Keras 3 API documentation / KerasNLP / Models / DistilBert / DistilBertBackbone model

DistilBertBackbone model

[source]

DistilBertBackbone class

keras_nlp.models.DistilBertBackbone(
    vocabulary_size,
    num_layers,
    num_heads,
    hidden_dim,
    intermediate_dim,
    dropout=0.1,
    max_sequence_length=512,
    dtype=None,
    **kwargs
)

A DistilBERT encoder network.

This network implements a bi-directional Transformer-based encoder as described in "DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter". It includes the embedding lookups and transformer layers, but not the masked language model or classification task networks.

The default constructor gives a fully customizable, randomly initialized DistilBERT encoder 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 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.
  • dropout: float. Dropout probability for the Transformer encoder.
  • max_sequence_length: int. The maximum sequence length that this encoder can consume. If None, max_sequence_length uses the value from sequence length. This determines the variable shape for positional embeddings.
  • 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 = {
    "token_ids": np.ones(shape=(1, 12), dtype="int32"),
    "padding_mask": np.array([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0]]),
}

# Pretrained DistilBERT encoder.
model = keras_nlp.models.DistilBertBackbone.from_preset(
    "distil_bert_base_en_uncased"
)
model(input_data)

# Randomly initialized DistilBERT encoder with custom config.
model = keras_nlp.models.DistilBertBackbone(
    vocabulary_size=30552,
    num_layers=4,
    num_heads=4,
    hidden_dim=256,
    intermediate_dim=512,
    max_sequence_length=128,
)
model(input_data)

[source]

from_preset method

DistilBertBackbone.from_preset()

Instantiate DistilBertBackbone model from preset architecture and weights.

Arguments

  • preset: string. Must be one of "distil_bert_base_en_uncased", "distil_bert_base_en", "distil_bert_base_multi".
  • load_weights: Whether to load pre-trained weights into model. Defaults to True.

Examples

# Load architecture and weights from preset
model = keras_nlp.models.DistilBertBackbone.from_preset(
    "distil_bert_base_en_uncased"
)

# Load randomly initialized model from preset architecture
model = keras_nlp.models.DistilBertBackbone.from_preset(
    "distil_bert_base_en_uncased",
    load_weights=False
)
Preset name Parameters Description
distil_bert_base_en_uncased 66.36M 6-layer DistilBERT model where all input is lowercased. Trained on English Wikipedia + BooksCorpus using BERT as the teacher model.
distil_bert_base_en 65.19M 6-layer DistilBERT model where case is maintained. Trained on English Wikipedia + BooksCorpus using BERT as the teacher model.
distil_bert_base_multi 134.73M 6-layer DistilBERT model where case is maintained. Trained on Wikipedias of 104 languages

token_embedding property

keras_nlp.models.DistilBertBackbone.token_embedding

A keras.layers.Embedding instance for embedding token ids.

This layer embeds integer token ids to the hidden dim of the model.