Instructions to use RichardErkhov/akhooli_-_gpt2-small-arabic-4bits with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use RichardErkhov/akhooli_-_gpt2-small-arabic-4bits with Transformers:

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

pipe = pipeline("text-generation", model="RichardErkhov/akhooli_-_gpt2-small-arabic-4bits")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("RichardErkhov/akhooli_-_gpt2-small-arabic-4bits")
model = AutoModelForCausalLM.from_pretrained("RichardErkhov/akhooli_-_gpt2-small-arabic-4bits")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use RichardErkhov/akhooli_-_gpt2-small-arabic-4bits with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "RichardErkhov/akhooli_-_gpt2-small-arabic-4bits"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "RichardErkhov/akhooli_-_gpt2-small-arabic-4bits",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/RichardErkhov/akhooli_-_gpt2-small-arabic-4bits

SGLang

How to use RichardErkhov/akhooli_-_gpt2-small-arabic-4bits 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 "RichardErkhov/akhooli_-_gpt2-small-arabic-4bits" \
    --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": "RichardErkhov/akhooli_-_gpt2-small-arabic-4bits",
		"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 "RichardErkhov/akhooli_-_gpt2-small-arabic-4bits" \
        --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": "RichardErkhov/akhooli_-_gpt2-small-arabic-4bits",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use RichardErkhov/akhooli_-_gpt2-small-arabic-4bits with Docker Model Runner:
```
docker model run hf.co/RichardErkhov/akhooli_-_gpt2-small-arabic-4bits
```

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Quantization made by Richard Erkhov.

Github

Discord

Request more models

gpt2-small-arabic - bnb 4bits

Model creator: https://huggingface.co/akhooli/
Original model: https://huggingface.co/akhooli/gpt2-small-arabic/

Original model description:

language: "ar" datasets: - Arabic Wikipedia metrics: - none

GPT2-Small-Arabic

Model description

GPT2 model from Arabic Wikipedia dataset based on gpt2-small (using Fastai2).

Intended uses & limitations

How to use

An example is provided in this colab notebook. Both text and poetry (fine-tuned model) generation are included.

Limitations and bias

GPT2-small-arabic (trained on Arabic Wikipedia) has several limitations in terms of coverage (Arabic Wikipeedia quality, no diacritics) and training performance. Use as demonstration or proof of concepts but not as production code.

Training data

This pretrained model used the Arabic Wikipedia dump (around 900 MB).

Training procedure

Training was done using Fastai2 library on Kaggle, using free GPU.

Eval results

Final perplexity reached was 72.19, loss: 4.28, accuracy: 0.307

BibTeX entry and citation info

@inproceedings{Abed Khooli,
  year={2020}
}

Downloads last month: 3

Safetensors

Model size

0.1B params

Tensor type

F32

F16