Instructions to use gagan3012/TrOCR-Ar-Small with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use gagan3012/TrOCR-Ar-Small with Transformers:

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

pipe = pipeline("image-text-to-text", model="gagan3012/TrOCR-Ar-Small")

# Load model directly
from transformers import AutoTokenizer, AutoModelForImageTextToText

tokenizer = AutoTokenizer.from_pretrained("gagan3012/TrOCR-Ar-Small")
model = AutoModelForImageTextToText.from_pretrained("gagan3012/TrOCR-Ar-Small")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use gagan3012/TrOCR-Ar-Small with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "gagan3012/TrOCR-Ar-Small"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "gagan3012/TrOCR-Ar-Small",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/gagan3012/TrOCR-Ar-Small

SGLang

How to use gagan3012/TrOCR-Ar-Small 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 "gagan3012/TrOCR-Ar-Small" \
    --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": "gagan3012/TrOCR-Ar-Small",
		"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 "gagan3012/TrOCR-Ar-Small" \
        --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": "gagan3012/TrOCR-Ar-Small",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use gagan3012/TrOCR-Ar-Small with Docker Model Runner:
```
docker model run hf.co/gagan3012/TrOCR-Ar-Small
```

what preprocessor should I use to train the handwritten arabic ocr on this base of ArOCR model? #1

by HGamal - opened Sep 4, 2022

Discussion

HGamal

Sep 4, 2022

Hey, I'm trying to train a handwritten arabic OCR using ArOCR and trOCR-Ar-small, but the preprocessor doesn't load so I used "giganticode/roberta-base-ar_miner" as the tokenizer.
When I check the validation set predictions it's all like a repeated garbage as shown.

here's my loadings:
def load_model(from_disk: bool) -> VisionEncoderDecoderModel:
model: VisionEncoderDecoderModel = VisionEncoderDecoderModel.from_pretrained('gagan3012/TrOCR-Ar-Small')#.from_encoder_decoder_pretrained("google/vit-base-patch16-224-in21k", "giganticode/roberta-base-ar_miner")#
print(f"Using device {device}.")
model.to(device)
return model

def init_model_for_training(model: VisionEncoderDecoderModel, processor: TrOCRProcessor):
model.config.decoder_start_token_id = processor.tokenizer.cls_token_id
model.config.pad_token_id = processor.tokenizer.pad_token_id
model.config.vocab_size = model.config.decoder.vocab_size
model.config.bos_token_id = processor.tokenizer.bos_token_id
model.config.decoder_start_token_id = 0
model.config.decoder.is_decoder = True
model.config.decoder.add_cross_attention = True

def load_processor() -> TrOCRProcessor:
feature_extractor=ViTFeatureExtractor.from_pretrained("google/vit-base-patch16-384")
model_path = "giganticode/roberta-base-ar_miner"
tokenizer = AutoTokenizer.from_pretrained(model_path)
return TrOCRProcessor(feature_extractor=feature_extractor, tokenizer=tokenizer)

adilJ

Nov 5, 2023

Try:
processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-handwritten")

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment