Args: model: nn.Module mlp: bool, if True, then use the MLP layer as the linear probe module freeze: bool

Toy Dataset for testing the audioset input with text labels Parameters ---------- index_path: str the link

r"""AudioSet evaluator. Args: audios_dir (str): directory of evaluation segments classes_num (int): the number of sou

r"""Clotho evaluator. Returns: None

r"""VGGSound evaluator. Args: data_recipe (str): dataset split, 'yan' Returns: None

r"""AudioCaps evaluator. Args: query (str): type of query, 'caption' or 'labels' Returns: None

r"""ESC-50 evaluator. Returns: None

add model config path or file and update registry

Forward one audio and produce the audio embedding Parameters ---------- audio: (audio_length) the time-domain au

r"""Collate mini-batch data to inputs and targets for training. Args: list_data_dict: e.g., [ { 'text': 'a so

Collate function for wdsdataloader. batch: a list of dict, each dict is a sample

r"""Configure optimizer.

r"""Get constant warm up scheduler for LambdaLR. Args: step (int): global step warm_up_steps (int): steps for warm up red

Check if dataset exists

r"""Forward data into the module. Args: input_tensor: (batch_size, input_feature_maps, time_steps, freq_bins) Returns:

r"""Forward data into the module. Args: input_tensor: (batch_size, input_feature_maps, time_steps, freq_bins) Returns:

r"""Forward data into the module. Args: input_tensor: (batch_size, input_feature_maps, downsampled_time_steps, downsampled_freq_b

Args: input: (batch_size, segment_samples, channels_num) Outputs: output_dict: { 'wav': (batch_size,

Input: (batch_size, data_length)

Input: (batch_size, data_length)

Input: (batch_size, data_length)

Args: x: waveform, torch.tensor [batch, t_samples] / batch of mel_spec and longer list mix_lambda: torch.tensor [batc

Args: x: input features with shape of (num_windows*B, N, C) mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh

x: B, H*W, C