# FSMT¶

DISCLAIMER: If you see something strange, file a Github Issue and assign @stas00.

## Overview¶

FSMT (FairSeq MachineTranslation) models were introduced in “Facebook FAIR’s WMT19 News Translation Task Submission” <this paper <https://arxiv.org/abs/1907.06616>__ by Nathan Ng, Kyra Yee, Alexei Baevski, Myle Ott, Michael Auli, Sergey Edunov.

The abstract of the paper is the following:

This paper describes Facebook FAIR’s submission to the WMT19 shared news translation task. We participate in two language pairs and four language directions, English <-> German and English <-> Russian. Following our submission from last year, our baseline systems are large BPE-based transformer models trained with the Fairseq sequence modeling toolkit which rely on sampled back-translations. This year we experiment with different bitext data filtering schemes, as well as with adding filtered back-translated data. We also ensemble and fine-tune our models on domain-specific data, then decode using noisy channel model reranking. Our submissions are ranked first in all four directions of the human evaluation campaign. On En->De, our system significantly outperforms other systems as well as human translations. This system improves upon our WMT’18 submission by 4.5 BLEU points.

The original code can be found here <https://github.com/pytorch/fairseq/tree/master/examples/wmt19>__.

## Implementation Notes¶

• FSMT uses source and target vocab pair, that aren’t combined into one. It doesn’t share embed tokens either. Its tokenizer is very similar to XLMTokenizer and the main model is derived from BartModel.

## FSMTForConditionalGeneration¶

class transformers.FSMTForConditionalGeneration(config: transformers.configuration_fsmt.FSMTConfig)[source]

The FSMT Model with a language modeling head. Can be used for summarization.

This model is a PyTorch torch.nn.Module sub-class. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matters related to general usage and behavior.

Parameters

config (FSMTConfig) – Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

forward(input_ids, attention_mask=None, encoder_outputs=None, decoder_input_ids=None, decoder_attention_mask=None, past_key_values=None, labels=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None, **unused)[source]

The FSMTForConditionalGeneration forward method, overrides the __call__() special method.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Parameters
• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) – Indices of input sequence tokens in the vocabulary. Use FSMTTokenizer.encode to produce them. Padding will be ignored by default should you provide it. Indices can be obtained using transformers.FSMTTokenizer.encode(text).

• attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional) – Mask to avoid performing attention on padding token indices in input_ids. Mask values selected in [0, 1]: 1 for tokens that are NOT MASKED, 0 for MASKED tokens.

• encoder_outputs (tuple(tuple(torch.FloatTensor), optional) – Tuple consists of (last_hidden_state, optional: hidden_states, optional: attentions) last_hidden_state of shape (batch_size, sequence_length, hidden_size), optional) is a sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention of the decoder.

• decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional) – Provide for translation and summarization training. By default, the model will create this tensor by shifting the input_ids right, following the paper.

• decoder_attention_mask (torch.BoolTensor of shape (batch_size, tgt_seq_len), optional) – Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default. If you want to change padding behavior, you should read _prepare_decoder_inputs() and modify. See diagram 1 in the paper for more info on the default strategy

• past_key_values (tuple(tuple(torch.FloatTensor)) of length config.n_layers with each tuple having 4 tensors of shape (batch_size, num_heads, sequence_length - 1, embed_size_per_head)) – Contains pre-computed key and value hidden-states of the attention blocks. Can be used to speed up decoding. If past_key_values are used, the user can optionally input only the last decoder_input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, 1) instead of all decoder_input_ids of shape (batch_size, sequence_length).

• use_cache (bool, optional, defaults to True) – If use_cache is True, past_key_values are returned and can be used to speed up decoding (see past_key_values).

• output_attentions (bool, optional) – If set to True, the attentions tensors of all attention layers are returned. See attentions under returned tensors for more detail.

• output_hidden_states (bool, optional) – If set to True, the hidden states of all layers are returned. See hidden_states under returned tensors for more detail.

• return_dict (bool, optional) – If set to True, the model will return a ModelOutput instead of a plain tuple.

• labels (torch.LongTensor of shape (batch_size, sequence_length), optional) – Labels for computing the masked language modeling loss. Indices should either be in [0, ..., config.vocab_size] or -100 (see input_ids docstring). Tokens with indices set to -100 are ignored (masked), the loss is only computed for the tokens with labels in [0, ..., config.vocab_size].

Returns

A Seq2SeqLMOutput (if return_dict=True is passed or when config.return_dict=True) or a tuple of torch.FloatTensor comprising various elements depending on the configuration (FSMTConfig) and inputs.

• loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) – Languaged modeling loss.

• logits (torch.FloatTensor of shape (batch_size, sequence_length, config.vocab_size)) – Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).

• past_key_values (List[torch.FloatTensor], optional, returned when use_cache=True is passed or when config.use_cache=True) – List of torch.FloatTensor of length config.n_layers, with each tensor of shape (2, batch_size, num_heads, sequence_length, embed_size_per_head)).

Contains pre-computed hidden-states (key and values in the attention blocks) of the decoder that can be used (see past_key_values input) to speed up sequential decoding.

• decoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) – Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

Hidden-states of the decoder at the output of each layer plus the initial embedding outputs.

• decoder_attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) – Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

Attentions weights of the decoder, after the attention softmax, used to compute the weighted average in the self-attention heads.

• encoder_last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) – Sequence of hidden-states at the output of the last layer of the encoder of the model.

• encoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) – Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

Hidden-states of the encoder at the output of each layer plus the initial embedding outputs.

• encoder_attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) – Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

Attentions weights of the encoder, after the attention softmax, used to compute the weighted average in the self-attention heads.

Return type

Seq2SeqLMOutput or tuple(torch.FloatTensor)

Translation example:

from transformers import FSMTTokenizer, FSMTForConditionalGeneration

mname = "facebook/wmt19-ru-en"
model = FSMTForConditionalGeneration.from_pretrained(mname)
tokenizer = FSMTTokenizer.from_pretrained(mname)

src_text = "Машинное обучение - это здорово, не так ли?"
input_ids = tokenizer.encode(src_text, return_tensors='pt')
outputs = model.generate(input_ids, num_beams=5, num_return_sequences=3)
for i, output in enumerate(outputs):
decoded = tokenizer.decode(output, skip_special_tokens=True)
print(f"{i}: {decoded})
# 1: Machine learning is great, isn't it? ...


## FSMTConfig¶

class transformers.FSMTConfig(langs, src_vocab_size, tgt_vocab_size, activation_function='relu', d_model=1024, max_length=200, max_position_embeddings=1024, encoder_ffn_dim=4096, encoder_layers=12, encoder_attention_heads=16, encoder_layerdrop=0.0, decoder_ffn_dim=4096, decoder_layers=12, decoder_attention_heads=16, decoder_layerdrop=0.0, attention_dropout=0.0, dropout=0.1, activation_dropout=0.0, init_std=0.02, pad_token_id=1, bos_token_id=0, eos_token_id=2, decoder_start_token_id=2, is_encoder_decoder=True, scale_embedding=True, tie_word_embeddings=False, num_beams=5, length_penalty=1.0, early_stopping=False, **common_kwargs)[source]

The FSMTConfig forward method, overrides the __call__() special method.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Parameters
• langs (List[str]) – source language, target_language (e.g. [‘en’, ‘ru’])

• src_vocab_size (int) – defines the different tokens that can be represented by inputs_ids passed to the forward method in the encoder.

• tgt_vocab_size (int) – defines the different tokens that can be represented by inputs_ids passed to the forward method in the decoder.

• d_model (int, optional, defaults to 1024) – Dimensionality of the layers and the pooler layer.

• encoder_layers (int, optional, defaults to 12) – Number of encoder layers, 16 for pegasus, 6 for bart-base and marian

• decoder_layers (int, optional, defaults to 12) – Number of decoder layers, 16 for pegasus, 6 for bart-base and marian

• encoder_attention_heads (int, optional, defaults to 16) – Number of attention heads for each attention layer in the Transformer encoder.

• decoder_attention_heads (int, optional, defaults to 16) – Number of attention heads for each attention layer in the Transformer decoder.

• decoder_ffn_dim (int, optional, defaults to 4096) – Dimensionality of the “intermediate” (i.e., feed-forward) layer in decoder.

• encoder_ffn_dim (int, optional, defaults to 4096) – Dimensionality of the “intermediate” (i.e., feed-forward) layer in decoder.

• activation_function (str or function, optional, defaults to “relu”) – The non-linear activation function (function or string) in the encoder and pooler. If string, “gelu”, “relu”, “swish” and “gelu_new” are supported.

• dropout (float, optional, defaults to 0.1) – The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.

• attention_dropout (float, optional, defaults to 0.0) – The dropout ratio for the attention probabilities.

• activation_dropout (float, optional, defaults to 0.0) – The dropout ratio for activations inside the fully connected layer.

• max_position_embeddings (int, optional, defaults to 1024) – 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.

• scale_embedding (bool, optional, defaults to True) – Scale embeddings by diving by sqrt(d_model).

• bos_token_id (int, optional, defaults to 0) – Beginning of stream token id.

• pad_token_id (int, optional, defaults to 1) – Padding token id.

• eos_token_id (int, optional, defaults to 2) – End of stream token id.

• decoder_start_token_id (int, optional) – This model starts decoding with eos_token_id

• encoder_layerdrop – (float, optional, defaults to 0.0): Google “layerdrop arxiv”, as its not explainable in one line.

• decoder_layerdrop – (float, optional, defaults to 0.0): Google “layerdrop arxiv”, as its not explainable in one line.

• is_encoder_decoder (bool, optional, defaults to True) – Whether this is an encoder/decoder model.

• tie_word_embeddings (bool, optional, defaults to False) – Whether to tie input and output embeddings.

• num_beams (int, optional, defaults to 5) – Number of beams for beam search that will be used by default in the generate method of the model. 1 means no beam search.

• length_penalty (float, optional, defaults to 1) – Exponential penalty to the length that will be used by default in the generate method of the model.

• early_stopping (bool, optional, defaults to False) – Flag that will be used by default in the generate method of the model. Whether to stop the beam search when at least num_beams sentences are finished per batch or not.

Configuration class for FSMT.

to_dict()[source]

Serializes this instance to a Python dictionary. Override the default to_dict() from PretrainedConfig.

Returns

Dictionary of all the attributes that make up this configuration instance,

Return type

Dict[str, any]

## FSMTTokenizer¶

class transformers.FSMTTokenizer(langs=None, src_vocab_file=None, tgt_vocab_file=None, merges_file=None, unk_token='<unk>', bos_token='<s>', sep_token='</s>', pad_token='<pad>', **kwargs)[source]

BPE tokenizer for FSMT (fairseq transformer) See: https://github.com/pytorch/fairseq/tree/master/examples/wmt19

• Moses preprocessing & tokenization for most supported languages

• (optionally) lower case & normalize all inputs text

• argument special_tokens and function set_special_tokens, can be used to add additional symbols (ex: “__classify__”) to a vocabulary

• langs defines a pair of languages

This tokenizer inherits from PreTrainedTokenizer which contains most of the methods. Users should refer to the superclass for more information regarding methods.

Parameters
• langs (List[str]) – a list of two languages to translate from and to, e.g. ["en", "ru"].

• src_vocab_file (string) – Source language vocabulary file.

• tgt_vocab_file (string) – Target language vocabulary file.

• merges_file (string) – Merges file.

• do_lower_case (bool, optional, defaults to True) – Whether to lowercase the input when tokenizing.

• unk_token (string, 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.

• bos_token (string, optional, defaults to “<s>”) –

The beginning of sequence token that was used during pre-training. Can be used a sequence classifier token.

Note

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.

• sep_token (string, 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.

• pad_token (string, optional, defaults to “<pad>”) – The token used for padding, for example when batching sequences of different lengths.

build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]] = None) → List[int][source]

Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and adding special tokens. A FAIRSEQ_TRANSFORMER sequence has the following format:

• single sequence: <s> X </s>

• pair of sequences: <s> A </s> B </s>

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 of input IDs with the appropriate special tokens.

Return type

List[int]

convert_tokens_to_string(tokens)[source]

Converts a sequence of tokens (string) in a single string.

create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]] = None) → List[int][source]

Creates a mask from the two sequences passed to be used in a sequence-pair classification task. An FAIRSEQ_TRANSFORMER sequence pair mask has the following format:

0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
| first sequence    | second sequence |


if token_ids_1 is None, only returns the first portion of the mask (0s).

Parameters
• token_ids_0 (List[int]) – List of ids.

• token_ids_1 (List[int], optional) – Optional second list of IDs for sequence pairs.

Returns

List of token type IDs according to the given sequence(s).

Return type

List[int]

get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False) → List[int][source]

Retrieves 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 methods.

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 to False) – Set to True if the token list is already formatted with special tokens for the model

Returns

A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.

Return type

List[int]

get_vocab() → Dict[str, int][source]

Returns the vocabulary as a dictionary of token to index.

tokenizer.get_vocab()[token] is equivalent to tokenizer.convert_tokens_to_ids(token) when token is in the vocab.

Returns

The vocabulary.

Return type

Dict[str, int]

prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]] = None, max_length: Optional[int] = None, max_target_length: Optional[int] = None, return_tensors: str = 'pt', truncation=True, padding='longest', **unused) → transformers.tokenization_utils_base.BatchEncoding[source]
Arguments:
src_texts: (list):

list of documents to summarize or source language texts

tgt_texts: (list, optional):

list of tgt language texts or summaries.

max_length (int, optional):

Controls the maximum length for encoder inputs (documents to summarize or source language texts) If left unset or set to None, this will use the predefined model maximum length if a maximum length is required by one of the truncation/padding parameters. If the model has no specific maximum input length (like XLNet) truncation/padding to a maximum length will be deactivated.

max_target_length (int, optional):

Controls the maximum length of decoder inputs (target language texts or summaries) If left unset or set to None, this will use the max_length value.

padding (bool, str or PaddingStrategy, optional, defaults to False):

Activates and controls padding. Accepts the following values:

• True or 'longest': Pad to the longest sequence in the batch (or no padding if only a single sequence if provided).

• 'max_length': Pad to a maximum length specified with the argument max_length or to the maximum acceptable input length for the model if that argument is not provided.

• False or 'do_not_pad' (default): No padding (i.e., can output a batch with sequences of different lengths).

return_tensors (str or TensorType, optional, defaults to “pt”):

If set, will return tensors instead of list of python integers. Acceptable values are:

• 'tf': Return TensorFlow tf.constant objects.

• 'pt': Return PyTorch torch.Tensor objects.

• 'np': Return Numpy np.ndarray objects.

truncation (bool, str or TruncationStrategy, optional, defaults to True):

Activates and controls truncation. Accepts the following values:

• True or 'longest_first': Truncate to a maximum length specified with the argument max_length or to the maximum acceptable input length for the model if that argument is not provided. This will truncate token by token, removing a token from the longest sequence in the pair if a pair of sequences (or a batch of pairs) is provided.

• 'only_first': Truncate to a maximum length specified with the argument max_length or to the maximum acceptable input length for the model if that argument is not provided. This will only truncate the first sequence of a pair if a pair of sequences (or a batch of pairs) is provided.

• 'only_second': Truncate to a maximum length specified with the argument max_length or to the maximum acceptable input length for the model if that argument is not provided. This will only truncate the second sequence of a pair if a pair of sequences (or a batch of pairs) is provided.

• False or 'do_not_truncate' (default): No truncation (i.e., can output batch with sequence lengths greater than the model maximum admissible input size).

Return:

BatchEncoding: A BatchEncoding with the following fields:

• input_ids – List of token ids to be fed to the encoder.

• attention_mask – List of indices specifying which tokens should be attended to by the model.

• decoder_input_ids – List of token ids to be fed to the decoder.

• decoder_attention_mask – List of indices specifying which tokens should be attended to by the decoder.

This does not include causal mask, which is built by the model.

The full set of keys [input_ids, attention_mask, decoder_input_ids,  decoder_attention_mask], will only be returned if tgt_texts is passed. Otherwise, input_ids, attention_mask will be the only keys.

Prepare model inputs for translation. For best performance, translate one sentence at a time.

save_vocabulary(save_directory)[source]

Save the vocabulary and special tokens file to a directory.

Parameters

save_directory (str) – The directory in which to save the vocabulary.

Returns

Paths to the files saved.

Return type

Tuple(str)

property vocab_size

Size of the base vocabulary (without the added tokens).

Type

int

## FSMTModel¶

class transformers.FSMTModel(config: transformers.configuration_fsmt.FSMTConfig)[source]

The bare FSMT Model outputting raw hidden-states without any specific head on top.

This model is a PyTorch torch.nn.Module sub-class. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matters related to general usage and behavior.

Parameters

config (FSMTConfig) – Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

forward(input_ids, attention_mask=None, decoder_input_ids=None, encoder_outputs: Optional[Tuple] = None, decoder_attention_mask=None, past_key_values=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None, **kwargs)[source]

The FSMTModel forward method, overrides the __call__() special method.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Parameters
• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) – Indices of input sequence tokens in the vocabulary. Use FSMTTokenizer.encode to produce them. Padding will be ignored by default should you provide it. Indices can be obtained using transformers.FSMTTokenizer.encode(text).

• attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional) – Mask to avoid performing attention on padding token indices in input_ids. Mask values selected in [0, 1]: 1 for tokens that are NOT MASKED, 0 for MASKED tokens.

• encoder_outputs (tuple(tuple(torch.FloatTensor), optional) – Tuple consists of (last_hidden_state, optional: hidden_states, optional: attentions) last_hidden_state of shape (batch_size, sequence_length, hidden_size), optional) is a sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention of the decoder.

• decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional) – Provide for translation and summarization training. By default, the model will create this tensor by shifting the input_ids right, following the paper.

• decoder_attention_mask (torch.BoolTensor of shape (batch_size, tgt_seq_len), optional) – Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default. If you want to change padding behavior, you should read _prepare_decoder_inputs() and modify. See diagram 1 in the paper for more info on the default strategy

• past_key_values (tuple(tuple(torch.FloatTensor)) of length config.n_layers with each tuple having 4 tensors of shape (batch_size, num_heads, sequence_length - 1, embed_size_per_head)) – Contains pre-computed key and value hidden-states of the attention blocks. Can be used to speed up decoding. If past_key_values are used, the user can optionally input only the last decoder_input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, 1) instead of all decoder_input_ids of shape (batch_size, sequence_length).

• use_cache (bool, optional, defaults to True) – If use_cache is True, past_key_values are returned and can be used to speed up decoding (see past_key_values).

• output_attentions (bool, optional) – If set to True, the attentions tensors of all attention layers are returned. See attentions under returned tensors for more detail.

• output_hidden_states (bool, optional) – If set to True, the hidden states of all layers are returned. See hidden_states under returned tensors for more detail.

• return_dict (bool, optional) – If set to True, the model will return a ModelOutput instead of a plain tuple.

Returns

A BaseModelOutputWithPast (if return_dict=True is passed or when config.return_dict=True) or a tuple of torch.FloatTensor comprising various elements depending on the configuration (FSMTConfig) and inputs.

• last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) – Sequence of hidden-states at the output of the last layer of the model.

If past_key_values is used only the last hidden-state of the sequences of shape (batch_size, 1, hidden_size) is output.

• past_key_values (List[torch.FloatTensor], optional, returned when use_cache=True is passed or when config.use_cache=True) – List of torch.FloatTensor of length config.n_layers, with each tensor of shape (2, batch_size, num_heads, sequence_length, embed_size_per_head)).

Contains pre-computed hidden-states (key and values in the attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

• hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) – Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

Hidden-states of the model at the output of each layer plus the initial embedding outputs.

• attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) – Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

Return type

BaseModelOutputWithPast or tuple(torch.FloatTensor)

Example:

>>> from transformers import FSMTTokenizer, FSMTModel
>>> import torch

>>> tokenizer = FSMTTokenizer.from_pretrained('facebook/wmt19-ru-en')
>>> model = FSMTModel.from_pretrained('facebook/wmt19-ru-en', return_dict=True)

>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> outputs = model(**inputs)

>>> last_hidden_states = outputs.last_hidden_state