<!-- 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. -->
Pegasus-x-large-arxiv-split_20percent
This model is a fine-tuned version of google/pegasus-x-large on the arxiv-summarization dataset. It achieves the following results on the evaluation set:
- Loss: 1.9209
{'rouge1': 0.4107,
'rouge2': 0.1659,
'rougeL': 0.2489,
'rougeLsum': 0.3334}
We used
- train = load_dataset("ccdv/arxiv-summarization", split='train[:10%]')
- val = load_dataset("ccdv/arxiv-summarization", split='validation[:20%]') #for evaluation in train-loop
- test = load_dataset("ccdv/arxiv-summarization", split='test[:20%]') #for benchmarking model after train.
DatasetDict({
train: Dataset({
features: ['article', 'abstract'],
num_rows: 20304
})
validation: Dataset({
features: ['article', 'abstract'],
num_rows: 1287
})
test: Dataset({
features: ['article', 'abstract'],
num_rows: 1288
})
})
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: 5e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 16
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 600
- num_epochs: 3
Training results
Training Loss | Epoch | Step | Validation Loss |
---|---|---|---|
2.1435 | 0.47 | 600 | 2.0177 |
2.0767 | 0.95 | 1200 | 1.9686 |
2.022 | 1.42 | 1800 | 1.9473 |
2.0329 | 1.89 | 2400 | 1.9334 |
1.9514 | 2.36 | 3000 | 1.9261 |
1.9913 | 2.84 | 3600 | 1.9209 |
Framework versions
- Transformers 4.30.2
- Pytorch 2.0.1+cu117
- Datasets 2.13.1
- Tokenizers 0.13.3