# Speech2Text¶

## Overview¶

The Speech2Text model was proposed in fairseq S2T: Fast Speech-to-Text Modeling with fairseq by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino. It’s a transformer-based seq2seq (encoder-decoder) model designed for end-to-end Automatic Speech Recognition (ASR) and Speech Translation (ST). It uses a convolutional downsampler to reduce the length of speech inputs by 3/4th before they are fed into the encoder. The model is trained with standard autoregressive cross-entropy loss and generates the transcripts/translations autoregressively. Speech2Text has been fine-tuned on several datasets for ASR and ST: LibriSpeech, CoVoST 2, MuST-C.

This model was contributed by valhalla. The original code can be found here.

## Inference¶

Speech2Text is a speech model that accepts a float tensor of log-mel filter-bank features extracted from the speech signal. It’s a transformer-based seq2seq model, so the transcripts/translations are generated autoregressively. The generate() method can be used for inference.

The Speech2TextFeatureExtractor class is responsible for extracting the log-mel filter-bank features. The Speech2TextProcessor wraps Speech2TextFeatureExtractor and Speech2TextTokenizer into a single instance to both extract the input features and decode the predicted token ids.

The feature extractor depends on torchaudio and the tokenizer depends on sentencepiece so be sure to install those packages before running the examples. You could either install those as extra speech dependancies with pip install transformers"[speech, sentencepiece]" or install the packages seperatly with pip install torchaudio sentencepiece. Also torchaudio requires the development version of the libsndfile package which can be installed via a system package manager. On Ubuntu it can be installed as follows: apt install libsndfile1-dev

• ASR and Speech Translation

>>> import torch
>>> from transformers import Speech2TextProcessor, Speech2TextForConditionalGeneration
>>> import soundfile as sf

>>> def map_to_array(batch):
...     batch["speech"] = speech
...     return batch

>>> ds = load_dataset("patrickvonplaten/librispeech_asr_dummy", "clean", split="validation")
>>> ds = ds.map(map_to_array)

>>> inputs = processor(ds["speech"][0], sampling_rate=16_000, return_tensors="pt")

>>> transcription = processor.batch_decode(generated_ids)

• Multilingual speech translation

For multilingual speech translation models, eos_token_id is used as the decoder_start_token_id and the target language id is forced as the first generated token. To force the target language id as the first generated token, pass the forced_bos_token_id parameter to the generate() method. The following example shows how to transate English speech to French text using the facebook/s2t-medium-mustc-multilingual-st checkpoint.

>>> import torch
>>> from transformers import Speech2TextProcessor, Speech2TextForConditionalGeneration
>>> import soundfile as sf

>>> def map_to_array(batch):
...     batch["speech"] = speech
...     return batch

>>> ds = load_dataset("patrickvonplaten/librispeech_asr_dummy", "clean", split="validation")
>>> ds = ds.map(map_to_array)

>>> inputs = processor(ds["speech"][0], sampling_rate=16_000, return_tensors="pt")

>>> translation = processor.batch_decode(generated_ids)


See the model hub to look for Speech2Text checkpoints.

## Speech2TextConfig¶

class transformers.Speech2TextConfig(vocab_size=10000, encoder_layers=12, encoder_ffn_dim=2048, encoder_attention_heads=4, decoder_layers=6, decoder_ffn_dim=2048, decoder_attention_heads=4, encoder_layerdrop=0.0, decoder_layerdrop=0.0, use_cache=True, is_encoder_decoder=True, activation_function='relu', d_model=256, dropout=0.1, attention_dropout=0.0, activation_dropout=0.0, init_std=0.02, decoder_start_token_id=2, classifier_dropout=0.0, scale_embedding=True, gradient_checkpointing=False, pad_token_id=1, bos_token_id=0, eos_token_id=2, max_source_positions=6000, max_target_positions=1024, num_conv_layers=2, conv_kernel_sizes=5, 5, conv_channels=1024, input_feat_per_channel=80, input_channels=1, **kwargs)[source]

This is the configuration class to store the configuration of a Speech2TextModel. It is used to instantiate an Speech2Text 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 Speech2Text facebook/s2t-small-librispeech-asr architecture.

Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.

Parameters
• vocab_size (int, optional, defaults to 50265) – Vocabulary size of the Speech2Text model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling Speech2TextModel

• 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.

• decoder_layers (int, optional, defaults to 12) – Number of decoder layers.

• 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” (often named feed-forward) layer in decoder.

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

• activation_function (str or function, 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.

• dropout (float, optional, defaults to 0.1) – The dropout probability 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.

• classifier_dropout (float, optional, defaults to 0.0) – The dropout ratio for classifier.

• init_std (float, optional, defaults to 0.02) – The standard deviation of the truncated_normal_initializer for initializing all weight matrices.

• encoder_layerdrop – (float, optional, defaults to 0.0): The LayerDrop probability for the encoder. See the LayerDrop paper for more details.

• decoder_layerdrop – (float, optional, defaults to 0.0): The LayerDrop probability for the decoder. See the LayerDrop paper for more details.

• use_cache (bool, optional, defaults to True) – Whether or not the model should return the last key/values attentions (not used by all models).

• max_source_positions (int, optional, defaults to 6000) – The maximum sequence length of log-mel filter-bank features that this model might ever be used with.

• max_target_positions – (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).

• num_conv_layers (int, optional, defaults to 2) – Number of 1D convolutional layers in the conv module.

• conv_kernel_sizes (Tuple[int], optional, defaults to (5, 5)) – A tuple of integers defining the kernel size of each 1D convolutional layer in the conv module. The length of conv_kernel_sizes has to match num_conv_layers.

• conv_channels (int, optional, defaults to 1024) – An integer defining the number of output channels of each convolution layers except the final one in the conv module.

• input_feat_per_channel (int, optional, defaults to 80) – An integer specifying the size of feature vector. This is also the dimensions of log-mel filter-bank features.

• input_channels (int, optional, defaults to 1) – An integer specifying number of input channels of the input feature vector.

• Example::

• from transformers import Speech2TextModel (>>>) –

• Speech2TextConfig

• # Initializing a Speech2Text s2t_transformer_s style configuration (>>>) –

• configuration = Speech2TextConfig() (>>>) –

• # Initializing a model from the s2t_transformer_s style configuration (>>>) –

• model = Speech2TextModel (>>>) –

• # Accessing the model configuration (>>>) –

• configuration = model.config (>>>) –

## Speech2TextTokenizer¶

class transformers.Speech2TextTokenizer(vocab_file, spm_file, bos_token='<s>', eos_token='</s>', pad_token='<pad>', unk_token='<unk>', do_upper_case=False, do_lower_case=False, tgt_lang=None, lang_codes=None, sp_model_kwargs: Optional[Dict[str, Any]] = None, **kwargs)[source]

Construct an Speech2Text tokenizer.

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

Parameters
• vocab_file (str) – File containing the vocabulary.

• spm_file (str) – Path to the SentencePiece model file

• bos_token (str, optional, defaults to "<s>") – The beginning of sentence token.

• eos_token (str, optional, defaults to "</s>") – The end of sentence token.

• 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.

• do_upper_case (bool, optional, defaults to False) – Whether or not to uppercase the output when decoding.

• do_lower_case (bool, optional, defaults to False) – Whether or not to lowercase the input when tokenizing.

• tgt_lang (str, optional) – A string representing the target language.

• sp_model_kwargs (dict, optional) –

Will be passed to the SentencePieceProcessor.__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.

• **kwargs – Additional keyword arguments passed along to PreTrainedTokenizer

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

Build model inputs from a sequence by appending eos_token_id.

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

Create the token type IDs corresponding to the sequences passed. What are token type IDs?

Should be overridden in a subclass if the model has a special way of building those.

Parameters
• token_ids_0 (List[int]) – The first tokenized sequence.

• token_ids_1 (List[int], optional) – The second tokenized sequence.

Returns

The token type ids.

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]

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.

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) – Whether or not 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]

save_vocabulary(save_directory: str, filename_prefix: Optional[str] = None) → Tuple[str][source]

Save only the vocabulary of the tokenizer (vocabulary + added tokens).

This method won’t save the configuration and special token mappings of the tokenizer. Use _save_pretrained() to save the whole state of the tokenizer.

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

• filename_prefix (str, optional) – An optional prefix to add to the named of the saved files.

Returns

Paths to the files saved.

Return type

Tuple(str)

## Speech2TextFeatureExtractor¶

class transformers.Speech2TextFeatureExtractor(feature_size=80, sampling_rate=16000, num_mel_bins=80, padding_value=0.0, do_ceptral_normalize=True, normalize_means=True, normalize_vars=True, **kwargs)[source]

Constructs a Speech2Text feature extractor.

This feature extractor inherits from Speech2TextFeatureExtractor which contains most of the main methods. Users should refer to this superclass for more information regarding those methods.

This class extracts mel-filter bank features from raw speech using TorchAudio and applies utterance-level cepstral mean and variance normalization to the extracted features.

Parameters
• feature_size (int, defaults to 80) – The feature dimension of the extracted features.

• sampling_rate (int, defaults to 16000) – The sampling rate at which the audio files should be digitalized expressed in Hertz per second (Hz).

• num_mel_bins (int, defaults to 80) – Number of Mel-frequency bins.

• padding_value (float, defaults to 0.0) – The value that is used to fill the padding vectors.

• do_ceptral_normalize (bool, optional, defaults to True) – Whether or not to apply utterance-level cepstral mean and variance normalization to extracted features.

• normalize_means (bool, optional, defaults to True) – Whether or not to zero-mean normalize the extracted features.

• normalize_vars (bool, optional, defaults to True) – Whether or not to unit-variance normalize the extracted features.

__call__(raw_speech: Union[numpy.ndarray, List[float], List[numpy.ndarray], List[List[float]]], padding: Union[bool, str, transformers.file_utils.PaddingStrategy] = False, max_length: Optional[int] = None, pad_to_multiple_of: Optional[int] = None, return_tensors: Optional[Union[str, transformers.file_utils.TensorType]] = None, sampling_rate: Optional[int] = None, return_attention_mask: Optional[bool] = None, **kwargs) → transformers.feature_extraction_utils.BatchFeature[source]

Main method to featurize and prepare for the model one or several sequence(s). sequences.

Parameters
• raw_speech (np.ndarray, List[float], List[np.ndarray], List[List[float]]) – The sequence or batch of sequences to be padded. Each sequence can be a numpy array, a list of float values, a list of numpy arrays or a list of list of float values.

• padding (bool, str or PaddingStrategy, optional, defaults to True) –

Select a strategy to pad the returned sequences (according to the model’s padding side and padding index) among:

• 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).

• max_length (int, optional) – Maximum length of the returned list and optionally padding length (see above).

• pad_to_multiple_of (int, optional) –

If set will pad the sequence to a multiple of the provided value.

This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >= 7.5 (Volta), or on TPUs which benefit from having sequence lengths be a multiple of 128.

• return_attention_mask (bool, optional) –

Whether to return the attention mask. If left to the default, will return the attention mask according to the specific feature_extractor’s default.

Note

For Speech2TextTransoformer models, attention_mask should alwys be passed for batched inference, to avoid subtle bugs.

• return_tensors (str or TensorType, optional) –

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.

• sampling_rate (int, optional) – The sampling rate at which the raw_speech input was sampled. It is strongly recommended to pass sampling_rate at the forward call to prevent silent errors.

• padding_value (float, defaults to 0.0) – The value that is used to fill the padding values / vectors.

## Speech2TextProcessor¶

class transformers.Speech2TextProcessor(feature_extractor, tokenizer)[source]

Constructs a Speech2Text processor which wraps a Speech2Text feature extractor and a Speech2Text tokenizer into a single processor.

Speech2TextProcessor offers all the functionalities of Speech2TextFeatureExtractor and Speech2TextTokenizer. See the __call__() and decode() for more information.

Parameters
__call__(*args, **kwargs)[source]

When used in normal mode, this method forwards all its arguments to Speech2TextFeatureExtractor’s __call__() and returns its output. If used in the context as_target_processor() this method forwards all its arguments to Speech2TextTokenizer’s __call__(). Please refer to the doctsring of the above two methods for more information.

as_target_processor()[source]

Temporarily sets the tokenizer for processing the input. Useful for encoding the labels when fine-tuning Speech2Text.

batch_decode(*args, **kwargs)[source]

This method forwards all its arguments to Speech2TextTokenizer’s batch_decode(). Please refer to the docstring of this method for more information.

decode(*args, **kwargs)[source]

This method forwards all its arguments to Speech2TextTokenizer’s decode(). Please refer to the docstring of this method for more information.

classmethod from_pretrained(pretrained_model_name_or_path, **kwargs)[source]

Instantiate a Speech2TextProcessor from a pretrained Speech2Text processor.

Note

This class method is simply calling Speech2TextFeatureExtractor’s from_pretrained() and Speech2TextTokenizer’s from_pretrained(). Please refer to the docstrings of the methods above for more information.

Parameters
• pretrained_model_name_or_path (str or os.PathLike) –

This can be either:

• a string, the model id of a pretrained feature_extractor 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 a feature extractor file saved using the save_pretrained() method, e.g., ./my_model_directory/.

• a path or url to a saved feature extractor JSON file, e.g., ./my_model_directory/preprocessor_config.json.

• **kwargs – Additional keyword arguments passed along to both PreTrainedFeatureExtractor and PreTrainedTokenizer

save_pretrained(save_directory)[source]

Save a Speech2Text feature extractor object and Speech2Text tokenizer object to the directory save_directory, so that it can be re-loaded using the from_pretrained() class method.

Note

This class method is simply calling save_pretrained() and save_pretrained(). Please refer to the docstrings of the methods above for more information.

Parameters

save_directory (str or os.PathLike) – Directory where the feature extractor JSON file and the tokenizer files will be saved (directory will be created if it does not exist).

## Speech2TextModel¶

class transformers.Speech2TextModel(config: transformers.models.speech_to_text.configuration_speech_to_text.Speech2TextConfig)[source]

The bare Speech2Text Model outputting raw hidden-states without any specific head on top. This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

Parameters

config (Speech2TextConfig) – 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_features=None, attention_mask=None, decoder_input_ids=None, decoder_attention_mask=None, head_mask=None, decoder_head_mask=None, cross_attn_head_mask=None, encoder_outputs=None, past_key_values=None, decoder_inputs_embeds=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None)[source]

The Speech2TextModel 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_features (torch.LongTensor of shape (batch_size, sequence_length, feature_size)) – Float values of fbank features extracted from the raw speech waveform. Raw speech waveform can be obtained by loading a .flac or .wav audio file into an array of type List[float] or a numpy.ndarray, e.g. via the soundfile library (pip install soundfile). To prepare the array into input_features, the Speech2TextTokenizer should be used for extracting the fbank features, padding and conversion into a tensor of type torch.FloatTensor. See __call__()

• attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional) –

Mask to avoid performing convolution and attention on padding token indices. Mask values selected in [0, 1]:

• 1 for tokens that are not masked,

• 0 for tokens that are masked.

• decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional) –

Indices of decoder input sequence tokens in the vocabulary.

Indices can be obtained using SpeechToTextTokenizer. See transformers.PreTrainedTokenizer.encode() and transformers.PreTrainedTokenizer.__call__() for details.

What are decoder input IDs?

SpeechToText uses the eos_token_id as the starting token for decoder_input_ids generation. If past_key_values is used, optionally only the last decoder_input_ids have to be input (see past_key_values).

• decoder_attention_mask (torch.LongTensor of shape (batch_size, target_sequence_length), optional) –

Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default. <<<<<<< HEAD

If you want to change padding behavior, you should read modeling_speech_to_text._prepare_decoder_inputs() and modify to your needs. See diagram 1 in the paper for more information on the default strategy.

• head_mask (torch.Tensor of shape (encoder_layers, encoder_attention_heads), optional) –

Mask to nullify selected heads of the attention modules in the encoder. Mask values selected in [0, 1]:

• decoder_head_mask (torch.Tensor of shape (decoder_layers, decoder_attention_heads), optional) –

Mask to nullify selected heads of the attention modules in the decoder. Mask values selected in [0, 1]:

• cross_attn_head_mask (torch.Tensor of shape (decoder_layers, decoder_attention_heads), optional) –

Mask to nullify selected heads of the cross-attention modules. Mask values selected in [0, 1]:

• 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.

• past_key_values (tuple(tuple(torch.FloatTensor)), optional, returned when use_cache=True is passed or when config.use_cache=True) –

Tuple of tuple(torch.FloatTensor) of length config.n_layers, with each tuple having 2 tensors of shape (batch_size, num_heads, sequence_length, embed_size_per_head)) and 2 additional tensors of shape (batch_size, num_heads, encoder_sequence_length, embed_size_per_head).

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

• decoder_inputs_embeds (torch.FloatTensor of shape (batch_size, target_sequence_length, hidden_size), optional) –

Optionally, instead of passing decoder_input_ids you can choose to directly pass an embedded representation. If past_key_values is used, optionally only the last decoder_inputs_embeds have to be input (see past_key_values). This is useful if you want more control over how to convert decoder_input_ids indices into associated vectors than the model’s internal embedding lookup matrix.

If decoder_input_ids and decoder_inputs_embeds are both unset, decoder_inputs_embeds takes the value of inputs_embeds.

• use_cache (bool, optional) – If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).

• output_attentions (bool, optional) – Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.

• output_hidden_states (bool, optional) – Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.

• return_dict (bool, optional) – Whether or not to return a ModelOutput instead of a plain tuple.

Returns

A Seq2SeqModelOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (Speech2TextConfig) 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 decoder 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 (tuple(tuple(torch.FloatTensor)), optional, returned when use_cache=True is passed or when config.use_cache=True) – Tuple of tuple(torch.FloatTensor) of length config.n_layers, with each tuple having 2 tensors of shape (batch_size, num_heads, sequence_length, embed_size_per_head)) and 2 additional tensors of shape (batch_size, num_heads, encoder_sequence_length, embed_size_per_head).

Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention blocks) 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.

• cross_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’s cross-attention layer, after the attention softmax, used to compute the weighted average in the cross-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

Seq2SeqModelOutput or tuple(torch.FloatTensor)

Example:

>>> from transformers import Speech2TextTokenizer, Speech2TextModel
>>> import torch

>>> tokenizer = Speech2TextTokenizer.from_pretrained('s2t_transformer_s')
>>> model = Speech2TextModel.from_pretrained('s2t_transformer_s')

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

>>> last_hidden_states = outputs.last_hidden_state


## Speech2TextForConditionalGeneration¶

class transformers.Speech2TextForConditionalGeneration(config: transformers.models.speech_to_text.configuration_speech_to_text.Speech2TextConfig)[source]

The Speech2Text Model with a language modeling head. Can be used for summarization. This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

Parameters

config (Speech2TextConfig) – 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_features=None, attention_mask=None, decoder_input_ids=None, decoder_attention_mask=None, head_mask=None, decoder_head_mask=None, cross_attn_head_mask=None, encoder_outputs=None, past_key_values=None, decoder_inputs_embeds=None, labels=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None)[source]

The Speech2TextForConditionalGeneration 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_features (torch.LongTensor of shape (batch_size, sequence_length, feature_size)) – Float values of fbank features extracted from the raw speech waveform. Raw speech waveform can be obtained by loading a .flac or .wav audio file into an array of type List[float] or a numpy.ndarray, e.g. via the soundfile library (pip install soundfile). To prepare the array into input_features, the Speech2TextTokenizer should be used for extracting the fbank features, padding and conversion into a tensor of type torch.FloatTensor. See __call__()

• attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional) –

Mask to avoid performing convolution and attention on padding token indices. Mask values selected in [0, 1]:

• 1 for tokens that are not masked,

• 0 for tokens that are masked.

• decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional) –

Indices of decoder input sequence tokens in the vocabulary.

Indices can be obtained using SpeechToTextTokenizer. See transformers.PreTrainedTokenizer.encode() and transformers.PreTrainedTokenizer.__call__() for details.

What are decoder input IDs?

SpeechToText uses the eos_token_id as the starting token for decoder_input_ids generation. If past_key_values is used, optionally only the last decoder_input_ids have to be input (see past_key_values).

• decoder_attention_mask (torch.LongTensor of shape (batch_size, target_sequence_length), optional) –

Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default. <<<<<<< HEAD

If you want to change padding behavior, you should read modeling_speech_to_text._prepare_decoder_inputs() and modify to your needs. See diagram 1 in the paper for more information on the default strategy.

• head_mask (torch.Tensor of shape (encoder_layers, encoder_attention_heads), optional) –

Mask to nullify selected heads of the attention modules in the encoder. Mask values selected in [0, 1]:

• decoder_head_mask (torch.Tensor of shape (decoder_layers, decoder_attention_heads), optional) –

Mask to nullify selected heads of the attention modules in the decoder. Mask values selected in [0, 1]:

• cross_attn_head_mask (torch.Tensor of shape (decoder_layers, decoder_attention_heads), optional) –

Mask to nullify selected heads of the cross-attention modules. Mask values selected in [0, 1]:

• 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.

• past_key_values (tuple(tuple(torch.FloatTensor)), optional, returned when use_cache=True is passed or when config.use_cache=True) –

Tuple of tuple(torch.FloatTensor) of length config.n_layers, with each tuple having 2 tensors of shape (batch_size, num_heads, sequence_length, embed_size_per_head)) and 2 additional tensors of shape (batch_size, num_heads, encoder_sequence_length, embed_size_per_head).

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

• decoder_inputs_embeds (torch.FloatTensor of shape (batch_size, target_sequence_length, hidden_size), optional) –

Optionally, instead of passing decoder_input_ids you can choose to directly pass an embedded representation. If past_key_values is used, optionally only the last decoder_inputs_embeds have to be input (see past_key_values). This is useful if you want more control over how to convert decoder_input_ids indices into associated vectors than the model’s internal embedding lookup matrix.

If decoder_input_ids and decoder_inputs_embeds are both unset, decoder_inputs_embeds takes the value of inputs_embeds.

• use_cache (bool, optional) – If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).

• output_attentions (bool, optional) – Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.

• output_hidden_states (bool, optional) – Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.

• return_dict (bool, optional) – Whether or not to return a ModelOutput instead of a plain tuple.

• labels (torch.LongTensor of shape (batch_size, sequence_length), optional) – Labels for computing the 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 or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (Speech2TextConfig) and inputs.

• loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) – Language 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 (tuple(tuple(torch.FloatTensor)), optional, returned when use_cache=True is passed or when config.use_cache=True) – Tuple of tuple(torch.FloatTensor) of length config.n_layers, with each tuple having 2 tensors of shape (batch_size, num_heads, sequence_length, embed_size_per_head)) and 2 additional tensors of shape (batch_size, num_heads, encoder_sequence_length, embed_size_per_head).

Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention blocks) 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.

• cross_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’s cross-attention layer, after the attention softmax, used to compute the weighted average in the cross-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.

Example:

>>> import torch
>>> from transformers import Speech2TextProcessor, Speech2TextForConditionalGeneration
>>> import soundfile as sf

>>> def map_to_array(batch):
>>>     batch["speech"] = speech
>>>     return batch

>>> ds = load_dataset("patrickvonplaten/librispeech_asr_dummy", "clean", split="validation")
>>> ds = ds.map(map_to_array)

>>> input_features = processor(ds["speech"][0], sampling_rate=16_000, return_tensors="pt").input_features  # Batch size 1
>>> generated_ids = model.generate(input_ids=input_features)

>>> transcription = processor.batch_decode(generated_ids)


Return type

Seq2SeqLMOutput or tuple(torch.FloatTensor)