Instructions to use datalab-to/surya-parsebench-rl with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use datalab-to/surya-parsebench-rl with Transformers:

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

pipe = pipeline("image-text-to-text", model="datalab-to/surya-parsebench-rl")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("datalab-to/surya-parsebench-rl")
model = AutoModelForMultimodalLM.from_pretrained("datalab-to/surya-parsebench-rl", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use datalab-to/surya-parsebench-rl with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "datalab-to/surya-parsebench-rl"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "datalab-to/surya-parsebench-rl",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/datalab-to/surya-parsebench-rl

SGLang

How to use datalab-to/surya-parsebench-rl 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 "datalab-to/surya-parsebench-rl" \
    --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": "datalab-to/surya-parsebench-rl",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "datalab-to/surya-parsebench-rl" \
        --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": "datalab-to/surya-parsebench-rl",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use datalab-to/surya-parsebench-rl with Docker Model Runner:
```
docker model run hf.co/datalab-to/surya-parsebench-rl
```

surya-parsebench-rl / code /bench.sh

vikp

Surya ParseBench RL checkpoint

6b4a31f 4 days ago

Raw History Blame Contribute Delete

1.97 kB

	#!/bin/bash
	# Score the Surya checkpoint currently served on :8000 on the four ParseBench
	# dimensions this project targets (charts excluded). Usage: [PORT=8000] [CONC=24] bench.sh <tag> [--test]
	# Results land in output/<tag>/ and a scorecard in output/<tag>/SCORECARD.txt.
	set -u
	cd ~/parsebench_rl
	TAG="$1"; shift
	EXTRA="${*:-}"
	export PYTHONPATH=$PWD SURYA_VLLM_URL=http://127.0.0.1:${PORT:-8000}/v1 SURYA_VLLM_MODEL=surya SURYA_PROMPT=${SURYA_PROMPT:-short} SURYA_FIGTEXT=${SURYA_FIGTEXT:-0}
	PB=.venv/bin/parse-bench
	OUT=output/$TAG
	for g in text_formatting text_content table layout; do
	$PB run surya_vllm --group $g --output_dir output_run_$TAG --max_concurrent ${CONC:-24} \
	--open_report False $EXTRA > output_run_$TAG.$g.log 2>&1
	done
	mkdir -p $OUT
	cp -r output_run_$TAG/surya_vllm/* $OUT/ 2>/dev/null
	for g in text_formatting text_content table; do
	$PB evaluation run $OUT --group $g --test_cases_dir data > /dev/null 2>&1
	cp $OUT/_evaluation_report.json $OUT/_rep_$g.json
	done
	$PB evaluation run $OUT --group layout --product_type layout_detection --test_cases_dir data > /dev/null 2>&1
	cp $OUT/_evaluation_report.json $OUT/_rep_layout.json
	.venv/bin/python - "$OUT" <<'PY' \| tee $OUT/SCORECARD.txt
	import json, sys
	out = sys.argv[1]
	keys = [("table", "avg_grits_trm_composite", "Tables"),
	("text_content", "avg_content_faithfulness", "Content"),
	("text_formatting", "avg_semantic_formatting", "SemFmt"),
	("layout", "avg_layout_element_rule_pass_rate", "VisGnd")]
	vals = []
	for g, k, lab in keys:
	try:
	r = json.load(open(f"{out}/_rep_{g}.json"))
	v = r["aggregate_metrics"].get(k, 0) * 100
	n = r.get("successful"); f = r.get("failed")
	except Exception as e:
	v, n, f = float("nan"), "?", str(e)[:40]
	vals.append(v)
	print(f"{lab:8s} {v:6.2f} (ok={n} failed={f})")
	print(f"{'MEAN4':8s} {sum(vals)/4:6.2f} (LlamaParse Agentic Plus = 89.21, Surya OCR 2 published = 75.55)")
	PY