XLM-ProphetNet
DISCLAIMER: If you see something strange, file a Github Issue and assign @patrickvonplaten
Overview
The XLM-ProphetNet model was proposed in ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training, by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang, Ming Zhou on 13 Jan, 2020.
XLM-ProphetNet is an encoder-decoder model and can predict n-future tokens for “ngram” language modeling instead of just the next token. Its architecture is identical to ProhpetNet, but the model was trained on the multi-lingual “wiki100” Wikipedia dump.
The abstract from the paper is the following:
In this paper, we present a new sequence-to-sequence pretraining model called ProphetNet, which introduces a novel self-supervised objective named future n-gram prediction and the proposed n-stream self-attention mechanism. Instead of the optimization of one-step ahead prediction in traditional sequence-to-sequence model, the ProphetNet is optimized by n-step ahead prediction which predicts the next n tokens simultaneously based on previous context tokens at each time step. The future n-gram prediction explicitly encourages the model to plan for the future tokens and prevent overfitting on strong local correlations. We pre-train ProphetNet using a base scale dataset (16GB) and a large scale dataset (160GB) respectively. Then we conduct experiments on CNN/DailyMail, Gigaword, and SQuAD 1.1 benchmarks for abstractive summarization and question generation tasks. Experimental results show that ProphetNet achieves new state-of-the-art results on all these datasets compared to the models using the same scale pretraining corpus.
The Authors’ code can be found here.
XLMProphetNetConfig
class transformers.XLMProphetNetConfig
< source >( activation_dropout: typing.Optional[float] = 0.1 activation_function: typing.Union[str, typing.Callable, NoneType] = 'gelu' vocab_size: typing.Optional[int] = 30522 hidden_size: typing.Optional[int] = 1024 encoder_ffn_dim: typing.Optional[int] = 4096 num_encoder_layers: typing.Optional[int] = 12 num_encoder_attention_heads: typing.Optional[int] = 16 decoder_ffn_dim: typing.Optional[int] = 4096 num_decoder_layers: typing.Optional[int] = 12 num_decoder_attention_heads: typing.Optional[int] = 16 attention_dropout: typing.Optional[float] = 0.1 dropout: typing.Optional[float] = 0.1 max_position_embeddings: typing.Optional[int] = 512 init_std: typing.Optional[float] = 0.02 is_encoder_decoder: typing.Optional[bool] = True add_cross_attention: typing.Optional[bool] = True decoder_start_token_id: typing.Optional[int] = 0 ngram: typing.Optional[int] = 2 num_buckets: typing.Optional[int] = 32 relative_max_distance: typing.Optional[int] = 128 disable_ngram_loss: typing.Optional[bool] = False eps: typing.Optional[float] = 0.0 use_cache: typing.Optional[bool] = True pad_token_id: typing.Optional[int] = 0 bos_token_id: typing.Optional[int] = 1 eos_token_id: typing.Optional[int] = 2 **kwargs )
Parameters
-
activation_dropout (
float, optional, defaults to 0.1) — The dropout ratio for activations inside the fully connected layer. -
activation_function (
strorfunction, optional, defaults to"gelu") — The non-linear activation function (function or string) in the encoder and pooler. If string,"gelu","relu","silu"and"gelu_new"are supported. -
vocab_size (
int, optional, defaults to 30522) — Vocabulary size of the ProphetNET model. Defines the number of different tokens that can be represented by theinputs_idspassed when calling XLMProphetNetModel. -
hidden_size (
int, optional, defaults to 1024) — Dimensionality of the layers and the pooler layer. -
encoder_ffn_dim (
int, optional, defaults to 4096) — Dimensionality of the “intermediate” (often named feed-forward) layer in decoder. -
num_encoder_layers (
int, optional, defaults to 12) — Number of encoder layers. -
num_encoder_attention_heads (
int, optional, defaults to 16) — Number of attention heads for each attention layer in the Transformer encoder. -
decoder_ffn_dim (
int, optional, defaults to 4096) — Dimensionality of theintermediate(often named feed-forward) layer in decoder. -
num_decoder_layers (
int, optional, defaults to 12) — Number of decoder layers. -
num_decoder_attention_heads (
int, optional, defaults to 16) — Number of attention heads for each attention layer in the Transformer decoder. -
attention_dropout (
float, optional, defaults to 0.1) — The dropout ratio for the attention probabilities. -
dropout (
float, optional, defaults to 0.1) — The dropout probability for all fully connected layers in the embeddings, encoder, and pooler. -
max_position_embeddings (
int, optional, defaults to 512) — The maximum sequence length that this model might ever be used with. Typically set this to something large just in case (e.g., 512 or 1024 or 2048). -
init_std (
float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices. -
add_cross_attention (
bool, optional, defaults toTrue) — Whether cross-attention layers should be added to the model. -
is_encoder_decoder (
bool, optional, defaults toTrue) — Whether this is an encoder/decoder model. -
pad_token_id (
int, optional, defaults to 1) — Padding token id. -
bos_token_id (
int, optional, defaults to 0) — Beginning of stream token id. -
eos_token_id (
int, optional, defaults to 2) — End of stream token id. -
ngram (
int, optional, defaults to 2) — Number of future tokens to predict. Set to 1 to be same as traditional Language model to predict next first token. -
num_buckets (
int, optional, defaults to 32) — The number of buckets to use for each attention layer. This is for relative position calculation. See the [T5 paper](see https://arxiv.org/abs/1910.10683) for more details. -
relative_max_distance (
int, optional, defaults to 128) — Relative distances greater than this number will be put into the last same bucket. This is for relative position calculation. See the [T5 paper](see https://arxiv.org/abs/1910.10683) for more details. -
disable_ngram_loss (
bool, optional, defaults toFalse) — Whether be trained predicting only the next first token. -
eps (
float, optional, defaults to 0.0) — Controls theepsilonparameter value for label smoothing in the loss calculation. If set to 0, no label smoothing is performed. -
use_cache (
bool, optional, defaults toTrue) — Whether or not the model should return the last key/values attentions (not used by all models).
This is the configuration class to store the configuration of a XLMProphetNetModel. It is used to instantiate a XLMProphetNet model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the XLMProphetNet microsoft/xprophetnet-large-wiki100-cased architecture.
Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.
XLMProphetNetTokenizer
class transformers.XLMProphetNetTokenizer
< source >( vocab_file bos_token = '[SEP]' eos_token = '[SEP]' sep_token = '[SEP]' unk_token = '[UNK]' pad_token = '[PAD]' cls_token = '[CLS]' mask_token = '[MASK]' sp_model_kwargs: typing.Union[typing.Dict[str, typing.Any], NoneType] = None **kwargs )
Parameters
-
vocab_file (
str) — Path to the vocabulary file. -
bos_token (
str, optional, defaults to"<s>") — The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token.When building a sequence using special tokens, this is not the token that is used for the beginning of sequence. The token used is the
cls_token. -
eos_token (
str, optional, defaults to"</s>") — The end of sequence token.When building a sequence using special tokens, this is not the token that is used for the end of sequence. The token used is the
sep_token. -
sep_token (
str, optional, defaults to"</s>") — The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for sequence classification or for a text and a question for question answering. It is also used as the last token of a sequence built with special tokens. -
cls_token (
str, optional, defaults to"<s>") — The classifier token which is used when doing sequence classification (classification of the whole sequence instead of per-token classification). It is the first token of the sequence when built with special tokens. -
unk_token (
str, optional, defaults to"<unk>") — The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this token instead. -
pad_token (
str, optional, defaults to"<pad>") — The token used for padding, for example when batching sequences of different lengths. -
mask_token (
str, optional, defaults to"<mask>") — The token used for masking values. This is the token used when training this model with masked language modeling. This is the token which the model will try to predict. -
additional_special_tokens (
List[str], optional, defaults to["<s>NOTUSED", "</s>NOTUSED"]) — Additional special tokens used by the tokenizer. -
sp_model_kwargs (
dict, optional) — Will be passed to theSentencePieceProcessor.__init__()method. The Python wrapper for SentencePiece can be used, among other things, to set:-
enable_sampling: Enable subword regularization. -
nbest_size: Sampling parameters for unigram. Invalid for BPE-Dropout.nbest_size = {0,1}: No sampling is performed.nbest_size > 1: samples from the nbest_size results.nbest_size < 0: assuming that nbest_size is infinite and samples from the all hypothesis (lattice) using forward-filtering-and-backward-sampling algorithm.
-
alpha: Smoothing parameter for unigram sampling, and dropout probability of merge operations for BPE-dropout.
-
-
sp_model (
SentencePieceProcessor) — The SentencePiece processor that is used for every conversion (string, tokens and IDs).
Adapted from RobertaTokenizer and XLNetTokenizer. Based on SentencePiece.
This tokenizer inherits from PreTrainedTokenizer which contains most of the main methods. Users should refer to this superclass for more information regarding those methods.
build_inputs_with_special_tokens
< source >(
token_ids_0: typing.List[int]
token_ids_1: typing.Optional[typing.List[int]] = None
)
→
List[int]
Parameters
-
token_ids_0 (
List[int]) — List of IDs to which the special tokens will be added -
token_ids_1 (
List[int], optional) — Optional second list of IDs for sequence pairs.
Returns
List[int]
list of input IDs with the appropriate special tokens.
Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and adding special tokens. A XLMProphetNet sequence has the following format:
- single sequence:
X [SEP] - pair of sequences:
A [SEP] B [SEP]
Converts a sequence of tokens (strings for sub-words) in a single string.
create_token_type_ids_from_sequences
< source >(
token_ids_0: typing.List[int]
token_ids_1: typing.Optional[typing.List[int]] = None
)
→
List[int]
Create a mask from the two sequences passed to be used in a sequence-pair classification task. XLMProphetNet does not make use of token type ids, therefore a list of zeros is returned.
get_special_tokens_mask
< source >(
token_ids_0: typing.List[int]
token_ids_1: typing.Optional[typing.List[int]] = None
already_has_special_tokens: bool = False
)
→
List[int]
Parameters
-
token_ids_0 (
List[int]) — List of IDs. -
token_ids_1 (
List[int], optional) — Optional second list of IDs for sequence pairs. -
already_has_special_tokens (
bool, optional, defaults toFalse) — Whether or not the token list is already formatted with special tokens for the model.
Returns
List[int]
A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
special tokens using the tokenizer prepare_for_model method.
XLMProphetNetModel
This class overrides ProphetNetModel. Please check the superclass for the appropriate documentation alongside usage examples.
Example:
>>> from transformers import XLMProphetNetTokenizer, XLMProphetNetModel
>>> tokenizer = XLMProphetNetTokenizer.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
>>> model = XLMProphetNetModel.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
>>> input_ids = tokenizer(
... "Studies have been shown that owning a dog is good for you", return_tensors="pt"
... ).input_ids # Batch size 1
>>> decoder_input_ids = tokenizer("Studies show that", return_tensors="pt").input_ids # Batch size 1
>>> outputs = model(input_ids=input_ids, decoder_input_ids=decoder_input_ids)
>>> last_hidden_states = outputs.last_hidden_state # main stream hidden states
>>> last_hidden_states_ngram = outputs.last_hidden_state_ngram # predict hidden statesXLMProphetNetEncoder
class transformers.XLMProphetNetEncoder
< source >( config: ProphetNetConfig word_embeddings: Embedding = None )
This class overrides ProphetNetEncoder. Please check the superclass for the appropriate documentation alongside usage examples.
Example:
>>> from transformers import XLMProphetNetTokenizer, XLMProphetNetEncoder
>>> import torch
>>> tokenizer = XLMProphetNetTokenizer.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
>>> model = XLMProphetNetEncoder.from_pretrained("patrickvonplaten/xprophetnet-large-uncased-standalone")
>>> assert model.config.is_decoder, f"{model.__class__} has to be configured as a decoder."
>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> outputs = model(**inputs)
>>> last_hidden_states = outputs.last_hidden_stateXLMProphetNetDecoder
class transformers.XLMProphetNetDecoder
< source >( config: ProphetNetConfig word_embeddings: typing.Optional[torch.nn.modules.sparse.Embedding] = None )
This class overrides ProphetNetDecoder. Please check the superclass for the appropriate documentation alongside usage examples.
Example:
>>> from transformers import XLMProphetNetTokenizer, XLMProphetNetDecoder
>>> import torch
>>> tokenizer = XLMProphetNetTokenizer.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
>>> model = XLMProphetNetDecoder.from_pretrained(
... "patrickvonplaten/xprophetnet-large-uncased-standalone", add_cross_attention=False
... )
>>> assert model.config.is_decoder, f"{model.__class__} has to be configured as a decoder."
>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> outputs = model(**inputs)
>>> last_hidden_states = outputs.last_hidden_stateXLMProphetNetForConditionalGeneration
This class overrides ProphetNetForConditionalGeneration. Please check the superclass for the appropriate documentation alongside usage examples.
Example:
>>> from transformers import XLMProphetNetTokenizer, XLMProphetNetForConditionalGeneration
>>> tokenizer = XLMProphetNetTokenizer.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
>>> model = XLMProphetNetForConditionalGeneration.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
>>> input_ids = tokenizer(
... "Studies have been shown that owning a dog is good for you", return_tensors="pt"
... ).input_ids # Batch size 1
>>> decoder_input_ids = tokenizer("Studies show that", return_tensors="pt").input_ids # Batch size 1
>>> outputs = model(input_ids=input_ids, decoder_input_ids=decoder_input_ids)
>>> logits_next_token = outputs.logits # logits to predict next token as usual
>>> logits_ngram_next_tokens = outputs.logits_ngram # logits to predict 2nd, 3rd, ... next tokensXLMProphetNetForCausalLM
This class overrides ProphetNetForCausalLM. Please check the superclass for the appropriate documentation alongside usage examples.
Example:
>>> from transformers import XLMProphetNetTokenizer, XLMProphetNetForCausalLM
>>> import torch
>>> tokenizer = XLMProphetNetTokenizer.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
>>> model = XLMProphetNetForCausalLM.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
>>> assert model.config.is_decoder, f"{model.__class__} has to be configured as a decoder."
>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> outputs = model(**inputs)
>>> logits = outputs.logits
>>> # Model can also be used with EncoderDecoder framework
>>> from transformers import EncoderDecoderModel, XLMProphetNetTokenizer, XLMRobertaTokenizer
>>> import torch
>>> tokenizer_enc = XLMRobertaTokenizer.from_pretrained("xlm-roberta-large")
>>> tokenizer_dec = XLMProphetNetTokenizer.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
>>> model = EncoderDecoderModel.from_encoder_decoder_pretrained(
... "xlm-roberta-large", "microsoft/xprophetnet-large-wiki100-cased"
... )
>>> ARTICLE = (
... "the us state department said wednesday it had received no "
... "formal word from bolivia that it was expelling the us ambassador there "
... "but said the charges made against him are `` baseless ."
... )
>>> input_ids = tokenizer_enc(ARTICLE, return_tensors="pt").input_ids
>>> labels = tokenizer_dec("us rejects charges against its ambassador in bolivia", return_tensors="pt").input_ids
>>> outputs = model(input_ids=input_ids, decoder_input_ids=labels[:, :-1], labels=labels[:, 1:])
>>> loss = outputs.loss