library_name: transformers
base_model: google/pegasus-cnn_dailymail
tags:
- generated_from_trainer
model-index:
- name: pegasus-samsum
results: []
pegasus-samsum
This model is a fine-tuned version of google/pegasus-cnn_dailymail on the SAMsum dataset. It achieves the following results on the evaluation set:
- Loss: 1.3986
Model description
The model uses PEGASUS pretrained on cnn-dailymail and it is fine-tuned on the SAMsum dataset in order to get summaries out of conversations
Intended uses & limitations
Summarization on conversations
Training and evaluation data
The training args are the following: num_train_epochs=1, warmup_steps=500, per_device_train_batch_size=1, per_gpu_eval_batch_size=1, weight_decay=0.01, logging_steps=10, push_to_hub=True, evaluation_strategy='steps', eval_steps=500, save_steps=1e6, gradient_accumulation_steps=16, remove_unused_columns=False,
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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 1.5449 | 0.5431 | 500 | 1.3986 |
Framework versions
- Transformers 4.46.3
- Pytorch 2.4.0
- Datasets 3.1.0
- Tokenizers 0.20.3