sima_utils.transformer.model.whisper_decoder_pre_model

Classes

WhisperDecoderPreModel

Implementation for the pre cache model of Whisper's decoder.

Module Contents

class sima_utils.transformer.model.whisper_decoder_pre_model.WhisperDecoderPreModel

Implementation for the pre cache model of Whisper’s decoder.

This implements a simplified version of the LanguagePreModel. This model is only used when generating new tokens so the num_tokens is assumed to be 1.

num_tokens

Number of tokens. Set to a value greater than 1 to consume multiple input tokens in one model.

layer_idx

Transformer layer index.

num_tokens: int
layer_idx: int
gen_onnx_files()

Generates ONNX files.