Instructions to use mustaphaelkady/arabic-historical-ocr-1.0 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mustaphaelkady/arabic-historical-ocr-1.0 with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-VL-3B-Instruct")
model = PeftModel.from_pretrained(base_model, "mustaphaelkady/arabic-historical-ocr-1.0")

Transformers

How to use mustaphaelkady/arabic-historical-ocr-1.0 with Transformers:

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

pipe = pipeline("text-generation", model="mustaphaelkady/arabic-historical-ocr-1.0")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("mustaphaelkady/arabic-historical-ocr-1.0", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use mustaphaelkady/arabic-historical-ocr-1.0 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mustaphaelkady/arabic-historical-ocr-1.0"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mustaphaelkady/arabic-historical-ocr-1.0",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/mustaphaelkady/arabic-historical-ocr-1.0

SGLang

How to use mustaphaelkady/arabic-historical-ocr-1.0 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 "mustaphaelkady/arabic-historical-ocr-1.0" \
    --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": "mustaphaelkady/arabic-historical-ocr-1.0",
		"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 "mustaphaelkady/arabic-historical-ocr-1.0" \
        --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": "mustaphaelkady/arabic-historical-ocr-1.0",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use mustaphaelkady/arabic-historical-ocr-1.0 with Docker Model Runner:
```
docker model run hf.co/mustaphaelkady/arabic-historical-ocr-1.0
```

arabic-historical-ocr-1.0

This model is a fine-tuned version of Qwen/Qwen2.5-VL-3B-Instruct on the arabhist_ocr_train_v1 dataset. It achieves the following results on the evaluation set:

Loss: 1.4923

Model description

More information needed

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: 5e-05
train_batch_size: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 4
total_train_batch_size: 8
total_eval_batch_size: 2
optimizer: Use OptimizerNames.ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 0.05
num_epochs: 10.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.3714	0.2535	50	1.8164
2.2474	0.5070	100	1.8791
2.0310	0.7605	150	1.7697
1.7838	1.0101	200	1.7236
1.6217	1.2636	250	1.6629
1.5266	1.5171	300	1.6241
1.4800	1.7706	350	1.5940
1.4212	2.0203	400	1.5697
1.2989	2.2738	450	1.5681
1.2542	2.5272	500	1.5486
1.1763	2.7807	550	1.5173
1.1718	3.0304	600	1.5278
1.0947	3.2839	650	1.5202
1.0830	3.5374	700	1.5131
1.0671	3.7909	750	1.4961
1.0030	4.0406	800	1.5113
0.9156	4.2940	850	1.5067
0.9285	4.5475	900	1.4923
0.9160	4.8010	950	1.4996
0.8849	5.0507	1000	1.5226
0.8052	5.3042	1050	1.5199
0.8333	5.5577	1100	1.5141
0.8189	5.8112	1150	1.5036
0.7346	6.0608	1200	1.5354
0.7169	6.3143	1250	1.5426
0.6911	6.5678	1300	1.5454
0.7211	6.8213	1350	1.5392
0.6831	7.0710	1400	1.5729
0.6381	7.3245	1450	1.5762
0.6018	7.5779	1500	1.5727
0.6316	7.8314	1550	1.5750
0.6091	8.0811	1600	1.6014
0.5753	8.3346	1650	1.6046
0.5923	8.5881	1700	1.6060
0.6013	8.8416	1750	1.6076
0.5749	9.0913	1800	1.6127
0.5576	9.3447	1850	1.6180
0.5575	9.5982	1900	1.6191
0.5533	9.8517	1950	1.6190
0.5809	10.0	1980	1.6190

Framework versions

PEFT 0.18.1
Transformers 5.8.0
Pytorch 2.8.0+cu128
Datasets 4.0.0
Tokenizers 0.22.2

Downloads last month: 437

Model tree for mustaphaelkady/arabic-historical-ocr-1.0

Base model

Qwen/Qwen2.5-VL-3B-Instruct

Adapter

(269)

this model