Longformer: The Long-Document Transformer
Paper • 2004.05150 • Published • 4
Summarization
Transformers
PyTorch
Safetensors
English
led
text2text-generation
Eval Results (legacy)
# Load model directly
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
tokenizer = AutoTokenizer.from_pretrained("AlgorithmicResearchGroup/led_base_16384_arxiv_summarization")
model = AutoModelForSeq2SeqLM.from_pretrained("AlgorithmicResearchGroup/led_base_16384_arxiv_summarization")Quick Links
Introduction
A led-base-16384 model to summarize ArXiv papers. Inputs are the abstracts of papers and full documents, and outputs are the summaries of the papers.
Allenai's Longformer Encoder-Decoder (LED).
As described in Longformer: The Long-Document Transformer by Iz Beltagy, Matthew E. Peters, Arman Cohan, led-base-16384 was initialized from bart-base since both models share the exact same architecture. To be able to process 16K tokens, bart-base's position embedding matrix was simply copied 16 times.
Rouge 2
| Type | Score |
|---|---|
precision |
0.1839148953011932 |
recall |
0.14904707945189774 |
fmeasure |
0.1580026685776864 |
- Downloads last month
- 37
Paper for AlgorithmicResearchGroup/led_base_16384_arxiv_summarization
Evaluation results
- ROUGE-1 on ccdv/arxiv-summarizationtest set verified37.325
- ROUGE-2 on ccdv/arxiv-summarizationtest set verified10.895
- ROUGE-L on ccdv/arxiv-summarizationtest set verified20.387
- ROUGE-LSUM on ccdv/arxiv-summarizationtest set verified33.301
- loss on ccdv/arxiv-summarizationtest set verified3.182
- gen_len on ccdv/arxiv-summarizationtest set verified145.590
# Use a pipeline as a high-level helper # Warning: Pipeline type "summarization" is no longer supported in transformers v5. # You must load the model directly (see below) or downgrade to v4.x with: # 'pip install "transformers<5.0.0' from transformers import pipeline pipe = pipeline("summarization", model="AlgorithmicResearchGroup/led_base_16384_arxiv_summarization")