Predictor
class Predictor
Performs model predictions on a batch of inputs.
__init__
def __init__(model, preprocessor)
Initializes a Predictor object with a trained transformer model a preprocessor.
Args
-
model (Model): Trained transformer model.
-
preprocessor (Preprocessor): Preprocessor corresponding to the model configuration.
__call__
def __call__(words, lang, batch_size)
Predicts phonemes for a list of words.
Args
-
words (list): List of words to predict.
-
lang (str): Language of texts.
-
batch_size (int): Size of batch for model input to speed up inference.
Returns
- List[Prediction]: A list of result objects containing (word, phonemes, phoneme_tokens, token_probs, confidence)
from_checkpoint
def from_checkpoint(cls, checkpoint_path, device)
Initializes the predictor from a checkpoint (.pt file).
Args
-
checkpoint_path (str): Path to the checkpoint file (.pt).
-
device (str): Device to load the model on ('cpu' or 'cuda'). (Default value = 'cpu').
Returns
- Predictor: Predictor object.