Artifact AI
commited on
Commit
•
5306ad9
1
Parent(s):
f725ba3
add readme
Browse files
README.md
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: en
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
## Introduction
|
| 6 |
+
|
| 7 |
+
A led-large-16384 model to summarize ArXiv papers. Inputs are the abstracts of papers and full documents, and outputs are the summaries of the papers.
|
| 8 |
+
|
| 9 |
+
[Allenai's Longformer Encoder-Decoder (LED)](https://github.com/allenai/longformer#longformer).
|
| 10 |
+
|
| 11 |
+
As described in [Longformer: The Long-Document Transformer](https://arxiv.org/pdf/2004.05150.pdf) by Iz Beltagy, Matthew E. Peters, Arman Cohan,
|
| 12 |
+
*led-base-16384* was initialized from [*bart-base*](https://huggingface.co/facebook/bart-base) since both models share the exact same architecture. To
|
| 13 |
+
be able to process 16K tokens, *bart-base*'s position embedding matrix was simply copied 16 times.
|