# 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.
"""TF general model utils."""
from __future__ import annotations
import functools
import gc
import inspect
import json
import os
import pickle
import re
import warnings
from collections.abc import Mapping
from pathlib import Path
from typing import TYPE_CHECKING, Any, Callable, Dict, List, Optional, Union
import h5py
import numpy as np
import tensorflow as tf
from huggingface_hub import Repository, list_repo_files
from keras import backend as K
from packaging.version import parse
from . import DataCollatorWithPadding, DefaultDataCollator
from .activations_tf import get_tf_activation
from .configuration_utils import PretrainedConfig
from .dynamic_module_utils import custom_object_save
from .generation import GenerationConfig, TFGenerationMixin
from .tf_utils import (
expand_1d,
load_attributes_from_hdf5_group,
save_attributes_to_hdf5_group,
shape_list,
)
from .utils import (
SAFE_WEIGHTS_INDEX_NAME,
SAFE_WEIGHTS_NAME,
TF2_WEIGHTS_INDEX_NAME,
TF2_WEIGHTS_NAME,
TF_WEIGHTS_NAME,
WEIGHTS_INDEX_NAME,
WEIGHTS_NAME,
ModelOutput,
PushToHubMixin,
cached_file,
download_url,
find_labels,
has_file,
is_offline_mode,
is_remote_url,
is_safetensors_available,
is_tf_symbolic_tensor,
logging,
requires_backends,
working_or_temp_dir,
)
from .utils.hub import convert_file_size_to_int, get_checkpoint_shard_files
if is_safetensors_available():
from safetensors import safe_open
from safetensors.tensorflow import save_file as safe_save_file
if TYPE_CHECKING:
from . import PreTrainedTokenizerBase
logger = logging.get_logger(__name__)
tf_logger = tf.get_logger()
TFModelInputType = Union[
List[tf.Tensor],
List[np.ndarray],
Dict[str, tf.Tensor],
Dict[str, np.ndarray],
tf.Tensor,
np.ndarray,
]
def dummy_loss(y_true, y_pred):
if y_pred.shape.rank <= 1:
return y_pred
else:
reduction_axes = list(range(1, y_pred.shape.rank))
return tf.reduce_mean(y_pred, axis=reduction_axes)
class TFModelUtilsMixin:
"""
A few utilities for `tf.keras.Model`, to be used as a mixin.
"""
def num_parameters(self, only_trainable: bool = False) -> int:
"""
Get the number of (optionally, trainable) parameters in the model.
Args:
only_trainable (`bool`, *optional*, defaults to `False`):
Whether or not to return only the number of trainable parameters
Returns:
`int`: The number of parameters.
"""
if only_trainable:
return int(sum(np.prod(w.shape.as_list()) for w in self.trainable_variables))
else:
return self.count_params()
def keras_serializable(cls):
"""
Decorate a Keras Layer class to support Keras serialization.
This is done by:
1. Adding a `transformers_config` dict to the Keras config dictionary in `get_config` (called by Keras at
serialization time.
2. Wrapping `__init__` to accept that `transformers_config` dict (passed by Keras at deserialization time) and
convert it to a config object for the actual layer initializer.
3. Registering the class as a custom object in Keras (if the Tensorflow version supports this), so that it does not
need to be supplied in `custom_objects` in the call to `tf.keras.models.load_model`.
Args:
cls (a `tf.keras.layers.Layers subclass`):
Typically a `TF.MainLayer` class in this project, in general must accept a `config` argument to its
initializer.
Returns:
The same class object, with modifications for Keras deserialization.
"""
initializer = cls.__init__
config_class = getattr(cls, "config_class", None)
if config_class is None:
raise AttributeError("Must set `config_class` to use @keras_serializable")
@functools.wraps(initializer)
def wrapped_init(self, *args, **kwargs):
config = args[0] if args and isinstance(args[0], PretrainedConfig) else kwargs.pop("config", None)
if isinstance(config, dict):
config = config_class.from_dict(config)
initializer(self, config, *args, **kwargs)
elif isinstance(config, PretrainedConfig):
if len(args) > 0:
initializer(self, *args, **kwargs)
else:
initializer(self, config, *args, **kwargs)
else:
raise ValueError("Must pass either `config` (PretrainedConfig) or `config` (dict)")
self._config = config
self._kwargs = kwargs
cls.__init__ = wrapped_init
if not hasattr(cls, "get_config"):
raise TypeError("Only use @keras_serializable on tf.keras.layers.Layer subclasses")
if hasattr(cls.get_config, "_is_default"):
def get_config(self):
cfg = super(cls, self).get_config()
cfg["config"] = self._config.to_dict()
cfg.update(self._kwargs)
return cfg
cls.get_config = get_config
cls._keras_serializable = True
if hasattr(tf.keras.utils, "register_keras_serializable"):
cls = tf.keras.utils.register_keras_serializable()(cls)
return cls
class TFCausalLanguageModelingLoss:
"""
Loss function suitable for causal language modeling (CLM), that is, the task of guessing the next token.
Any label of -100 will be ignored (along with the corresponding logits) in the loss computation.
"""
def hf_compute_loss(self, labels, logits):
loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(
from_logits=True, reduction=tf.keras.losses.Reduction.NONE
)
if self.config.tf_legacy_loss:
# make sure only labels that are not equal to -100 affect the loss
active_loss = tf.not_equal(tf.reshape(labels, (-1,)), -100)
reduced_logits = tf.boolean_mask(tf.reshape(logits, (-1, shape_list(logits)[2])), active_loss)
labels = tf.boolean_mask(tf.reshape(labels, (-1,)), active_loss)
return loss_fn(labels, reduced_logits)
# Clip negative labels to zero here to avoid NaNs and errors - those positions will get masked later anyway
unmasked_loss = loss_fn(tf.nn.relu(labels), logits)
# make sure only labels that are not equal to -100 affect the loss
loss_mask = tf.cast(labels != -100, dtype=unmasked_loss.dtype)
masked_loss = unmasked_loss * loss_mask
reduced_masked_loss = tf.reduce_sum(masked_loss) / tf.reduce_sum(loss_mask)
return tf.reshape(reduced_masked_loss, (1,))
class TFQuestionAnsweringLoss:
"""
Loss function suitable for question answering.
"""
def hf_compute_loss(self, labels, logits):
loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(
from_logits=True, reduction=tf.keras.losses.Reduction.NONE
)
start_loss = loss_fn(labels["start_position"], logits[0])
end_loss = loss_fn(labels["end_position"], logits[1])
return (start_loss + end_loss) / 2.0
class TFTokenClassificationLoss:
"""
Loss function suitable for token classification.
Any label of -100 will be ignored (along with the corresponding logits) in the loss computation.
"""
def hf_compute_loss(self, labels, logits):
loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(
from_logits=True, reduction=tf.keras.losses.Reduction.NONE
)
if tf.executing_eagerly(): # Data-dependent conditionals are forbidden in XLA
if tf.math.reduce_any(labels == -1):
tf.print("Using `-1` to mask the loss for the token is deprecated. Please use `-100` instead.")
if self.config.tf_legacy_loss:
# make sure only labels that are not equal to -100
# are taken into account as loss
if tf.math.reduce_any(labels == -1):
tf.print("Using `-1` to mask the loss for the token is deprecated. Please use `-100` instead.")
active_loss = tf.reshape(labels, (-1,)) != -1
else:
active_loss = tf.reshape(labels, (-1,)) != -100
reduced_logits = tf.boolean_mask(tf.reshape(logits, (-1, shape_list(logits)[2])), active_loss)
labels = tf.boolean_mask(tf.reshape(labels, (-1,)), active_loss)
return loss_fn(labels, reduced_logits)
# Clip negative labels to zero here to avoid NaNs and errors - those positions will get masked later anyway
unmasked_loss = loss_fn(tf.nn.relu(labels), logits)
# make sure only labels that are not equal to -100 or -1
# are taken into account as loss
loss_mask = tf.cast(labels >= 0, dtype=unmasked_loss.dtype)
# Avoid possible division by zero later
# Masked positions will have a loss of NaN because -100 and -1 are not valid labels
masked_loss = unmasked_loss * loss_mask
reduced_masked_loss = tf.reduce_sum(masked_loss) / tf.reduce_sum(loss_mask)
return tf.reshape(reduced_masked_loss, (1,))
class TFSequenceClassificationLoss:
"""
Loss function suitable for sequence classification.
"""
def hf_compute_loss(self, labels, logits):
if logits.shape.rank == 1 or logits.shape[1] == 1:
loss_fn = tf.keras.losses.MeanSquaredError(reduction=tf.keras.losses.Reduction.NONE)
if labels.shape.rank == 1:
# MeanSquaredError returns a scalar loss if the labels are 1D, so avoid that
labels = tf.expand_dims(labels, axis=-1)
else:
loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(
from_logits=True, reduction=tf.keras.losses.Reduction.NONE
)
return loss_fn(labels, logits)
class TFMultipleChoiceLoss:
"""Loss function suitable for multiple choice tasks."""
def hf_compute_loss(self, labels, logits):
loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(
from_logits=True, reduction=tf.keras.losses.Reduction.NONE
)
return loss_fn(labels, logits)
class TFMaskedLanguageModelingLoss(TFCausalLanguageModelingLoss):
"""
Loss function suitable for masked language modeling (MLM), that is, the task of guessing the masked tokens.
Any label of -100 will be ignored (along with the corresponding logits) in the loss computation.
"""
class TFNextSentencePredictionLoss:
"""
Loss function suitable for next sentence prediction (NSP), that is, the task of guessing the next sentence.
Any label of -100 will be ignored (along with the corresponding logits) in the loss computation.
"""
def hf_compute_loss(self, labels, logits):
loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(
from_logits=True, reduction=tf.keras.losses.Reduction.NONE
)
if self.config.tf_legacy_loss:
# make sure only labels that are not equal to -100
# are taken into account as loss
next_sentence_active_loss = tf.not_equal(tf.reshape(labels, (-1,)), -100)
next_sentence_reduced_logits = tf.boolean_mask(tf.reshape(logits, (-1, 2)), next_sentence_active_loss)
next_sentence_label = tf.boolean_mask(tf.reshape(labels, (-1,)), next_sentence_active_loss)
return loss_fn(next_sentence_label, next_sentence_reduced_logits)
# make sure only labels that are not equal to -100
# are taken into account as loss
# Clip negative labels to zero here to avoid NaNs and errors - those positions will get masked later anyway
unmasked_ns_loss = loss_fn(y_true=tf.nn.relu(labels), y_pred=logits)
ns_loss_mask = tf.cast(labels != -100, dtype=unmasked_ns_loss.dtype)
# Just zero out samples where label is -100, no reduction
masked_ns_loss = unmasked_ns_loss * ns_loss_mask
return masked_ns_loss
def booleans_processing(config, **kwargs):
"""
Process the input booleans of each model.
Args:
config ([`PretrainedConfig`]):
The config of the running model.
**kwargs:
The boolean parameters
Returns:
A dictionary with the proper values for each boolean
"""
final_booleans = {}
# Pure conv models (such as ConvNext) do not have `output_attentions`. If the signature has
# `output_attentions`, it will be present here in `kwargs`, even if unset (in that case, as `None`)
if "output_attentions" in kwargs:
final_booleans["output_attentions"] = (
kwargs["output_attentions"] if kwargs["output_attentions"] is not None else config.output_attentions
)
final_booleans["output_hidden_states"] = (
kwargs["output_hidden_states"] if kwargs["output_hidden_states"] is not None else config.output_hidden_states
)
final_booleans["return_dict"] = kwargs["return_dict"] if kwargs["return_dict"] is not None else config.return_dict
if "use_cache" in kwargs:
final_booleans["use_cache"] = (
kwargs["use_cache"] if kwargs["use_cache"] is not None else getattr(config, "use_cache", None)
)
return final_booleans
def unpack_inputs(func):
"""
Decorator that processes the inputs to a Keras layer, passing them to the layer as keyword arguments. This enables
downstream use of the inputs by their variable name, even if they arrive packed as a dictionary in the first input
(common case in Keras).
Args:
func (`callable`):
The callable function of the TensorFlow model.
Returns:
A callable that wraps the original `func` with the behavior described above.
"""
original_signature = inspect.signature(func)
@functools.wraps(func)
def run_call_with_unpacked_inputs(self, *args, **kwargs):
# isolates the actual `**kwargs` for the decorated function
kwargs_call = {key: val for key, val in kwargs.items() if key not in dict(original_signature.parameters)}
fn_args_and_kwargs = {key: val for key, val in kwargs.items() if key not in kwargs_call}
fn_args_and_kwargs.update({"kwargs_call": kwargs_call})
# move any arg into kwargs, if they exist
fn_args_and_kwargs.update(dict(zip(func.__code__.co_varnames[1:], args)))
# Encoder Decoder models delegate the application of the configuration options to their inner models.
if "EncoderDecoder" in self.__class__.__name__:
config = None
else:
config = self.config
unpacked_inputs = input_processing(func, config, **fn_args_and_kwargs)
return func(self, **unpacked_inputs)
# Keras enforces the first layer argument to be passed, and checks it through `inspect.getfullargspec()`. This
# function does not follow wrapper chains (i.e. ignores `functools.wraps()`), meaning that without the line below
# Keras would attempt to check the first argument against the literal signature of the wrapper.
run_call_with_unpacked_inputs.__signature__ = original_signature
return run_call_with_unpacked_inputs
def input_processing(func, config, **kwargs):
"""
Process the input of each TensorFlow model including the booleans. In case of a list of symbolic inputs, each input
has to be named accordingly to the parameters name, i.e. `input_ids = tf.keras.Input(shape=(128,), dtype='int32',
name="input_ids")` otherwise the order of the tensors will not be guaranteed during the training.
Args:
func (`callable`):
The callable function of the TensorFlow model.
config ([`PretrainedConfig`]):
The config of the running model.
**kwargs:
The inputs of the model.
Returns:
Two lists, one for the missing layers, and another one for the unexpected layers.
"""
signature = dict(inspect.signature(func).parameters)
has_kwargs = bool(signature.pop("kwargs", None))
signature.pop("self", None)
parameter_names = list(signature.keys())
main_input_name = parameter_names[0]
main_input = kwargs.pop(main_input_name, None)
output = {}
allowed_types = (tf.Tensor, bool, int, ModelOutput, tuple, list, dict, np.ndarray)
if "inputs" in kwargs["kwargs_call"]:
warnings.warn(
"The `inputs` argument is deprecated and will be removed in a future version, use `input_ids` instead.",
FutureWarning,
)
output["input_ids"] = kwargs["kwargs_call"].pop("inputs")
if "decoder_cached_states" in kwargs["kwargs_call"]:
warnings.warn(
"The `decoder_cached_states` argument is deprecated and will be removed in a future version, use"
" `past_key_values` instead.",
FutureWarning,
)
output["past_key_values"] = kwargs["kwargs_call"].pop("decoder_cached_states")
if "past" in kwargs["kwargs_call"] and "past_key_values" in parameter_names:
warnings.warn(
"The `past` argument is deprecated and will be removed in a future version, use `past_key_values`"
" instead.",
FutureWarning,
)
kwargs["past_key_values"] = kwargs["kwargs_call"].pop("past")
elif "past_key_values" in kwargs["kwargs_call"] and "past" in parameter_names:
kwargs["past"] = kwargs["kwargs_call"].pop("past_key_values")
if has_kwargs:
output["kwargs"] = kwargs.pop("kwargs_call", {})
else:
if len(kwargs["kwargs_call"]) > 0:
raise ValueError(
"The following keyword arguments are not supported by this model:"
f" {list(kwargs['kwargs_call'].keys())}."
)
kwargs.pop("kwargs_call")
for k, v in kwargs.items():
if isinstance(v, allowed_types) or tf.is_tensor(v) or v is None:
output[k] = v
else:
raise ValueError(f"Data of type {type(v)} is not allowed only {allowed_types} is accepted for {k}.")
if isinstance(main_input, (tuple, list)):
for i, input in enumerate(main_input):
# EagerTensors don't allow to use the .name property so we check for a real Tensor
if is_tf_symbolic_tensor(input):
# Tensor names have always the pattern `name:id` then we check only the
# `name` part
tensor_name = input.name.split(":")[0]
if tensor_name in parameter_names:
output[tensor_name] = input
else:
output[parameter_names[i]] = input
elif isinstance(input, allowed_types) or input is None:
output[parameter_names[i]] = input
else:
raise ValueError(
f"Data of type {type(input)} is not allowed only {allowed_types} is accepted for"
f" {parameter_names[i]}."
)
elif isinstance(main_input, Mapping):
if "inputs" in main_input:
warnings.warn(
"The `inputs` argument is deprecated and will be removed in a future version, use `input_ids`"
" instead.",
FutureWarning,
)
output["input_ids"] = main_input.pop("inputs")
if "decoder_cached_states" in main_input:
warnings.warn(
"The `decoder_cached_states` argument is deprecated and will be removed in a future version, use"
" `past_key_values` instead.",
FutureWarning,
)
output["past_key_values"] = main_input.pop("decoder_cached_states")
for k, v in dict(main_input).items():
if isinstance(v, allowed_types) or v is None:
output[k] = v
elif k not in parameter_names and "args" not in parameter_names:
logger.warning(
f"The parameter {k} does not belongs to the parameter list {parameter_names} and will be ignored."
)
continue
else:
raise ValueError(f"Data of type {type(v)} is not allowed only {allowed_types} is accepted for {k}.")
else:
if tf.is_tensor(main_input) or main_input is None:
output[main_input_name] = main_input
else:
raise ValueError(
f"Data of type {type(main_input)} is not allowed only {allowed_types} is accepted for"
f" {main_input_name}."
)
# Populates any unspecified argument with their default value, according to the signature.
for name in parameter_names:
if name not in list(output.keys()) and name != "args":
output[name] = kwargs.pop(name, signature[name].default)
# When creating a SavedModel TF calls the method with LayerCall.__call__(args, **kwargs)
# So to respect the proper output we have to add this exception
if "args" in output:
if output["args"] is not None and is_tf_symbolic_tensor(output["args"]):
tensor_name = output["args"].name.split(":")[0]
output[tensor_name] = output["args"]
else:
# `args` in this case is always the first parameter, then `input_ids`
output["input_ids"] = output["args"]
del output["args"]
if "kwargs" in output:
del output["kwargs"]
cast_output = {}
for key, val in output.items():
if isinstance(val, tf.Tensor) and val.dtype == tf.int64:
cast_output[key] = tf.cast(val, tf.int32)
elif isinstance(val, np.ndarray) and val.dtype == np.int64:
cast_output[key] = val.astype(np.int32)
else:
cast_output[key] = val
output = cast_output
del cast_output
if config is not None:
boolean_dict = {
k: v
for k, v in output.items()
if k in ["return_dict", "output_attentions", "output_hidden_states", "use_cache"]
}
output.update(
booleans_processing(
config=config,
**boolean_dict,
)
)
return output
def dtype_byte_size(dtype):
"""
Returns the size (in bytes) occupied by one parameter of type `dtype`.
Example:
```py
>>> dtype_byte_size(tf.float32)
4
```
"""
if dtype == tf.bool:
return 1 / 8
bit_search = re.search(r"[^\d](\d+)$", dtype.name)
if bit_search is None:
raise ValueError(f"`dtype` is not a valid dtype: {dtype}.")
bit_size = int(bit_search.groups()[0])
return bit_size // 8
def strip_model_name_and_prefix(name, _prefix=None):
if _prefix is not None and name.startswith(_prefix):
name = name[len(_prefix) :]
if name.startswith("/"):
name = name[1:]
if "model." not in name and len(name.split("/")) > 1:
name = "/".join(name.split("/")[1:])
return name
def tf_shard_checkpoint(weights, max_shard_size="10GB"):
"""
Splits a model state dictionary in sub-checkpoints so that the final size of each sub-checkpoint does not exceed a
given size.
The sub-checkpoints are determined by iterating through the `state_dict` in the order of its keys, so there is no
optimization made to make each sub-checkpoint as close as possible to the maximum size passed. For example, if the
limit is 10GB and we have weights of sizes [6GB, 6GB, 2GB, 6GB, 2GB, 2GB] they will get sharded as [6GB], [6+2GB],
[6+2+2GB] and not [6+2+2GB], [6+2GB], [6GB].
If one of the model's weight is bigger that `max_shard_size`, it will end up in its own sub-checkpoint which will
have a size greater than `max_shard_size`.
Args:
weights (`Dict[str, tf.RessourceVariable]`): The list of tf.RessourceVariable of a model to save.
max_shard_size (`int` or `str`, *optional*, defaults to `"10GB"`):
The maximum size of each sub-checkpoint. If expressed as a string, needs to be digits followed by a unit
(like `"5MB"`).
"""
max_shard_size = convert_file_size_to_int(max_shard_size)
sharded_state_dicts = []
current_block = []
current_block_size = 0
total_size = 0
for item in weights:
weight_size = item.numpy().size * dtype_byte_size(item.dtype)
# If this weight is going to tip up over the maximal size, we split.
if current_block_size + weight_size > max_shard_size:
sharded_state_dicts.append(current_block)
current_block = []
current_block_size = 0
current_block.append(item)
current_block_size += weight_size
total_size += weight_size
# Add the last block
sharded_state_dicts.append(current_block)
# If we only have one shard, we return it
if len(sharded_state_dicts) == 1:
return {TF2_WEIGHTS_NAME: sharded_state_dicts[0]}, None
# Otherwise, let's build the index
weight_map = {}
shards = {}
for idx, shard in enumerate(sharded_state_dicts):
shard_file = TF2_WEIGHTS_NAME.replace(".h5", f"-{idx+1:05d}-of-{len(sharded_state_dicts):05d}.h5")
shards[shard_file] = shard
for weight in shard:
weight_name = weight.name
weight_map[weight_name] = shard_file
# Add the metadata
metadata = {"total_size": total_size}
index = {"metadata": metadata, "weight_map": weight_map}
return shards, index
def load_tf_sharded_weights(model, shard_files, ignore_mismatched_sizes=False, strict=False, _prefix=None):
"""
This is the same as `load_tf_weights` but for a sharded checkpoint. Detect missing and unexpected layers and load
the TF weights from the shard file accordingly to their names and shapes.
This load is performed efficiently: each checkpoint shard is loaded one by one in RAM and deleted after being
loaded in the model.
Args:
model (`tf.keras.models.Model`): The model in which to load the checkpoint.
shard_files (`str` or `os.PathLike`): A list containing the sharded checkpoint names.
ignore_mismatched_sizes`bool`, *optional`, defaults to `True`):
Whether or not to ignore the mismatch between the sizes
strict (`bool`, *optional*, defaults to `True`):
Whether to strictly enforce that the keys in the model state dict match the keys in the sharded checkpoint.
Returns:
Three lists, one for the missing layers, another one for the unexpected layers, and a last one for the
mismatched layers.
"""
# Load the index
unexpected_keys = set()
saved_keys = set()
mismatched_keys = set()
# Since TF adds the name of the class to its weights, and uses the index and not the name of the layer to load
# the weight, we have to get rid of the first prefix of the name of the layer.
model_keys = set()
model_layer_map = {}
for i, k in enumerate(model.weights):
layer_name = k.name
if _prefix is not None and layer_name.startswith(_prefix):
layer_name = layer_name[len(_prefix) :]
layer_name = layer_name.lstrip("/")
if not ("model." in layer_name or len(layer_name.split("/")) == 1):
layer_name = "/".join(layer_name.split("/")[1:])
model_keys.add(layer_name)
model_layer_map[layer_name] = i
for shard_file in shard_files:
saved_weight_names_set, unexpected_keys_set, mismatched_keys_set = load_tf_shard(
model,
model_layer_map,
shard_file,
ignore_mismatched_sizes=ignore_mismatched_sizes,
_prefix=_prefix,
)
saved_keys.update(saved_weight_names_set)
unexpected_keys.update(unexpected_keys_set)
mismatched_keys.update(mismatched_keys_set)
gc.collect()
missing_keys = model_keys - saved_keys
if strict and (len(missing_keys) > 0 or len(unexpected_keys) > 0):
error_message = f"Error(s) in loading state_dict for {model.__class__.__name__}"
if len(missing_keys) > 0:
str_missing_keys = ",".join([f'"{k}"' for k in missing_keys])
error_message += f"\nMissing key(s): {str_missing_keys}."
if len(unexpected_keys) > 0:
str_unexpected_keys = ",".join([f'"{k}"' for k in unexpected_keys])
error_message += f"\nMissing key(s): {str_unexpected_keys}."
raise RuntimeError(error_message)
return missing_keys, unexpected_keys, mismatched_keys
def load_tf_shard(model, model_layer_map, resolved_archive_file, ignore_mismatched_sizes=False, _prefix=None):
"""
Loads a shard from a sharded checkpoint file. Handles the missing keys and unexpected keys.
Args:
model (`tf.keras.models.Model`): Model in which the weights are loaded
model_layer_map (`Dict`): A dictionary mapping the layer name to the index of the layer in the model.
resolved_archive_file (`str`): Path to the checkpoint file from which the weights will be loaded
ignore_mismatched_sizes (`bool`, *optional*, defaults to `False`): Whether to ignore the mismatched keys
Returns:
`tf.keras.models.Model`: Three lists, one for the layers that were found and succesfully restored (from the
shard file), one for the mismatched layers, and another one for the unexpected layers.
"""
saved_weight_names_set = set()
saved_weights = {}
mismatched_keys = set()
unexpected_keys = set()
# Read the H5 file
try:
with h5py.File(resolved_archive_file, "r") as sharded_checkpoint_file:
# Retrieve the name of each layer from the H5 file
saved_h5_model_layers_name = set(load_attributes_from_hdf5_group(sharded_checkpoint_file, "layer_names"))
weight_value_tuples = []
# Compute missing and unexpected sub layers
# Store the weights in list of tuples that looks like [(weight_object, value_of_weight),...]
for layer_name in saved_h5_model_layers_name:
h5_layer_object = sharded_checkpoint_file[layer_name]
saved_weights[layer_name] = np.asarray(h5_layer_object)
saved_weight_names_set.add(layer_name)
if layer_name not in model_layer_map:
unexpected_keys.add(layer_name)
else:
symbolic_weight = model.weights[model_layer_map[layer_name]]
saved_weight_value = saved_weights[layer_name]
# If the current weight is found
if saved_weight_value is not None:
# Check if the shape of the current weight and the one from the H5 file are different
if K.int_shape(symbolic_weight) != saved_weight_value.shape:
# If yes we reshape the weight from the H5 file accordingly to the current weight
# If the two shapes are not compatible we raise an issue
try:
array = np.reshape(saved_weight_value, K.int_shape(symbolic_weight))
except ValueError as e:
if ignore_mismatched_sizes:
mismatched_keys.add(
(layer_name, saved_weight_value.shape, K.int_shape(symbolic_weight))
)
continue
else:
raise e
else:
array = saved_weight_value
# We create the tuple that will be loaded and add it to the final list
weight_value_tuples.append((symbolic_weight, array))
K.batch_set_value(weight_value_tuples)
return saved_weight_names_set, unexpected_keys, mismatched_keys
except Exception as e:
try:
with open(resolved_archive_file) as f:
if f.read().startswith("version"):
raise OSError(
"You seem to have cloned a repository without having git-lfs installed. Please install "
"git-lfs and run `git lfs install` followed by `git lfs pull` in the folder "
"you cloned."
)
else:
raise ValueError(
f"Unable to locate the file {resolved_archive_file} which is necessary to load this pretrained"
" model. Make sure you have saved the model properly."
) from e
except (UnicodeDecodeError, ValueError):
raise OSError(
f"Unable to load weights from TF checkpoint file for '{resolved_archive_file}' "
f"at '{resolved_archive_file}'. "
"If you tried to load a TF model from a sharded checkpoint, you should try converting the model "
"by loading it in pytorch and saving it localy. A convertion script should be realeased soon."
)
def load_tf_weights(model, resolved_archive_file, ignore_mismatched_sizes=False, _prefix=None):
"""
Detect missing and unexpected layers and load the TF weights from the shard file accordingly to their names and
shapes.
Args:
model (`tf.keras.models.Model`):
The model to load the weights into.
resolved_archive_file (`str`):
The location of the H5 file.
ignore_mismatched_sizes (`bool`, *optional*, defaults to `False`):
Whether or not to ignore weights with shapes that don't match between the checkpoint of the model.
Returns:
Three lists, one for the missing layers, another one for the unexpected layers, and a last one for the
mismatched layers.
"""
if resolved_archive_file.endswith(".safetensors"):
load_function = load_tf_weights_from_safetensors
else:
load_function = load_tf_weights_from_h5
return load_function(
model, resolved_archive_file, ignore_mismatched_sizes=ignore_mismatched_sizes, _prefix=_prefix
)
def load_tf_weights_from_h5(model, resolved_archive_file, ignore_mismatched_sizes=False, _prefix=None):
mismatched_layers = []
# Read the H5 file
with h5py.File(resolved_archive_file, "r") as sharded_checkpoint_file:
# Retrieve the name of each layer from the H5 file
saved_h5_model_layers_name = set(load_attributes_from_hdf5_group(sharded_checkpoint_file, "layer_names"))
# Find the missing layers from the high level list of layers
missing_layers = list({layer.name for layer in model.layers} - saved_h5_model_layers_name)
# Find the unexpected layers from the high level list of layers
unexpected_layers = list(saved_h5_model_layers_name - {layer.name for layer in model.layers})
saved_weight_names_set = set()
symbolic_weights_names = set()
weight_value_tuples = []
# Compute missing and unexpected sub layers
# Store the weights in list of tuples that looks like [(weight_object, value_of_weight),...]
for layer in model.layers:
# if layer_name from the H5 file belongs to the layers from the instantiated model
if layer.name in saved_h5_model_layers_name:
# Get the H5 layer object from its name
h5_layer_object = sharded_checkpoint_file[layer.name]
# Get all the weights as a list from the layer object
symbolic_weights = layer.trainable_weights + layer.non_trainable_weights
saved_weights = {}
# Create a dict from the H5 saved model that looks like {"weight_name": weight_value}
# And a set with only the names
for weight_name in load_attributes_from_hdf5_group(h5_layer_object, "weight_names"):
# TF names always start with the model name so we ignore it
name = "/".join(weight_name.split("/")[1:])
if _prefix is not None:
name = _prefix + "/" + name
saved_weights[name] = np.asarray(h5_layer_object[weight_name])
# Add the updated name to the final list for computing missing/unexpected values
saved_weight_names_set.add(name)
# Loop over each weights from the instantiated model and compare with the weights from the H5 file
for symbolic_weight in symbolic_weights:
# TF names always start with the model name so we ignore it
if _prefix is not None:
delimeter = len(_prefix.split("/"))
symbolic_weight_name = "/".join(
symbolic_weight.name.split("/")[:delimeter]
+ symbolic_weight.name.split("/")[delimeter + 1 :]
)
else:
symbolic_weight_name = "/".join(symbolic_weight.name.split("/")[1:])
# here we check if the current weight is among the weights from the H5 file
# If yes, get the weight_value of the corresponding weight from the H5 file
# If not, make the value to None
saved_weight_value = saved_weights.get(symbolic_weight_name, None)
# Retrocompatibility patch: some embeddings are stored with the weights name (e.g. Bart's
# `model.shared/embeddings:0` are stored as `model.shared/weights:0`)
if saved_weight_value is None and symbolic_weight_name.endswith("embeddings:0"):
symbolic_weight_name = symbolic_weight_name[:-12] + "weight:0"
saved_weight_value = saved_weights.get(symbolic_weight_name, None)
# Add the updated name to the final list for computing missing/unexpected values
symbolic_weights_names.add(symbolic_weight_name)
# If the current weight is found
if saved_weight_value is not None:
# Check if the shape of the current weight and the one from the H5 file are different
if K.int_shape(symbolic_weight) != saved_weight_value.shape:
# If yes we reshape the weight from the H5 file accordingly to the current weight
# If the two shapes are not compatible we raise an issue
try:
array = np.reshape(saved_weight_value, K.int_shape(symbolic_weight))
except ValueError as e:
if ignore_mismatched_sizes:
mismatched_layers.append(
(symbolic_weight_name, saved_weight_value.shape, K.int_shape(symbolic_weight))
)
continue
else:
raise e
else:
array = saved_weight_value
# We create the tuple that will be loaded and add it to the final list
weight_value_tuples.append((symbolic_weight, array))
# Load all the weights
K.batch_set_value(weight_value_tuples)
# Compute the missing and unexpected layers
missing_layers.extend(list(symbolic_weights_names - saved_weight_names_set))
unexpected_layers.extend(list(saved_weight_names_set - symbolic_weights_names))
return missing_layers, unexpected_layers, mismatched_layers
def load_tf_weights_from_safetensors(model, resolved_archive_file, ignore_mismatched_sizes=False, _prefix=None):
# Read the safetensors file
with safe_open(resolved_archive_file, framework="tf") as safetensors_archive:
mismatched_layers = []
weight_names = [strip_model_name_and_prefix(w.name, _prefix=_prefix) for w in model.weights]
loaded_weight_names = list(safetensors_archive.keys())
# Find the missing layers from the high level list of layers
missing_layers = list(set(weight_names) - set(loaded_weight_names))
# Find the unexpected layers from the high level list of layers
unexpected_layers = list(set(loaded_weight_names) - set(weight_names))
for weight in model.weights:
weight_name = strip_model_name_and_prefix(weight.name, _prefix=_prefix)
if weight_name in loaded_weight_names:
weight_value = safetensors_archive.get_tensor(weight_name)
# Check if the shape of the current weight and the one from the H5 file are different
if K.int_shape(weight) != weight_value.shape:
# If yes we reshape the weight from the H5 file accordingly to the current weight
# If the two shapes are not compatible we raise an issue
try:
weight_value = tf.reshape(weight_value, K.int_shape(weight))
except (ValueError, tf.errors.InvalidArgumentError) as e:
if ignore_mismatched_sizes:
mismatched_layers.append((weight_name, weight_value.shape, K.int_shape(weight)))
continue
else:
raise e
K.set_value(weight, weight_value) # weight.assign() might break if weight is a DTensor
return missing_layers, unexpected_layers, mismatched_layers
def init_copy_embeddings(old_embeddings, new_num_tokens):
r"""
This function aims to reduce the embeddings in case new_num_tokens < old_num_tokens or to pad with -1 in case
new_num_tokens > old_num_tokens. A mask is also computed in order to know which weight in the embeddings should be
kept or not. Example:
- if new_num_tokens=5 and old_num_tokens=4 and old_embeddings=[w1,w2,w3,w4]
- mask=[True,True,True,True,False] and current_weights=[w1,w2,w3,w4,-1]
- if new_num_tokens=4 and old_num_tokens=5 and old_embeddings=[w1,w2,w3,w4,w5]
- mask=[True,True,True,True] and current_weights=[w1,w2,w3,w4]
"""
old_num_tokens, old_embedding_dim = shape_list(old_embeddings)
size_diff = new_num_tokens - old_num_tokens
# initialize new embeddings
# Copy token embeddings from the previous ones
if tf.math.greater(size_diff, 0):
# if the new size is greater than the old one, we extend the current embeddings with a padding until getting new size
# and we create a mask to properly identify the padded values and be replaced by the values of the newly created
# embeddings
current_weights = tf.pad(
old_embeddings.value(), tf.convert_to_tensor([[0, size_diff], [0, 0]]), constant_values=-1
)
num_tokens_to_copy = min(old_num_tokens, new_num_tokens)
mask = tf.fill(tf.convert_to_tensor([num_tokens_to_copy, 1]), True)
mask = tf.pad(mask, tf.convert_to_tensor([[0, size_diff], [0, 0]]), constant_values=False)
else:
# if the new size if lower than the old one, we take the current embeddings until the new size
current_weights = tf.slice(
old_embeddings.value(),
tf.convert_to_tensor([0, 0]),
tf.convert_to_tensor([new_num_tokens, old_embedding_dim]),
)
mask = tf.fill(tf.convert_to_tensor([new_num_tokens, 1]), True)
return mask, current_weights
class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin, TFGenerationMixin, PushToHubMixin):
r"""
Base class for all TF models.
[`TFPreTrainedModel`] takes care of storing the configuration of the models and handles methods for loading,
downloading and saving models as well as a few methods common to all models to:
- resize the input embeddings,
- prune heads in the self-attention heads.
Class attributes (overridden by derived classes):
- **config_class** ([`PretrainedConfig`]) -- A subclass of [`PretrainedConfig`] to use as configuration class
for this model architecture.
- **base_model_prefix** (`str`) -- A string indicating the attribute associated to the base model in derived
classes of the same architecture adding modules on top of the base model.
- **main_input_name** (`str`) -- The name of the principal input to the model (often `input_ids` for NLP
models, `pixel_values` for vision models and `input_values` for speech models).
"""
config_class = None
base_model_prefix = ""
main_input_name = "input_ids"
_auto_class = None
_using_dummy_loss = None
_label_to_output_map = None
# a list of re pattern of tensor names to ignore from the model when loading the model weights
# (and avoid unnecessary warnings).
_keys_to_ignore_on_load_missing = None
# a list of re pattern of tensor names to ignore from the weights when loading the model weights
# (and avoid unnecessary warnings).
_keys_to_ignore_on_load_unexpected = None
_requires_load_weight_prefix = False
@property
def dummy_inputs(self) -> Dict[str, tf.Tensor]:
"""
Dummy inputs to build the network.
Returns:
`Dict[str, tf.Tensor]`: The dummy inputs.
"""
dummies = {}
for key, spec in self.input_signature.items():
# 2 is the most correct arbitrary size. I will not be taking questions
dummy_shape = [dim if dim is not None else 2 for dim in spec.shape]
if spec.shape[0] is None:
# But let's make the batch size 1 to save memory anyway
dummy_shape[0] = 1
dummies[key] = tf.ones(shape=dummy_shape, dtype=spec.dtype)
if key == "token_type_ids":
# Some models have token_type_ids but with a vocab_size of 1
dummies[key] = tf.zeros_like(dummies[key])
if self.config.add_cross_attention and "encoder_hidden_states" in inspect.signature(self.call).parameters:
if "encoder_hidden_states" not in dummies:
if self.main_input_name == "input_ids":
dummies["encoder_hidden_states"] = tf.ones(
shape=(1, 2, self.config.hidden_size), dtype=tf.float32, name="encoder_hidden_states"
)
else:
raise NotImplementedError(
"Model has cross-attention but we couldn't infer the shape for the encoder hidden states. Please manually override dummy_inputs!"
)
return dummies
def build_in_name_scope(self):
with tf.name_scope(self.name):
self.build(input_shape=None)
@property
def framework(self) -> str:
"""
:str: Identifies that this is a TensorFlow model.
"""
return "tf"
def build(self, input_shape=None):
pass # This is just here to make sure we don't call the superclass build()
def __init__(self, config, *inputs, **kwargs):
super().__init__(*inputs, **kwargs)
if not isinstance(config, PretrainedConfig):
raise ValueError(
f"Parameter config in `{self.__class__.__name__}(config)` should be an instance of class "
"`PretrainedConfig`. To create a model from a pretrained model use "
f"`model = {self.__class__.__name__}.from_pretrained(PRETRAINED_MODEL_NAME)`"
)
# Save config and origin of the pretrained weights if given in model
self.config = config
self.name_or_path = config.name_or_path
self.generation_config = GenerationConfig.from_model_config(config) if self.can_generate() else None
self._set_save_spec(self.input_signature)
def get_config(self):
return self.config.to_dict()
@classmethod
def from_config(cls, config, **kwargs):
if isinstance(config, PretrainedConfig):
return cls._from_config(config, **kwargs)
return cls._from_config(cls.config_class.from_dict(config, **kwargs))
@classmethod
def _from_config(cls, config, **kwargs):
"""
All context managers that the model should be initialized under go here.
"""
return cls(config, **kwargs)
def get_head_mask(self, head_mask: tf.Tensor | None, num_hidden_layers: int) -> tf.Tensor:
"""
Prepare the head mask if needed.
Args:
head_mask (`tf.Tensor` with shape `[num_heads]` or `[num_hidden_layers x num_heads]`, *optional*):
The mask indicating if we should keep the heads or not (1.0 for keep, 0.0 for discard).
num_hidden_layers (`int`):
The number of hidden layers in the model.
Returns:
`tf.Tensor` with shape `[num_hidden_layers x batch x num_heads x seq_length x seq_length]` or list with
`[None]` for each layer.
"""
if head_mask is not None:
head_mask = self._convert_head_mask_to_5d(head_mask, num_hidden_layers)
else:
head_mask = [None] * num_hidden_layers
return head_mask
def _convert_head_mask_to_5d(self, head_mask, num_hidden_layers):
"""-> [num_hidden_layers x batch x num_heads x seq_length x seq_length]"""
if head_mask.shape.rank == 1:
head_mask = head_mask[None, None, :, None, None]
head_mask = tf.repeat(head_mask, repeats=num_hidden_layers, axis=0)
elif head_mask.shape.rank == 2:
head_mask = head_mask[:, None, :, None, None]
assert head_mask.shape.rank == 5, f"head_mask.dim != 5, instead {head_mask.dim()}"
head_mask = tf.cast(head_mask, tf.float32) # switch to float if need + fp16 compatibility
return head_mask
@tf.function
def serving(self, inputs):
"""
Args:
Method used for serving the model. Does not have a specific signature, but will be specialized as concrete
functions when saving with `save_pretrained`.
inputs (`Dict[str, tf.Tensor]`):
The input of the saved model as a dictionary of tensors.
"""
output = self.call(inputs)
return self.serving_output(output)
def eager_serving(self, inputs):
"""
Method used for serving the model. This method is deprecated, and will be removed.
Args:
inputs (`Dict[str, tf.Tensor]`):
The input of the saved model as a dictionary of tensors.
"""
warnings.warn(
"The function `eager_serving` is deprecated and will be removed in version 4.32.0 of Transformers",
FutureWarning,
)
output = self.call(inputs)
return self.serving_output(output)
@property
def input_signature(self) -> Dict[str, tf.TensorSpec]:
"""
This property should return a dict mapping input names to tf.TensorSpec objects, representing the expected
shape and dtype for model inputs. It is used for both serving and for generating the dummy inputs used to build
the model.
"""
model_inputs = list(inspect.signature(self.call).parameters)
sig = {}
if "input_ids" in model_inputs:
if self.__class__.__name__.endswith("ForMultipleChoice"):
text_dims = 3
else:
text_dims = 2
for input_name in (
"input_ids",
"attention_mask",
"token_type_ids",
"decoder_input_ids",
"decoder_attention_mask",
):
if input_name in model_inputs:
sig[input_name] = tf.TensorSpec([None] * text_dims, tf.int32, name=input_name)
if "pixel_values" in model_inputs:
pixel_values_shape = [None, None, None, None]
if hasattr(self.config, "vision_config"):
vision_config = self.config.vision_config
else:
vision_config = self.config
if hasattr(vision_config, "num_channels"):
pixel_values_shape[1] = vision_config.num_channels
else:
raise NotImplementedError(
"Could not infer number of channels from config, please override input_signature to specify input shapes."
)
if hasattr(vision_config, "image_size"):
pixel_values_shape[2] = pixel_values_shape[3] = vision_config.image_size
elif hasattr(vision_config, "input_size"):
pixel_values_shape[2] = pixel_values_shape[3] = vision_config.input_size
else:
raise NotImplementedError(
"Could not infer input image shape from config, please override input_signature to specify input shapes."
)
sig["pixel_values"] = tf.TensorSpec(pixel_values_shape, tf.float32, name="pixel_values")
if "input_features" in model_inputs:
raise NotImplementedError("Audio models need a manually defined input_signature")
return sig
def serving_output(self, output):
"""
Prepare the output of the saved model. Can be overridden if specific serving modifications are required.
"""
if not isinstance(output, ModelOutput):
return output
for key in output:
if key.endswith("hidden_states") and not getattr(self.config, "output_hidden_states", False):
output[key] = None
elif key.endswith("attentions") and not getattr(self.config, "output_attentions", False):
output[key] = None
elif key == "past_key_values" and not getattr(self.config, "use_cache", False):
output[key] = None
elif key == "cross_attentions" and not (
getattr(self.config, "output_attentions", False) and getattr(self.config, "add_cross_attention", False)
):
output[key] = None
if isinstance(output[key], (tuple, list)):
try:
output[key] = tf.convert_to_tensor(output[key])
except (ValueError, tf.errors.InvalidArgumentError):
pass # Layers may not have the same dimensions
return output
@classmethod
def can_generate(cls) -> bool:
"""
Returns whether this model can generate sequences with `.generate()`.
Returns:
`bool`: Whether this model can generate sequences with `.generate()`.
"""
# Detects whether `prepare_inputs_for_generation` has been overwritten, which is a requirement for generation.
# Alternativelly, the model can also have a custom `generate` function.
if "GenerationMixin" in str(cls.prepare_inputs_for_generation) and "GenerationMixin" in str(cls.generate):
return False
return True
def get_input_embeddings(self) -> tf.keras.layers.Layer:
"""
Returns the model's input embeddings layer.
Returns:
`tf.Variable`: The embeddings layer mapping vocabulary to hidden states.
"""
main_layer = getattr(self, self.base_model_prefix, self)
if main_layer is not self:
return main_layer.get_input_embeddings()
else:
raise NotImplementedError
def _save_checkpoint(self, checkpoint_dir, epoch):
if not os.path.isdir(checkpoint_dir):
os.mkdir(checkpoint_dir)
# We avoid tf.train.checkpoint or saving weights in TF format, even though that includes optimizer
# state for us, because it requires special handling for objects like custom losses, which we use
# internally and which users are likely to use too
weights_path = os.path.join(checkpoint_dir, "weights.h5")
self.save_weights(weights_path)
extra_data = {"epoch": epoch, "optimizer_state": self.optimizer.get_weights()}
extra_data_path = os.path.join(checkpoint_dir, "extra_data.pickle")
with open(extra_data_path, "wb") as f:
pickle.dump(extra_data, f)
def load_repo_checkpoint(self, repo_path_or_name):
"""
Loads a saved checkpoint (model weights and optimizer state) from a repo. Returns the current epoch count when
the checkpoint was made.
Args:
repo_path_or_name (`str`):
Can either be a repository name for your {object} in the Hub or a path to a local folder (in which case
the repository will have the name of that local folder).
Returns:
`dict`: A dictionary of extra metadata from the checkpoint, most commonly an "epoch" count.
"""
if getattr(self, "optimizer", None) is None:
raise RuntimeError(
"Checkpoint loading failed as no optimizer is attached to the model. "
"This is most likely caused by the model not being compiled."
)
if os.path.isdir(repo_path_or_name):
local_dir = repo_path_or_name
else:
# If this isn't a local path, check that the remote repo exists and has a checkpoint in it
repo_files = list_repo_files(repo_path_or_name)
for file in ("checkpoint/weights.h5", "checkpoint/extra_data.pickle"):
if file not in repo_files:
raise FileNotFoundError(f"Repo {repo_path_or_name} does not contain checkpoint file {file}!")
repo = Repository(repo_path_or_name.split("/")[-1], clone_from=repo_path_or_name)
local_dir = repo.local_dir
# Now make sure the repo actually has a checkpoint in it.
checkpoint_dir = os.path.join(local_dir, "checkpoint")
weights_file = os.path.join(checkpoint_dir, "weights.h5")
if not os.path.isfile(weights_file):
raise FileNotFoundError(f"Could not find checkpoint file weights.h5 in repo {repo_path_or_name}!")
extra_data_file = os.path.join(checkpoint_dir, "extra_data.pickle")
if not os.path.isfile(extra_data_file):
raise FileNotFoundError(f"Could not find checkpoint file extra_data.pickle in repo {repo_path_or_name}!")
# Assuming the repo is real and we got a checkpoint, load the weights and the optimizer state into the model.
# The optimizer state includes the iteration count, so learning rate schedules should resume as normal too.
self.load_weights(weights_file)
with open(extra_data_file, "rb") as f:
extra_data = pickle.load(f)
self.optimizer.set_weights(extra_data["optimizer_state"])
# Finally, return the epoch number from the checkpoint. This isn't a property of the model, so we can't
# set it directly, but the user can pass it to fit().
return {"epoch": extra_data["epoch"]}
def prepare_tf_dataset(
self,
dataset: "datasets.Dataset", # noqa:F821
batch_size: int = 8,
shuffle: bool = True,
tokenizer: Optional["PreTrainedTokenizerBase"] = None,
collate_fn: Optional[Callable] = None,
collate_fn_args: Optional[Dict[str, Any]] = None,
drop_remainder: Optional[bool] = None,
prefetch: bool = True,
):
"""
Wraps a HuggingFace [`~datasets.Dataset`] as a `tf.data.Dataset` with collation and batching. This method is
designed to create a "ready-to-use" dataset that can be passed directly to Keras methods like `fit()` without
further modification. The method will drop columns from the dataset if they don't match input names for the
model. If you want to specify the column names to return rather than using the names that match this model, we
recommend using `Dataset.to_tf_dataset()` instead.
Args:
dataset (`Any`):
A [~`datasets.Dataset`] to be wrapped as a `tf.data.Dataset`.
batch_size (`int`, defaults to 8):
The size of batches to return.
shuffle (`bool`, defaults to `True`):
Whether to return samples from the dataset in random order. Usually `True` for training datasets and
`False` for validation/test datasets.
tokenizer ([`PreTrainedTokenizerBase`], *optional*):
A `PreTrainedTokenizer` that will be used to pad samples to create batches. Has no effect if a specific
`collate_fn` is passed instead.
collate_fn (`Callable`, *optional*):
A function that collates samples from the dataset into a single batch. Defaults to
`DefaultDataCollator` if no `tokenizer` is supplied or `DataCollatorWithPadding` if a `tokenizer` is
passed.
collate_fn_args (`Dict[str, Any]`, *optional*):
A dict of arguments to pass to the `collate_fn` alongside the list of samples.
drop_remainder (`bool`, *optional*):
Whether to drop the final batch, if the batch_size does not evenly divide the dataset length. Defaults
to the same setting as `shuffle`.
prefetch (`bool`, defaults to `True`):
Whether to add prefetching to the end of the `tf.data` pipeline. This is almost always beneficial for
performance, but can be disabled in edge cases.
Returns:
`Dataset`: A `tf.data.Dataset` which is ready to pass to the Keras API.
"""
requires_backends(self, ["datasets"])
import datasets
if collate_fn is None:
if tokenizer is None:
collate_fn = DefaultDataCollator(return_tensors="np")
else:
collate_fn = DataCollatorWithPadding(tokenizer=tokenizer, return_tensors="np")
if collate_fn_args is None:
collate_fn_args = {}
if not isinstance(dataset, datasets.Dataset):
raise TypeError("Dataset argument should be a datasets.Dataset!")
model_inputs = list(inspect.signature(self.call).parameters)
model_labels = find_labels(self.__class__)
if "cols_to_retain" in list(inspect.signature(dataset._get_output_signature).parameters.keys()):
output_signature, _ = dataset._get_output_signature(
dataset,
batch_size=None,
collate_fn=collate_fn,
collate_fn_args=collate_fn_args,
cols_to_retain=model_inputs,
)
else:
# TODO Matt: This is a workaround for older versions of datasets that are missing the `cols_to_retain`
# argument. We should remove this once the minimum supported version of datasets is > 2.3.2
unwanted_columns = [
feature
for feature in dataset.features
if feature not in model_inputs and feature not in ("label_ids", "label")
]
dataset = dataset.remove_columns(unwanted_columns)
output_signature, _ = dataset._get_output_signature(
dataset, batch_size=None, collate_fn=collate_fn, collate_fn_args=collate_fn_args
)
output_columns = list(output_signature.keys())
feature_cols = [col for col in output_columns if col in model_inputs and col not in model_labels]
label_cols = [col for col in output_columns if col in model_labels]
# Backwards compatibility for older versions of datasets. Previously, if `columns` or `label_cols`
# were a single element list, the returned element spec would be a single element. Now, passing [feature]
# will return a dict structure {"feature": feature}, and passing a single string will return a single element.
feature_cols = feature_cols[0] if len(feature_cols) == 1 else feature_cols
label_cols = label_cols[0] if len(label_cols) == 1 else label_cols
if drop_remainder is None:
drop_remainder = shuffle
tf_dataset = dataset.to_tf_dataset(
columns=feature_cols,
label_cols=label_cols,
batch_size=batch_size,
shuffle=shuffle,
drop_remainder=drop_remainder,
collate_fn=collate_fn,
collate_fn_args=collate_fn_args,
prefetch=prefetch,
)
return tf_dataset
def compile(
self,
optimizer="rmsprop",
loss="auto_with_warning",
metrics=None,
loss_weights=None,
weighted_metrics=None,
run_eagerly=None,
steps_per_execution=None,
**kwargs,
):
"""
This is a thin wrapper that sets the model's loss output head as the loss if the user does not specify a loss
function themselves.
"""
if loss in ("auto_with_warning", "passthrough"): # "passthrough" for workflow backward compatibility
logger.info(
"No loss specified in compile() - the model's internal loss computation will be used as the "
"loss. Don't panic - this is a common way to train TensorFlow models in Transformers! "
"To disable this behaviour please pass a loss argument, or explicitly pass "
"`loss=None` if you do not want your model to compute a loss. You can also specify `loss='auto'` to "
"get the internal loss without printing this info string."
)
loss = "auto"
if loss == "auto":
loss = dummy_loss
self._using_dummy_loss = True
else:
self._using_dummy_loss = False
parent_args = list(inspect.signature(tf.keras.Model.compile).parameters.keys())
# This argument got renamed, we need to support both versions
if "steps_per_execution" in parent_args:
super().compile(
optimizer=optimizer,
loss=loss,
metrics=metrics,
loss_weights=loss_weights,
weighted_metrics=weighted_metrics,
run_eagerly=run_eagerly,
steps_per_execution=steps_per_execution,
**kwargs,
)
else:
super().compile(
optimizer=optimizer,
loss=loss,
metrics=metrics,
loss_weights=loss_weights,
weighted_metrics=weighted_metrics,
run_eagerly=run_eagerly,
experimental_steps_per_execution=steps_per_execution,
**kwargs,
)
def compute_loss(self, *args, **kwargs):
if hasattr(tf.keras.Model, "compute_loss"):
# This will be true in TF 2.8 or greater
return super().compute_loss(*args, **kwargs)
else:
warnings.warn(
"The old compute_loss method is deprecated as it conflicts with the Keras compute_loss "
"method added in TF 2.8. If you want the original HF compute_loss, please call "
"hf_compute_loss() instead. From TF versions >= 2.8, or Transformers versions >= 5, "
"calling compute_loss() will get the Keras method instead.",
FutureWarning,
)
return self.hf_compute_loss(*args, **kwargs)
def get_label_to_output_name_mapping(self):
arg_names = list(inspect.signature(self.call).parameters)
if self._label_to_output_map is not None:
return self._label_to_output_map
elif "start_positions" in arg_names:
return {"start_positions": "start_logits", "end_positions": "end_logits"}
elif "sentence_order_label" in arg_names:
return {"labels": "prediction_logits", "sentence_order_label": "sop_logits"}
elif "next_sentence_label" in arg_names:
return {"labels": "prediction_logits", "next_sentence_label": "seq_relationship_logits"}
elif "mc_labels" in arg_names:
return {"labels": "logits", "mc_labels": "mc_logits"}
else:
return {}
def train_step(self, data):
"""
A modification of Keras's default `train_step` that correctly handles matching outputs to labels for our models
and supports directly training on the loss output head. In addition, it ensures input keys are copied to the
labels where appropriate. It will also copy label keys into the input dict when using the dummy loss, to ensure
that they are available to the model during the forward pass.
"""
# We hardcode the most common renamings; models with weirder names can set `self._label_to_output_map`
arg_names = list(inspect.signature(self.call).parameters)
label_kwargs = find_labels(self.__class__)
label_to_output = self.get_label_to_output_name_mapping()
output_to_label = {val: key for key, val in label_to_output.items()}
if not self._using_dummy_loss and parse(tf.__version__) < parse("2.11.0"):
# Newer TF train steps leave this out
data = expand_1d(data)
x, y, sample_weight = tf.keras.utils.unpack_x_y_sample_weight(data)
# If the inputs are mutable dictionaries, make a shallow copy of them because we will modify
# them during input/label pre-processing. This avoids surprising the user by wrecking their data.
# In addition, modifying mutable Python inputs makes XLA compilation impossible.
if isinstance(x, dict):
x = x.copy()
if isinstance(y, dict):
y = y.copy()
# When using a dummy loss, we ensure that separate labels are copied to the correct model arguments,
# if those keys are not already present in the input dict
if self._using_dummy_loss and y is not None:
# If y is a tensor and the model only has one label-like input, map y to that input
if len(label_kwargs) == 1 and isinstance(y, tf.Tensor):
if isinstance(x, tf.Tensor):
x = {arg_names[0]: x}
label_kwarg = next(iter(label_kwargs))
if label_kwarg not in x:
x[label_kwarg] = y
# Otherwise, copy keys from y to x as long as they weren't already present in x
elif isinstance(y, dict):
if isinstance(x, tf.Tensor):
x = {arg_names[0]: x}
for key, val in y.items():
if key in arg_names and key not in x:
x[key] = val
elif output_to_label.get(key, None) in arg_names and key not in x:
x[output_to_label[key]] = val
if y is None:
y = {key: val for key, val in x.items() if key in label_kwargs}
if not y and not self._using_dummy_loss:
raise ValueError("Could not find label column(s) in input dict and no separate labels were provided!")
if isinstance(y, dict):
# Rename labels at this point to match output heads
y = {label_to_output.get(key, key): val for key, val in y.items()}
# Run forward pass.
with tf.GradientTape() as tape:
if self._using_dummy_loss and "return_loss" in arg_names:
y_pred = self(x, training=True, return_loss=True)
else:
y_pred = self(x, training=True)
if self._using_dummy_loss:
loss = self.compiled_loss(y_pred.loss, y_pred.loss, sample_weight, regularization_losses=self.losses)
else:
loss = None
# This next block matches outputs to label keys. Tensorflow's standard method for doing this
# can get very confused if any of the keys contain nested values (e.g. lists/tuples of Tensors)
if isinstance(y, dict) and len(y) == 1:
if list(y.keys())[0] in y_pred.keys():
y_pred = y_pred[list(y.keys())[0]]
elif list(y_pred.keys())[0] == "loss":
y_pred = y_pred[1]
else:
y_pred = y_pred[0]
_, y = y.popitem()
elif isinstance(y, dict):
# If the labels are a dict, match keys from the output by name
y_pred = {key: val for key, val in y_pred.items() if key in y}
elif isinstance(y, tuple) or isinstance(y, list):
# If the labels are a tuple/list, match keys to the output by order, skipping the loss.
if list(y_pred.keys())[0] == "loss":
y_pred = y_pred.to_tuple()[1:]
else:
y_pred = y_pred.to_tuple()
y_pred = y_pred[: len(y)] # Remove unused fields in case those cause problems
else:
# If the labels are a single tensor, match them to the first non-loss tensor in the output
if list(y_pred.keys())[0] == "loss":
y_pred = y_pred[1]
else:
y_pred = y_pred[0]
if loss is None:
loss = self.compiled_loss(y, y_pred, sample_weight, regularization_losses=self.losses)
# Run backwards pass.
self.optimizer.minimize(loss, self.trainable_variables, tape=tape)
self.compiled_metrics.update_state(y, y_pred, sample_weight)
# Collect metrics to return
return_metrics = {}
for metric in self.metrics:
result = metric.result()
if isinstance(result, dict):
return_metrics.update(result)
else:
return_metrics[metric.name] = result
return return_metrics
def test_step(self, data):
"""
A modification of Keras's default `train_step` that correctly handles matching outputs to labels for our models
and supports directly training on the loss output head. In addition, it ensures input keys are copied to the
labels where appropriate. It will also copy label keys into the input dict when using the dummy loss, to ensure
that they are available to the model during the forward pass.
"""
# We hardcode the most common renamings; models with weirder names can set `self._label_to_output_map`
arg_names = list(inspect.signature(self.call).parameters)
label_kwargs = find_labels(self.__class__)
label_to_output = self.get_label_to_output_name_mapping()
output_to_label = {val: key for key, val in label_to_output.items()}
if not self._using_dummy_loss and parse(tf.__version__) < parse("2.11.0"):
# Newer versions leave this out
data = expand_1d(data)
x, y, sample_weight = tf.keras.utils.unpack_x_y_sample_weight(data)
# If the inputs are mutable dictionaries, make a shallow copy of them because we will modify
# them during input/label pre-processing. This avoids surprising the user by wrecking their data.
# In addition, modifying mutable Python inputs makes XLA compilation impossible.
if isinstance(x, dict):
x = x.copy()
if isinstance(y, dict):
y = y.copy()
# When using a dummy loss, we ensure that separate labels are copied to the correct model arguments,
# if those keys are not already present in the input dict
if self._using_dummy_loss and y is not None:
arg_names = list(inspect.signature(self.call).parameters)
# If y is a tensor and the model only has one label-like input, map y to that input
if len(label_kwargs) == 1 and isinstance(y, tf.Tensor):
if isinstance(x, tf.Tensor):
x = {arg_names[0]: x}
label_kwarg = next(iter(label_kwargs))
if label_kwarg not in x:
x[label_kwarg] = y
# Otherwise, copy keys from y to x as long as they weren't already present in x
elif isinstance(y, dict):
if isinstance(x, tf.Tensor):
x = {arg_names[0]: x}
for key, val in y.items():
if key in arg_names and key not in x:
x[key] = val
elif output_to_label.get(key, None) in arg_names and key not in x:
x[output_to_label[key]] = val
if y is None:
y = {key: val for key, val in x.items() if key in label_kwargs}
if not y and not self._using_dummy_loss:
raise ValueError("Could not find label column(s) in input dict and no separate labels were provided!")
if isinstance(y, dict):
# Rename labels at this point to match output heads
y = {label_to_output.get(key, key): val for key, val in y.items()}
# Run forward pass.
if self._using_dummy_loss and "return_loss" in arg_names:
y_pred = self(x, return_loss=True, training=False)
else:
y_pred = self(x, training=False)
if self._using_dummy_loss:
loss = self.compiled_loss(y_pred.loss, y_pred.loss, sample_weight, regularization_losses=self.losses)
else:
loss = None
# This next block matches outputs to label keys. Tensorflow's standard method for doing this
# can get very confused if any of the keys contain nested values (e.g. lists/tuples of Tensors)
if isinstance(y, dict) and len(y) == 1:
if list(y.keys())[0] in y_pred.keys():
y_pred = y_pred[list(y.keys())[0]]
elif list(y_pred.keys())[0] == "loss":
y_pred = y_pred[1]
else:
y_pred = y_pred[0]
_, y = y.popitem()
elif isinstance(y, dict):
# If the labels are a dict, match keys from the output by name
y_pred = {key: val for key, val in y_pred.items() if key in y}
elif isinstance(y, tuple) or isinstance(y, list):
# If the labels are a tuple/list, match keys to the output by order, skipping the loss.
if list(y_pred.keys())[0] == "loss":
y_pred = y_pred.to_tuple()[1:]
else:
y_pred = y_pred.to_tuple()
y_pred = y_pred[: len(y)] # Remove unused fields in case those cause problems
else:
# If the labels are a single tensor, match them to the first non-loss tensor in the output
if list(y_pred.keys())[0] == "loss":
y_pred = y_pred[1]
else:
y_pred = y_pred[0]
if loss is None:
loss = self.compiled_loss(y, y_pred, sample_weight, regularization_losses=self.losses)
self.compiled_metrics.update_state(y, y_pred, sample_weight)
# Collect metrics to return
return_metrics = {}
for metric in self.metrics:
result = metric.result()
if isinstance(result, dict):
return_metrics.update(result)
else:
return_metrics[metric.name] = result
return return_metrics
def create_model_card(
self,
output_dir,
model_name: str,
language: Optional[str] = None,
license: Optional[str] = None,
tags: Optional[str] = None,
finetuned_from: Optional[str] = None,
tasks: Optional[str] = None,
dataset_tags: Optional[Union[str, List[str]]] = None,
dataset: Optional[Union[str, List[str]]] = None,
dataset_args: Optional[Union[str, List[str]]] = None,
):
"""
Creates a draft of a model card using the information available to the `Trainer`.
Args:
output_dir (`str` or `os.PathLike`):
The folder in which to create the model card.
model_name (`str`, *optional*):
The name of the model.
language (`str`, *optional*):
The language of the model (if applicable)
license (`str`, *optional*):
The license of the model. Will default to the license of the pretrained model used, if the original
model given to the `Trainer` comes from a repo on the Hub.
tags (`str` or `List[str]`, *optional*):
Some tags to be included in the metadata of the model card.
finetuned_from (`str`, *optional*):
The name of the model used to fine-tune this one (if applicable). Will default to the name of the repo
of the original model given to the `Trainer` (if it comes from the Hub).
tasks (`str` or `List[str]`, *optional*):
One or several task identifiers, to be included in the metadata of the model card.
dataset_tags (`str` or `List[str]`, *optional*):
One or several dataset tags, to be included in the metadata of the model card.
dataset (`str` or `List[str]`, *optional*):
One or several dataset identifiers, to be included in the metadata of the model card.
dataset_args (`str` or `List[str]`, *optional*):
One or several dataset arguments, to be included in the metadata of the model card.
"""
# Avoids a circular import by doing this when necessary.
from .modelcard import TrainingSummary # tests_ignore
training_summary = TrainingSummary.from_keras(
self,
keras_history=self.history,
language=language,
license=license,
tags=tags,
model_name=model_name,
finetuned_from=finetuned_from,
tasks=tasks,
dataset_tags=dataset_tags,
dataset=dataset,
dataset_args=dataset_args,
)
model_card = training_summary.to_model_card()
with open(os.path.join(output_dir, "README.md"), "w") as f:
f.write(model_card)
def set_input_embeddings(self, value):
"""
Set model's input embeddings
Args:
value (`tf.Variable`):
The new weights mapping hidden states to vocabulary.
"""
main_layer = getattr(self, self.base_model_prefix)
if main_layer is None:
raise NotImplementedError("The model does not implements the base_model_prefix attribute.")
try:
main_layer.set_input_embeddings(value)
except AttributeError:
logger.info("Building the model")
self.build_in_name_scope()
main_layer.set_input_embeddings(value)
def get_output_embeddings(self) -> Union[None, tf.keras.layers.Layer]:
"""
Returns the model's output embeddings
Returns:
`tf.Variable`: The new weights mapping vocabulary to hidden states.
"""
if self.get_lm_head() is not None:
lm_head = self.get_lm_head()
try:
return lm_head.get_output_embeddings()
except AttributeError:
logger.info("Building the model")
self.build_in_name_scope()
return lm_head().get_output_embeddings()
return None # Overwrite for models with output embeddings
def set_output_embeddings(self, value):
"""
Set model's output embeddings
Args:
value (`tf.Variable`):
The new weights mapping hidden states to vocabulary.
"""
if self.get_lm_head() is not None:
lm_head = self.get_lm_head()
try:
lm_head.set_output_embeddings(value)
except AttributeError:
logger.info("Building the model")
self.build_in_name_scope()
lm_head.set_output_embeddings(value)
def get_output_layer_with_bias(self) -> Union[None, tf.keras.layers.Layer]:
"""
Get the layer that handles a bias attribute in case the model has an LM head with weights tied to the
embeddings
Return:
`tf.keras.layers.Layer`: The layer that handles the bias, None if not an LM model.
"""
warnings.warn(
"The method get_output_layer_with_bias is deprecated. Please use `get_lm_head` instead.", FutureWarning
)
return self.get_lm_head()
def get_prefix_bias_name(self) -> Union[None, str]:
"""
Get the concatenated _prefix name of the bias from the model name to the parent layer
Return:
`str`: The _prefix name of the bias.
"""
warnings.warn("The method get_prefix_bias_name is deprecated. Please use `get_bias` instead.", FutureWarning)
return None
def get_bias(self) -> Union[None, Dict[str, tf.Variable]]:
"""
Dict of bias attached to an LM head. The key represents the name of the bias attribute.
Return:
`tf.Variable`: The weights representing the bias, None if not an LM model.
"""
if self.get_lm_head() is not None:
lm_head = self.get_lm_head()
try:
return lm_head.get_bias()
except AttributeError:
self.build_in_name_scope()
return lm_head.get_bias()
return None
def set_bias(self, value):
"""
Set all the bias in the LM head.
Args:
value (`Dict[tf.Variable]`):
All the new bias attached to an LM head.
"""
if self.get_lm_head() is not None:
lm_head = self.get_lm_head()
try:
lm_head.set_bias(value)
except AttributeError:
self.build_in_name_scope()
lm_head.set_bias(value)
def get_lm_head(self) -> tf.keras.layers.Layer:
"""
The LM Head layer. This method must be overwritten by all the models that have a lm head.
Return:
`tf.keras.layers.Layer`: The LM head layer if the model has one, None if not.
"""
return None
def resize_token_embeddings(
self, new_num_tokens: Optional[int] = None
) -> Union[tf.keras.layers.Embedding, tf.Variable]:
"""
Resizes input token embeddings matrix of the model if `new_num_tokens != config.vocab_size`.
Takes care of tying weights embeddings afterwards if the model class has a `tie_weights()` method.
Arguments:
new_num_tokens (`int`, *optional*):
The number of new tokens in the embedding matrix. Increasing the size will add newly initialized
vectors at the end. Reducing the size will remove vectors from the end. If not provided or `None`, just
returns a pointer to the input tokens without doing anything.
Return:
`tf.Variable` or `tf.keras.layers.Embedding`: Pointer to the input tokens of the model.
"""
# TODO (joao): flagged for replacement (by `_v2_resized_token_embeddings`) due to embeddings refactor
# Run the new code path if the model has a keras embeddings layer
if isinstance(self.get_input_embeddings(), tf.keras.layers.Embedding):
return self._v2_resized_token_embeddings(new_num_tokens)
if new_num_tokens is None or new_num_tokens == self.config.vocab_size:
return self._get_word_embedding_weight(self.get_input_embeddings())
model_embeds = self._resize_token_embeddings(new_num_tokens)
# Update base model and current model config
self.config.vocab_size = new_num_tokens
return model_embeds
def _v2_resized_token_embeddings(self, new_num_tokens: Optional[int] = None) -> tf.keras.layers.Embedding:
"""
Resizes input token embeddings matrix of the model if `new_num_tokens != config.vocab_size`.
Arguments:
new_num_tokens (`int`, *optional*):
The number of new tokens in the embedding matrix. Increasing the size will add newly initialized
vectors at the end. Reducing the size will remove vectors from the end. If not provided or `None`, just
returns a pointer to the input tokens without doing anything.
Return:
`tf.keras.layers.Embedding`: Pointer to the input tokens of the model.
"""
if new_num_tokens is None or new_num_tokens == self.config.vocab_size:
return self.get_input_embeddings()
model_embeds = self._v2_resize_token_embeddings(new_num_tokens)
# Update base model and current model config
self.config.vocab_size = new_num_tokens
return model_embeds
def _get_word_embedding_weight(model, embedding_layer):
# TODO (joao): flagged for delection due to embeddings refactor
# If the variable holds the weights themselves, return them
if isinstance(embedding_layer, tf.Tensor):
return embedding_layer
# Otherwise, try to get them from the layer's attributes
embeds = getattr(embedding_layer, "weight", None)
if embeds is not None:
return embeds
embeds = getattr(embedding_layer, "decoder", None)
if embeds is not None:
return embeds
# The reason why the attributes don't exist might be
# because the model is not built, so retry getting
# the argument after building the model
model.build_in_name_scope()
embeds = getattr(embedding_layer, "weight", None)
if embeds is not None:
return embeds
embeds = getattr(embedding_layer, "decoder", None)
if embeds is not None:
return embeds
return None
def _resize_token_embeddings(self, new_num_tokens):
# TODO (joao): flagged for replacement (by `_v2_resize_token_embeddings`) due to embeddings refactor
old_embeddings = self._get_word_embedding_weight(self.get_input_embeddings())
new_embeddings = self._get_resized_embeddings(old_embeddings, new_num_tokens)
# if word embeddings are not tied, make sure that lm head bias is resized as well
if self.get_bias() is not None:
old_lm_head_bias = self.get_bias()
new_lm_head_bias = self._get_resized_lm_head_bias(old_lm_head_bias, new_num_tokens)
self.set_bias(new_lm_head_bias)
# if word embeddings are not tied, make sure that lm head decoder is resized as well
if self.get_output_embeddings() is not None:
old_lm_head_decoder = self._get_word_embedding_weight(self.get_output_embeddings())
new_lm_head_decoder = self._get_resized_lm_head_decoder(old_lm_head_decoder, new_num_tokens)
self.set_output_embeddings(new_lm_head_decoder)
self.set_input_embeddings(new_embeddings)
return self.get_input_embeddings()
def _v2_resize_token_embeddings(self, new_num_tokens):
old_embeddings = self.get_input_embeddings()
new_embeddings = self._v2_get_resized_embeddings(old_embeddings, new_num_tokens)
self.set_input_embeddings(new_embeddings)
# If word embeddings are not tied, make sure that lm head bias is resized as well
if self.get_bias() is not None:
old_lm_head_bias = self.get_bias()
new_lm_head_bias = self._v2_get_resized_lm_head_bias(old_lm_head_bias, new_num_tokens)
self.set_bias(new_lm_head_bias)
# If word embeddings are not tied, make sure that lm head decoder is resized as well.
tied_weights = self.get_input_embeddings() == self.get_output_embeddings()
if self.get_output_embeddings() is not None and not tied_weights:
old_lm_head_decoder = self._get_word_embedding_weight(self.get_output_embeddings())
# TODO (joao): this one probably needs a v2 version with other models
new_lm_head_decoder = self._get_resized_lm_head_decoder(old_lm_head_decoder, new_num_tokens)
self.set_output_embeddings(new_lm_head_decoder)
return self.get_input_embeddings()
def _get_resized_lm_head_bias(self, old_lm_head_bias, new_num_tokens):
"""
Build a resized bias from the old ones. Increasing the size will add newly initialized vectors at the end.
Reducing the size will remove vectors from the end
Args:
old_lm_head_bias (`tf.Variable`):
Old lm head bias to be resized.
new_num_tokens (`int`, *optional*):
New number of tokens in the linear matrix.
Increasing the size will add newly initialized vectors at the end. Reducing the size will remove
vectors from the end. If not provided or `None`, just returns None
Return:
`tf.Variable`: Pointer to the resized bias.
"""
# TODO (joao): flagged for replacement (by `_v2_get_resized_lm_head_bias`) due to embeddings refactor
new_lm_head_bias = {}
for attr, weight in old_lm_head_bias.items():
first_dim, old_num_tokens = (None, shape_list(weight)[0]) if tf.rank(weight) == 1 else shape_list(weight)
size_diff = new_num_tokens - old_num_tokens
final_shape = [new_num_tokens] if first_dim is None else [first_dim, new_num_tokens]
# initialize new bias
if tf.math.greater(size_diff, 0):
padding_shape = [[0, size_diff]] if first_dim is None else [[0, 0], [0, size_diff]]
current_bias = tf.pad(weight.value(), tf.convert_to_tensor(padding_shape), constant_values=-1)
num_tokens_to_copy = min(old_num_tokens, new_num_tokens)
mask_shape = [num_tokens_to_copy] if first_dim is None else [1, num_tokens_to_copy]
bias_mask = tf.fill(tf.convert_to_tensor(mask_shape), True)
bias_mask = tf.pad(bias_mask, tf.convert_to_tensor(padding_shape), constant_values=False)
else:
slice_from = [0] if first_dim is None else [0, 0]
current_bias = tf.slice(
weight.value(), tf.convert_to_tensor(slice_from), tf.convert_to_tensor(final_shape)
)
bias_mask = tf.fill(tf.convert_to_tensor(final_shape), True)
new_bias = self.add_weight(
shape=final_shape,
initializer="zeros",
trainable=True,
name=weight.name.split(":")[0],
)
init_bias = tf.where(bias_mask, current_bias, new_bias.value())
new_bias.assign(init_bias)
new_lm_head_bias[attr] = new_bias
return new_lm_head_bias
def _v2_get_resized_lm_head_bias(
self, old_lm_head_bias: Dict[str, tf.Variable], new_num_tokens: int
) -> Dict[str, tf.Tensor]:
"""
Build a resized bias from the old ones. Increasing the size will add newly initialized vectors at the end.
Reducing the size will remove vectors from the end
Args:
old_lm_head_bias (`Dict[str, tf.Variable]`):
Old lm head bias to be resized.
new_num_tokens (`int`):
New number of tokens in the linear matrix. Increasing the size will add newly initialized vectors at
the end. Reducing the size will remove vectors from the end.
Return:
`tf.Tensor`: Values for the resized bias.
"""
new_lm_head_bias = {}
for attr, weight in old_lm_head_bias.items():
# Determine the size difference (depending on the shape)
first_dim, old_num_tokens = (None, shape_list(weight)[0]) if tf.rank(weight) == 1 else shape_list(weight)
size_diff = new_num_tokens - old_num_tokens
# Copy the old bias values to the new bias
if old_num_tokens > new_num_tokens:
new_bias = weight.value()[..., :new_num_tokens]
else:
padding_shape = [[0, size_diff]] if first_dim is None else [[0, 0], [0, size_diff]]
new_bias = tf.pad(weight.value(), tf.convert_to_tensor(padding_shape))
new_lm_head_bias[attr] = new_bias
return new_lm_head_bias
def _get_resized_lm_head_decoder(self, old_lm_head_decoder, new_num_tokens):
"""
Build a resized decoder from the old ones. Increasing the size will add newly initialized vectors at the end.
Reducing the size will remove vectors from the end
Args:
old_lm_head_decoder (`tf.Variable`):
Old lm head decoder to be resized.
new_num_tokens (`int`, *optional*):
New number of tokens in the linear matrix.
Increasing the size will add newly initialized vectors at the end. Reducing the size will remove
vectors from the end. If not provided or `None`, just returns None
Return:
`tf.Variable`: Pointer to the resized decoder or None if the output embeddings are different from the input
ones.
"""
new_lm_head_decoder = old_lm_head_decoder
is_input_output_equals = tf.reduce_any(
self._get_word_embedding_weight(self.get_input_embeddings()) == old_lm_head_decoder
)
if old_lm_head_decoder is not None and not is_input_output_equals:
old_embedding_dim = shape_list(old_lm_head_decoder)[1]
decoder_mask, current_decoder = init_copy_embeddings(old_lm_head_decoder, new_num_tokens)
new_lm_head_decoder = self.add_weight(
shape=(new_num_tokens, old_embedding_dim),
initializer="zeros",
trainable=True,
name=old_lm_head_decoder.name.split(":")[0],
)
init_decoder = tf.where(decoder_mask, current_decoder, new_lm_head_decoder.value())
new_lm_head_decoder.assign(init_decoder)
return new_lm_head_decoder
def _get_resized_embeddings(self, old_embeddings, new_num_tokens=None) -> tf.Variable:
"""
Build a resized Embedding weights from a provided token Embedding weights. Increasing the size will add newly
initialized vectors at the end. Reducing the size will remove vectors from the end
Args:
old_embeddings (`tf.Variable`):
Old embeddings to be resized.
new_num_tokens (`int`, *optional*):
New number of tokens in the embedding matrix.
Increasing the size will add newly initialized vectors at the end. Reducing the size will remove
vectors from the end. If not provided or `None`, just returns a pointer to the input tokens
`tf.Variable` module of the model without doing anything.
Return:
`tf.Variable`: Pointer to the resized Embedding Module or the old Embedding Module if `new_num_tokens` is
`None`
"""
# TODO (joao): flagged for replacement (by `_v2_get_resized_embeddings`) due to embeddings refactor
old_embedding_dim = shape_list(old_embeddings)[1]
init_range = getattr(self.config, "initializer_range", 0.02)
embeddings_mask, current_embeddings = init_copy_embeddings(old_embeddings, new_num_tokens)
new_embeddings = self.add_weight(
name=old_embeddings.name.split(":")[0],
shape=[new_num_tokens, old_embedding_dim],
initializer=get_initializer(init_range),
dtype=tf.float32,
)
init_embeddings = tf.where(embeddings_mask, current_embeddings, new_embeddings.value())
new_embeddings.assign(init_embeddings)
return new_embeddings
def _v2_get_resized_embeddings(
self, old_embeddings: tf.keras.layers.Embedding, new_num_tokens: int
) -> tf.keras.layers.Embedding:
"""
Build a resized Embedding layer from a provided Embedding layer. Increasing the size will add newly initialized
vectors at the end. Reducing the size will remove vectors from the end.
Args:
old_embeddings (`tf.keras.layers.Embedding`):
Old embeddings to be resized.
new_num_tokens (`int`, *optional*):
New number of tokens in the embedding matrix.
Return:
`tf.keras.layers.Embedding`: Resized Embedding layer.
"""
# Get the initialization range for the embeddings
init_range = 0.02 # default value
potential_initialization_variable_names = [
"initializer_range", # most common
"initializer_factor", # e.g. T5
"init_std", # e.g BART
]
for var_name in potential_initialization_variable_names:
if hasattr(self.config, var_name):
init_range = getattr(self.config, var_name)
# Get a new (initialized) embeddings layer
new_embeddings = tf.keras.layers.Embedding(
input_dim=new_num_tokens,
output_dim=old_embeddings.output_dim,
embeddings_initializer=tf.keras.initializers.TruncatedNormal(stddev=init_range),
name=old_embeddings.embeddings.name[:-13], # exact same scoped name except "/embeddings:0"
)
new_embeddings(tf.constant([[0]]))
# Copy the old embeddings to the new embeddings
if old_embeddings.input_dim >= new_num_tokens:
init_embeddings = old_embeddings.embeddings[:new_num_tokens]
else:
init_embeddings = tf.concat(
[old_embeddings.embeddings, new_embeddings.embeddings[old_embeddings.input_dim :]], axis=0
)
new_embeddings.embeddings.assign(init_embeddings)
return new_embeddings
def prune_heads(self, heads_to_prune):
"""
Prunes heads of the base model.
Arguments:
heads_to_prune (`Dict[int, List[int]]`):
Dictionary with keys being selected layer indices (`int`) and associated values being the list of heads
to prune in said layer (list of `int`). For instance {1: [0, 2], 2: [2, 3]} will prune heads 0 and 2 on
layer 1 and heads 2 and 3 on layer 2.
"""
raise NotImplementedError
def save_pretrained(
self,
save_directory,
saved_model=False,
version=1,
push_to_hub=False,
signatures=None,
max_shard_size: Union[int, str] = "10GB",
create_pr: bool = False,
safe_serialization: bool = False,
token: Optional[Union[str, bool]] = None,
**kwargs,
):
"""
Save a model and its configuration file to a directory, so that it can be re-loaded using the
[`~TFPreTrainedModel.from_pretrained`] class method.
Arguments:
save_directory (`str`):
Directory to which to save. Will be created if it doesn't exist.
saved_model (`bool`, *optional*, defaults to `False`):
If the model has to be saved in saved model format as well or not.
version (`int`, *optional*, defaults to 1):
The version of the saved model. A saved model needs to be versioned in order to be properly loaded by
TensorFlow Serving as detailed in the official documentation
https://www.tensorflow.org/tfx/serving/serving_basic
push_to_hub (`bool`, *optional*, defaults to `False`):
Whether or not to push your model to the Hugging Face model hub after saving it. You can specify the
repository you want to push to with `repo_id` (will default to the name of `save_directory` in your
namespace).
signatures (`dict` or `tf.function`, *optional*):
Model's signature used for serving. This will be passed to the `signatures` argument of model.save().
max_shard_size (`int` or `str`, *optional*, defaults to `"10GB"`):
The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size
lower than this size. If expressed as a string, needs to be digits followed by a unit (like `"5MB"`).
If a single weight of the model is bigger than `max_shard_size`, it will be in its own checkpoint shard
which will be bigger than `max_shard_size`.
create_pr (`bool`, *optional*, defaults to `False`):
Whether or not to create a PR with the uploaded files or directly commit.
safe_serialization (`bool`, *optional*, defaults to `False`):
Whether to save the model using `safetensors` or the traditional TensorFlow way (that uses `h5`).
token (`str` or `bool`, *optional*):
The token to use as HTTP bearer authorization for remote files. If `True`, or not specified, will use
the token generated when running `huggingface-cli login` (stored in `~/.huggingface`).
kwargs (`Dict[str, Any]`, *optional*):
Additional key word arguments passed along to the [`~utils.PushToHubMixin.push_to_hub`] method.
"""
use_auth_token = kwargs.pop("use_auth_token", None)
if use_auth_token is not None:
warnings.warn(
"The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.",
FutureWarning,
)
if token is not None:
raise ValueError(
"`token` and `use_auth_token` are both specified. Please set only the argument `token`."
)
token = use_auth_token
if token is not None:
kwargs["token"] = token
if os.path.isfile(save_directory):
logger.error(f"Provided path ({save_directory}) should be a directory, not a file")
return
os.makedirs(save_directory, exist_ok=True)
if push_to_hub:
commit_message = kwargs.pop("commit_message", None)
repo_id = kwargs.pop("repo_id", save_directory.split(os.path.sep)[-1])
repo_id = self._create_repo(repo_id, **kwargs)
files_timestamps = self._get_files_timestamps(save_directory)
if saved_model:
# If `torch_dtype` is in the config with a torch dtype class as the value, we need to change it to string.
# (Although TF doesn't care about this attribute, we can't just remove it or set it to `None`.)
if getattr(self.config, "torch_dtype", None) is not None and not isinstance(self.config.torch_dtype, str):
self.config.torch_dtype = str(self.config.torch_dtype).split(".")[1]
if signatures is None:
serving_default = self.serving.get_concrete_function(self.input_signature)
if any(spec.dtype == tf.int32 for spec in self.input_signature.values()):
int64_spec = {
key: tf.TensorSpec(
shape=spec.shape, dtype=tf.int64 if spec.dtype == tf.int32 else spec.dtype, name=spec.name
)
for key, spec in self.input_signature.items()
}
int64_serving = self.serving.get_concrete_function(int64_spec)
signatures = {"serving_default": serving_default, "int64_serving": int64_serving}
else:
signatures = serving_default
saved_model_dir = os.path.join(save_directory, "saved_model", str(version))
self.save(saved_model_dir, include_optimizer=False, signatures=signatures)
logger.info(f"Saved model created in {saved_model_dir}")
# Save configuration file
self.config.architectures = [self.__class__.__name__[2:]]
# If we have a custom model, we copy the file defining it in the folder and set the attributes so it can be
# loaded from the Hub.
if self._auto_class is not None:
custom_object_save(self, save_directory, config=self.config)
self.config.save_pretrained(save_directory)
if self.can_generate():
self.generation_config.save_pretrained(save_directory)
# If we save using the predefined names, we can load using `from_pretrained`
weights_name = SAFE_WEIGHTS_NAME if safe_serialization else TF2_WEIGHTS_NAME
output_model_file = os.path.join(save_directory, weights_name)
shards, index = tf_shard_checkpoint(self.weights, max_shard_size)
# Clean the folder from a previous save
for filename in os.listdir(save_directory):
full_filename = os.path.join(save_directory, filename)
# If we have a shard file that is not going to be replaced, we delete it, but only from the main process
# in distributed settings to avoid race conditions.
weights_no_suffix = weights_name.replace(".bin", "").replace(".safetensors", "")
if (
filename.startswith(weights_no_suffix)
and os.path.isfile(full_filename)
and filename not in shards.keys()
):
os.remove(full_filename)
if index is None:
if safe_serialization:
state_dict = {strip_model_name_and_prefix(w.name): w.value() for w in self.weights}
safe_save_file(state_dict, output_model_file, metadata={"format": "tf"})
else:
self.save_weights(output_model_file)
logger.info(f"Model weights saved in {output_model_file}")
else:
save_index_file = os.path.join(save_directory, TF2_WEIGHTS_INDEX_NAME)
# Save the index as well
with open(save_index_file, "w", encoding="utf-8") as index_file:
content = json.dumps(index, indent=2, sort_keys=True) + "\n"
index_file.write(content)
logger.info(
f"The model is bigger than the maximum size per checkpoint ({max_shard_size}) and is going to be "
f"split in {len(shards)} checkpoint shards. You can find where each parameters has been saved in the "
f"index located at {save_index_file}."
)
for shard_file, shard in shards.items():
with h5py.File(os.path.join(save_directory, shard_file), mode="w") as shard_file:
layers = []
for layer in sorted(shard, key=lambda x: x.name):
if "model." in layer.name or len(layer.name.split("/")) == 1:
layer_name = layer.name
else:
layer_name = "/".join(layer.name.split("/")[1:])
param_dset = shard_file.create_dataset(
layer_name, layer.numpy().shape, dtype=layer.numpy().dtype
)
param_dset[:] = layer.numpy()
layers.append(layer_name.encode("utf8"))
save_attributes_to_hdf5_group(shard_file, "layer_names", layers)
if push_to_hub:
self._upload_modified_files(
save_directory,
repo_id,
files_timestamps,
commit_message=commit_message,
token=token,
)
@classmethod
def from_pretrained(
cls,
pretrained_model_name_or_path: Optional[Union[str, os.PathLike]],
*model_args,
config: Optional[Union[PretrainedConfig, str, os.PathLike]] = None,
cache_dir: Optional[Union[str, os.PathLike]] = None,
ignore_mismatched_sizes: bool = False,
force_download: bool = False,
local_files_only: bool = False,
token: Optional[Union[str, bool]] = None,
revision: str = "main",
**kwargs,
):
r"""
Instantiate a pretrained TF 2.0 model from a pre-trained model configuration.
The warning *Weights from XXX not initialized from pretrained model* means that the weights of XXX do not come
pretrained with the rest of the model. It is up to you to train those weights with a downstream fine-tuning
task.
The warning *Weights from XXX not used in YYY* means that the layer XXX is not used by YYY, therefore those
weights are discarded.
Parameters:
pretrained_model_name_or_path (`str`, *optional*):
Can be either:
- A string, the *model id* of a pretrained model hosted inside a model repo on huggingface.co.
Valid model ids can be located at the root-level, like `bert-base-uncased`, or namespaced under a
user or organization name, like `dbmdz/bert-base-german-cased`.
- A path to a *directory* containing model weights saved using
[`~TFPreTrainedModel.save_pretrained`], e.g., `./my_model_directory/`.
- A path or url to a *PyTorch state_dict save file* (e.g, `./pt_model/pytorch_model.bin`). In this
case, `from_pt` should be set to `True` and a configuration object should be provided as `config`
argument. This loading path is slower than converting the PyTorch model in a TensorFlow model
using the provided conversion scripts and loading the TensorFlow model afterwards.
- `None` if you are both providing the configuration and state dictionary (resp. with keyword
arguments `config` and `state_dict`).
model_args (sequence of positional arguments, *optional*):
All remaining positional arguments will be passed to the underlying model's `__init__` method.
config (`Union[PretrainedConfig, str]`, *optional*):
Can be either:
- an instance of a class derived from [`PretrainedConfig`],
- a string valid as input to [`~PretrainedConfig.from_pretrained`].
Configuration for the model to use instead of an automatically loaded configuration. Configuration can
be automatically loaded when:
- The model is a model provided by the library (loaded with the *model id* string of a pretrained
model).
- The model was saved using [`~TFPreTrainedModel.save_pretrained`] and is reloaded by supplying the
save directory.
- The model is loaded by supplying a local directory as `pretrained_model_name_or_path` and a
configuration JSON file named *config.json* is found in the directory.
from_pt (`bool`, *optional*, defaults to `False`):
Load the model weights from a PyTorch state_dict save file (see docstring of
`pretrained_model_name_or_path` argument).
ignore_mismatched_sizes (`bool`, *optional*, defaults to `False`):
Whether or not to raise an error if some of the weights from the checkpoint do not have the same size
as the weights of the model (if for instance, you are instantiating a model with 10 labels from a
checkpoint with 3 labels).
cache_dir (`str`, *optional*):
Path to a directory in which a downloaded pretrained model configuration should be cached if the
standard cache should not be used.
force_download (`bool`, *optional*, defaults to `False`):
Whether or not to force the (re-)download of the model weights and configuration files, overriding the
cached versions if they exist.
resume_download (`bool`, *optional*, defaults to `False`):
Whether or not to delete incompletely received files. Will attempt to resume the download if such a
file exists.
proxies:
(`Dict[str, str], `optional`): A dictionary of proxy servers to use by protocol or endpoint, e.g.,
`{'http': 'foo.bar:3128', 'http://hostname': 'foo.bar:4012'}`. The proxies are used on each request.
output_loading_info(`bool`, *optional*, defaults to `False`): Whether ot not to also return a
dictionary containing missing keys, unexpected keys and error messages.
local_files_only(`bool`, *optional*, defaults to `False`):
Whether or not to only look at local files (e.g., not try downloading the model).
token (`str` or `bool`, *optional*):
The token to use as HTTP bearer authorization for remote files. If `True`, or not specified, will use
the token generated when running `huggingface-cli login` (stored in `~/.huggingface`).
revision (`str`, *optional*, defaults to `"main"`):
The specific model version to use. It can be a branch name, a tag name, or a commit id, since we use a
git-based system for storing models and other artifacts on huggingface.co, so `revision` can be any
identifier allowed by git.
To test a pull request you made on the Hub, you can pass `revision="refs/pr/".
mirror (`str`, *optional*):
Mirror source to accelerate downloads in China. If you are from China and have an accessibility
problem, you can set this option to resolve it. Note that we do not guarantee the timeliness or safety.
Please refer to the mirror site for more information.
subfolder (`str`, *optional*, defaults to `""`):
In case the relevant files are located inside a subfolder of the model repo on huggingface.co, you can
specify the folder name here.
tf_to_pt_weight_rename (`Callable`, *optional*):
A function that is called to transform the names of weights during the PyTorch to TensorFlow
crossloading process. This is not necessary for most models, but is useful to allow composite models to
be crossloaded correctly.
kwargs (remaining dictionary of keyword arguments, *optional*):
Can be used to update the configuration object (after it being loaded) and initiate the model (e.g.,
`output_attentions=True`). Behaves differently depending on whether a `config` is provided or
automatically loaded:
- If a configuration is provided with `config`, `**kwargs` will be directly passed to the
underlying model's `__init__` method (we assume all relevant updates to the configuration have
already been done)
- If a configuration is not provided, `kwargs` will be first passed to the configuration class
initialization function ([`~PretrainedConfig.from_pretrained`]). Each key of `kwargs` that
corresponds to a configuration attribute will be used to override said attribute with the
supplied `kwargs` value. Remaining keys that do not correspond to any configuration attribute
will be passed to the underlying model's `__init__` function.
Examples:
```python
>>> from transformers import BertConfig, TFBertModel
>>> # Download model and configuration from huggingface.co and cache.
>>> model = TFBertModel.from_pretrained("bert-base-uncased")
>>> # Model was saved using *save_pretrained('./test/saved_model/')* (for example purposes, not runnable).
>>> model = TFBertModel.from_pretrained("./test/saved_model/")
>>> # Update configuration during loading.
>>> model = TFBertModel.from_pretrained("bert-base-uncased", output_attentions=True)
>>> assert model.config.output_attentions == True
>>> # Loading from a Pytorch model file instead of a TensorFlow checkpoint (slower, for example purposes, not runnable).
>>> config = BertConfig.from_json_file("./pt_model/my_pt_model_config.json")
>>> model = TFBertModel.from_pretrained("./pt_model/my_pytorch_model.bin", from_pt=True, config=config)
```"""
from_pt = kwargs.pop("from_pt", False)
resume_download = kwargs.pop("resume_download", False)
proxies = kwargs.pop("proxies", None)
output_loading_info = kwargs.pop("output_loading_info", False)
use_auth_token = kwargs.pop("use_auth_token", None)
trust_remote_code = kwargs.pop("trust_remote_code", None)
_ = kwargs.pop("mirror", None)
load_weight_prefix = kwargs.pop("load_weight_prefix", None)
from_pipeline = kwargs.pop("_from_pipeline", None)
from_auto_class = kwargs.pop("_from_auto", False)
subfolder = kwargs.pop("subfolder", "")
commit_hash = kwargs.pop("_commit_hash", None)
tf_to_pt_weight_rename = kwargs.pop("tf_to_pt_weight_rename", None)
# Not relevant for TF models
_ = kwargs.pop("adapter_kwargs", None)
if use_auth_token is not None:
warnings.warn(
"The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.",
FutureWarning,
)
if token is not None:
raise ValueError(
"`token` and `use_auth_token` are both specified. Please set only the argument `token`."
)
token = use_auth_token
if trust_remote_code is True:
logger.warning(
"The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is"
" ignored."
)
user_agent = {"file_type": "model", "framework": "tensorflow", "from_auto_class": from_auto_class}
if from_pipeline is not None:
user_agent["using_pipeline"] = from_pipeline
if is_offline_mode() and not local_files_only:
logger.info("Offline mode: forcing local_files_only=True")
local_files_only = True
# Load config if we don't provide a configuration
if not isinstance(config, PretrainedConfig):
config_path = config if config is not None else pretrained_model_name_or_path
config, model_kwargs = cls.config_class.from_pretrained(
config_path,
cache_dir=cache_dir,
return_unused_kwargs=True,
force_download=force_download,
resume_download=resume_download,
proxies=proxies,
local_files_only=local_files_only,
token=token,
revision=revision,
_from_auto=from_auto_class,
_from_pipeline=from_pipeline,
_commit_hash=commit_hash,
**kwargs,
)
else:
model_kwargs = kwargs
if commit_hash is None:
commit_hash = getattr(config, "_commit_hash", None)
# This variable will flag if we're loading a sharded checkpoint. In this case the archive file is just the
# index of the files.
is_sharded = False
# Load model
if pretrained_model_name_or_path is not None:
pretrained_model_name_or_path = str(pretrained_model_name_or_path)
is_local = os.path.isdir(pretrained_model_name_or_path)
if is_local:
if from_pt and os.path.isfile(os.path.join(pretrained_model_name_or_path, WEIGHTS_NAME)):
# Load from a PyTorch checkpoint in priority if from_pt
archive_file = os.path.join(pretrained_model_name_or_path, WEIGHTS_NAME)
elif from_pt and os.path.isfile(os.path.join(pretrained_model_name_or_path, WEIGHTS_INDEX_NAME)):
# Load from a sharded PyTorch checkpoint
archive_file = os.path.join(pretrained_model_name_or_path, WEIGHTS_INDEX_NAME)
is_sharded = True
elif is_safetensors_available() and os.path.isfile(
os.path.join(pretrained_model_name_or_path, SAFE_WEIGHTS_NAME)
):
# Load from a safetensors checkpoint
archive_file = os.path.join(pretrained_model_name_or_path, SAFE_WEIGHTS_NAME)
elif os.path.isfile(os.path.join(pretrained_model_name_or_path, TF2_WEIGHTS_NAME)):
# Load from a TF 2.0 checkpoint
archive_file = os.path.join(pretrained_model_name_or_path, TF2_WEIGHTS_NAME)
elif os.path.isfile(os.path.join(pretrained_model_name_or_path, TF2_WEIGHTS_INDEX_NAME)):
# Load from a sharded TF 2.0 checkpoint
archive_file = os.path.join(pretrained_model_name_or_path, TF2_WEIGHTS_INDEX_NAME)
is_sharded = True
elif is_safetensors_available() and os.path.isfile(
os.path.join(pretrained_model_name_or_path, SAFE_WEIGHTS_INDEX_NAME)
):
# Load from a sharded safetensors checkpoint
archive_file = os.path.join(pretrained_model_name_or_path, SAFE_WEIGHTS_INDEX_NAME)
is_sharded = True
raise NotImplementedError("Support for sharded checkpoints using safetensors is coming soon!")
# At this stage we don't have a weight file so we will raise an error.
elif os.path.isfile(os.path.join(pretrained_model_name_or_path, WEIGHTS_NAME)) or os.path.isfile(
os.path.join(pretrained_model_name_or_path, WEIGHTS_INDEX_NAME)
):
raise EnvironmentError(
f"Error no file named {TF2_WEIGHTS_NAME} found in directory {pretrained_model_name_or_path} "
"but there is a file for PyTorch weights. Use `from_pt=True` to load this model from those "
"weights."
)
else:
raise EnvironmentError(
f"Error no file named {TF2_WEIGHTS_NAME} or {WEIGHTS_NAME} found in directory "
f"{pretrained_model_name_or_path}."
)
elif os.path.isfile(pretrained_model_name_or_path):
archive_file = pretrained_model_name_or_path
is_local = True
elif os.path.isfile(pretrained_model_name_or_path + ".index"):
archive_file = pretrained_model_name_or_path + ".index"
is_local = True
elif is_remote_url(pretrained_model_name_or_path):
filename = pretrained_model_name_or_path
resolved_archive_file = download_url(pretrained_model_name_or_path)
else:
# set correct filename
if from_pt:
filename = WEIGHTS_NAME
elif is_safetensors_available():
filename = SAFE_WEIGHTS_NAME
else:
filename = TF2_WEIGHTS_NAME
try:
# Load from URL or cache if already cached
cached_file_kwargs = {
"cache_dir": cache_dir,
"force_download": force_download,
"proxies": proxies,
"resume_download": resume_download,
"local_files_only": local_files_only,
"token": token,
"user_agent": user_agent,
"revision": revision,
"subfolder": subfolder,
"_raise_exceptions_for_missing_entries": False,
"_commit_hash": commit_hash,
}
resolved_archive_file = cached_file(pretrained_model_name_or_path, filename, **cached_file_kwargs)
# Since we set _raise_exceptions_for_missing_entries=False, we don't get an exception but a None
# result when internet is up, the repo and revision exist, but the file does not.
if resolved_archive_file is None and filename == SAFE_WEIGHTS_NAME:
# Did not find the safetensors file, let's fallback to TF.
# No support for sharded safetensors yet, so we'll raise an error if that's all we find.
filename = TF2_WEIGHTS_NAME
resolved_archive_file = cached_file(
pretrained_model_name_or_path, TF2_WEIGHTS_NAME, **cached_file_kwargs
)
if resolved_archive_file is None and filename == TF2_WEIGHTS_NAME:
# Maybe the checkpoint is sharded, we try to grab the index name in this case.
resolved_archive_file = cached_file(
pretrained_model_name_or_path, TF2_WEIGHTS_INDEX_NAME, **cached_file_kwargs
)
if resolved_archive_file is not None:
is_sharded = True
if resolved_archive_file is None and filename == WEIGHTS_NAME:
# Maybe the checkpoint is sharded, we try to grab the index name in this case.
resolved_archive_file = cached_file(
pretrained_model_name_or_path, WEIGHTS_INDEX_NAME, **cached_file_kwargs
)
if resolved_archive_file is not None:
is_sharded = True
if resolved_archive_file is None:
# Otherwise, maybe there is a PyTorch or Flax model file. We try those to give a helpful error
# message.
has_file_kwargs = {
"revision": revision,
"proxies": proxies,
"token": token,
}
if has_file(pretrained_model_name_or_path, SAFE_WEIGHTS_INDEX_NAME, **has_file_kwargs):
is_sharded = True
raise NotImplementedError(
"Support for sharded checkpoints using safetensors is coming soon!"
)
elif has_file(pretrained_model_name_or_path, WEIGHTS_NAME, **has_file_kwargs):
raise EnvironmentError(
f"{pretrained_model_name_or_path} does not appear to have a file named"
f" {TF2_WEIGHTS_NAME} but there is a file for PyTorch weights. Use `from_pt=True` to"
" load this model from those weights."
)
else:
raise EnvironmentError(
f"{pretrained_model_name_or_path} does not appear to have a file named {WEIGHTS_NAME},"
f" {TF2_WEIGHTS_NAME} or {TF_WEIGHTS_NAME}"
)
except EnvironmentError:
# Raise any environment error raise by `cached_file`. It will have a helpful error message adapted
# to the original exception.
raise
except Exception:
# For any other exception, we throw a generic error.
raise EnvironmentError(
f"Can't load the model for '{pretrained_model_name_or_path}'. If you were trying to load it"
" from 'https://huggingface.co/models', make sure you don't have a local directory with the"
f" same name. Otherwise, make sure '{pretrained_model_name_or_path}' is the correct path to a"
f" directory containing a file named {WEIGHTS_NAME}, {TF2_WEIGHTS_NAME} or {TF_WEIGHTS_NAME}"
)
if is_local:
logger.info(f"loading weights file {archive_file}")
resolved_archive_file = archive_file
filename = resolved_archive_file.split(os.path.sep)[-1]
else:
logger.info(f"loading weights file {filename} from cache at {resolved_archive_file}")
else:
resolved_archive_file = None
# We'll need to download and cache each checkpoint shard if the checkpoint is sharded.
if is_sharded:
# resolved_archive_file becomes a list of files that point to the different checkpoint shards in this case.
resolved_archive_file, _ = get_checkpoint_shard_files(
pretrained_model_name_or_path,
resolved_archive_file,
cache_dir=cache_dir,
force_download=force_download,
proxies=proxies,
resume_download=resume_download,
local_files_only=local_files_only,
token=token,
user_agent=user_agent,
revision=revision,
_commit_hash=commit_hash,
)
safetensors_from_pt = False
if filename == SAFE_WEIGHTS_NAME:
with safe_open(resolved_archive_file, framework="tf") as f:
safetensors_metadata = f.metadata()
if safetensors_metadata is None or safetensors_metadata.get("format") not in ["pt", "tf", "flax"]:
raise OSError(
f"The safetensors archive passed at {resolved_archive_file} does not contain the valid metadata."
" Make sure you save your model with the `save_pretrained` method."
)
safetensors_from_pt = safetensors_metadata.get("format") == "pt"
config.name_or_path = pretrained_model_name_or_path
# composed models, *e.g.* TFRag, require special treatment when it comes to loading
# pre-trained weights.
if cls._requires_load_weight_prefix and model_kwargs.get("name") is not None:
model_kwargs["load_weight_prefix"] = load_weight_prefix + "/" + model_kwargs.get("name")
# Instantiate model.
model = cls(config, *model_args, **model_kwargs)
if tf_to_pt_weight_rename is None and hasattr(model, "tf_to_pt_weight_rename"):
# TODO Matt: This is a temporary workaround to allow weight renaming, but requires a method
# to be defined for each class that requires a rename. We can probably just have a class-level
# dict and a single top-level method or something and cut down a lot of boilerplate code
tf_to_pt_weight_rename = model.tf_to_pt_weight_rename
if from_pt:
from .modeling_tf_pytorch_utils import load_pytorch_checkpoint_in_tf2_model
# Load from a PyTorch checkpoint
return load_pytorch_checkpoint_in_tf2_model(
model,
resolved_archive_file,
allow_missing_keys=True,
output_loading_info=output_loading_info,
_prefix=load_weight_prefix,
tf_to_pt_weight_rename=tf_to_pt_weight_rename,
)
# we might need to extend the variable scope for composite models
if load_weight_prefix is not None:
with tf.compat.v1.variable_scope(load_weight_prefix):
model.build_in_name_scope() # build the network with dummy inputs
else:
model.build_in_name_scope() # build the network with dummy inputs
if safetensors_from_pt:
from .modeling_tf_pytorch_utils import load_pytorch_state_dict_in_tf2_model
with safe_open(resolved_archive_file, framework="tf") as safetensors_archive:
# Load from a PyTorch checkpoint
# We load in TF format here because PT weights often need to be transposed, and this is much
# faster on GPU. Loading as numpy and transposing on CPU adds several seconds to load times.
return load_pytorch_state_dict_in_tf2_model(
model,
safetensors_archive,
tf_inputs=False, # No need to build the model again
allow_missing_keys=True,
output_loading_info=output_loading_info,
_prefix=load_weight_prefix,
ignore_mismatched_sizes=ignore_mismatched_sizes,
tf_to_pt_weight_rename=tf_to_pt_weight_rename,
)
# 'by_name' allow us to do transfer learning by skipping/adding layers
# see https://github.com/tensorflow/tensorflow/blob/00fad90125b18b80fe054de1055770cfb8fe4ba3/tensorflow/python/keras/engine/network.py#L1339-L1357
try:
if is_sharded:
for file in resolved_archive_file:
os.path.isfile(file), f"Error retrieving files {file}"
missing_keys, unexpected_keys, mismatched_keys = load_tf_sharded_weights(
model,
resolved_archive_file,
ignore_mismatched_sizes=ignore_mismatched_sizes,
_prefix=load_weight_prefix,
)
else:
missing_keys, unexpected_keys, mismatched_keys = load_tf_weights(
model,
resolved_archive_file,
ignore_mismatched_sizes=ignore_mismatched_sizes,
_prefix=load_weight_prefix,
)
except OSError as e:
try:
with open(resolved_archive_file) as f:
if f.read().startswith("version"):
raise OSError(
"You seem to have cloned a repository without having git-lfs installed. Please install "
"git-lfs and run `git lfs install` followed by `git lfs pull` in the folder "
"you cloned."
)
else:
raise ValueError from e
except (UnicodeDecodeError, ValueError):
raise OSError(
"Unable to load weights from h5 file. "
"If you tried to load a TF 2.0 model from a PyTorch checkpoint, please set from_pt=True. "
)
if cls._keys_to_ignore_on_load_missing is not None:
for pat in cls._keys_to_ignore_on_load_missing:
missing_keys = [k for k in missing_keys if re.search(pat, k) is None]
if cls._keys_to_ignore_on_load_unexpected is not None:
for pat in cls._keys_to_ignore_on_load_unexpected:
unexpected_keys = [k for k in unexpected_keys if re.search(pat, k) is None]
if len(unexpected_keys) > 0:
logger.warning(
f"Some layers from the model checkpoint at {pretrained_model_name_or_path} were not used when"
f" initializing {model.__class__.__name__}: {unexpected_keys}\n- This IS expected if you are"
f" initializing {model.__class__.__name__} from the checkpoint of a model trained on another task or"
" with another architecture (e.g. initializing a BertForSequenceClassification model from a"
" BertForPreTraining model).\n- This IS NOT expected if you are initializing"
f" {model.__class__.__name__} from the checkpoint of a model that you expect to be exactly identical"
" (initializing a BertForSequenceClassification model from a BertForSequenceClassification model)."
)
else:
logger.warning(f"All model checkpoint layers were used when initializing {model.__class__.__name__}.\n")
if len(missing_keys) > 0:
logger.warning(
f"Some layers of {model.__class__.__name__} were not initialized from the model checkpoint at"
f" {pretrained_model_name_or_path} and are newly initialized: {missing_keys}\nYou should probably"
" TRAIN this model on a down-stream task to be able to use it for predictions and inference."
)
elif len(mismatched_keys) == 0:
logger.warning(
f"All the layers of {model.__class__.__name__} were initialized from the model checkpoint at"
f" {pretrained_model_name_or_path}.\nIf your task is similar to the task the model of the checkpoint"
f" was trained on, you can already use {model.__class__.__name__} for predictions without further"
" training."
)
if len(mismatched_keys) > 0:
mismatched_warning = "\n".join(
[
f"- {key}: found shape {shape1} in the checkpoint and {shape2} in the model instantiated"
for key, shape1, shape2 in mismatched_keys
]
)
logger.warning(
f"Some weights of {model.__class__.__name__} were not initialized from the model checkpoint at"
f" {pretrained_model_name_or_path} and are newly initialized because the shapes did not"
f" match:\n{mismatched_warning}\nYou should probably TRAIN this model on a down-stream task to be able"
" to use it for predictions and inference."
)
# If it is a model with generation capabilities, attempt to load the generation config
if model.can_generate():
try:
model.generation_config = GenerationConfig.from_pretrained(
pretrained_model_name_or_path,
cache_dir=cache_dir,
force_download=force_download,
resume_download=resume_download,
proxies=proxies,
local_files_only=local_files_only,
token=token,
revision=revision,
subfolder=subfolder,
_from_auto=from_auto_class,
_from_pipeline=from_pipeline,
**kwargs,
)
except OSError:
logger.info(
"Generation config file not found, using a generation config created from the model config."
)
pass
if output_loading_info:
loading_info = {
"missing_keys": missing_keys,
"unexpected_keys": unexpected_keys,
"mismatched_keys": mismatched_keys,
}
return model, loading_info
return model
def push_to_hub(
self,
repo_id: str,
use_temp_dir: Optional[bool] = None,
commit_message: Optional[str] = None,
private: Optional[bool] = None,
max_shard_size: Optional[Union[int, str]] = "10GB",
token: Optional[Union[bool, str]] = None,
# (`use_auth_token` is deprecated: we have to keep it here as we don't have **kwargs)
use_auth_token: Optional[Union[bool, str]] = None,
create_pr: bool = False,
**base_model_card_args,
) -> str:
"""
Upload the model files to the 🤗 Model Hub while synchronizing a local clone of the repo in `repo_path_or_name`.
Parameters:
repo_id (`str`):
The name of the repository you want to push your model to. It should contain your organization name
when pushing to a given organization.
use_temp_dir (`bool`, *optional*):
Whether or not to use a temporary directory to store the files saved before they are pushed to the Hub.
Will default to `True` if there is no directory named like `repo_id`, `False` otherwise.
commit_message (`str`, *optional*):
Message to commit while pushing. Will default to `"Upload model"`.
private (`bool`, *optional*):
Whether or not the repository created should be private.
token (`bool` or `str`, *optional*):
The token to use as HTTP bearer authorization for remote files. If `True`, will use the token generated
when running `huggingface-cli login` (stored in `~/.huggingface`). Will default to `True` if `repo_url`
is not specified.
max_shard_size (`int` or `str`, *optional*, defaults to `"10GB"`):
Only applicable for models. The maximum size for a checkpoint before being sharded. Checkpoints shard
will then be each of size lower than this size. If expressed as a string, needs to be digits followed
by a unit (like `"5MB"`).
create_pr (`bool`, *optional*, defaults to `False`):
Whether or not to create a PR with the uploaded files or directly commit.
Examples:
```python
from transformers import TFAutoModel
model = TFAutoModel.from_pretrained("bert-base-cased")
# Push the model to your namespace with the name "my-finetuned-bert".
model.push_to_hub("my-finetuned-bert")
# Push the model to an organization with the name "my-finetuned-bert".
model.push_to_hub("huggingface/my-finetuned-bert")
```
"""
if use_auth_token is not None:
warnings.warn(
"The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.",
FutureWarning,
)
if token is not None:
raise ValueError(
"`token` and `use_auth_token` are both specified. Please set only the argument `token`."
)
token = use_auth_token
if "repo_path_or_name" in base_model_card_args:
warnings.warn(
"The `repo_path_or_name` argument is deprecated and will be removed in v5 of Transformers. Use "
"`repo_id` instead."
)
repo_id = base_model_card_args.pop("repo_path_or_name")
# Deprecation warning will be sent after for repo_url and organization
repo_url = base_model_card_args.pop("repo_url", None)
organization = base_model_card_args.pop("organization", None)
if os.path.isdir(repo_id):
working_dir = repo_id
repo_id = repo_id.split(os.path.sep)[-1]
else:
working_dir = repo_id.split("/")[-1]
repo_id = self._create_repo(
repo_id, private=private, token=token, repo_url=repo_url, organization=organization
)
if use_temp_dir is None:
use_temp_dir = not os.path.isdir(working_dir)
with working_or_temp_dir(working_dir=working_dir, use_temp_dir=use_temp_dir) as work_dir:
files_timestamps = self._get_files_timestamps(work_dir)
# Save all files.
self.save_pretrained(work_dir, max_shard_size=max_shard_size)
if hasattr(self, "history") and hasattr(self, "create_model_card"):
# This is a Keras model and we might be able to fish out its History and make a model card out of it
base_model_card_args = {
"output_dir": work_dir,
"model_name": Path(repo_id).name,
}
base_model_card_args.update(base_model_card_args)
self.create_model_card(**base_model_card_args)
self._upload_modified_files(
work_dir,
repo_id,
files_timestamps,
commit_message=commit_message,
token=token,
create_pr=create_pr,
)
@classmethod
def register_for_auto_class(cls, auto_class="TFAutoModel"):
"""
Register this class with a given auto class. This should only be used for custom models as the ones in the
library are already mapped with an auto class.
This API is experimental and may have some slight breaking changes in the next releases.
Args:
auto_class (`str` or `type`, *optional*, defaults to `"TFAutoModel"`):
The auto class to register this new model with.
"""
if not isinstance(auto_class, str):
auto_class = auto_class.__name__
import transformers.models.auto as auto_module
if not hasattr(auto_module, auto_class):
raise ValueError(f"{auto_class} is not a valid auto class.")
cls._auto_class = auto_class
class TFConv1D(tf.keras.layers.Layer):
"""
1D-convolutional layer as defined by Radford et al. for OpenAI GPT (and also used in GPT-2).
Basically works like a linear layer but the weights are transposed.
Args:
nf (`int`):
The number of output features.
nx (`int`):
The number of input features.
initializer_range (`float`, *optional*, defaults to 0.02):
The standard deviation to use to initialize the weights.
kwargs (`Dict[str, Any]`, *optional*):
Additional keyword arguments passed along to the `__init__` of `tf.keras.layers.Layer`.
"""
def __init__(self, nf, nx, initializer_range=0.02, **kwargs):
super().__init__(**kwargs)
self.nf = nf
self.nx = nx
self.initializer_range = initializer_range
def build(self, input_shape):
if self.built:
return
self.built = True
self.weight = self.add_weight(
"weight", shape=[self.nx, self.nf], initializer=get_initializer(self.initializer_range)
)
self.bias = self.add_weight("bias", shape=[1, self.nf], initializer=tf.zeros_initializer())
def call(self, x):
bz, sl = shape_list(x)[:2]
x = tf.reshape(x, [-1, self.nx])
x = tf.matmul(x, self.weight) + self.bias
x = tf.reshape(x, [bz, sl, self.nf])
return x
class TFSharedEmbeddings(tf.keras.layers.Layer):
r"""
Construct shared token embeddings.
The weights of the embedding layer is usually shared with the weights of the linear decoder when doing language
modeling.
Args:
vocab_size (`int`):
The size of the vocabulary, e.g., the number of unique tokens.
hidden_size (`int`):
The size of the embedding vectors.
initializer_range (`float`, *optional*):
The standard deviation to use when initializing the weights. If no value is provided, it will default to
\\(1/\sqrt{hidden\_size}\\).
kwargs (`Dict[str, Any]`, *optional*):
Additional keyword arguments passed along to the `__init__` of `tf.keras.layers.Layer`.
"""
# TODO (joao): flagged for delection due to embeddings refactor
def __init__(self, vocab_size: int, hidden_size: int, initializer_range: Optional[float] = None, **kwargs):
super().__init__(**kwargs)
self.vocab_size = vocab_size
self.hidden_size = hidden_size
self.initializer_range = hidden_size**-0.5 if initializer_range is None else initializer_range
warnings.warn(
"`TFSharedEmbeddings` is scheduled for deletion in v4.32, use `tf.keras.layers.Embedding` instead.",
DeprecationWarning,
)
def build(self, input_shape):
"""
Build shared token embedding layer Shared weights logic adapted from
https://github.com/tensorflow/models/blob/a009f4fb9d2fc4949e32192a944688925ef78659/official/transformer/v2/embedding_layer.py#L24
"""
self.weight = self.add_weight(
"weight", shape=[self.vocab_size, self.hidden_size], initializer=get_initializer(self.initializer_range)
)
super().build(input_shape)
def get_config(self):
config = {
"vocab_size": self.vocab_size,
"hidden_size": self.hidden_size,
"initializer_range": self.initializer_range,
}
base_config = super().get_config()
return dict(list(base_config.items()) + list(config.items()))
def call(self, inputs: tf.Tensor, mode: str = "embedding") -> tf.Tensor:
"""
Get token embeddings of inputs or decode final hidden state.
Args:
inputs (`tf.Tensor`):
In embedding mode, should be an int64 tensor with shape `[batch_size, length]`.
In linear mode, should be a float tensor with shape `[batch_size, length, hidden_size]`.
mode (`str`, defaults to `"embedding"`):
A valid value is either `"embedding"` or `"linear"`, the first one indicates that the layer should be
used as an embedding layer, the second one that the layer should be used as a linear decoder.
Returns:
`tf.Tensor`: In embedding mode, the output is a float32 embedding tensor, with shape `[batch_size, length,
embedding_size]`.
In linear mode, the output is a float32 with shape `[batch_size, length, vocab_size]`.
Raises:
ValueError: if `mode` is not valid.
Shared weights logic is adapted from
[here](https://github.com/tensorflow/models/blob/a009f4fb9d2fc4949e32192a944688925ef78659/official/transformer/v2/embedding_layer.py#L24).
"""
if mode == "embedding":
return self._embedding(inputs)
elif mode == "linear":
return self._linear(inputs)
else:
raise ValueError(f"mode {mode} is not valid.")
def _embedding(self, input_ids):
"""Applies embedding based on inputs tensor."""
return tf.gather(self.weight, input_ids)
def _linear(self, inputs):
"""
Computes logits by running inputs through a linear layer.
Args:
inputs: A float32 tensor with shape [..., hidden_size]
Returns:
float32 tensor with shape [..., vocab_size].
"""
first_dims = shape_list(inputs)[:-1]
x = tf.reshape(inputs, [-1, self.hidden_size])
logits = tf.matmul(x, self.weight, transpose_b=True)
return tf.reshape(logits, first_dims + [self.vocab_size])
class TFSequenceSummary(tf.keras.layers.Layer):
"""
Compute a single vector summary of a sequence hidden states.
Args:
config ([`PretrainedConfig`]):
The config used by the model. Relevant arguments in the config class of the model are (refer to the actual
config class of your model for the default values it uses):
- **summary_type** (`str`) -- The method to use to make this summary. Accepted values are:
- `"last"` -- Take the last token hidden state (like XLNet)
- `"first"` -- Take the first token hidden state (like Bert)
- `"mean"` -- Take the mean of all tokens hidden states
- `"cls_index"` -- Supply a Tensor of classification token position (GPT/GPT-2)
- `"attn"` -- Not implemented now, use multi-head attention
- **summary_use_proj** (`bool`) -- Add a projection after the vector extraction.
- **summary_proj_to_labels** (`bool`) -- If `True`, the projection outputs to `config.num_labels` classes
(otherwise to `config.hidden_size`).
- **summary_activation** (`Optional[str]`) -- Set to `"tanh"` to add a tanh activation to the output,
another string or `None` will add no activation.
- **summary_first_dropout** (`float`) -- Optional dropout probability before the projection and activation.
- **summary_last_dropout** (`float`)-- Optional dropout probability after the projection and activation.
initializer_range (`float`, defaults to 0.02): The standard deviation to use to initialize the weights.
kwargs (`Dict[str, Any]`, *optional*):
Additional keyword arguments passed along to the `__init__` of `tf.keras.layers.Layer`.
"""
def __init__(self, config: PretrainedConfig, initializer_range: float = 0.02, **kwargs):
super().__init__(**kwargs)
self.summary_type = config.summary_type if hasattr(config, "summary_use_proj") else "last"
if self.summary_type == "attn":
# We should use a standard multi-head attention module with absolute positional embedding for that.
# Cf. https://github.com/zihangdai/xlnet/blob/master/modeling.py#L253-L276
# We can probably just use the multi-head attention module of PyTorch >=1.1.0
raise NotImplementedError
self.has_summary = hasattr(config, "summary_use_proj") and config.summary_use_proj
if self.has_summary:
if hasattr(config, "summary_proj_to_labels") and config.summary_proj_to_labels and config.num_labels > 0:
num_classes = config.num_labels
else:
num_classes = config.hidden_size
self.summary = tf.keras.layers.Dense(
num_classes, kernel_initializer=get_initializer(initializer_range), name="summary"
)
self.has_activation = False
activation_string = getattr(config, "summary_activation", None)
if activation_string is not None:
self.has_activation = True
self.activation = get_tf_activation(activation_string)
self.has_first_dropout = hasattr(config, "summary_first_dropout") and config.summary_first_dropout > 0
if self.has_first_dropout:
self.first_dropout = tf.keras.layers.Dropout(config.summary_first_dropout)
self.has_last_dropout = hasattr(config, "summary_last_dropout") and config.summary_last_dropout > 0
if self.has_last_dropout:
self.last_dropout = tf.keras.layers.Dropout(config.summary_last_dropout)
self.hidden_size = config.hidden_size
def call(self, inputs, cls_index=None, training=False):
if not isinstance(inputs, (dict, tuple, list)):
hidden_states = inputs
elif isinstance(inputs, (tuple, list)):
hidden_states = inputs[0]
cls_index = inputs[1] if len(inputs) > 1 else None
assert len(inputs) <= 2, "Too many inputs."
else:
hidden_states = inputs.get("hidden_states")
cls_index = inputs.get("cls_index", None)
if self.summary_type == "last":
output = hidden_states[:, -1]
elif self.summary_type == "first":
output = hidden_states[:, 0]
elif self.summary_type == "mean":
output = tf.reduce_mean(hidden_states, axis=1)
elif self.summary_type == "cls_index":
hidden_shape = shape_list(hidden_states) # e.g. [batch, num choices, seq length, hidden dims]
if cls_index is None:
cls_index = tf.fill(
hidden_shape[:-2], hidden_shape[-2] - 1
) # A tensor full of shape [batch] or [batch, num choices] full of sequence length
cls_shape = shape_list(cls_index)
if len(cls_shape) <= len(hidden_shape) - 2:
cls_index = tf.expand_dims(cls_index, axis=-1)
# else:
# cls_index = cls_index[..., tf.newaxis]
# cls_index = cls_index.expand((-1,) * (cls_index.dim()-1) + (hidden_states.size(-1),))
# shape of cls_index: (bsz, XX, 1, hidden_size) where XX are optional leading dim of hidden_states
output = tf.gather(hidden_states, cls_index, batch_dims=len(hidden_shape) - 2)
output = tf.squeeze(
output, axis=len(hidden_shape) - 2
) # shape of output: (batch, num choices, hidden_size)
elif self.summary_type == "attn":
raise NotImplementedError
if self.has_first_dropout:
output = self.first_dropout(output, training=training)
if self.has_summary:
output = self.summary(output)
if self.has_activation:
output = self.activation(output)
if self.has_last_dropout:
output = self.last_dropout(output, training=training)
return output
def build(self, input_shape):
if self.built:
return
self.built = True
if getattr(self, "summary", None) is not None:
with tf.name_scope("summary"):
self.summary.build(self.hidden_size)
def get_initializer(initializer_range: float = 0.02) -> tf.keras.initializers.TruncatedNormal:
"""
Creates a `tf.keras.initializers.TruncatedNormal` with the given range.
Args:
initializer_range (*float*, defaults to 0.02): Standard deviation of the initializer range.
Returns:
`tf.keras.initializers.TruncatedNormal`: The truncated normal initializer.
"""
return tf.keras.initializers.TruncatedNormal(stddev=initializer_range)