---
library_name: transformers
license: mit
base_model: openai-community/gpt2
tags:
  - generated_from_trainer
model-index:
  - name: arabic-nano-gpt-v2
    results: []
datasets:
  - wikimedia/wikipedia
language:
  - ar
---

# arabic-nano-gpt-v2

This model is a fine-tuned version of [openai-community/gpt2](https://huggingface.co/openai-community/gpt2) on the arabic [wikimedia/wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia) dataset.

Repository on GitHub: [e-hossam96/arabic-nano-gpt](https://github.com/e-hossam96/arabic-nano-gpt.git)

The model achieves the following results on the held-out test set:

- Loss: 3.25564

## How to Use

```python
import torch
from transformers import pipeline

model_ckpt = "e-hossam96/arabic-nano-gpt-v2"
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")


lm = pipeline(task="text-generation", model=model_ckpt, device=device)

prompt = """المحرك النفاث هو محرك ينفث الموائع (الماء أو الهواء) بسرعة فائقة \
لينتج قوة دافعة اعتمادا على مبدأ قانون نيوتن الثالث للحركة. \
هذا التعريف الواسع للمحركات النفاثة يتضمن أيضا"""

output = lm(prompt, max_new_tokens=128)

print(output[0]["generated_text"])
```

## Model description

- Embedding Size: 384
- Attention Heads: 6
- Attention Layers: 8

## Training and evaluation data

The entire wikipedia dataset was split into three splits based on the 90-5-5 ratios.

## Training hyperparameters

The following hyperparameters were used during training:

- learning_rate: 0.0001
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.01
- num_epochs: 8

## Training Loss

![Training Loss](assets/arabic-nano-gpt-v2-train-loss.png)

## Validation Loss

![Validation Loss](assets/arabic-nano-gpt-v2-eval-loss.png)

## Framework versions

- Transformers 4.45.2
- Pytorch 2.5.0
- Datasets 3.0.1
- Tokenizers 0.20.1