INNER CODE UNIT · Python

loss

barissayil/SentimentAnalysis · analyzer.py:107

            loss = criterion(input=logits.squeeze(-1), target=labels.float())
            # Backpropagate the loss.
            loss.backward()
            # Optimize the model.
            optimizer.step()

    # Saves analyzer.
    def save(self):
        # Save model.
        self.model.save_pretrained(save_directory=f"models/{self.output_dir}/")
        # Save configuration.
        self.config.save_pretrained(save_directory=f"models/{self.output_dir}/")
        # Save tokenizer.
        self.tokenizer.save_pretrained(save_directory=f"models/{self.output_dir}/")

    # Classifies sentiment as positve or negative.
    def classify_sentiment(self, text):
        # Don't track gradient.

View source record →