def predict(self, data_csv_path: str, model_state_file):
torch.cuda.empty_cache()
saved_params = torch.load(model_state_file)
model = EncoderDecoderVisualTransformer(
sequence_length=self.sequence_length, vocab=self.vocab).to(device)
model.load_state_dict(saved_params['state_dict'])
model.eval()
dataframe = pd.read_csv(data_csv_path)
results_file_name = data_csv_path.replace(".csv", "_results.csv")
TrainingUtils.predict_on_dataset(model, dataframe, self.batch_size,
self.vocab, results_file_name)
def predict(self, data_csv_path: str, model_state_file):
torch.cuda.empty_cache()
saved_params = torch.load(model_state_file)
self.embed_size = saved_params['embed_size']
self.attention_dim = saved_params['attention_dim']
self.encoder_dim = saved_params['encoder_dim']
self.decoder_dim = saved_params['decoder_dim']
model = EncoderDecoder(embed_size=self.embed_size,
vocab_size=len(self.vocab),
attention_dim=self.attention_dim,
encoder_dim=self.encoder_dim,
decoder_dim=self.decoder_dim,
sequence_length=self.sequence_length,
vocab=self.vocab).to(device)
model.load_state_dict(saved_params['state_dict'])
model.eval()
dataframe = pd.read_csv(data_csv_path)
results_file_name = data_csv_path.replace(".csv", "_results.csv")
TrainingUtils.predict_on_dataset(model, dataframe, self.batch_size,
self.vocab, results_file_name)