---
license: apache-2.0
base_model: distilgpt2
tags:
- generated_from_trainer
model-index:
- name: wikipedia-20230601.ace
  results: []
datasets:
- graelo/wikipedia
metrics:
- perplexity
---

<!-- 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. -->

# wikipedia-20230601.ace

This model is a fine-tuned version of [distilgpt2](https://huggingface.co/distilgpt2) on the graelo/wikipedia-20230601.ace dataset.
It achieves the following results on the evaluation set:
- Loss: 1.0119

## Model description

This model finetune distilgpt2 to Acehnese just for experiment purpose

## 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: 2e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3.0

### Training results

| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 2.2459        | 1.0   | 673  | 1.0254          |
| 1.7159        | 2.0   | 1346 | 1.0161          |
| 1.6392        | 3.0   | 2019 | 1.0119          |

### Perplexity

Datatest: `load_dataset("graelo/wikipedia", "20230601.ace", split="train[-10%:]")`

- original distilgpt2 : 40.5980
- this model (finetuned) : 3.9992

### Framework versions

- Transformers 4.32.0.dev0
- Pytorch 2.0.1+cu117
- Datasets 2.13.1
- Tokenizers 0.13.3