# Whisper max audio length 
MAX_DURATION_IN_SECONDS = 30.0
MAX_INPUT_LENGTH = MAX_DURATION_IN_SECONDS * 16000
MAX_LABEL_LENGTH = 448

# Add these functions to your code

def prepare_dataset(batch):

    audio = batch["audio"]

    batch["input_features"] = feature_extractor(audio["array"], sampling_rate=audio["sampling_rate"]).input_features[0]

    batch["labels"] = tokenizer(batch["transcript"]).input_ids

    batch["input_length"] = len(batch["audio"])

    batch["labels_length"] = len(tokenizer(batch["transcript"], add_special_tokens=False).input_ids)

    return batch

def filter_inputs(input_length):

    """Filter inputs with zero input length or longer than 30s"""

    return 0 < input_length < MAX_INPUT_LENGTH

def filter_labels(labels_length):

    """Filter label sequences longer than max length (448)"""

    return labels_length < MAX_LABEL_LENGTH 


...

afrispeech = afrispeech.map(prepare_dataset)
afrispeech = afrispeech.filter(filter_inputs, input_columns=["input_length"])
afrispeech = afrispeech.filter(filter_labels, input_columns=["labels_length"])
afrispeech = afrispeech.remove_columns(['labels_length', 'input_length'])

...