BART (Bidirectional and Auto-Regressive Transformers) represents a robust sequence-to-sequence architecture ideal for document summarization. The facebook/bart-large-cnn variant, fine-tuned on the CNN/DailyMail corpus, delivers exceptional performance for news article condensation. Its bidirectional encoder and autoregressive decoder structure efficiently captures key information while maintaining contextual coherence.

Install required dependencies using the following command:

pip install transformers torch

Load the pre-trained model and tokenizer with standardized Auto classes:

from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

model_identifier = 'facebook/bart-large-cnn'
model = AutoModelForSeq2SeqLM.from_pretrained(model_identifier)
tokenizer = AutoTokenizer.from_pretrained(model_identifier)

Process source content with tokenization parameters optimized for summarization tasks:

input_text = """
Modern document summarization systems address information overload by distilling essential content from extended texts. These solutions leverage transformer architectures with attention mechanisms to identify salient phrases and maintain semantic relationships. The resulting summaries enable efficient knowledge extraction without loss of critical context.
"""

tokenized_input = tokenizer(
    input_text,
    max_length=1024,
    return_tensors='pt',
    truncation=True
)

Generate concise summaries using beam search parameters:

summary_tokens = model.generate(
    tokenized_input['input_ids'],
    max_length=140,
    min_length=35,
    length_penalty=1.8,
    num_beams=5,
    early_stopping=True
)

Convert generated token IDs to readable text:

generated_summary = tokenizer.decode(
    summary_tokens[0],
    skip_special_tokens=True
)

Parameter explanations for the generation process:

• max_length: Maximum output token count (140 tokens in this example)
• min_length: Minimum required summary length (35 tokens)
• length_penalty: Controls output length preference (higher values produce shorter summaries)
• num_beams: Beam width for search optimization (5 beams balance quality and speed)
• early_stopping: Halts generation once all beam paths complete

Complete implementation example:

from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

model_identifier = 'facebook/bart-large-cnn'
model = AutoModelForSeq2SeqLM.from_pretrained(model_identifier)
tokenizer = AutoTokenizer.from_pretrained(model_identifier)

input_text = """
Modern document summarization systems address information overload by distilling essential content from extended texts. These solutions leverage transformer architectures with attention mechanisms to identify salient phrases and maintain semantic relationships. The resulting summaries enable efficient knowledge extraction without loss of critical context.
"""

tokenized_input = tokenizer(
    input_text,
    max_length=1024,
    return_tensors='pt',
    truncation=True
)

summary_tokens = model.generate(
    tokenized_input['input_ids'],
    max_length=140,
    min_length=35,
    length_penalty=1.8,
    num_beams=5,
    early_stopping=True
)

generated_summary = tokenizer.decode(
    summary_tokens[0],
    skip_special_tokens=True
)