echarlaix/bart-base-cnn-r2-19.4-d35-hybrid

🤗 Hugging Face sourcesummarizationapache-2.01.1 GBother✓ 2 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 echarlaix/bart-base-cnn-r2-19.4-d35-hybrid ./model-folder

Needs a seeder →

facebook/bart-base model fine-tuned on CNN/DailyMail

This model was created using the nn_pruning python library: the linear layers contains 35% of the original weights.

The model contains 53% of the original weights overall (the embeddings account for a significant part of the model, and they are not pruned by this method).

Fine-Pruning details

This model was fine-tuned from the HuggingFace model.

A side-effect of the block pruning is that some of the attention heads are completely removed: 38 heads were removed on a total of 216 (17.6%).

Details of the CNN/DailyMail dataset

Dataset	Split	# samples
CNN/DailyMail	train	287K
CNN/DailyMail	eval	13K

Results

Metric	# Value
Rouge 1	42.18
Rouge 2	19.44
Rouge L	39.17