Edit model card

bart-large-cnn-samsum

This model was trained using Amazon SageMaker and the new Hugging Face Deep Learning container. For more information look at:

Hyperparameters

{
"dataset_name": "samsum",
"do_eval": true,
"do_predict": true,
"do_train": true,
"fp16": true,
"learning_rate": 5e-05,
"model_name_or_path": "facebook/bart-large-cnn",
"num_train_epochs": 3,
"output_dir": "/opt/ml/model",
"per_device_eval_batch_size": 4,
"per_device_train_batch_size": 4,
"predict_with_generate": true,
"seed": 7

}

Usage

from transformers import pipeline
summarizer = pipeline("summarization", model="slauw87/bart-large-cnn-samsum")
conversation = '''Sugi: I am tired of everything in my life. 
Tommy: What? How happy you life is! I do envy you.
Sugi: You don't know that I have been over-protected by my mother these years. I am really about to leave the family and spread my wings.
Tommy: Maybe you are right.                                           
'''
nlp(conversation)

Results

key value
eval_rouge1 43.2111
eval_rouge2 22.3519
eval_rougeL 33.3153
eval_rougeLsum 40.0527
predict_rouge1 41.8283
predict_rouge2 20.9857
predict_rougeL 32.3602
predict_rougeLsum 38.7316
Downloads last month
4,025

Dataset used to train slauw87/bart_summarisation

Spaces using slauw87/bart_summarisation 3

Evaluation results

  • Validation ROGUE-1 on SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization
    self-reported
    43.211
  • Validation ROGUE-2 on SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization
    self-reported
    22.352
  • Validation ROGUE-L on SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization
    self-reported
    33.315
  • Test ROGUE-1 on SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization
    self-reported
    41.828
  • Test ROGUE-2 on SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization
    self-reported
    20.986
  • Test ROGUE-L on SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization
    self-reported
    32.360