---
library_name: peft
license: apache-2.0
base_model: google/mt5-small
tags:
- base_model:adapter:google/mt5-small
- lora
- transformers
model-index:
- name: mt5-small-GEC-spanish-LORA-merged
  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. -->

# mt5-small-GEC-spanish-LORA-merged

This model is a fine-tuned version of [google/mt5-small](https://huggingface.co/google/mt5-small) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1016
- Gleu: 0.7352

## 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: 0.001635475313478006
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Use OptimizerNames.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: 0.0993688388184371
- num_epochs: 2

### Training results

| Training Loss | Epoch | Step  | Validation Loss | Gleu   |
|:-------------:|:-----:|:-----:|:---------------:|:------:|
| 0.2301        | 1.0   | 11213 | 0.1175          | 0.7102 |
| 0.1762        | 2.0   | 22426 | 0.1016          | 0.7352 |


### Framework versions

- PEFT 0.18.1
- Transformers 5.0.0
- Pytorch 2.7.0+cu126
- Datasets 4.8.5
- Tokenizers 0.22.2