoder", False)
# Parameters for sequence generation
self.max_length = kwargs.pop("max_length", 20)
self.min_length = kwargs.pop("min_length", 0)
self.do_sample = kwargs.pop("do_sample", False)
self.early_stopping = kwargs.pop("early_stopping", False)
self.num_beams = kwargs.pop("num_beams", 1)
self.temperature = kwargs.pop("temperature", 1.0)
self.top_k = kwargs.pop("top_k", 50)
self.top_p = kwargs.pop("top_p", 1.0)
self.repetition_penalty = kwargs.pop("repetition_penalty", 1.0)
self.length_penalty = kwargs.pop("length_penalty", 1.0)
self.no_repeat_ngram_size = kwargs.pop("no_repeat_ngram_size", 0)
self.bad_words_ids = kwargs.pop("bad_words_ids", None)
self.num_return_sequences = kwargs.pop("num_return_sequences", 1)
# Fine-tuning task arguments
self.architectures = kwargs.pop("architectures", None)
self.finetuning_task = kwargs.pop("finetuning_task", None)
self.id2label = kwargs.pop("id2label", None)
self.label2id = kwargs.pop("label2id", None)
if self.id2label is not None:
kwargs.pop("num_labels", None)
self.id2label = dict((int(key), value) for key, value in self.id2label.items())
# Keys are always strings in JSON so convert ids to int here.
else:
self.num_labels = kwargs.pop("num_labels", 2)
# Tokenizer arguments TODO: eventually tokenizer and models should share the same config
self.prefix = kwargs.pop("prefix", None)
self.bos_token_id = kwargs.pop("bos_token_id", None)
self.pad_token_id = kwargs.pop("pad_token_id", None)
self.eos_token_id = kwargs.pop("eos_token_id", None)
self.decoder_start_token_id = kwargs.pop("decoder_start_token_id", None)
# task specific arguments
self.task_specific_params = kwargs.pop("task_specific_params", None)
# TPU arguments
self.xla_device = kwargs.pop("xla_device", None)
# Additional attributes without default values
for key, value in kwargs.items():
try:
setattr(self, key, value)
except AttributeError as err:
logger.error("Can't set {} with value {} for {}".format(key, value, self))
raise err
@property
def num_labels(self):
return len(self.id2label)
@num_labels.setter
def num_labels(self, num_labels):
self.id2label = {i: "LABEL_{}".format(i) for i in range(num_labels)}
self.label2id = dict(zip(self.id2label.values(), self.id2label.keys()))
def save_pretrained(self, save_directory):
"""
Save a configuration object to the directory `save_directory`, so that it
can be re-loaded using the :func:`~transformers.PretrainedConfig.from_pretrained` class method.
Args:
save_directory (:obj:`string`):
Directory where the configuration JSON file will be saved.
"""
if os.path.isfile(save_directory):
raise AssertionError("Provided path ({}) should be a directory, not a file".format(save_directory))
os.makedirs(save_directory, exist_ok=True)
# If we save using the predefined names, we can load using `from_pretrained`
output_config_file = os.path.join(save_directory, CONFIG_NAME)
self.to_json_file(output_config_file, use_diff=True)
logger.info("Configuration saved in {}".format(output_config_file))
@classmethod
def from_pretrained(cls, pretrained_model_name_or_path, **kwargs) -> "PretrainedConfig":
r"""
Instantiate a :class:`~transformers.PretrainedConfig` (or a derived class) from a pre-trained model configuration.
Args:
pretrained_model_name_or_path (:obj:`string`):
either:
- a string with the `shortcut name` of a pre-trained model configuration to load from cache or
download, e.g.: ``bert-base-uncased``.
- a string with the `identifier name` of a pre-trained model configuration that was user-uploaded to
our S3, e.g.: ``dbmdz/bert-base-german-cased``.
- a path to a `directory` containing a configuration file saved using the
:func:`~transformers.PretrainedConfig.save_pretrained` method, e.g.: ``./my_model_directory/``.
- a path or url to a saved configuration JSON `file`, e.g.:
``./my_model_directory/configuration.json``.
cache_dir (:obj:`string`, `optional`):
Path to a directory in which a downloaded pre-trained model
configuration should be cached if the standard cache should not be used.
kwargs (:obj:`Dict[str, any]`, `optional`):
The values in kwargs of any keys which are configuration attributes will be used to override the loaded
values. Behavior concerning key/value pairs whose keys are *not* configuration attributes is
controlled by the `return_unused_kwargs` keyword parameter.
force_download (:obj:`bool`, `optional`, defaults to :obj:`False`):
Force to (re-)download the model weights and configuration files and override the cached versions if they exist.
resume_download (:obj:`bool`, `optional`, defaults to :obj:`False`):
Do not delete incompletely recieved file. Attempt to resume the download if such a file exists.
proxies (:obj:`Dict`, `optional`):
A dictionary of proxy servers to use by protocol or endpoint, e.g.:
:obj:`{'http': 'foo.bar:3128', 'http://hostname': 'foo.bar:4012'}.`
The proxies are used on each request.
return_unused_kwargs: (`optional`) bool:
If False, then this function returns just the final configuration object.
If True, then this functions returns a :obj:`Tuple(config, unused_kwargs)` where `unused_kwargs` is a
dictionary consisting of the key/value pairs whose keys are not configuration attributes: ie the part
of kwargs which has not been used to update `config` and is otherwise ignored.
Returns:
:class:`PretrainedConfig`: An instance of a configuration object
Examples::
# We can't instantiate directly the base class `PretrainedConfig` so let's show the examples on a
# derived class: BertConfig
config = BertConfig.from_pretrained('bert-base-uncased') # Download configuration from S3 and cache.
config = BertConfig.from_pretrained('./test/saved_model/') # E.g. config (or model) was saved using `save_pretrained('./test/saved_model/')`
config = BertConfig.from_pretrained('./test/saved_model/my_configuration.json')
config = BertConfig.from_pretrained('bert-base-uncased', output_attention=True, foo=False)
assert config.output_attention == True
config, unused_kwargs = BertConfig.from_pretrained('bert-base-uncased', output_attention=True,
foo=False, return_unused_kwargs=True)
assert config.output_attention == True
assert unused_kwargs == {'foo': False}
"""
config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
return cls.from_dict(config_dict, **kwargs)
@classmethod
def get_config_dict(cls, pretrained_model_name_or_path: str, **kwargs) -> Tuple[Dict, Dict]:
"""
From a `pretrained_model_name_or_path`, resolve to a dictionary of parameters, to be used
for instantiating a Config using `from_dict`.
Parameters:
pretrained_model_name_or_path (:obj:`string`):
The identifier of the pre-trained checkpoint from which we want the dictionary of parameters.
Returns:
:obj:`Tuple[Dict, Dict]`: The dictionary that will be used to instantiate the configuration object.
"""
cache_dir = kwargs.pop("cache_dir", None)
force_download = kwargs.pop("force_download", False)
resume_download = kwargs.pop("resume_download", False)
proxies = kwargs.pop("proxies", None)
local_files_only = kwargs.pop("local_files_only", False)
if os.path.isdir(pretrained_model_name_or_path):
config_file = os.path.join(pretrained_model_name_or_path, CONFIG_NAME)
elif os.path.isfile(pretrained_model_name_or_path) or is_remote_url(pretrained_model_name_or_path):
config_file = pretrained_model_name_or_path
else:
config_file = hf_bucket_url(pretrained_model_name_or_path, filename=CONFIG_NAME, use_cdn=False)
try:
# Load from URL or cache if already cached
resolved_config_file = cached_path(
config_file,
cache_dir=cache_dir,
force_download=force_download,
proxies=proxies,
resume_download=resume_download,
local_files_only=local_files_only,
)
# Load config dict
if resolved_config_file is None:
raise EnvironmentError
config_dict = cls._dict_from_json_file(resolved_config_file)
except EnvironmentError:
msg = (
f"Can't load config for '{pretrained_model_name_or_path}'. Make sure that:\n\n"
f"- '{pretrained_model_name_or_path}' is a correct model identifier listed on 'https://huggingface.co/models'\n\n"
f"- or '{pretrained_model_name_or_path}' is the correct path to a directory containing a {CONFIG_NAME} file\n\n"
)
raise EnvironmentError(msg)
except json.JSONDecodeError:
msg = (
"Couldn't reach server at '{}' to download configuration file or "
"configuration file is not a valid JSON file. "
"Please check network or file content here: {}.".format(config_file, resolved_config_file)
)
raise EnvironmentError(msg)
if resolved_config_file == config_file:
logger.info("loading configuration file {}".format(config_file))
else:
logger.info("loading configuration file {} from cache at {}".format(config_file, resolved_config_file))
return config_dict, kwargs
@classmethod
def from_dict(cls, config_dict: Dict, **kwargs) -> "PretrainedConfig":
"""
Constructs a `Config` from a Python dictionary of parameters.
Args:
config_dict (:obj:`Dict[str, any]`):
Dictionary that will be used to instantiate the configuration object. Such a dictionary can be retrieved
from a pre-trained checkpoint by leveraging the :func:`~transformers.PretrainedConfig.get_config_dict`
method.
kwargs (:obj:`Dict[str, any]`):
Additional parameters from which to initialize the configuration object.
Returns:
:class:`PretrainedConfig`: An instance of a configuration object
"""
return_unused_kwargs = kwargs.pop("return_unused_kwargs", False)
config = cls(**config_dict)
if hasattr(config, "pruned_heads"):
config.pruned_heads = dict((int(key), value) for key, value in config.pruned_heads.items())
# Update config with kwargs if needed
to_remove = []
for key, value in kwargs.items():
if hasattr(config, key):
setattr(config, key, value)
to_remove.append(key)
for key in to_remove:
kwargs.pop(key, None)
logger.info("Model config %s", str(config))
if return_unused_kwargs:
return config, kwargs
else:
return config
@classmethod
def from_json_file(cls, json_file: str) -> "PretrainedConfig":
"""
Constructs a `Config` from the path to a json file of parameters.
Args:
json_file (:obj:`string`):
Path to the JSON file containing the parameters.
Returns:
:class:`PretrainedConfig`: An instance of a configuration object
"""
config_dict = cls._dict_from_json_file(json_file)
return cls(**config_dict)
@classmethod
def _dict_from_json_file(cls, json_file: str):
with open(json_file, "r", encoding="utf-8") as reader:
text = reader.read()
return json.loads(text)
def __eq__(self, other):
return self.__dict__ == other.__dict__
def __repr__(self):
return "{} {}".format(self.__class__.__name__, self.to_json_string())
def to_diff_dict(self):
"""
Removes all attributes from config which correspond to the default
config attributes for better readability and serializes to a Python
dictionary.
Returns:
:obj:`Dict[str, any]`: Dictionary of all the attributes that make up this configuration instance,
"""
config_dict = self.to_dict()
# get the default config dict
default_config_dict = PretrainedConfig().to_dict()
serializable_config_dict = {}
# only serialize values that differ from the default config
for key, value in config_dict.items():
if key not in default_config_dict or value != default_config_dict[key]:
serializable_config_dict[key] = value
return serializable_config_dict
def to_dict(self):
"""
Serializes this instance to a Python dictionary.
Returns:
:obj:`Dict[str, any]`: Dictionary of all the attributes that make up this configuration instance,
"""
output = copy.deepcopy(self.__dict__)
if hasattr(self.__class__, "model_type"):
output["model_type"] = self.__class__.model_type
return output
def to_json_string(self, use_diff=True):
"""
Serializes this instance to a JSON string.
Args:
use_diff (:obj:`bool`):
If set to True, only the difference between the config instance and the default PretrainedConfig() is serialized to JSON string.
Returns:
:obj:`string`: String containing all the attributes that make up this configuration instance in JSON format.
"""
if use_diff is True:
config_dict = self.to_diff_dict()
else:
config_dict = self.to_dict()
return json.dumps(config_dict, indent=2, sort_keys=True) + "\n"
def to_json_file(self, json_file_path, use_diff=True):
"""
Save this instance to a json file.
Args:
json_file_path (:obj:`string`):
Path to the JSON file in which this configuration instance's parameters will be saved.
use_diff (:obj:`bool`):
If set to True, only the difference between the config instance and the default PretrainedConfig() is serialized to JSON file.
"""
with open(json_file_path, "w", encoding="utf-8") as writer:
writer.write(self.to_json_string(use_diff=use_diff))
def update(self, config_dict: Dict):
"""
Updates attributes of this class
with attributes from `config_dict`.
Args:
:obj:`Dict[str, any]`: Dictionary of attributes that shall be updated for this class.
"""
for key, value in config_dict.items():
setattr(self, key, value)