WhisperBackbone model

[source]

WhisperBackbone class

keras_hub.models.WhisperBackbone(
    vocabulary_size,
    num_layers,
    num_heads,
    hidden_dim,
    intermediate_dim,
    num_mels=80,
    dropout=0.0,
    max_encoder_sequence_length=3000,
    max_decoder_sequence_length=448,
    dtype=None,
    **kwargs
)

A Whisper encoder-decoder network for speech.

This class implements a Transformer-based encoder-decoder model as described in "Robust Speech Recognition via Large-Scale Weak Supervision". It includes the embedding lookups and transformer layers, but not the head for predicting the next token.

The default constructor gives a fully customizable, randomly initialized Whisper model with any number of layers, heads, and embedding dimensions. To load preset architectures and weights, use the from_preset() constructor.

Disclaimer: Pre-trained models are provided on an "as is" basis, without warranties or conditions of any kind. The underlying model is provided by a third party and subject to a separate license, available here.

Arguments

• vocabulary_size: int. The size of the token vocabulary.
• num_layers: int. The number of transformer encoder layers and transformer decoder layers.
• num_heads: int. The number of attention heads for each transformer. The hidden size must be divisible by the number of attention heads.
• hidden_dim: int. The size of the transformer encoding and pooler layers.
• intermediate_dim: int. The output dimension of the first Dense layer in a two-layer feedforward network for each transformer.
• num_mels: int. The number of mel-frequency filters. Defaults to 80.
• dropout: float. Dropout probability for the Transformer encoder.
• max_encoder_sequence_length: int. The maximum sequence length that the audio encoder can consume. Since the second convolutional layer in the encoder reduces the sequence length by half (stride of 2), we use max_encoder_sequence_length // 2 as the sequence length for the positional embedding layer.
• max_decoder_sequence_length: int. The maximum sequence length that the text decoder can consume.
• dtype: string or keras.mixed_precision.DTypePolicy. The dtype to use for model computations and weights. Note that some computations, such as softmax and layer normalization, will always be done at float32 precision regardless of dtype.

Examples

input_data = {
    "encoder_features": np.ones(shape=(1, 12, 80), dtype="int32"),
    "decoder_token_ids": np.ones(shape=(1, 12), dtype="int32"),
    "decoder_padding_mask": np.array(
        [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0]]
    ),
}

# Randomly initialized Whisper encoder-decoder model with a custom config.
model = keras_hub.models.WhisperBackbone(
    vocabulary_size=51864,
    num_layers=4,
    num_heads=4,
    hidden_dim=256,
    intermediate_dim=512,
    max_encoder_sequence_length=128,
    max_decoder_sequence_length=128,
)
model(input_data)

[source]

from_preset method

WhisperBackbone.from_preset(preset, load_weights=True, **kwargs)

Instantiate a keras_hub.models.Backbone from a model preset.

A preset is a directory of configs, weights and other file assets used to save and load a pre-trained model. The preset can be passed as a one of:

1. a built-in preset identifier like 'bert_base_en'
2. a Kaggle Models handle like 'kaggle://user/bert/keras/bert_base_en'
3. a Hugging Face handle like 'hf://user/bert_base_en'
4. a path to a local preset directory like './bert_base_en'

This constructor can be called in one of two ways. Either from the base class like keras_hub.models.Backbone.from_preset(), or from a model class like keras_hub.models.GemmaBackbone.from_preset(). If calling from the base class, the subclass of the returning object will be inferred from the config in the preset directory.

For any Backbone subclass, you can run cls.presets.keys() to list all built-in presets available on the class.

Arguments

• preset: string. A built-in preset identifier, a Kaggle Models handle, a Hugging Face handle, or a path to a local directory.
• load_weights: bool. If True, the weights will be loaded into the model architecture. If False, the weights will be randomly initialized.

Examples

# Load a Gemma backbone with pre-trained weights.
model = keras_hub.models.Backbone.from_preset(
    "gemma_2b_en",
)

# Load a Bert backbone with a pre-trained config and random weights.
model = keras_hub.models.Backbone.from_preset(
    "bert_base_en",
    load_weights=False,
)