mirrors/docling
1
0
Fork 0
mirror of https://github.com/DS4SD/docling.git synced 2025-12-11 22:28:31 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat: Support for Python 3.14 (#2530)

Browse Source 
* fix dependencies for py314

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* add metadata and CI tests

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* add back gliner

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* update error message about python 3.14 availability

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* skip tests which cannot run on py 3.14

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* fix lint

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* remove vllm from py 3.14 deps

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* safe import for vllm

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* update lock

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* remove torch.compile()

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* update checkbox results after docling-core changes

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* cannot run mlx example in CI

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* add test for rapidocr backends and skip onnxruntime on py3.14

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* fix other occurances of torch.compile()

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* allow torch.compile for Python <3.14. proper support will be introduced with new torch releases

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

---------

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>
This commit is contained in:
Michele Dolfi
2025-10-28 14:32:15 +01:00
committed by GitHub
parent 9a6fdf936b
commit cdffb47b9a
13 changed files with 2278 additions and 1361 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
docling/cli/main.py
View File

@@ -738,10 +738,15 @@ def convert( # noqa: C901
pipeline_options.vlm_options = SMOLDOCLING_MLX
except ImportError:
_log.warning(
"To run SmolDocling faster, please install mlx-vlm:\n"
"pip install mlx-vlm"
)
if sys.version_info < (3, 14):
_log.warning(
"To run SmolDocling faster, please install mlx-vlm:\n"
"pip install mlx-vlm"
)
else:
_log.warning(
"You can run SmolDocling faster with MLX support, but it is unfortunately not yet available on Python 3.14."
)
elif vlm_model == VlmModelType.GRANITEDOCLING:
pipeline_options.vlm_options = GRANITEDOCLING_TRANSFORMERS
@@ -751,10 +756,16 @@ def convert( # noqa: C901
pipeline_options.vlm_options = GRANITEDOCLING_MLX
except ImportError:
_log.warning(
"To run GraniteDocling faster, please install mlx-vlm:\n"
"pip install mlx-vlm"
)
if sys.version_info < (3, 14):
_log.warning(
"To run GraniteDocling faster, please install mlx-vlm:\n"
"pip install mlx-vlm"
)
else:
_log.warning(
"You can run GraniteDocling faster with MLX support, but it is unfortunately not yet available on Python 3.14."
)
elif vlm_model == VlmModelType.SMOLDOCLING_VLLM:
pipeline_options.vlm_options = SMOLDOCLING_VLLM

6
docling/models/picture_description_vlm_model.py
View File

@@ -1,3 +1,4 @@
import sys
import threading
from collections.abc import Iterable
from pathlib import Path
@@ -75,7 +76,10 @@ class PictureDescriptionVlmModel(
else "sdpa"
),
)
self.model = torch.compile(self.model) # type: ignore
if sys.version_info < (3, 14):
self.model = torch.compile(self.model) # type: ignore
else:
self.model.eval()
self.provenance = f"{self.options.repo_id}"

6
docling/models/vlm_models_inline/hf_transformers_model.py
View File

@@ -1,5 +1,6 @@
import importlib.metadata
import logging
import sys
import time
from collections.abc import Iterable
from pathlib import Path
@@ -129,7 +130,10 @@ class HuggingFaceTransformersVlmModel(BaseVlmPageModel, HuggingFaceModelDownload
trust_remote_code=vlm_options.trust_remote_code,
revision=vlm_options.revision,
)
self.vlm_model = torch.compile(self.vlm_model) # type: ignore
if sys.version_info < (3, 14):
self.vlm_model = torch.compile(self.vlm_model) # type: ignore
else:
self.vlm_model.eval()
# Load generation config
self.generation_config = GenerationConfig.from_pretrained(

11
docling/models/vlm_models_inline/mlx_model.py
View File

@@ -50,9 +50,14 @@ class HuggingFaceMlxModel(BaseVlmPageModel, HuggingFaceModelDownloadMixin):
from mlx_vlm.prompt_utils import apply_chat_template # type: ignore
from mlx_vlm.utils import load_config # type: ignore
except ImportError:
raise ImportError(
"mlx-vlm is not installed. Please install it via `pip install mlx-vlm` to use MLX VLM models."
)
if sys.version_info < (3, 14):
raise ImportError(
"mlx-vlm is not installed. Please install it via `pip install mlx-vlm` to use MLX VLM models."
)
else:
raise ImportError(
"mlx-vlm is not installed. It is not yet available on Python 3.14."
)
repo_cache_folder = vlm_options.repo_id.replace("/", "--")

6
docling/models/vlm_models_inline/nuextract_transformers_model.py
View File

@@ -1,4 +1,5 @@
import logging
import sys
import time
from collections.abc import Iterable
from pathlib import Path
@@ -153,7 +154,10 @@ class NuExtractTransformersModel(BaseVlmModel, HuggingFaceModelDownloadMixin):
),
trust_remote_code=vlm_options.trust_remote_code,
)
self.vlm_model = torch.compile(self.vlm_model) # type: ignore
if sys.version_info < (3, 14):
self.vlm_model = torch.compile(self.vlm_model) # type: ignore
else:
self.vlm_model.eval()
# Load generation config
self.generation_config = GenerationConfig.from_pretrained(artifacts_path)

14
docling/models/vlm_models_inline/vllm_model.py
View File

@@ -1,4 +1,5 @@
import logging
import sys
import time
from collections.abc import Iterable
from pathlib import Path
@@ -100,7 +101,18 @@ class VllmVlmModel(BaseVlmPageModel, HuggingFaceModelDownloadMixin):
return
from transformers import AutoProcessor
from vllm import LLM, SamplingParams
try:
from vllm import LLM, SamplingParams
except ImportError:
if sys.version_info < (3, 14):
raise ImportError(
"vllm is not installed. Please install it via `pip install vllm`."
)
else:
raise ImportError(
"vllm is not installed. It is not yet available on Python 3.14."
)
# Device selection
self.device = decide_device(

13
docling/pipeline/asr_pipeline.py
View File

@@ -1,6 +1,7 @@
import logging
import os
import re
import sys
import tempfile
from io import BytesIO
from pathlib import Path
@@ -117,9 +118,15 @@ class _NativeWhisperModel:
try:
import whisper # type: ignore
except ImportError:
raise ImportError(
"whisper is not installed. Please install it via `pip install openai-whisper` or do `uv sync --extra asr`."
)
if sys.version_info < (3, 14):
raise ImportError(
"whisper is not installed. Please install it via `pip install openai-whisper` or do `uv sync --extra asr`."
)
else:
raise ImportError(
"whisper is not installed. Unfortunately its dependencies are not yet available for Python 3.14."
)
self.asr_options = asr_options
self.max_tokens = asr_options.max_new_tokens
self.temperature = asr_options.temperature