initial reference to granite-doclong

Signed-off-by: Peter Staar <taa@zurich.ibm.com>
This commit is contained in:
Peter Staar 2025-06-20 07:47:12 +02:00
parent 64ac043786
commit 90da15f611
2 changed files with 32 additions and 0 deletions

View File

@ -50,6 +50,7 @@ from docling.datamodel.settings import settings
from docling.datamodel.vlm_model_specs import (
GRANITE_VISION_OLLAMA,
GRANITE_VISION_TRANSFORMERS,
GRANITEDOCLING_TRANSFORMERS,
SMOLDOCLING_MLX,
SMOLDOCLING_TRANSFORMERS,
VlmModelType,
@ -595,6 +596,20 @@ def convert( # noqa: C901
"To run SmolDocling faster, please install mlx-vlm:\n"
"pip install mlx-vlm"
)
elif vlm_model == VlmModelType.GRANITE_DOCLING:
pipeline_options.vlm_options = GRANITEDOCLING_TRANSFORMERS
"""
if sys.platform == "darwin":
try:
import mlx_vlm
pipeline_options.vlm_options = GRANITEDOCLING_MLX
except ImportError:
_log.warning(
"To run SmolDocling faster, please install mlx-vlm:\n"
"pip install mlx-vlm"
)
"""
pdf_format_option = PdfFormatOption(
pipeline_cls=VlmPipeline, pipeline_options=pipeline_options

View File

@ -43,6 +43,22 @@ SMOLDOCLING_TRANSFORMERS = InlineVlmOptions(
temperature=0.0,
)
# GraniteDocling
GRANITEDOCLING_TRANSFORMERS = InlineVlmOptions(
repo_id="ibm-granite/granite-docling-256m-preview",
prompt="Convert this page to docling.",
response_format=ResponseFormat.DOCTAGS,
inference_framework=InferenceFramework.TRANSFORMERS,
transformers_model_type=TransformersModelType.AUTOMODEL_VISION2SEQ,
supported_devices=[
AcceleratorDevice.CPU,
AcceleratorDevice.CUDA,
AcceleratorDevice.MPS,
],
scale=2.0,
temperature=0.0,
)
# GraniteVision
GRANITE_VISION_TRANSFORMERS = InlineVlmOptions(
repo_id="ibm-granite/granite-vision-3.2-2b",
@ -140,5 +156,6 @@ GEMMA3_27B_MLX = InlineVlmOptions(
class VlmModelType(str, Enum):
SMOLDOCLING = "smoldocling"
GRANITE_DOCLING = "granite_docling"
GRANITE_VISION = "granite_vision"
GRANITE_VISION_OLLAMA = "granite_vision_ollama"