Instructions to use monsoon-nlp/dialect-ar-gpt-2021 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use monsoon-nlp/dialect-ar-gpt-2021 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="monsoon-nlp/dialect-ar-gpt-2021")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("monsoon-nlp/dialect-ar-gpt-2021")
model = AutoModelForCausalLM.from_pretrained("monsoon-nlp/dialect-ar-gpt-2021")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use monsoon-nlp/dialect-ar-gpt-2021 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "monsoon-nlp/dialect-ar-gpt-2021"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "monsoon-nlp/dialect-ar-gpt-2021",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/monsoon-nlp/dialect-ar-gpt-2021

SGLang

How to use monsoon-nlp/dialect-ar-gpt-2021 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "monsoon-nlp/dialect-ar-gpt-2021" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "monsoon-nlp/dialect-ar-gpt-2021",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "monsoon-nlp/dialect-ar-gpt-2021" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "monsoon-nlp/dialect-ar-gpt-2021",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use monsoon-nlp/dialect-ar-gpt-2021 with Docker Model Runner:
```
docker model run hf.co/monsoon-nlp/dialect-ar-gpt-2021
```

Dialect-AR-GPT-2021

Finetuned AraGPT-2 demo

This model started with AraGPT2-Medium, from AUB MIND Lab.

This model was then finetuned on dialect datasets from Qatar University, University of British Columbia / NLP, and Johns Hopkins University / LREC for 10 epochs.

You can use special tokens to prompt five dialects: [EGYPTIAN], [GULF], [LEVANTINE], [MAGHREBI], or [MSA], followed by a space.

from simpletransformers.language_generation import LanguageGenerationModel
model = LanguageGenerationModel("gpt2", "monsoon-nlp/dialect-ar-gpt-2021")
model.generate('[GULF] ' + "Ù…Ø¯ÙŠÙ†ØªÙŠ Ù‡ÙŠ", { 'max_length': 100 })

There is NO content filtering in the current version; do not use for public-facing text generation!

Training and Finetuning details

Original model: https://huggingface.co/aubmindlab/aragpt2-medium

I inserted new tokens into the tokenizer, finetuned the model on the dialect samples, and exported the new model.

Notebook: https://colab.research.google.com/drive/19C0zbkSCt5ncVCa4kY-ik9hSEiJcjI-F

Citations

AraGPT2 model:

@misc{antoun2020aragpt2,
      title={AraGPT2: Pre-Trained Transformer for Arabic Language Generation},
      author={Wissam Antoun and Fady Baly and Hazem Hajj},
      year={2020},
      eprint={2012.15520},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

Dialect data sources:

Downloads last month: 1,959

Safetensors

Model size

0.4B params

Tensor type

F32

Paper for monsoon-nlp/dialect-ar-gpt-2021

AraGPT2: Pre-Trained Transformer for Arabic Language Generation

Paper • 2012.15520 • Published Dec 31, 2020