espnet2.asr_transducer.encoder.validation.validate_input_block

source

espnet2.asr_transducer.encoder.validation.validate_input_block(configuration: Dict[str, Any], body_first_conf: Dict[str, Any], input_size: int) → int

Validate the input block configuration for the encoder architecture.

This function checks the validity of the input block configuration against the first body block configuration. It ensures that the parameters are correctly set and compatible with the specified block types and subsampling factors.

espnet2.asr_transducer.encoder.validation.configuration

Encoder input block configuration.

• Type: Dict[str, Any]

espnet2.asr_transducer.encoder.validation.body_first_conf

Encoder first body block configuration.

• Type: Dict[str, Any]

espnet2.asr_transducer.encoder.validation.input_size

Encoder input block input size.

• Type: int

• Parameters:

  • configuration – A dictionary containing the configuration for the encoder input block.
  • body_first_conf – A dictionary containing the configuration for the first body block of the encoder.
  • input_size – An integer representing the input size for the encoder input block.

• Returns: The output size of the encoder input block.

• Return type: int

• Raises:ValueError – If the subsampling factor is invalid for the specified block type or if the configuration parameters are missing or incompatible.

Examples

>>> input_config = {"vgg_like": True, "subsampling_factor": 4, "conv_size": 64}
>>> body_config = {"block_type": "conformer", "hidden_size": 128}
>>> output_size = validate_input_block(input_config, body_config, 256)
>>> print(output_size)
128

NOTE

• The function supports both VGG-like and standard Conv2D input modules.
• Valid subsampling factors differ based on the input module type.