<!-- 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-newsroom-cnn-adam8bit-bs16x64acc
This model is a fine-tuned version of oMateos2020/pegasus-newsroom-cnn_full-adafactor-bs6 on the cnn_dailymail dataset. It achieves the following results on the evaluation set:
- Loss: 2.8664
- Rouge1: 44.1788
- Rouge2: 21.453
- Rougel: 31.2643
- Rougelsum: 41.1116
- Gen Len: 71.7213
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: 6.4e-05
- train_batch_size: 4
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 64
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.25
- num_epochs: 1
- mixed_precision_training: Native AMP
- label_smoothing_factor: 0.1
Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
|---|---|---|---|---|---|---|---|---|
| 2.9449 | 1.0 | 1121 | 2.8664 | 44.1788 | 21.453 | 31.2643 | 41.1116 | 71.7213 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.12.0+cu113
- Datasets 2.4.0
- Tokenizers 0.12.1
