VLM comparison
Compare different VLM models by running the VLM pipeline and timing outputs.
What this example does
- Iterates through a list of VLM model configurations and converts the same file.
- Prints per-page generation times and saves JSON/MD/HTML to
scratch/. - Summarizes total inference time and pages processed in a table.
Requirements
- Install
tabulatefor pretty printing (pip install tabulate).
Prerequisites
- Install Docling with VLM extras. Ensure models can be downloaded or are available.
How to run
- From the repo root:
python docs/examples/compare_vlm_models.py. - Results are saved to
scratch/with filenames including the model and framework.
Notes
- MLX models are skipped automatically on non-macOS platforms.
- On CUDA systems, you can enable flash_attention_2 (see commented lines).
- Running multiple VLMs can be GPU/CPU intensive and time-consuming; ensure enough VRAM/system RAM and close other memory-heavy apps.
In [ ]:
Copied!
import json
import sys
import time
from pathlib import Path
from docling_core.types.doc import DocItemLabel, ImageRefMode
from docling_core.types.doc.document import DEFAULT_EXPORT_LABELS
from tabulate import tabulate
from docling.datamodel import vlm_model_specs
from docling.datamodel.accelerator_options import AcceleratorDevice
from docling.datamodel.base_models import InputFormat
from docling.datamodel.pipeline_options import (
VlmPipelineOptions,
)
from docling.datamodel.pipeline_options_vlm_model import (
InferenceFramework,
InlineVlmOptions,
ResponseFormat,
TransformersModelType,
TransformersPromptStyle,
)
from docling.document_converter import DocumentConverter, PdfFormatOption
from docling.pipeline.vlm_pipeline import VlmPipeline
def convert(sources: list[Path], converter: DocumentConverter):
# Note: this helper assumes a single-item `sources` list. It returns after
# processing the first source to keep runtime/output focused.
model_id = pipeline_options.vlm_options.repo_id.replace("/", "_")
framework = pipeline_options.vlm_options.inference_framework
for source in sources:
print("================================================")
print("Processing...")
print(f"Source: {source}")
print("---")
print(f"Model: {model_id}")
print(f"Framework: {framework}")
print("================================================")
print("")
res = converter.convert(source)
print("")
fname = f"{res.input.file.stem}-{model_id}-{framework}"
inference_time = 0.0
for i, page in enumerate(res.pages):
inference_time += page.predictions.vlm_response.generation_time
print("")
print(
f" ---------- Predicted page {i} in {pipeline_options.vlm_options.response_format} in {page.predictions.vlm_response.generation_time} [sec]:"
)
print(page.predictions.vlm_response.text)
print(" ---------- ")
print("===== Final output of the converted document =======")
# Manual export for illustration. Below, `save_as_json()` writes the same
# JSON again; kept intentionally to show both approaches.
with (out_path / f"{fname}.json").open("w") as fp:
fp.write(json.dumps(res.document.export_to_dict()))
res.document.save_as_json(
out_path / f"{fname}.json",
image_mode=ImageRefMode.PLACEHOLDER,
)
print(f" => produced {out_path / fname}.json")
res.document.save_as_markdown(
out_path / f"{fname}.md",
image_mode=ImageRefMode.PLACEHOLDER,
)
print(f" => produced {out_path / fname}.md")
res.document.save_as_html(
out_path / f"{fname}.html",
image_mode=ImageRefMode.EMBEDDED,
labels=[*DEFAULT_EXPORT_LABELS, DocItemLabel.FOOTNOTE],
split_page_view=True,
)
print(f" => produced {out_path / fname}.html")
pg_num = res.document.num_pages()
print("")
print(
f"Total document prediction time: {inference_time:.2f} seconds, pages: {pg_num}"
)
print("====================================================")
return [
source,
model_id,
str(framework),
pg_num,
inference_time,
]
if __name__ == "__main__":
sources = [
"tests/data/pdf/2305.03393v1-pg9.pdf",
]
out_path = Path("scratch")
out_path.mkdir(parents=True, exist_ok=True)
## Definiton of more inline models
llava_qwen = InlineVlmOptions(
repo_id="llava-hf/llava-interleave-qwen-0.5b-hf",
# prompt="Read text in the image.",
prompt="Convert this page to markdown. Do not miss any text and only output the bare markdown!",
# prompt="Parse the reading order of this document.",
response_format=ResponseFormat.MARKDOWN,
inference_framework=InferenceFramework.TRANSFORMERS,
transformers_model_type=TransformersModelType.AUTOMODEL_IMAGETEXTTOTEXT,
supported_devices=[AcceleratorDevice.CUDA, AcceleratorDevice.CPU],
scale=2.0,
temperature=0.0,
)
# Note that this is not the expected way of using the Dolphin model, but it shows the usage of a raw prompt.
dolphin_oneshot = InlineVlmOptions(
repo_id="ByteDance/Dolphin",
prompt="<s>Read text in the image. <Answer/>",
response_format=ResponseFormat.MARKDOWN,
inference_framework=InferenceFramework.TRANSFORMERS,
transformers_model_type=TransformersModelType.AUTOMODEL_IMAGETEXTTOTEXT,
transformers_prompt_style=TransformersPromptStyle.RAW,
supported_devices=[AcceleratorDevice.CUDA, AcceleratorDevice.CPU],
scale=2.0,
temperature=0.0,
)
## Use VlmPipeline
pipeline_options = VlmPipelineOptions()
pipeline_options.generate_page_images = True
## On GPU systems, enable flash_attention_2 with CUDA:
# pipeline_options.accelerator_options.device = AcceleratorDevice.CUDA
# pipeline_options.accelerator_options.cuda_use_flash_attention2 = True
vlm_models = [
## DocTags / SmolDocling models
vlm_model_specs.SMOLDOCLING_MLX,
vlm_model_specs.SMOLDOCLING_TRANSFORMERS,
## Markdown models (using MLX framework)
vlm_model_specs.QWEN25_VL_3B_MLX,
vlm_model_specs.PIXTRAL_12B_MLX,
vlm_model_specs.GEMMA3_12B_MLX,
## Markdown models (using Transformers framework)
vlm_model_specs.GRANITE_VISION_TRANSFORMERS,
vlm_model_specs.PHI4_TRANSFORMERS,
vlm_model_specs.PIXTRAL_12B_TRANSFORMERS,
## More inline models
dolphin_oneshot,
llava_qwen,
]
# Remove MLX models if not on Mac
if sys.platform != "darwin":
vlm_models = [
m for m in vlm_models if m.inference_framework != InferenceFramework.MLX
]
rows = []
for vlm_options in vlm_models:
pipeline_options.vlm_options = vlm_options
## Set up pipeline for PDF or image inputs
converter = DocumentConverter(
format_options={
InputFormat.PDF: PdfFormatOption(
pipeline_cls=VlmPipeline,
pipeline_options=pipeline_options,
),
InputFormat.IMAGE: PdfFormatOption(
pipeline_cls=VlmPipeline,
pipeline_options=pipeline_options,
),
},
)
row = convert(sources=sources, converter=converter)
rows.append(row)
print(
tabulate(
rows, headers=["source", "model_id", "framework", "num_pages", "time"]
)
)
print("see if memory gets released ...")
time.sleep(10)
import json
import sys
import time
from pathlib import Path
from docling_core.types.doc import DocItemLabel, ImageRefMode
from docling_core.types.doc.document import DEFAULT_EXPORT_LABELS
from tabulate import tabulate
from docling.datamodel import vlm_model_specs
from docling.datamodel.accelerator_options import AcceleratorDevice
from docling.datamodel.base_models import InputFormat
from docling.datamodel.pipeline_options import (
VlmPipelineOptions,
)
from docling.datamodel.pipeline_options_vlm_model import (
InferenceFramework,
InlineVlmOptions,
ResponseFormat,
TransformersModelType,
TransformersPromptStyle,
)
from docling.document_converter import DocumentConverter, PdfFormatOption
from docling.pipeline.vlm_pipeline import VlmPipeline
def convert(sources: list[Path], converter: DocumentConverter):
# Note: this helper assumes a single-item `sources` list. It returns after
# processing the first source to keep runtime/output focused.
model_id = pipeline_options.vlm_options.repo_id.replace("/", "_")
framework = pipeline_options.vlm_options.inference_framework
for source in sources:
print("================================================")
print("Processing...")
print(f"Source: {source}")
print("---")
print(f"Model: {model_id}")
print(f"Framework: {framework}")
print("================================================")
print("")
res = converter.convert(source)
print("")
fname = f"{res.input.file.stem}-{model_id}-{framework}"
inference_time = 0.0
for i, page in enumerate(res.pages):
inference_time += page.predictions.vlm_response.generation_time
print("")
print(
f" ---------- Predicted page {i} in {pipeline_options.vlm_options.response_format} in {page.predictions.vlm_response.generation_time} [sec]:"
)
print(page.predictions.vlm_response.text)
print(" ---------- ")
print("===== Final output of the converted document =======")
# Manual export for illustration. Below, `save_as_json()` writes the same
# JSON again; kept intentionally to show both approaches.
with (out_path / f"{fname}.json").open("w") as fp:
fp.write(json.dumps(res.document.export_to_dict()))
res.document.save_as_json(
out_path / f"{fname}.json",
image_mode=ImageRefMode.PLACEHOLDER,
)
print(f" => produced {out_path / fname}.json")
res.document.save_as_markdown(
out_path / f"{fname}.md",
image_mode=ImageRefMode.PLACEHOLDER,
)
print(f" => produced {out_path / fname}.md")
res.document.save_as_html(
out_path / f"{fname}.html",
image_mode=ImageRefMode.EMBEDDED,
labels=[*DEFAULT_EXPORT_LABELS, DocItemLabel.FOOTNOTE],
split_page_view=True,
)
print(f" => produced {out_path / fname}.html")
pg_num = res.document.num_pages()
print("")
print(
f"Total document prediction time: {inference_time:.2f} seconds, pages: {pg_num}"
)
print("====================================================")
return [
source,
model_id,
str(framework),
pg_num,
inference_time,
]
if __name__ == "__main__":
sources = [
"tests/data/pdf/2305.03393v1-pg9.pdf",
]
out_path = Path("scratch")
out_path.mkdir(parents=True, exist_ok=True)
## Definiton of more inline models
llava_qwen = InlineVlmOptions(
repo_id="llava-hf/llava-interleave-qwen-0.5b-hf",
# prompt="Read text in the image.",
prompt="Convert this page to markdown. Do not miss any text and only output the bare markdown!",
# prompt="Parse the reading order of this document.",
response_format=ResponseFormat.MARKDOWN,
inference_framework=InferenceFramework.TRANSFORMERS,
transformers_model_type=TransformersModelType.AUTOMODEL_IMAGETEXTTOTEXT,
supported_devices=[AcceleratorDevice.CUDA, AcceleratorDevice.CPU],
scale=2.0,
temperature=0.0,
)
# Note that this is not the expected way of using the Dolphin model, but it shows the usage of a raw prompt.
dolphin_oneshot = InlineVlmOptions(
repo_id="ByteDance/Dolphin",
prompt="Read text in the image.