中文Sign up

AdamCodd/bart-large-xsum-samsum

🤗 Hugging Face sourcesummarizationapache-2.0406M params1.6 GBsafetensors✓ 6 checksumsupdated today
Submit in one command

Run it next to your model folder. It makes the torrent, checks your files against Hugging Face, and submits it. You just start seeding and paste your key from your account. It only reads your files and never changes them. Read the script first if you like.

curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo AdamCodd/bart-large-xsum-samsum ./model-folder
Needs a seeder →

bart-large-xsum-samsum

This model is a fine-tuned version of facebook/bart-large-xsum on the samsum dataset. It achieves the following results on the evaluation set:

  • Loss: 0.759
  • Rouge1: 54.3073
  • Rouge2: 29.0947
  • Rougel: 44.4676
  • Rougelsum: 49.895

Model description

This model tends to generate less verbose summaries compared to AdamCodd/bart-large-cnn-samsum, yet I find its quality to be superior (which is reflected in the metrics).

Intended uses & limitations

Suitable for summarizing dialogue-style text, it may not perform as well with other types of text formats.

from transformers import pipeline
summarizer = pipeline("summarization", model="AdamCodd/bart-large-xsum-samsum")

conversation = '''Emily: Hey Alex, have you heard about the new restaurant that opened downtown?
Alex: No, I haven't. What's it called?
Emily: It's called "Savory Bites." They say it has the best pasta in town.
Alex: That sounds delicious. When are you thinking of checking it out?
Emily: How about this Saturday? We can make it a dinner date.
Alex: Sounds like a plan, Emily. I'm looking forward to it.                                       
'''
result = summarizer(conversation)
print(result)

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 3e-05
  • train_batch_size: 4
  • eval_batch_size: 4
  • seed: 1270
  • optimizer: AdamW with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 150
  • num_epochs: 1

Training results

key value
eval_rouge1 54.3073
eval_rouge2 29.0947
eval_rougeL 44.4676
eval_rougeLsum 49.895

Framework versions

  • Transformers 4.35.0
  • Accelerate 0.24.1
  • Datasets 2.14.6
  • Tokenizers 0.14.3

If you want to support me, you can here.