> For the complete documentation index, see [llms.txt](https://boinc-ai.gitbook.io/transformers/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://boinc-ai.gitbook.io/transformers/api/models/audio-models/whisper.md).

# Whisper

## Whisper

### Overview

The Whisper model was proposed in [Robust Speech Recognition via Large-Scale Weak Supervision](https://cdn.openai.com/papers/whisper.pdf) by Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever.

The abstract from the paper is the following:

*We study the capabilities of speech processing systems trained simply to predict large amounts of transcripts of audio on the internet. When scaled to 680,000 hours of multilingual and multitask supervision, the resulting models generalize well to standard benchmarks and are often competitive with prior fully supervised results but in a zeroshot transfer setting without the need for any finetuning. When compared to humans, the models approach their accuracy and robustness. We are releasing models and inference code to serve as a foundation for further work on robust speech processing.*

Tips:

* The model usually performs well without requiring any finetuning.
* The architecture follows a classic encoder-decoder architecture, which means that it relies on the [generate()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/text_generation#transformers.GenerationMixin.generate) function for inference.
* Inference is currently only implemented for short-form i.e. audio is pre-segmented into <=30s segments. Long-form (including timestamps) will be implemented in a future release.
* One can use [WhisperProcessor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperProcessor) to prepare audio for the model, and decode the predicted ID’s back into text.

This model was contributed by [Arthur Zucker](https://huggingface.co/ArthurZ). The Tensorflow version of this model was contributed by [amyeroberts](https://huggingface.co/amyeroberts). The original code can be found [here](https://github.com/openai/whisper).

### WhisperConfig

#### class transformers.WhisperConfig

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/configuration_whisper.py#L62)

( vocab\_size = 51865num\_mel\_bins = 80encoder\_layers = 6encoder\_attention\_heads = 4decoder\_layers = 6decoder\_attention\_heads = 4decoder\_ffn\_dim = 1536encoder\_ffn\_dim = 1536encoder\_layerdrop = 0.0decoder\_layerdrop = 0.0decoder\_start\_token\_id = 50257use\_cache = Trueis\_encoder\_decoder = Trueactivation\_function = 'gelu'd\_model = 256dropout = 0.0attention\_dropout = 0.0activation\_dropout = 0.0init\_std = 0.02scale\_embedding = Falsemax\_source\_positions = 1500max\_target\_positions = 448pad\_token\_id = 50256bos\_token\_id = 50256eos\_token\_id = 50256suppress\_tokens = Nonebegin\_suppress\_tokens = \[220, 50256]use\_weighted\_layer\_sum = Falseclassifier\_proj\_size = 256apply\_spec\_augment = Falsemask\_time\_prob = 0.05mask\_time\_length = 10mask\_time\_min\_masks = 2mask\_feature\_prob = 0.0mask\_feature\_length = 10mask\_feature\_min\_masks = 0median\_filter\_width = 7\*\*kwargs )

Parameters

* **vocab\_size** (`int`, *optional*, defaults to 51865) — Vocabulary size of the Whisper model. Defines the number of different tokens that can be represented by the `decoder_input_ids` passed when calling [WhisperModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperModel)
* **num\_mel\_bins** (`int`, *optional*, defaults to 80) — Number of mel features used per input features. Should correspond to the value used in the `WhisperProcessor` class.
* **encoder\_layers** (`int`, *optional*, defaults to 6) — Number of encoder layers.
* **decoder\_layers** (`int`, *optional*, defaults to 6) — Number of decoder layers.
* **encoder\_attention\_heads** (`int`, *optional*, defaults to 4) — Number of attention heads for each attention layer in the Transformer encoder.
* **decoder\_attention\_heads** (`int`, *optional*, defaults to 4) — Number of attention heads for each attention layer in the Transformer decoder.
* **encoder\_ffn\_dim** (`int`, *optional*, defaults to 1536) — Dimensionality of the “intermediate” (often named feed-forward) layer in encoder.
* **decoder\_ffn\_dim** (`int`, *optional*, defaults to 1536) — Dimensionality of the “intermediate” (often named feed-forward) layer in decoder.
* **encoder\_layerdrop** (`float`, *optional*, defaults to 0.0) — The LayerDrop probability for the encoder. See the \[LayerDrop paper]\(see <https://arxiv.org/abs/1909.11556>) for more details.
* **decoder\_layerdrop** (`float`, *optional*, defaults to 0.0) — The LayerDrop probability for the decoder. See the \[LayerDrop paper]\(see <https://arxiv.org/abs/1909.11556>) for more details.
* **decoder\_start\_token\_id** (`int`, *optional*, defaults to 50257) — Corresponds to the ”<|startoftranscript|>” token, which is automatically used when no `decoder_input_ids` are provided to the `generate` function. It is used to guide the model\`s generation process depending on the task.
* **use\_cache** (`bool`, *optional*, defaults to `True`) — Whether or not the model should return the last key/values attentions (not used by all models).
* **is\_encoder\_decoder** (`bool`, *optional*, defaults to `True`) — Whether the model is used as an encoder/decoder or not.
* **activation\_function** (`str`, *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.
* **d\_model** (`int`, *optional*, defaults to 256) — Dimensionality of the layers.
* **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.
* **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 False) — Scale embeddings by diving by sqrt(d\_model).
* **max\_source\_positions** (`int`, *optional*, defaults to 1500) — 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 448) — 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).
* **pad\_token\_id** (`int`, *optional*, defaults to 50256) — Padding token id.
* **bos\_token\_id** (`int`, *optional*, defaults to 50256) — Begin of stream token id.
* **eos\_token\_id** (`int`, *optional*, defaults to 50256) — End of stream token id.
* **suppress\_tokens** (`List[int]`, *optional*) — A list containing the non-speech tokens that will be used by the logit processor in the `generate` function. NON\_SPEECH\_TOKENS and NON\_SPEECH\_TOKENS\_MULTI each correspond to the `english-only` and the `multilingual` model.
* **begin\_suppress\_tokens** (`List[int]`, *optional*, defaults to `[220,50256]`) — A list containing tokens that will be supressed at the beginning of the sampling process. Initialized as the token for `" "` (`blank_token_id`) and the `eos_token_id`
* **use\_weighted\_layer\_sum** (`bool`, *optional*, defaults to `False`) — Whether to use a weighted average of layer outputs with learned weights. Only relevant when using an instance of [WhisperForAudioClassification](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperForAudioClassification).
* **classifier\_proj\_size** (`int`, *optional*, defaults to 256) — Dimensionality of the projection before token mean-pooling for classification. Only relevant when using an instance of [WhisperForAudioClassification](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperForAudioClassification).
* **apply\_spec\_augment** (`bool`, *optional*, defaults to `False`) — Whether to apply *SpecAugment* data augmentation to the outputs of the feature encoder. For reference see [SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition](https://arxiv.org/abs/1904.08779).
* **mask\_time\_prob** (`float`, *optional*, defaults to 0.05) — Percentage (between 0 and 1) of all feature vectors along the time axis which will be masked. The masking procecure generates `mask_time_prob*len(time_axis)/mask_time_length` independent masks over the axis. If reasoning from the propability of each feature vector to be chosen as the start of the vector span to be masked, *mask\_time\_prob* should be `prob_vector_start*mask_time_length`. Note that overlap may decrease the actual percentage of masked vectors. This is only relevant if `apply_spec_augment == True`.
* **mask\_time\_length** (`int`, *optional*, defaults to 10) — Length of vector span along the time axis.
* **mask\_time\_min\_masks** (`int`, *optional*, defaults to 2), — The minimum number of masks of length `mask_feature_length` generated along the time axis, each time step, irrespectively of `mask_feature_prob`. Only relevant if ”mask\_time\_prob\*len(time\_axis)/mask\_time\_length < mask\_time\_min\_masks”
* **mask\_feature\_prob** (`float`, *optional*, defaults to 0.0) — Percentage (between 0 and 1) of all feature vectors along the feature axis which will be masked. The masking procecure generates `mask_feature_prob*len(feature_axis)/mask_time_length` independent masks over the axis. If reasoning from the propability of each feature vector to be chosen as the start of the vector span to be masked, *mask\_feature\_prob* should be `prob_vector_start*mask_feature_length`. Note that overlap may decrease the actual percentage of masked vectors. This is only relevant if `apply_spec_augment is True`.
* **mask\_feature\_length** (`int`, *optional*, defaults to 10) — Length of vector span along the feature axis.
* **mask\_feature\_min\_masks** (`int`, *optional*, defaults to 0), — The minimum number of masks of length `mask_feature_length` generated along the feature axis, each time step, irrespectively of `mask_feature_prob`. Only relevant if `mask_feature_prob*len(feature_axis)/mask_feature_length < mask_feature_min_masks`.
* **median\_filter\_width** (`int`, *optional*, defaults to 7) — Width of the median filter used to smoothen to cross-attention outputs when computing token timestamps. Should be an odd number.

This is the configuration class to store the configuration of a [WhisperModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperModel). It is used to instantiate a Whisper 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 Whisper [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) architecture.

Configuration objects inherit from [PretrainedConfig](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/configuration#transformers.PretrainedConfig) and can be used to control the model outputs. Read the documentation from [PretrainedConfig](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/configuration#transformers.PretrainedConfig) for more information.

Example:

Copied

```
>>> from transformers import WhisperConfig, WhisperModel

>>> # Initializing a Whisper tiny style configuration
>>> configuration = WhisperConfig()

>>> # Initializing a model (with random weights) from the tiny style configuration
>>> model = WhisperModel(configuration)

>>> # Accessing the model configuration
>>> configuration = model.config
```

### WhisperTokenizer

#### class transformers.WhisperTokenizer

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/tokenization_whisper.py#L215)

( vocab\_filemerges\_filenormalizer\_file = Noneerrors = 'replace'unk\_token = '<|endoftext|>'bos\_token = '<|endoftext|>'eos\_token = '<|endoftext|>'pad\_token = Noneadd\_prefix\_space = Falselanguage = Nonetask = Nonepredict\_timestamps = False\*\*kwargs )

Parameters

* **vocab\_file** (`str`) — Path to the vocabulary file.
* **merges\_file** (`str`) — Path to the merges file.
* **normalizer\_file** (`str`, *optional*, defaults to `None`) — Path to the normalizer\_file file.
* **errors** (`str`, *optional*, defaults to `"replace"`) — Paradigm to follow when decoding bytes to UTF-8. See [bytes.decode](https://docs.python.org/3/library/stdtypes.html#bytes.decode) for more information.
* **unk\_token** (`str`, *optional*, defaults to `"<|endoftext|>"`) — 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** (`str`, *optional*, defaults to `"<|endoftext|>"`) — The beginning of sequence token. The `decoder_start_token_id` is used to set the first token as `"<|startoftranscript|>"` when generating.
* **eos\_token** (`str`, *optional*, defaults to `"<|endoftext|>"`) — The end of sequence token.
* **add\_prefix\_space** (`bool`, *optional*, defaults to `False`) — Whether or not to add an initial space to the input. This allows to treat the leading word just as any other word.
* **language** (`str`, *optional*) — The language of the transcription text. The corresponding language id token is appended to the start of the sequence for multilingual speech recognition and speech translation tasks, e.g. for Spanish the token `"<|es|>"` is appended to the start of sequence. This should be used for multilingual fine-tuning only.
* **task** (`str`, *optional*) — Task identifier to append at the start of sequence (if any). This should be used for mulitlingual fine-tuning, with `"transcribe"` for speech recognition and `"translate"` for speech translation.
* **predict\_timestamps** (`bool`, *optional*, defaults to `False`) — Whether to omit the `<|notimestamps|>` token at the start of the sequence.

Construct a Whisper tokenizer.

This tokenizer inherits from [PreTrainedTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizer) which contains some of the main methods. Users should refer to the superclass for more information regarding such methods.

**set\_prefix\_tokens**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/tokenization_whisper.py#L385)

( language: str = Nonetask: str = Nonepredict\_timestamps: bool = None )

Parameters

* **language** (`str`, *optional*, defaults to `None`) — The language of the transcription text.
* **task** (`str`, *optional*, defaults to `None`) — Task identifier to append at the start of sequence (if any).
* **predict\_timestamps** (`bool`, *optional*, defaults to `None`) — Whether to omit the `<|notimestamps|>` token at the start of the sequence.

Override the prefix tokens appended to the start of the label sequence. This method can be used standalone to

update the prefix tokens as required when fine-tuning. Example:

Copied

```
>>> # instantiate the tokenizer and set the prefix token to Spanish
>>> tokenizer = WhisperTokenizer.from_pretrained("openai/whisper-tiny", language="spanish")
>>> # now switch the prefix token from Spanish to French
>>> tokenizer.set_prefix_tokens(language="french")
```

**build\_inputs\_with\_special\_tokens**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/tokenization_whisper.py#L444)

( token\_ids\_0token\_ids\_1 = None )

Build model inputs from a sequence by appending eos\_token\_id.

**get\_special\_tokens\_mask**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/tokenization_whisper.py#L452)

( token\_ids\_0: typing.List\[int]token\_ids\_1: typing.Optional\[typing.List\[int]] = Nonealready\_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 to `False`) — 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.

**create\_token\_type\_ids\_from\_sequences**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/tokenization_utils_base.py#L3289)

( token\_ids\_0: typing.List\[int]token\_ids\_1: typing.Optional\[typing.List\[int]] = None ) → `List[int]`

Parameters

* **token\_ids\_0** (`List[int]`) — The first tokenized sequence.
* **token\_ids\_1** (`List[int]`, *optional*) — The second tokenized sequence.

Returns

`List[int]`

The token type ids.

Create the token type IDs corresponding to the sequences passed. [What are token type IDs?](https://huggingface.co/docs/transformers/glossary#token-type-ids)

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

**save\_vocabulary**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/tokenization_whisper.py#L718)

( save\_directory: strfilename\_prefix: typing.Optional\[str] = None )

### WhisperTokenizerFast

#### class transformers.WhisperTokenizerFast

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/tokenization_whisper_fast.py#L90)

( vocab\_file = Nonemerges\_file = Nonenormalizer\_file = Nonetokenizer\_file = Noneunk\_token = '<|endoftext|>'bos\_token = '<|endoftext|>'eos\_token = '<|endoftext|>'add\_prefix\_space = Falselanguage = Nonetask = Nonepredict\_timestamps = False\*\*kwargs )

Parameters

* **vocab\_file** (`str`) — Path to the vocabulary file.
* **merges\_file** (`str`) — Path to the merges file.
* **normalizer\_file** (`str`, *optional*, defaults to `None`) — Path to the normalizer\_file file.
* **errors** (`str`, *optional*, defaults to `"replace"`) — Paradigm to follow when decoding bytes to UTF-8. See [bytes.decode](https://docs.python.org/3/library/stdtypes.html#bytes.decode) for more information.
* **unk\_token** (`str`, *optional*, defaults to `<|endoftext|>`) — 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** (`str`, *optional*, defaults to `"<|endoftext|>"`) — The beginning of sequence token. The `decoder_start_token_id` is used to set the first token as `"<|startoftranscript|>"` when generating.
* **eos\_token** (`str`, *optional*, defaults to `<|endoftext|>`) — The end of sequence token.
* **add\_prefix\_space** (`bool`, *optional*, defaults to `False`) — Whether or not to add an initial space to the input. This allows to treat the leading word just as any other word. (Whisper tokenizer detect beginning of words by the preceding space).
* **trim\_offsets** (`bool`, *optional*, defaults to `True`) — Whether or not the post-processing step should trim offsets to avoid including whitespaces.
* **language** (`str`, *optional*) — The language of the transcription text. The corresponding language id token is appended to the start of the sequence for multilingual speech recognition and speech translation tasks, e.g. for Spanish the token `"<|es|>"` is appended to the start of sequence. This should be used for multilingual fine-tuning only.
* **task** (`str`, *optional*) — Task identifier to append at the start of sequence (if any). This should be used for mulitlingual fine-tuning, with `"transcribe"` for speech recognition and `"translate"` for speech translation.
* **predict\_timestamps** (`bool`, *optional*, defaults to `False`) — Whether to omit the `<|notimestamps|>` token at the start of the sequence.

Construct a “fast” Whisper tokenizer (backed by BOINC AI’s *tokenizers* library).

This tokenizer inherits from [PreTrainedTokenizerFast](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast) which contains most of the main methods. Users should refer to this superclass for more information regarding those methods.

**set\_prefix\_tokens**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/tokenization_whisper_fast.py#L421)

( language: str = Nonetask: str = Nonepredict\_timestamps: bool = None )

Parameters

* **language** (`str`, *optional*, defaults to `None`) — The language of the transcription text.
* **task** (`str`, *optional*, defaults to `None`) — Task identifier to append at the start of sequence (if any).
* **predict\_timestamps** (`bool`, *optional*, defaults to `None`) — Whether to omit the `<|notimestamps|>` token at the start of the sequence.

Override the prefix tokens appended to the start of the label sequence. This method can be used standalone to

update the prefix tokens as required when fine-tuning. Example:

Copied

```
>>> # instantiate the tokenizer and set the prefix token to Spanish
>>> tokenizer = WhisperTokenizerFast.from_pretrained("openai/whisper-tiny", language="spanish")
>>> # now switch the prefix token from Spanish to French
>>> tokenizer.set_prefix_tokens(language="french")
```

**build\_inputs\_with\_special\_tokens**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/tokenization_whisper_fast.py#L495)

( token\_ids\_0token\_ids\_1 = None )

Build model inputs from a sequence by appending eos\_token\_id.

**get\_special\_tokens\_mask**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/tokenization_whisper_fast.py#L503)

( token\_ids\_0: typing.List\[int]token\_ids\_1: typing.Optional\[typing.List\[int]] = Nonealready\_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 to `False`) — 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.

**create\_token\_type\_ids\_from\_sequences**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/tokenization_utils_base.py#L3289)

( token\_ids\_0: typing.List\[int]token\_ids\_1: typing.Optional\[typing.List\[int]] = None ) → `List[int]`

Parameters

* **token\_ids\_0** (`List[int]`) — The first tokenized sequence.
* **token\_ids\_1** (`List[int]`, *optional*) — The second tokenized sequence.

Returns

`List[int]`

The token type ids.

Create the token type IDs corresponding to the sequences passed. [What are token type IDs?](https://huggingface.co/docs/transformers/glossary#token-type-ids)

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

**save\_vocabulary**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/tokenization_whisper_fast.py#L406)

( save\_directory: strfilename\_prefix: typing.Optional\[str] = None )

### WhisperFeatureExtractor

#### class transformers.WhisperFeatureExtractor

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/feature_extraction_whisper.py#L32)

( feature\_size = 80sampling\_rate = 16000hop\_length = 160chunk\_length = 30n\_fft = 400padding\_value = 0.0return\_attention\_mask = False\*\*kwargs )

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 (Hz).
* **hop\_length** (`int`, defaults to 160) — Length of the overlaping windows for the STFT used to obtain the Mel Frequency coefficients.
* **chunk\_length** (`int`, defaults to 30) — The maximum number of chuncks of `sampling_rate` samples used to trim and pad longer or shorter audio sequences.
* **n\_fft** (`int`, defaults to 400) — Size of the Fourier transform.
* **padding\_value** (`float`, *optional*, defaults to 0.0) — Padding value used to pad the audio. Should correspond to silences.

Constructs a Whisper feature extractor.

This feature extractor inherits from [SequenceFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/feature_extractor#transformers.SequenceFeatureExtractor) 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 a custom numpy implementation of the `Short Time Fourier Transform` which should match pytorch’s `torch.stft` equivalent.

**\_\_call\_\_**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/feature_extraction_whisper.py#L136)

( raw\_speech: typing.Union\[numpy.ndarray, typing.List\[float], typing.List\[numpy.ndarray], typing.List\[typing.List\[float]]]truncation: bool = Truepad\_to\_multiple\_of: typing.Optional\[int] = Nonereturn\_tensors: typing.Union\[str, transformers.utils.generic.TensorType, NoneType] = Nonereturn\_attention\_mask: typing.Optional\[bool] = Nonepadding: typing.Optional\[str] = 'max\_length'max\_length: typing.Optional\[int] = Nonesampling\_rate: typing.Optional\[int] = Nonedo\_normalize: typing.Optional\[bool] = None\*\*kwargs )

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. Must be mono channel audio, not stereo, i.e. single float per timestep.
* **truncation** (`bool`, *optional*, default to `True`) — Activates truncation to cut input sequences longer than *max\_length* to *max\_length*.
* **pad\_to\_multiple\_of** (`int`, *optional*, defaults to None) — 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.

  [What are attention masks?](https://huggingface.co/docs/transformers/glossary#attention-mask)

  For Whisper models, `attention_mask` should always be passed for batched inference, to avoid subtle bugs.
* **return\_tensors** (`str` or [TensorType](https://huggingface.co/docs/transformers/v4.34.1/en/internal/file_utils#transformers.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 and allow automatic speech recognition pipeline.
* **padding\_value** (`float`, defaults to 0.0) — The value that is used to fill the padding values / vectors.
* **do\_normalize** (`bool`, *optional*, defaults to `False`) — Whether or not to zero-mean unit-variance normalize the input. Normalizing can help to significantly improve the performance of the model.

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

### WhisperProcessor

#### class transformers.WhisperProcessor

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/processing_whisper.py#L23)

( feature\_extractortokenizer )

Parameters

* **feature\_extractor** (`WhisperFeatureExtractor`) — An instance of [WhisperFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor). The feature extractor is a required input.
* **tokenizer** (`WhisperTokenizer`) — An instance of [WhisperTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperTokenizer). The tokenizer is a required input.

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

[WhisperProcessor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperProcessor) offers all the functionalities of [WhisperFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor) and [WhisperTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperTokenizer). See the [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperProcessor.__call__) and [decode()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperProcessor.decode) for more information.

**\_\_call\_\_**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/processing_whisper.py#L48)

( \*args\*\*kwargs )

Forwards the `audio` argument to WhisperFeatureExtractor’s [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor.__call__) and the `text` argument to [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__). Please refer to the doctsring of the above two methods for more information.

**from\_pretrained**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/processing_utils.py#L167)

( pretrained\_model\_name\_or\_path: typing.Union\[str, os.PathLike]cache\_dir: typing.Union\[str, os.PathLike, NoneType] = Noneforce\_download: bool = Falselocal\_files\_only: bool = Falsetoken: typing.Union\[bool, str, NoneType] = Nonerevision: str = 'main'\*\*kwargs )

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 boincai.com. 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/feature_extractor#transformers.FeatureExtractionMixin.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 [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/feature_extractor#transformers.FeatureExtractionMixin.from_pretrained) and `~tokenization_utils_base.PreTrainedTokenizer.from_pretrained`.

Instantiate a processor associated with a pretrained model.

This class method is simply calling the feature extractor [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/feature_extractor#transformers.FeatureExtractionMixin.from_pretrained), image processor [ImageProcessingMixin](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/image_processor#transformers.ImageProcessingMixin) and the tokenizer `~tokenization_utils_base.PreTrainedTokenizer.from_pretrained` methods. Please refer to the docstrings of the methods above for more information.

**save\_pretrained**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/processing_utils.py#L93)

( save\_directorypush\_to\_hub: bool = False\*\*kwargs )

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).
* **push\_to\_hub** (`bool`, *optional*, defaults to `False`) — Whether or not to push your model to the BOINC AI 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).
* **kwargs** (`Dict[str, Any]`, *optional*) — Additional key word arguments passed along to the [push\_to\_hub()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/processors#transformers.ProcessorMixin.push_to_hub) method.

Saves the attributes of this processor (feature extractor, tokenizer…) in the specified directory so that it can be reloaded using the [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/nougat#transformers.NougatProcessor.from_pretrained) method.

This class method is simply calling [save\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/feature_extractor#transformers.FeatureExtractionMixin.save_pretrained) and [save\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.save_pretrained). Please refer to the docstrings of the methods above for more information.

**batch\_decode**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/processing_whisper.py#L82)

( \*args\*\*kwargs )

This method forwards all its arguments to WhisperTokenizer’s [batch\_decode()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/speecht5#transformers.SpeechT5Tokenizer.batch_decode). Please refer to the docstring of this method for more information.

**decode**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/processing_whisper.py#L89)

( \*args\*\*kwargs )

This method forwards all its arguments to WhisperTokenizer’s [decode()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/speecht5#transformers.SpeechT5Tokenizer.decode). Please refer to the docstring of this method for more information.

### WhisperModel

#### class transformers.WhisperModel

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_whisper.py#L1227)

( config: WhisperConfig )

Parameters

* **config** ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) — 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.PreTrainedModel.from_pretrained) method to load the model weights.

The bare Whisper Model outputting raw hidden-states without any specific head on top. This model inherits from [PreTrainedModel](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.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](https://pytorch.org/docs/stable/nn.html#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.

**forward**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_whisper.py#L1298)

( input\_features: typing.Optional\[torch.FloatTensor] = Noneattention\_mask: typing.Optional\[torch.LongTensor] = Nonedecoder\_input\_ids: typing.Optional\[torch.LongTensor] = Nonedecoder\_attention\_mask: typing.Optional\[torch.LongTensor] = Nonehead\_mask: typing.Optional\[torch.Tensor] = Nonedecoder\_head\_mask: typing.Optional\[torch.Tensor] = Nonecross\_attn\_head\_mask: typing.Optional\[torch.Tensor] = Noneencoder\_outputs: typing.Optional\[typing.Tuple\[typing.Tuple\[torch.FloatTensor]]] = Nonepast\_key\_values: typing.Optional\[typing.Tuple\[typing.Tuple\[torch.FloatTensor]]] = Nonedecoder\_inputs\_embeds: typing.Optional\[typing.Tuple\[torch.FloatTensor]] = Noneuse\_cache: typing.Optional\[bool] = Noneoutput\_attentions: typing.Optional\[bool] = Noneoutput\_hidden\_states: typing.Optional\[bool] = Nonereturn\_dict: typing.Optional\[bool] = None ) → [transformers.modeling\_outputs.Seq2SeqModelOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.Seq2SeqModelOutput) or `tuple(torch.FloatTensor)`

Parameters

* **input\_features** (`torch.FloatTensor` of shape `(batch_size, feature_size, sequence_length)`) — Float values mel 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 [AutoFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.AutoFeatureExtractor) should be used for extracting the mel features, padding and conversion into a tensor of type `torch.FloatTensor`. See [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor.__call__)
* **attention\_mask** (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*) — Mask to avoid performing *SpecAugment* data augmentation on padding token indices. Mask values selected in `[0, 1]`:

  * 1 for tokens that are **not masked**,
  * 0 for tokens that are **masked**.

  [What are attention masks?](https://huggingface.co/docs/transformers/glossary#attention-mask)
* **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 [WhisperTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperTokenizer). See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details.

  [What are decoder input IDs?](https://huggingface.co/docs/transformers/glossary#decoder-input-ids)

  Whisper uses the `decoder_start_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.

  If you want to change padding behavior, you should read `modeling_whisper._prepare_decoder_attention_mask` and modify to your needs. See diagram 1 in [the BART paper](https://arxiv.org/abs/1910.13461) 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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **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.
* **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](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.utils.ModelOutput) instead of a plain tuple.

Returns

[transformers.modeling\_outputs.Seq2SeqModelOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.Seq2SeqModelOutput) or `tuple(torch.FloatTensor)`

A [transformers.modeling\_outputs.Seq2SeqModelOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.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 ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) 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, if the model has an embedding layer, + 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 optional 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, if the model has an embedding layer, + 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 optional 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.

The [WhisperModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperModel) forward method, overrides the `__call__` special method.

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.

Example:

Copied

```
>>> import torch
>>> from transformers import AutoFeatureExtractor, WhisperModel
>>> from datasets import load_dataset

>>> model = WhisperModel.from_pretrained("openai/whisper-base")
>>> feature_extractor = AutoFeatureExtractor.from_pretrained("openai/whisper-base")
>>> ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
>>> inputs = feature_extractor(ds[0]["audio"]["array"], return_tensors="pt")
>>> input_features = inputs.input_features
>>> decoder_input_ids = torch.tensor([[1, 1]]) * model.config.decoder_start_token_id
>>> last_hidden_state = model(input_features, decoder_input_ids=decoder_input_ids).last_hidden_state
>>> list(last_hidden_state.shape)
[1, 2, 512]
```

**\_mask\_input\_features**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_whisper.py#L1255)

( input\_features: FloatTensorattention\_mask: typing.Optional\[torch.LongTensor] = None )

Masks extracted features along time axis and/or along feature axis according to [SpecAugment](https://arxiv.org/abs/1904.08779).

### WhisperForConditionalGeneration

#### class transformers.WhisperForConditionalGeneration

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_whisper.py#L1395)

( config: WhisperConfig )

Parameters

* **config** ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) — 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.PreTrainedModel.from_pretrained) method to load the model weights.

The Whisper Model with a language modeling head. Can be used for automatic speech recognition. This model inherits from [PreTrainedModel](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.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](https://pytorch.org/docs/stable/nn.html#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.

**forward**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_whisper.py#L1429)

( input\_features: typing.Optional\[torch.FloatTensor] = Noneattention\_mask: typing.Optional\[torch.LongTensor] = Nonedecoder\_input\_ids: typing.Optional\[torch.LongTensor] = Nonedecoder\_attention\_mask: typing.Optional\[torch.LongTensor] = Nonehead\_mask: typing.Optional\[torch.Tensor] = Nonedecoder\_head\_mask: typing.Optional\[torch.Tensor] = Nonecross\_attn\_head\_mask: typing.Optional\[torch.Tensor] = Noneencoder\_outputs: typing.Optional\[typing.Tuple\[typing.Tuple\[torch.FloatTensor]]] = Nonepast\_key\_values: typing.Optional\[typing.Tuple\[typing.Tuple\[torch.FloatTensor]]] = Nonedecoder\_inputs\_embeds: typing.Optional\[typing.Tuple\[torch.FloatTensor]] = Nonelabels: typing.Optional\[torch.LongTensor] = Noneuse\_cache: typing.Optional\[bool] = Noneoutput\_attentions: typing.Optional\[bool] = Noneoutput\_hidden\_states: typing.Optional\[bool] = Nonereturn\_dict: typing.Optional\[bool] = None ) → [transformers.modeling\_outputs.Seq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.Seq2SeqLMOutput) or `tuple(torch.FloatTensor)`

Parameters

* **input\_features** (`torch.FloatTensor` of shape `(batch_size, feature_size, sequence_length)`) — Float values mel 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 [AutoFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.AutoFeatureExtractor) should be used for extracting the mel features, padding and conversion into a tensor of type `torch.FloatTensor`. See [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor.__call__)
* **attention\_mask** (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*) — Mask to avoid performing *SpecAugment* data augmentation on padding token indices. Mask values selected in `[0, 1]`:

  * 1 for tokens that are **not masked**,
  * 0 for tokens that are **masked**.

  [What are attention masks?](https://huggingface.co/docs/transformers/glossary#attention-mask)
* **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 [WhisperTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperTokenizer). See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details.

  [What are decoder input IDs?](https://huggingface.co/docs/transformers/glossary#decoder-input-ids)

  Whisper uses the `decoder_start_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.

  If you want to change padding behavior, you should read `modeling_whisper._prepare_decoder_attention_mask` and modify to your needs. See diagram 1 in [the BART paper](https://arxiv.org/abs/1910.13461) 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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **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.
* **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](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.utils.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

[transformers.modeling\_outputs.Seq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.Seq2SeqLMOutput) or `tuple(torch.FloatTensor)`

A [transformers.modeling\_outputs.Seq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.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 ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) 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, if the model has an embedding layer, + 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, if the model has an embedding layer, + 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.

The [WhisperForConditionalGeneration](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperForConditionalGeneration) forward method, overrides the `__call__` special method.

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.

Example:

Copied

```
>>> import torch
>>> from transformers import AutoProcessor, WhisperForConditionalGeneration
>>> from datasets import load_dataset

>>> processor = AutoProcessor.from_pretrained("openai/whisper-tiny.en")
>>> model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-tiny.en")

>>> ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")

>>> inputs = processor(ds[0]["audio"]["array"], return_tensors="pt")
>>> input_features = inputs.input_features

>>> generated_ids = model.generate(inputs=input_features)

>>> transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
>>> transcription
' Mr. Quilter is the apostle of the middle classes, and we are glad to welcome his gospel.'
```

### WhisperForAudioClassification

#### class transformers.WhisperForAudioClassification

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_whisper.py#L1866)

( config )

Parameters

* **input\_features** (`torch.FloatTensor` of shape `(batch_size, feature_size, sequence_length)`) — Float values mel 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 [AutoFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.AutoFeatureExtractor) should be used for extracting the mel features, padding and conversion into a tensor of type `torch.FloatTensor`. See [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor.__call__)
* **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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **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.
* **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](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.utils.ModelOutput) instead of a plain tuple.

Whisper Encoder Model with a sequence classification head on top (a linear layer over the pooled output) for tasks like SUPERB Keyword Spotting.

**forward**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_whisper.py#L1893)

( input\_features: typing.Optional\[torch.LongTensor] = Nonehead\_mask: typing.Optional\[torch.Tensor] = Noneencoder\_outputs: typing.Optional\[typing.Tuple\[typing.Tuple\[torch.FloatTensor]]] = Nonelabels: typing.Optional\[torch.LongTensor] = Noneoutput\_attentions: typing.Optional\[bool] = Noneoutput\_hidden\_states: typing.Optional\[bool] = Nonereturn\_dict: typing.Optional\[bool] = None ) → [transformers.modeling\_outputs.SequenceClassifierOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.SequenceClassifierOutput) or `tuple(torch.FloatTensor)`

Parameters

* **input\_features** (`torch.FloatTensor` of shape `(batch_size, feature_size, sequence_length)`) — Float values mel 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 [AutoFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.AutoFeatureExtractor) should be used for extracting the mel features, padding and conversion into a tensor of type `torch.FloatTensor`. See [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor.__call__)
* **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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **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.
* **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](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.utils.ModelOutput) instead of a plain tuple.
* **labels** (`torch.LongTensor` of shape `(batch_size,)`, *optional*) — Labels for computing the sequence classification/regression loss. Indices should be in `[0, ..., config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If `config.num_labels > 1` a classification loss is computed (Cross-Entropy).

Returns

[transformers.modeling\_outputs.SequenceClassifierOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.SequenceClassifierOutput) or `tuple(torch.FloatTensor)`

A [transformers.modeling\_outputs.SequenceClassifierOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.SequenceClassifierOutput) 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 ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) and inputs.

* **loss** (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided) — Classification (or regression if config.num\_labels==1) loss.
* **logits** (`torch.FloatTensor` of shape `(batch_size, config.num_labels)`) — Classification (or regression if config.num\_labels==1) scores (before SoftMax).
* **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, if the model has an embedding layer, + 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 optional 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.

The [WhisperForAudioClassification](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperForAudioClassification) forward method, overrides the `__call__` special method.

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.

Example:

Copied

```
>>> import torch
>>> from transformers import AutoFeatureExtractor, WhisperForAudioClassification
>>> from datasets import load_dataset

>>> feature_extractor = AutoFeatureExtractor.from_pretrained("sanchit-gandhi/whisper-medium-fleurs-lang-id")
>>> model = WhisperForAudioClassification.from_pretrained("sanchit-gandhi/whisper-medium-fleurs-lang-id")

>>> ds = load_dataset("google/fleurs", "all", split="validation", streaming=True)
>>> sample = next(iter(ds))

>>> inputs = feature_extractor(
...     sample["audio"]["array"], sampling_rate=sample["audio"]["sampling_rate"], return_tensors="pt"
... )
>>> input_features = inputs.input_features

>>> with torch.no_grad():
...     logits = model(input_features).logits

>>> predicted_class_ids = torch.argmax(logits).item()
>>> predicted_label = model.config.id2label[predicted_class_ids]
>>> predicted_label
'Afrikaans'
```

### TFWhisperModel

#### class transformers.TFWhisperModel

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_tf_whisper.py#L1093)

( \*args\*\*kwargs )

Parameters

* **config** ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) — 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.TFPreTrainedModel.from_pretrained) method to load the model weights.

The bare Whisper Model outputting raw hidden-states without any specific head on top. This model inherits from [TFPreTrainedModel](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.TFPreTrainedModel). 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 [tf.keras.Model](https://www.tensorflow.org/api_docs/python/tf/keras/Model) subclass. Use it as a regular TF 2.0 Keras Model and refer to the TF 2.0 documentation for all matter related to general usage and behavior.

**call**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_tf_whisper.py#L1117)

( input\_features: TFModelInputType | None = Nonedecoder\_input\_ids: np.ndarray | tf.Tensor | None = Nonedecoder\_attention\_mask: np.ndarray | tf.Tensor | None = Nonedecoder\_position\_ids: np.ndarray | tf.Tensor | None = Nonehead\_mask: np.ndarray | tf.Tensor | None = Nonedecoder\_head\_mask: np.ndarray | tf.Tensor | None = Nonecross\_attn\_head\_mask: np.ndarray | tf.Tensor | None = Noneencoder\_outputs: Optional\[Tuple\[Tuple\[Union\[np.ndarray, tf.Tensor]]]] = Nonepast\_key\_values: Optional\[Tuple\[Tuple\[Union\[np.ndarray, tf.Tensor]]]] = Nonedecoder\_inputs\_embeds: Optional\[Tuple\[Union\[np.ndarray, tf.Tensor]]] = Noneuse\_cache: Optional\[bool] = Noneoutput\_attentions: Optional\[bool] = Noneoutput\_hidden\_states: Optional\[bool] = Nonereturn\_dict: Optional\[bool] = Nonetraining: bool = False ) → [transformers.modeling\_tf\_outputs.TFSeq2SeqModelOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_tf_outputs.TFSeq2SeqModelOutput) or `tuple(tf.Tensor)`

Parameters

* **input\_features** (`tf.Tensor` of shape `(batch_size, feature_size, sequence_length)`) — 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 [AutoFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.AutoFeatureExtractor) should be used for extracting the fbank features, padding and conversion into a tensor of type `tf.Tensor`. See [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor.__call__)
* **decoder\_input\_ids** (`tf.Tensor` of shape `(batch_size, target_sequence_length)`, *optional*) — Indices of decoder input sequence tokens in the vocabulary.

  Indices can be obtained using `SpeechToTextTokenizer`. See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details.

  [What are decoder input IDs?](https://huggingface.co/docs/transformers/glossary#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** (`tf.Tensor` 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.

  If you want to change padding behavior, you should read `modeling_whisper._prepare_decoder_attention_mask` and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more information on the default strategy.
* **head\_mask** (`tf.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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **decoder\_head\_mask** (`tf.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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **cross\_attn\_head\_mask** (`tf.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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **encoder\_outputs** (`tuple(tuple(tf.Tensor)`, *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(tf.Tensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`) — Tuple of `tuple(tf.Tensor)` 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** (`tf.Tensor` 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.
* **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](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.utils.ModelOutput) instead of a plain tuple.

Returns

[transformers.modeling\_tf\_outputs.TFSeq2SeqModelOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_tf_outputs.TFSeq2SeqModelOutput) or `tuple(tf.Tensor)`

A [transformers.modeling\_tf\_outputs.TFSeq2SeqModelOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_tf_outputs.TFSeq2SeqModelOutput) or a tuple of `tf.Tensor` (if `return_dict=False` is passed or when `config.return_dict=False`) comprising various elements depending on the configuration ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) and inputs.

* **last\_hidden\_state** (`tf.Tensor` 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** (`List[tf.Tensor]`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`) — List of `tf.Tensor` 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(tf.Tensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `tf.Tensor` (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(tf.Tensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `tf.Tensor` (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(tf.Tensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `tf.Tensor` (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** (`tf.Tensor` 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(tf.Tensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `tf.Tensor` (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(tf.Tensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `tf.Tensor` (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.

The [TFWhisperModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.TFWhisperModel) forward method, overrides the `__call__` special method.

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.

Example:

Copied

```
>>> import tensorflow as tf
>>> from transformers import TFWhisperModel, AutoFeatureExtractor
>>> from datasets import load_dataset

>>> model = TFWhisperModel.from_pretrained("openai/whisper-base")
>>> feature_extractor = AutoFeatureExtractor.from_pretrained("openai/whisper-base")
>>> ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
>>> inputs = feature_extractor(ds[0]["audio"]["array"], return_tensors="tf")
>>> input_features = inputs.input_features
>>> decoder_input_ids = tf.convert_to_tensor([[1, 1]]) * model.config.decoder_start_token_id
>>> last_hidden_state = model(input_features, decoder_input_ids=decoder_input_ids).last_hidden_state
>>> list(last_hidden_state.shape)
[1, 2, 512]
```

### TFWhisperForConditionalGeneration

#### class transformers.TFWhisperForConditionalGeneration

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_tf_whisper.py#L1201)

( \*args\*\*kwargs )

Parameters

* **config** ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) — 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.TFPreTrainedModel.from_pretrained) method to load the model weights.

The Whisper Model with a language modeling head. Can be used for automatic speech recognition. This model inherits from [TFPreTrainedModel](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.TFPreTrainedModel). 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 [tf.keras.Model](https://www.tensorflow.org/api_docs/python/tf/keras/Model) subclass. Use it as a regular TF 2.0 Keras Model and refer to the TF 2.0 documentation for all matter related to general usage and behavior.

**call**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_tf_whisper.py#L1232)

( input\_features: TFModelInputType | None = Nonedecoder\_input\_ids: np.ndarray | tf.Tensor | None = Nonedecoder\_attention\_mask: np.ndarray | tf.Tensor | None = Nonedecoder\_position\_ids: np.ndarray | tf.Tensor | None = Nonehead\_mask: np.ndarray | tf.Tensor | None = Nonedecoder\_head\_mask: np.ndarray | tf.Tensor | None = Nonecross\_attn\_head\_mask: np.ndarray | tf.Tensor | None = Noneencoder\_outputs: Optional\[Tuple\[Tuple\[Union\[np.ndarray, tf.Tensor]]]] = Nonepast\_key\_values: Optional\[Tuple\[Tuple\[Union\[np.ndarray, tf.Tensor]]]] = Nonedecoder\_inputs\_embeds: Optional\[Tuple\[Union\[np.ndarray, tf.Tensor]]] = Nonelabels: np.ndarray | tf.Tensor | None = Noneuse\_cache: Optional\[bool] = Noneoutput\_attentions: Optional\[bool] = Noneoutput\_hidden\_states: Optional\[bool] = Nonereturn\_dict: Optional\[bool] = Nonetraining: bool = False ) → [transformers.modeling\_tf\_outputs.TFSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_tf_outputs.TFSeq2SeqLMOutput) or `tuple(tf.Tensor)`

Parameters

* **input\_features** (`tf.Tensor` of shape `(batch_size, feature_size, sequence_length)`) — 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 [AutoFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.AutoFeatureExtractor) should be used for extracting the fbank features, padding and conversion into a tensor of type `tf.Tensor`. See [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor.__call__)
* **decoder\_input\_ids** (`tf.Tensor` of shape `(batch_size, target_sequence_length)`, *optional*) — Indices of decoder input sequence tokens in the vocabulary.

  Indices can be obtained using `SpeechToTextTokenizer`. See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details.

  [What are decoder input IDs?](https://huggingface.co/docs/transformers/glossary#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** (`tf.Tensor` 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.

  If you want to change padding behavior, you should read `modeling_whisper._prepare_decoder_attention_mask` and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more information on the default strategy.
* **head\_mask** (`tf.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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **decoder\_head\_mask** (`tf.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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **cross\_attn\_head\_mask** (`tf.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]`:
  * 1 indicates the head is **not masked**,
  * 0 indicates the head is **masked**.
* **encoder\_outputs** (`tuple(tuple(tf.Tensor)`, *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(tf.Tensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`) — Tuple of `tuple(tf.Tensor)` 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** (`tf.Tensor` 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.
* **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](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.utils.ModelOutput) instead of a plain tuple.
* **labels** (`tf.Tensor` 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

[transformers.modeling\_tf\_outputs.TFSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_tf_outputs.TFSeq2SeqLMOutput) or `tuple(tf.Tensor)`

A [transformers.modeling\_tf\_outputs.TFSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_tf_outputs.TFSeq2SeqLMOutput) or a tuple of `tf.Tensor` (if `return_dict=False` is passed or when `config.return_dict=False`) comprising various elements depending on the configuration ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) and inputs.

* **loss** (`tf.Tensor` of shape `(n,)`, *optional*, where n is the number of non-masked labels, returned when `labels` is provided) — Language modeling loss.
* **logits** (`tf.Tensor` 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[tf.Tensor]`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`) — List of `tf.Tensor` 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(tf.Tensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `tf.Tensor` (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(tf.Tensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `tf.Tensor` (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(tf.Tensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `tf.Tensor` (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** (`tf.Tensor` 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(tf.Tensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `tf.Tensor` (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(tf.Tensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `tf.Tensor` (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.

The [TFWhisperForConditionalGeneration](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.TFWhisperForConditionalGeneration) forward method, overrides the `__call__` special method.

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.

Example:

Copied

```
>>> import tensorflow as tf
>>> from transformers import AutoProcessor, TFWhisperForConditionalGeneration
>>> from datasets import load_dataset

>>> processor = AutoProcessor.from_pretrained("openai/whisper-tiny.en")
>>> model = TFWhisperForConditionalGeneration.from_pretrained("openai/whisper-tiny.en")

>>> ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")

>>> inputs = processor(ds[0]["audio"]["array"], return_tensors="tf")
>>> input_features = inputs.input_features

>>> generated_ids = model.generate(input_features=input_features)

>>> transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
>>> transcription
' Mr. Quilter is the apostle of the middle classes, and we are glad to welcome his gospel.'
```

### FlaxWhisperModel

#### class transformers.FlaxWhisperModel

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_flax_whisper.py#L1165)

( config: WhisperConfiginput\_shape: typing.Tuple\[int] = (1, 80, 3000)seed: int = 0dtype: dtype = \<class 'jax.numpy.float32'>\_do\_init: bool = Truegradient\_checkpointing: bool = False\*\*kwargs )

Parameters

* **config** ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) — 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.from_pretrained) method to load the model weights.
* **dtype** (`jax.numpy.dtype`, *optional*, defaults to `jax.numpy.float32`) — The data type of the computation. Can be one of `jax.numpy.float32`, `jax.numpy.float16` (on GPUs) and `jax.numpy.bfloat16` (on TPUs). This can be used to enable mixed-precision training or half-precision inference on GPUs or TPUs. If specified all the computation will be performed with the given `dtype`. **Note that this only specifies the dtype of the computation and does not influence the dtype of model parameters.** If you wish to change the dtype of the model parameters, see [to\_fp16()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.to_fp16) and [to\_bf16()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.to_bf16).

The bare Whisper Model transformer outputting raw hidden-states without any specific head on top. This model inherits from [FlaxPreTrainedModel](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel). Check the superclass documentation for the generic methods the library implements for all its models (such as downloading or saving, resizing the input embeddings, pruning heads etc.) This model is also a Flax Linen [flax.nn.Module](https://flax.readthedocs.io/en/latest/_autosummary/flax.nn.module.html) subclass. Use it as a regular Flax Module and refer to the Flax documentation for all matter related to general usage and behavior. Finally, this model supports inherent JAX features such as:

* [Just-In-Time (JIT) compilation](https://jax.readthedocs.io/en/latest/jax.html#just-in-time-compilation-jit)
* [Automatic Differentiation](https://jax.readthedocs.io/en/latest/jax.html#automatic-differentiation)
* [Vectorization](https://jax.readthedocs.io/en/latest/jax.html#vectorization-vmap)
* [Parallelization](https://jax.readthedocs.io/en/latest/jax.html#parallelization-pmap)

**\_\_call\_\_**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_flax_whisper.py#L1110)

( input\_features: Arraydecoder\_input\_ids: Arrayattention\_mask: typing.Optional\[jax.Array] = Nonedecoder\_attention\_mask: typing.Optional\[jax.Array] = Noneposition\_ids: typing.Optional\[jax.Array] = Nonedecoder\_position\_ids: typing.Optional\[jax.Array] = Noneoutput\_attentions: typing.Optional\[bool] = Noneoutput\_hidden\_states: typing.Optional\[bool] = Nonereturn\_dict: typing.Optional\[bool] = Nonetrain: bool = Falseparams: dict = Nonedropout\_rng: PRNGKey = None ) → [transformers.modeling\_flax\_outputs.FlaxSeq2SeqModelOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSeq2SeqModelOutput) or `tuple(torch.FloatTensor)`

Parameters

* **input\_features** (`numpy.ndarray` of shape `(batch_size, feature_size, sequence_length)`) — Float values mel 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 [WhisperFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor) should be used for extracting the features, padding and conversion into a tensor of type `numpy.ndarray`. See [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor.__call__)
* **attention\_mask** (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*) — Whisper does not support masking of the `input_features`, this argument is preserved for compatibility, but is not used. By default the silence in the input log mel spectrogram are ignored.
* **decoder\_input\_ids** (`numpy.ndarray` of shape `(batch_size, target_sequence_length)`, *optional*) — Indices of decoder input sequence tokens in the vocabulary. Indices can be obtained using [WhisperTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperTokenizer). See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details. [What are decoder input IDs?](https://huggingface.co/docs/transformers/glossary#decoder-input-ids) Whisper uses the `decoder_start_token_id` as the starting token for `decoder_input_ids` generation.
* **decoder\_attention\_mask** (`numpy.ndarray` 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. If you want to change padding behavior, you should modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more information on the default strategy.
* **position\_ids** (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*) — Whisper does not use `position_ids` in the encoder as `input_features` is always the same size and doesn’t use masking, but this argument is preserved for compatibility. By default the silence in the input log mel spectrogram are ignored.
* **decoder\_position\_ids** (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*) — Indices of positions of each decoder input sequence tokens in the position embeddings. Selected in the range `[0, config.max_position_embeddings - 1]`.
* **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](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.utils.ModelOutput) instead of a plain tuple.

Returns

[transformers.modeling\_flax\_outputs.FlaxSeq2SeqModelOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSeq2SeqModelOutput) or `tuple(torch.FloatTensor)`

A [transformers.modeling\_flax\_outputs.FlaxSeq2SeqModelOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSeq2SeqModelOutput) 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 ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) and inputs.

* **last\_hidden\_state** (`jnp.ndarray` 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(jnp.ndarray))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`) — Tuple of `tuple(jnp.ndarray)` 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(jnp.ndarray)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `jnp.ndarray` (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(jnp.ndarray)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `jnp.ndarray` (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(jnp.ndarray)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `jnp.ndarray` (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** (`jnp.ndarray` 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(jnp.ndarray)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `jnp.ndarray` (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(jnp.ndarray)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `jnp.ndarray` (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.

The `FlaxWhisperPreTrainedModel` forward method, overrides the `__call__` special method.

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.

Example:

Copied

```
>>> from transformers import AutoTokenizer, FlaxWhisperModel

>>> tokenizer = AutoTokenizer.from_pretrained("openai/whisper-tiny")
>>> model = FlaxWhisperModel.from_pretrained("openai/whisper-tiny")

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

>>> last_hidden_states = outputs.last_hidden_state
```

### FlaxWhisperForConditionalGeneration

#### class transformers.FlaxWhisperForConditionalGeneration

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_flax_whisper.py#L1244)

( config: WhisperConfiginput\_shape: typing.Tuple\[int] = (1, 80, 3000)seed: int = 0dtype: dtype = \<class 'jax.numpy.float32'>\_do\_init: bool = Truegradient\_checkpointing: bool = False\*\*kwargs )

Parameters

* **config** ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) — 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.from_pretrained) method to load the model weights.
* **dtype** (`jax.numpy.dtype`, *optional*, defaults to `jax.numpy.float32`) — The data type of the computation. Can be one of `jax.numpy.float32`, `jax.numpy.float16` (on GPUs) and `jax.numpy.bfloat16` (on TPUs). This can be used to enable mixed-precision training or half-precision inference on GPUs or TPUs. If specified all the computation will be performed with the given `dtype`. **Note that this only specifies the dtype of the computation and does not influence the dtype of model parameters.** If you wish to change the dtype of the model parameters, see [to\_fp16()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.to_fp16) and [to\_bf16()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.to_bf16).

The Whisper Model with a language modeling head. This model inherits from [FlaxPreTrainedModel](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel). Check the superclass documentation for the generic methods the library implements for all its models (such as downloading or saving, resizing the input embeddings, pruning heads etc.) This model is also a Flax Linen [flax.nn.Module](https://flax.readthedocs.io/en/latest/_autosummary/flax.nn.module.html) subclass. Use it as a regular Flax Module and refer to the Flax documentation for all matter related to general usage and behavior. Finally, this model supports inherent JAX features such as:

* [Just-In-Time (JIT) compilation](https://jax.readthedocs.io/en/latest/jax.html#just-in-time-compilation-jit)
* [Automatic Differentiation](https://jax.readthedocs.io/en/latest/jax.html#automatic-differentiation)
* [Vectorization](https://jax.readthedocs.io/en/latest/jax.html#vectorization-vmap)
* [Parallelization](https://jax.readthedocs.io/en/latest/jax.html#parallelization-pmap)

**\_\_call\_\_**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_flax_whisper.py#L1110)

( input\_features: Arraydecoder\_input\_ids: Arrayattention\_mask: typing.Optional\[jax.Array] = Nonedecoder\_attention\_mask: typing.Optional\[jax.Array] = Noneposition\_ids: typing.Optional\[jax.Array] = Nonedecoder\_position\_ids: typing.Optional\[jax.Array] = Noneoutput\_attentions: typing.Optional\[bool] = Noneoutput\_hidden\_states: typing.Optional\[bool] = Nonereturn\_dict: typing.Optional\[bool] = Nonetrain: bool = Falseparams: dict = Nonedropout\_rng: PRNGKey = None ) → [transformers.modeling\_flax\_outputs.FlaxSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSeq2SeqLMOutput) or `tuple(torch.FloatTensor)`

Parameters

* **input\_features** (`numpy.ndarray` of shape `(batch_size, feature_size, sequence_length)`) — Float values mel 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 [WhisperFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor) should be used for extracting the features, padding and conversion into a tensor of type `numpy.ndarray`. See [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor.__call__)
* **attention\_mask** (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*) — Whisper does not support masking of the `input_features`, this argument is preserved for compatibility, but is not used. By default the silence in the input log mel spectrogram are ignored.
* **decoder\_input\_ids** (`numpy.ndarray` of shape `(batch_size, target_sequence_length)`, *optional*) — Indices of decoder input sequence tokens in the vocabulary. Indices can be obtained using [WhisperTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperTokenizer). See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details. [What are decoder input IDs?](https://huggingface.co/docs/transformers/glossary#decoder-input-ids) Whisper uses the `decoder_start_token_id` as the starting token for `decoder_input_ids` generation.
* **decoder\_attention\_mask** (`numpy.ndarray` 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. If you want to change padding behavior, you should modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more information on the default strategy.
* **position\_ids** (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*) — Whisper does not use `position_ids` in the encoder as `input_features` is always the same size and doesn’t use masking, but this argument is preserved for compatibility. By default the silence in the input log mel spectrogram are ignored.
* **decoder\_position\_ids** (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*) — Indices of positions of each decoder input sequence tokens in the position embeddings. Selected in the range `[0, config.max_position_embeddings - 1]`.
* **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](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.utils.ModelOutput) instead of a plain tuple.

Returns

[transformers.modeling\_flax\_outputs.FlaxSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSeq2SeqLMOutput) or `tuple(torch.FloatTensor)`

A [transformers.modeling\_flax\_outputs.FlaxSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSeq2SeqLMOutput) 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 ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) and inputs.

* **logits** (`jnp.ndarray` 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(jnp.ndarray))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`) — Tuple of `tuple(jnp.ndarray)` 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(jnp.ndarray)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `jnp.ndarray` (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(jnp.ndarray)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `jnp.ndarray` (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(jnp.ndarray)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `jnp.ndarray` (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** (`jnp.ndarray` 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(jnp.ndarray)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `jnp.ndarray` (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(jnp.ndarray)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `jnp.ndarray` (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.

The `FlaxWhisperPreTrainedModel` forward method, overrides the `__call__` special method.

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.

Transcription example:

Copied

```
>>> from transformers import WhisperProcessor, FlaxWhisperForConditionalGeneration
>>> from datasets import load_dataset

>>> processor = WhisperProcessor.from_pretrained("openai/whisper-tiny.en")
>>> model = FlaxWhisperForConditionalGeneration.from_pretrained("openai/whisper-tiny.en", from_pt=True)
>>> ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
>>> inputs = processor(ds[0]["audio"]["array"], return_tensors="np")
>>> input_features = inputs.input_features
>>> generated_ids = model.generate(input_ids=input_features)
>>> transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
>>> transcription
' Mr. Quilter is the apostle of the middle classes, and we are glad to welcome his gospel.'
```

### FlaxWhisperForAudioClassification

#### class transformers.FlaxWhisperForAudioClassification

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_flax_whisper.py#L1574)

( config: WhisperConfiginput\_shape: typing.Tuple\[int] = (1, 80, 3000)seed: int = 0dtype: dtype = \<class 'jax.numpy.float32'>\_do\_init: bool = Truegradient\_checkpointing: bool = False\*\*kwargs )

Parameters

* **config** ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) — 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.from_pretrained) method to load the model weights.
* **dtype** (`jax.numpy.dtype`, *optional*, defaults to `jax.numpy.float32`) — The data type of the computation. Can be one of `jax.numpy.float32`, `jax.numpy.float16` (on GPUs) and `jax.numpy.bfloat16` (on TPUs). This can be used to enable mixed-precision training or half-precision inference on GPUs or TPUs. If specified all the computation will be performed with the given `dtype`. **Note that this only specifies the dtype of the computation and does not influence the dtype of model parameters.** If you wish to change the dtype of the model parameters, see [to\_fp16()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.to_fp16) and [to\_bf16()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.to_bf16).

The Whisper Model with an audio classification head on top. This model inherits from [FlaxPreTrainedModel](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel). Check the superclass documentation for the generic methods the library implements for all its models (such as downloading or saving, resizing the input embeddings, pruning heads etc.) This model is also a Flax Linen [flax.nn.Module](https://flax.readthedocs.io/en/latest/_autosummary/flax.nn.module.html) subclass. Use it as a regular Flax Module and refer to the Flax documentation for all matter related to general usage and behavior. Finally, this model supports inherent JAX features such as:

* [Just-In-Time (JIT) compilation](https://jax.readthedocs.io/en/latest/jax.html#just-in-time-compilation-jit)
* [Automatic Differentiation](https://jax.readthedocs.io/en/latest/jax.html#automatic-differentiation)
* [Vectorization](https://jax.readthedocs.io/en/latest/jax.html#vectorization-vmap)
* [Parallelization](https://jax.readthedocs.io/en/latest/jax.html#parallelization-pmap)

**\_\_call\_\_**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/whisper/modeling_flax_whisper.py#L1601)

( input\_features: Arrayattention\_mask: typing.Optional\[jax.Array] = Noneoutput\_attentions: typing.Optional\[bool] = Noneoutput\_hidden\_states: typing.Optional\[bool] = Nonereturn\_dict: typing.Optional\[bool] = Nonetrain: bool = Falseparams: dict = Nonedropout\_rng: PRNGKey = None\*\*kwargs ) → [transformers.modeling\_flax\_outputs.FlaxSequenceClassifierOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSequenceClassifierOutput) or `tuple(torch.FloatTensor)`

Parameters

* **input\_features** (`numpy.ndarray` of shape `(batch_size, feature_size, sequence_length)`) — Float values mel 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 [WhisperFeatureExtractor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor) should be used for extracting the features, padding and conversion into a tensor of type `numpy.ndarray`. See [**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperFeatureExtractor.__call__)
* **attention\_mask** (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*) — Whisper does not support masking of the `input_features`, this argument is preserved for compatibility, but is not used. By default the silence in the input log mel spectrogram are ignored.
* **decoder\_input\_ids** (`numpy.ndarray` of shape `(batch_size, target_sequence_length)`, *optional*) — Indices of decoder input sequence tokens in the vocabulary. Indices can be obtained using [WhisperTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperTokenizer). See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details. [What are decoder input IDs?](https://huggingface.co/docs/transformers/glossary#decoder-input-ids) Whisper uses the `decoder_start_token_id` as the starting token for `decoder_input_ids` generation.
* **decoder\_attention\_mask** (`numpy.ndarray` 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. If you want to change padding behavior, you should modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more information on the default strategy.
* **position\_ids** (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*) — Whisper does not use `position_ids` in the encoder as `input_features` is always the same size and doesn’t use masking, but this argument is preserved for compatibility. By default the silence in the input log mel spectrogram are ignored.
* **decoder\_position\_ids** (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*) — Indices of positions of each decoder input sequence tokens in the position embeddings. Selected in the range `[0, config.max_position_embeddings - 1]`.
* **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](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.utils.ModelOutput) instead of a plain tuple.

Returns

[transformers.modeling\_flax\_outputs.FlaxSequenceClassifierOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSequenceClassifierOutput) or `tuple(torch.FloatTensor)`

A [transformers.modeling\_flax\_outputs.FlaxSequenceClassifierOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSequenceClassifierOutput) 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 ([WhisperConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.WhisperConfig)) and inputs.

* **logits** (`jnp.ndarray` of shape `(batch_size, config.num_labels)`) — Classification (or regression if config.num\_labels==1) scores (before SoftMax).
* **hidden\_states** (`tuple(jnp.ndarray)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `jnp.ndarray` (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(jnp.ndarray)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `jnp.ndarray` (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.

The [FlaxWhisperForAudioClassification](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/whisper#transformers.FlaxWhisperForAudioClassification) forward method, overrides the `__call__` special method.

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.

Transcription example:

Copied

```
>>> import jax.numpy as jnp
>>> from transformers import AutoFeatureExtractor, FlaxWhisperForAudioClassification
>>> from datasets import load_dataset

>>> feature_extractor = AutoFeatureExtractor.from_pretrained("sanchit-gandhi/whisper-medium-fleurs-lang-id")
>>> model = FlaxWhisperForAudioClassification.from_pretrained(
...     "sanchit-gandhi/whisper-medium-fleurs-lang-id", from_pt=True
... )
>>> ds = load_dataset("google/fleurs", "all", split="validation", streaming=True)

>>> sample = next(iter(ds))

>>> inputs = feature_extractor(
...     sample["audio"]["array"], sampling_rate=sample["audio"]["sampling_rate"], return_tensors="np"
... )
>>> input_features = inputs.input_features

>>> logits = model(input_features).logits

>>> predicted_class_ids = jnp.argmax(logits).item()
>>> predicted_label = model.config.id2label[predicted_class_ids]
>>> predicted_label
'af_za'
```