metadata

license: apache-2.0
base_model: facebook/bart-large
tags:
  - generated_from_trainer
metrics:
  - rouge
  - wer
model-index:
  - name: bart_extractive_512_500
    results: []

bart_extractive_512_500

This model is a fine-tuned version of facebook/bart-large on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.9749
Rouge1: 0.7
Rouge2: 0.4441
Rougel: 0.6408
Rougelsum: 0.6409
Wer: 0.4458

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e-05
train_batch_size: 6
eval_batch_size: 6
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 2
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Wer
No log	0.13	250	1.2262	0.6523	0.3774	0.5876	0.5877	0.5064
2.0992	0.27	500	1.1233	0.6736	0.4029	0.6091	0.6091	0.4868
2.0992	0.4	750	1.1033	0.6826	0.4152	0.6187	0.6188	0.4768
1.1914	0.53	1000	1.0645	0.6812	0.4159	0.6178	0.618	0.4713
1.1914	0.66	1250	1.0493	0.6845	0.4206	0.6217	0.6219	0.4673
1.1319	0.8	1500	1.0348	0.6906	0.427	0.6292	0.6292	0.4649
1.1319	0.93	1750	1.0227	0.6893	0.4289	0.6286	0.6287	0.4596
1.0853	1.06	2000	1.0093	0.6898	0.4297	0.6298	0.6298	0.4575
1.0853	1.2	2250	1.0045	0.6981	0.4381	0.6376	0.6377	0.4547
0.9975	1.33	2500	0.9967	0.6964	0.4394	0.6368	0.6369	0.4511
0.9975	1.46	2750	0.9863	0.6995	0.4419	0.6401	0.6403	0.4495
0.997	1.6	3000	0.9844	0.7016	0.4441	0.642	0.6421	0.4483
0.997	1.73	3250	0.9819	0.6982	0.4431	0.6399	0.64	0.4476
0.9651	1.86	3500	0.9746	0.6994	0.4441	0.6404	0.6406	0.4456
0.9651	1.99	3750	0.9749	0.7	0.4441	0.6408	0.6409	0.4458

Framework versions

Transformers 4.38.2
Pytorch 2.2.1+cu121
Datasets 2.18.0
Tokenizers 0.15.2