metadata
library_name: keras-hub
This is a Moonshine model uploaded using the KerasHub library and can be used with JAX, TensorFlow, and PyTorch backends.
This model is related to a AudioToText task.
Model config:
- name: moonshine_backbone_1
- trainable: True
- vocabulary_size: 32768
- filter_dim: 288
- encoder_num_layers: 6
- decoder_num_layers: 6
- hidden_dim: 288
- intermediate_dim: 1152
- encoder_num_heads: 8
- decoder_num_heads: 8
- feedforward_expansion_factor: 4
- encoder_use_swiglu_activation: False
- decoder_use_swiglu_activation: True
- max_position_embeddings: 194
- pad_head_dim_to_multiple_of: None
- partial_rotary_factor: 0.9
- dropout: 0.0
- initializer_range: 0.02
- rope_theta: 10000.0
- attention_bias: False
- attention_dropout: 0.0
- dtype: float32
This model card has been generated automatically and should be completed by the model author. See Model Cards documentation for more information.