Pegasus

This model was released on 2019-12-18 and added to Hugging Face Transformers on 2020-11-16.

Pegasus

Pegasus is an encoder-decoder (sequence-to-sequence) transformer model pretrained on unlabeled text to perform abstractive summarization. Pegasus is trained jointly on two self-supervised objective functions, masked language modeling (MLM) and gap sentence generation (GSG). Whole sentences are masked and the model has to fill in the gaps in the document. It can be fine-tuned with good performance even on small datasets with only 1000 examples.

You can find all the original Pegasus checkpoints under the Google organization.

Tip

Click on the Pegasus models in the right sidebar for more examples of how to apply Pegasus to different language tasks.

The example below demonstrates how to summarize text with Pipeline, AutoModel, and from the command line.

import torch
from transformers import pipeline

pipeline = pipeline(
    task="summarization",
    model="google/pegasus-xsum",
    dtype=torch.float16,
    device=0
)
pipeline("""Plants are remarkable organisms that produce their own food using a method called photosynthesis.
This process involves converting sunlight, carbon dioxide, and water into glucose, which provides energy for growth.
Plants play a crucial role in sustaining life on Earth by generating oxygen and serving as the foundation of most ecosystems.""")

import torch
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained(
    "google/pegasus-xsum"
)
model = AutoModelForSeq2SeqLM.from_pretrained(
    "google/pegasus-xsum",
    dtype=torch.float16,
    device_map="auto",
    attn_implementation="sdpa"
)

input_text = """Plants are remarkable organisms that produce their own food using a method called photosynthesis.
This process involves converting sunlight, carbon dioxide, and water into glucose, which provides energy for growth.
Plants play a crucial role in sustaining life on Earth by generating oxygen and serving as the foundation of most ecosystems."""
input_ids = tokenizer(input_text, return_tensors="pt").to(model.device)

output = model.generate(**input_ids, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))

echo -e "Plants are remarkable organisms that produce their own food using a method called photosynthesis. This process involves converting sunlight, carbon dioxide, and water into glucose, which provides energy for growth. Plants play a crucial role in sustaining life on Earth by generating oxygen and serving as the foundation of most ecosystems." | transformers run --task summarization --model google/pegasus-xsum --device 0

Quantization reduces the memory burden of large models by representing the weights in a lower precision. Refer to the Quantization overview for more available quantization backends.

The example below uses bitsandbytes to only quantize the weights to int4.

import torch
from transformers import BitsAndBytesConfig, AutoModelForSeq2SeqLM, AutoTokenizer

quantization_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_compute_dtype=torch.bfloat16,
    bnb_4bit_quant_type="nf4"
)
model = AutoModelForSeq2SeqLM.from_pretrained(
    "google/pegasus-xsum",
    dtype=torch.bfloat16,
    device_map="auto",
    quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained(
    "google/pegasus-xsum"
)
input_text = """Plants are remarkable organisms that produce their own food using a method called photosynthesis.
This process involves converting sunlight, carbon dioxide, and water into glucose, which provides energy for growth.
Plants play a crucial role in sustaining life on Earth by generating oxygen and serving as the foundation of most ecosystems."""
input_ids = tokenizer(input_text, return_tensors="pt").to(model.device)

output = model.generate(**input_ids, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))

Notes

• AdaFactor is the recommended optimizer for fine-tuning Pegasus.
• This implementation of Pegasus inherits from BartForConditionalGeneration but it uses static/sinusoidal positional embeddings instead. Pegasus also starts generating with pad_token_id as the prefix and uses num_beams=8.

PegasusConfig

[[autodoc]] PegasusConfig

PegasusTokenizer

warning: add_tokens does not work at the moment.

[[autodoc]] PegasusTokenizer

PegasusTokenizerFast

[[autodoc]] PegasusTokenizerFast

PegasusModel

[[autodoc]] PegasusModel - forward

PegasusForConditionalGeneration

[[autodoc]] PegasusForConditionalGeneration - forward

PegasusForCausalLM

[[autodoc]] PegasusForCausalLM - forward