led-base-16384-finetuned-govreport

This model is a fine-tuned version of allenai/led-base-16384 on the pszemraj/govreport-summarization-8192 dataset. It achieves the following results on the evaluation set:

The rouge metrics calculations were processed later down the line (final notebook can be found HERE).

It achieved the following results on the validation set:

It achieved the following results on the test set:

Model description

As described in Longformer: The Long-Document Transformer by Iz Beltagy, Matthew E. Peters, Arman Cohan, Allenai's Longformer Encoder-Decoder (LED) 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.

This model is especially interesting for long-range summarization and question answering.

pszemraj/govreport-summarization-8192 is a pre-processed version of the dataset ccdv/govreport-summarization, which is a dataset for summarization of long documents adapted from this repository and this paper.

The Allenai's LED model was fine-tuned to this dataset, allowing the summarization of documents up to 16384 tokens.

The following hyperparameters were used during training:

Training Loss	Epoch	Step	Validation Loss
1.1492	0.24	250	1.4233
1.0077	0.49	500	1.3813
1.0069	0.73	750	1.3499
0.9639	0.98	1000	1.3216
0.7996	1.22	1250	1.3172
0.9395	1.46	1500	1.3003
0.913	1.71	1750	1.2919
0.8843	1.95	2000	1.2887