summarization

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

Transformers >= 4.23.1
This model relies on a custom modeling file, you need to add trust_remote_code=True
See #13467

LSG ArXiv paper.
Github/conversion script is available at this link.

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline

tokenizer = AutoTokenizer.from_pretrained("ccdv/lsg-bart-base-4096-pubmed", trust_remote_code=True)
model = AutoModelForSeq2SeqLM.from_pretrained("ccdv/lsg-bart-base-4096-pubmed", trust_remote_code=True)

text = "Replace by what you want."
pipe = pipeline("text2text-generation", model=model, tokenizer=tokenizer, device=0)
generated_text = pipe(
  text, 
  truncation=True, 
  max_length=64, 
  no_repeat_ngram_size=7,
  num_beams=2,
  early_stopping=True
  )

ccdv/lsg-bart-base-4096-pubmed

This model is a fine-tuned version of ccdv/lsg-bart-base-4096 on the scientific_papers pubmed dataset.
It achieves the following results on the test set:

Length Sparse Type Block Size Sparsity Connexions R1 R2 RL RLsum
4096 Local 256 0 768 47.37 21.74 28.59 43.67
4096 Local 128 0 384 47.02 21.33 28.34 43.31
4096 Pooling 128 4 644 47.11 21.42 28.43 43.40
4096 Stride 128 4 644 47.16 21.49 28.38 43.44
4096 Block Stride 128 4 644 47.13 21.46 28.39 43.42
4096 Norm 128 4 644 47.09 21.44 28.40 43.36
4096 LSH 128 4 644 47.11 21.41 28.41 43.42

With smaller block size (lower ressources):

Length Sparse Type Block Size Sparsity Connexions R1 R2 RL RLsum
4096 Local 64 0 192 45.74 20.26 27.51 41.99
4096 Local 32 0 96 42.69 17.83 25.62 38.89
4096 Pooling 32 4 160 44.60 19.35 26.83 40.85
4096 Stride 32 4 160 45.52 20.07 27.39 41.75
4096 Block Stride 32 4 160 45.30 19.89 27.22 41.54
4096 Norm 32 4 160 44.30 19.05 26.57 40.47
4096 LSH 32 4 160 44.53 19.27 26.84 40.74

Model description

The model relies on Local-Sparse-Global attention to handle long sequences: attn

The model has about ~145 millions parameters (6 encoder layers - 6 decoder layers).
The model is warm started from BART-base, converted to handle long sequences (encoder only) and fine tuned.

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

Generate hyperparameters

The following hyperparameters were used during generation:

Framework versions