metadata

pipeline_tag: summarization
datasets:
  - samsum
language:
  - en
metrics:
  - rouge
library_name: transformers
widget:
  - text: >
      Rita: I'm so bloody tired. Falling asleep at work. :-(

      Tina: I know what you mean.

      Tina: I keep on nodding off at my keyboard hoping that the boss doesn't
      notice..

      Rita: The time just keeps on dragging on and on and on.... 

      Rita: I keep on looking at the clock and there's still 4 hours of this
      drudgery to go.

      Tina: Times like these I really hate my work.

      Rita: I'm really not cut out for this level of boredom.

      Tina: Neither am I.
  - text: >
      Beatrice: I am in town, shopping. They have nice scarfs in the shop next
      to the church. Do you want one?

      Leo: No, thanks

      Beatrice: But you don't have a scarf.

      Leo: Because I don't need it.

      Beatrice: Last winter you had a cold all the time. A scarf could help.

      Leo: I don't like them.

      Beatrice: Actually, I don't care. You will get a scarf.

      Leo: How understanding of you!

      Beatrice: You were complaining the whole winter that you're going to die.
      I've had enough.

      Leo: Eh.
  - text: |
      Jack: Cocktails later?
      May: YES!!!
      May: You read my mind...
      Jack: Possibly a little tightly strung today?
      May: Sigh... without question.
      Jack: Thought so.
      May: A little drink will help!
      Jack: Maybe two!
model-index:
  - name: bart-finetuned-samsum
    results:
      - task:
          name: Text Summarization
          type: summarization
        dataset:
          name: SamSum
          type: samsum
        metrics:
          - name: Validation ROUGE-1
            type: rouge-1
            value: 53.6163
          - name: Validation ROUGE-2
            type: rouge-2
            value: 28.914
          - name: Validation ROUGE-L
            type: rougeL
            value: 44.1443
          - name: Validation ROUGE-L Sum
            type: rougeLsum
            value: 49.2995

Description

This model was trained by fine-tuning the facebook/bart-large-xsum model using these parameters and the samsum dataset.

Development

Jupyter Notebook: Text Summarization With BART

Usage

from transformers import pipeline

model = pipeline("summarization", model="adedamolade26/bart-finetuned-samsum")

conversation = '''Jack: Cocktails later?
May: YES!!!
May: You read my mind...
Jack: Possibly a little tightly strung today?
May: Sigh... without question.
Jack: Thought so.
May: A little drink will help!
Jack: Maybe two!
'''
model(conversation)

Training Parameters

evaluation_strategy = "epoch",
save_strategy = 'epoch',
load_best_model_at_end = True,
metric_for_best_model = 'eval_loss',
seed = 42,
learning_rate=2e-5,
per_device_train_batch_size=4,
per_device_eval_batch_size=4,
gradient_accumulation_steps=2,
weight_decay=0.01,
save_total_limit=2,
num_train_epochs=4,
predict_with_generate=True,
fp16=True,
report_to="none"

References

Model Training process was adapted from Luis Fernando Torres's Kaggle Notebook: 📝 Text Summarization with Large Language Models