---
license: apache-2.0
base_model: google/mt5-small
tags:
- summarization
- generated_from_trainer
metrics:
- rouge
model-index:
- name: mt5-small-finetuned-genius
  results: []
pipeline_tag: summarization
datasets:
- miscjose/genius
---

# mt5-small-finetuned-genius

This model is a fine-tuned version of [google/mt5-small](https://huggingface.co/google/mt5-small) on the [Genius](https://genius.com/) Music dataset found [here](https://www.cs.cornell.edu/~arb/data/genius-expertise/). 
The song lyrics and song titles were preprocessed and used for fine-tuning. 

You can view more examples of this model's inference on the following [Space](https://huggingface.co/spaces/miscjose/genius_summarization_space).

## Model description

Please visit: [google/mt5-small](https://huggingface.co/google/mt5-small)

## Intended uses & limitations

- Intended Uses: Given song lyrics, generate a summary. 
- Limitations: Due to the nature of music, the model can generate summaries containing hate speech.

## Training and evaluation data

- 27.6K Training Samples
- 3.45 Validation Samples

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 4e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5

### Training results

| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum |
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|:------:|:---------:|
| 7.9304        | 1.0   | 863  | 3.5226          | 14.235 | 6.78   | 14.206 | 14.168    |
| 3.8394        | 2.0   | 1726 | 3.0382          | 22.97  | 13.166 | 22.981 | 22.944    |
| 3.3799        | 3.0   | 2589 | 2.9010          | 24.932 | 14.54  | 24.929 | 24.919    |
| 3.2204        | 4.0   | 3452 | 2.8441          | 26.678 | 15.587 | 26.624 | 26.665    |
| 3.1498        | 5.0   | 4315 | 2.8363          | **26.827** | **15.696** | **26.773** | **26.793**    |


### Framework versions

- Transformers 4.31.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.1
- Tokenizers 0.13.3