---
library_name: transformers
license: apache-2.0
base_model: facebook/bart-large
tags:
- generated_from_trainer
metrics:
- rouge
model-index:
- name: bart-large-xsum-rouge-3-loss-differentiable-100-cnt-supervised
  results: []
---

<!-- 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. -->

# bart-large-xsum-rouge-3-loss-differentiable-100-cnt-supervised

This model is a fine-tuned version of [facebook/bart-large](https://huggingface.co/facebook/bart-large) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 1.8086
- Rouge1: 0.376
- Rouge2: 0.1514
- Rougel: 0.29
- Rougelsum: 0.2903

## 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: 1e-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.06
- num_epochs: 20

### Training results

| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum |
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|:------:|:---------:|
| 2.7975        | 0.8   | 20   | 2.0561          | 0.2159 | 0.0478 | 0.1473 | 0.1473    |
| 2.0853        | 1.6   | 40   | 1.8086          | 0.376  | 0.1514 | 0.29   | 0.2903    |
| 1.7718        | 2.4   | 60   | 1.8573          | 0.3679 | 0.1464 | 0.2872 | 0.2876    |
| 1.4964        | 3.2   | 80   | 1.9286          | 0.3675 | 0.1456 | 0.2857 | 0.286     |
| 1.1544        | 4.0   | 100  | 1.9857          | 0.3638 | 0.1439 | 0.2855 | 0.2858    |


### Framework versions

- Transformers 4.57.3
- Pytorch 2.9.1+cu128
- Datasets 3.6.0
- Tokenizers 0.22.1