Error summarizing batch x: 'numpy.int32' object is not callable

Question

I am having trouble with transformer's FlauBERT module. I have been trying to summarize some texts in French in order to add them to a database but I've been getting this "'numpy.int32' object is not callable." error each time. As I am dealing with long texts, I am summarizing them in batches

I'm not sure what piece of the function calling for summarization is causing the problem. I imported both NumPy and TensorFlow for my code. Here's the snippets for both the function definition and the main loop in question:

Function definition:

...
def summarize_text_batch(batch, language):
    """
    Summarizes cleaned text. Procedure done by batches as mentionned below.

    Parameters:
    - batch (list): List of cleaned text segments.
    - language (str): Language identifier ('en' for English, 'fr' for French).

    Returns:
    - summaries (list): List of summarized text segments.
    """

    try:
        summaries = []

        if language == 'fr':
            tokenizer = FlaubertTokenizer.from_pretrained("jplu/tf-flaubert-base-cased")

        for i, text_segment in enumerate(tqdm(batch, desc="Summarizing")):
            if language == 'en':
                summarizer = pipeline("summarization", device=0, batch_size=1600)
                current_summary = summarizer(text_segment, max_length=13, min_length=2, length_penalty=2.0, num_beams=4, early_stopping=True)
            elif language == 'fr':
                inputs = tokenizer(text_segment, return_tensors='tf', max_length=512, truncation=True)
                outputs = flaubert_model(**inputs)
                summary_logits = outputs.logits
                summary_ids = tf.argmax(summary_logits, axis=-1)
                current_summary = tokenizer.decode(tf.convert_to_tensor(summary_ids[0].numpy().tolist()), skip_special_tokens=True)

            else:
                raise ValueError(f"Unsupported language: {language}")
            
            print(f"Batch {i + 1} summary: {current_summary}")
            summaries.append(current_summary)
    
    except Exception as e:
        print(f"Error summarizing batch {i + 1}: {e}.")
        summaries.append(None)
...

Main loop:

...
flaubert_model = load_flaubert_model("jplu/tf-flaubert-base-cased")

while True:
    pdf_path_str = validate_input("PDF filepath")
    organization = validate_input("Organization")
    document_type = validate_input("Document Type")
    category = validate_input("Category")
    clientele = validate_input("Clientele")
    language = validate_input("Language")

    try:
        pdf_path = Path(pdf_path_str)
        content = extract_text_from_pdf(pdf_path)
        
        # Cleaning text
        cleaned_content = clean_text(content)
        '''print(cleaned_content)''' # Debugging

        # Preprocessing text
        cleaned_content = preprocess_text(cleaned_content)
        
        # Split the cleaned content into smaller batches
        batch_size = 1600
        batches = [cleaned_content[i:i+batch_size] for i in range(0, len(cleaned_content), batch_size)]

        # Summarize each batch
        summaries = []

        for i, batch in enumerate(batches):
            print(f"Processing batch {i + 1}/{len(batches)}")
            current_summary = summarize_text_batch([batch], language)
            
            if current_summary and None not in current_summary[0]:
                summaries.extend(current_summary)

        # Flatten the nested list and then join
        summary = ' '.join([item for sublist in summaries for item in sublist and None not in sublist])

        with connect_to_database() as connection:
            insert_data(connection, organization, document_type, category, clientele, language, str(pdf_path), content, summary)

        print("Operation successful. Please proceed with the next document.")
    
    except OSError as e:
        print(f"Error processing PDF: {e}. Please enter a valid PDF path.")

I tried changing the current_summary variable inside the function definition in order to accept numpy variables.

Below are the modifications I made:

current_summary = tokenizer.decode(np.array(summary_ids[0]), skip_special_tokens=True)

current_summary = tokenizer.decode(summary_ids[0].numpy()), skip_special_tokens=True)

current_summary = tokenizer.decode(list(summary_ids[0]), skip_special_tokens=True)

current_summary = tokenizer.decode(summary_ids[0].numpy().tolist(), skip_special_tokens=True)

current_summary = tokenizer.decode(tf.convert_to_tensor(summary_ids[0].numpy().tolist()), skip_special_tokens=True)

All of them gave me the same error.