| from .speech_encoder import WhisperWrappedEncoder, DualWrappedEncoder |
| import torch.nn as nn |
|
|
| def build_speech_encoder(config): |
| speech_encoder_type = getattr(config, 'speech_encoder_type', None) |
|
|
| print(f"Building speech encoder: {speech_encoder_type}") |
| if "whisper" in speech_encoder_type.lower(): |
| return WhisperWrappedEncoder.load(config) |
| elif "dual" in speech_encoder_type.lower(): |
| return DualWrappedEncoder(config) |
| elif "none" in speech_encoder_type.lower(): |
| return None |
| |
| raise ValueError(f'Unknown speech encoder: {speech_encoder_type}') |
|
|
