Instructions to use mohammedkhas/customized-ar-translator with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mohammedkhas/customized-ar-translator with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
model = PeftModel.from_pretrained(base_model, "mohammedkhas/customized-ar-translator")

Transformers

How to use mohammedkhas/customized-ar-translator with Transformers:

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

pipe = pipeline("text-generation", model="mohammedkhas/customized-ar-translator")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("mohammedkhas/customized-ar-translator", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use mohammedkhas/customized-ar-translator with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mohammedkhas/customized-ar-translator"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mohammedkhas/customized-ar-translator",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/mohammedkhas/customized-ar-translator

SGLang

How to use mohammedkhas/customized-ar-translator 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 "mohammedkhas/customized-ar-translator" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mohammedkhas/customized-ar-translator",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "mohammedkhas/customized-ar-translator" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mohammedkhas/customized-ar-translator",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use mohammedkhas/customized-ar-translator with Docker Model Runner:
```
docker model run hf.co/mohammedkhas/customized-ar-translator
```

customized-ar-translator

This model is a fine-tuned version of Qwen/Qwen2.5-0.5B-Instruct on the trans_finetune_train dataset. It achieves the following results on the evaluation set:

Loss: 0.8289

Model description

language: - ar - en tags: - translation - technical-translation - education - nlp license: apache-2.0

🤖 Tech-En-Ar-Translator-Glossary

Model Description

This model is a specialized English-to-Arabic translator designed specifically for the IT and Technology domain.

Unlike standard translators, this model does not translate technical jargon (like "Cloud Computing," "Servers," "Infrastructure") into obscure Arabic terms. Instead, it:

Preserves Technical Terms: Keeps key English terminology within the Arabic sentence structure.
HTML-Ready Output: Wraps the text in HTML tags (dir="rtl", <span>) for immediate web rendering.
Auto-Glossary: Generates a separate "Explaining" section that defines the preserved English terms in Arabic.

It is designed for educational platforms, technical documentation, and learning management systems where students need to learn standard English IT terminology while reading in Arabic.

✨ Key Features

Smart Code-Switching: Fluent Arabic grammar mixed with English technical nouns.
Dual Output: Returns a JSON object containing the translated text and the explaining (glossary) text.
Frontend Friendly: Outputs sanitized HTML strings ready for integration into web apps (React, Vue, plain HTML).

🚀 Example Output

Input Text: > "Cloud computing is the delivery of IT resources including servers, storage, and databases over the internet."

Model Output (JSON):

'''json { "translated": "<div dir="rtl">إنترنت كمبيوتر (<span dir="ltr">Cloud computing) هو توصيل <span dir="ltr">IT resources، بما في ذلك موارد <span dir="ltr">servers و<span dir="ltr">storage و<span dir="ltr">databases، عبر الإنترنت مع أسعار <span dir="ltr">pay-as-you-go.", "explaining": "<div dir="rtl">Cloud computing: هي تقنية تسمح للمستخدمين ب Retrieving (توصيل) برامج وتطبيقات وبيانات...<div dir="rtl">servers: هي مكونات أساسية في نظام تشغيل كمبيوتر..." }'''

Intended uses & limitations

This model is fine-tuned specifically for IT contexts. It may not perform well on general conversational English (e.g., translating a novel or a poem).

The output includes HTML tags; if you need plain text, you will need to strip the tags post-processing.

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0001
train_batch_size: 3
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 3
total_train_batch_size: 9
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.251	0.2	10	1.1727
1.0906	0.4	20	1.0423
0.9856	0.6	30	0.9799
1.0287	0.8	40	0.9357
0.998	1.0	50	0.9081
0.7978	1.2	60	0.8900
0.7939	1.4	70	0.8629
0.7864	1.6	80	0.8529
0.7682	1.8	90	0.8424
0.7936	2.0	100	0.8330
0.6651	2.2	110	0.8357
0.6441	2.4	120	0.8340
0.6665	2.6	130	0.8300
0.6833	2.8	140	0.8287
0.691	3.0	150	0.8289

Framework versions

PEFT 0.17.1
Transformers 4.57.1
Pytorch 2.9.0+cu126
Datasets 4.0.0
Tokenizers 0.22.1

Downloads last month: 14

Model tree for mohammedkhas/customized-ar-translator

Base model

Qwen/Qwen2.5-0.5B

Finetuned

Qwen/Qwen2.5-0.5B-Instruct

Adapter

(855)

this model