metadata
license: apache-2.0
base_model: google/mt5-small
tags:
- generated_from_trainer
datasets:
- kde4
metrics:
- rouge
- sacrebleu
model-index:
- name: mt5_small_kde4_en_ko
results:
- task:
name: Sequence-to-sequence Language Modeling
type: text2text-generation
dataset:
name: kde4
type: kde4
config: en-ko
split: train
args: en-ko
metrics:
- name: Rouge1
type: rouge
value: 0.0832
- name: Sacrebleu
type: sacrebleu
value: 3.3559
mt5_small_kde4_en_ko
This model is a fine-tuned version of google/mt5-small on the kde4 dataset. It achieves the following results on the evaluation set:
- Loss: 3.1644
- Rouge1: 0.0832
- Rouge2: 0.0195
- Rougel: 0.0826
- Sacrebleu: 3.3559
Model description
This model tries to achieve translation from English to Korean using google's mt5 multilingual model.
Intended uses & limitations
Translation from English to Korean
Usage
You can use this model directly with a pipeline for translation language modeling:
>>> from transformers import pipeline
>>> translator = pipeline('translation', model='chunwoolee0/ke_t5_base_bongsoo_en_ko')
>>> translator("Let us go for a walk after lunch.")
[{'translation_text': '오류를 방문하십시오.'}]
The translation fails completely.
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3
### Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Sacrebleu |
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|:------:|:---------:|
| 15.8735 | 0.46 | 500 | 6.5322 | 0.0101 | 0.0004 | 0.0102 | 0.464 |
| 7.183 | 0.93 | 1000 | 4.2298 | 0.0203 | 0.0012 | 0.02 | 0.6102 |
| 5.4447 | 1.39 | 1500 | 3.5600 | 0.0399 | 0.005 | 0.0396 | 1.5798 |
| 4.8372 | 1.85 | 2000 | 3.3343 | 0.0537 | 0.0088 | 0.0533 | 3.0115 |
| 4.5579 | 2.32 | 2500 | 3.2131 | 0.0732 | 0.016 | 0.0729 | 3.3743 |
| 4.4532 | 2.78 | 3000 | 3.1644 | 0.0832 | 0.0195 | 0.0826 | 3.3559 |
### Framework versions
- Transformers 4.32.0
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3