RWKV7Backbone model

[source]

RWKV7Backbone class

keras_hub.models.RWKV7Backbone(
    hidden_size,
    head_size,
    num_layers,
    vocabulary_size,
    intermediate_dim,
    gate_lora=128,
    mv_lora=32,
    aaa_lora=64,
    decay_lora=64,
    dtype=None,
    dropout_rate=0,
    **kwargs
)

The RWKV7 Transformer core architecture with hyperparameters.

This network implements a RNN-based decoder network, Goose, as described in RWKV-7.

This network implements a Modern RNN architecture based on linear attention mechanisms with recurrent processing, as described in the RWKV papers. It includes the embedding lookups and RWKV-7 blocks.

The default constructor gives a fully customizable, randomly initialized RWKV-7 model with any number of layers, heads, and embedding dimensions. To load preset architectures and weights, use the from_preset constructor.

Arguments

• hidden_size: int. The size of the transformer encoding and pooling layers.
• head_size: int. The size of each attention head.
• num_layers: int. The number of transformer layers.
• vocabulary_size: int. The size of the token vocabulary.
• intermediate_dim: int. The output dimension of the first Dense layer in a two-layer feedforward network for each transformer.
• gate_lora: int. LoRA dimension for gating. Defaults to 128 .
• mv_lora: int. LoRA dimension for value mixing. Defaults to 32 .
• aaa_lora: int. LoRA dimension for alpha parameters.Defaults to 64 .
• decay_lora: int. LoRA dimension for decay parameters.Defaults to 64 .
• 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.
• dropout_rate: float. Dropout rate for the dropout layer.

Examples

input_data = np.ones(shape=(1, 12), dtype="int32")


# Randomly initialized RWKV-7 decoder with custom config.
model = keras_hub.models.RWKV7Backbone(
    vocabulary_size=10,
    hidden_size=512,
    num_layers=2,
    head_size=64,
    intermediate_dim=1024,
    dtype="float32"
)
model(input_data)

[source]

from_preset method

RWKV7Backbone.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 ModelScope handle like 'modelscope://user/bert_base_en'
5. 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,
)

token_embedding property

keras_hub.models.RWKV7Backbone.token_embedding

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

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