请教关于torch.load报错问题

[postman0001]

(MinerU) PS C:\Users\49631> magic-pdf -p small_ocr.pdf -o ./output
2025-05-31 15:59:53.551 | INFO | magic_pdf.data.dataset:init:157 - lang: None
2025-05-31 15:59:54.345 | INFO | magic_pdf.libs.pdf_check:detect_invalid_chars:67 - cid_count: 0, text_len: 8, cid_chars_radio: 0.0
2025-05-31 15:59:54.347 | WARNING | magic_pdf.filter.pdf_classify_by_type:classify:335 - OCR needed based on classification result, by_image_area: False, by_text: False, by_avg_words: False, by_img_num: True, by_img_narrow_strips: False, by_invalid_chars: True
2025-05-31 15:59:54.926 | INFO | magic_pdf.model.doc_analyze_by_custom_model:doc_analyze:162 - Batch 1/1: 8 pages/8 pages
2025-05-31 15:59:54.930 | INFO | magic_pdf.model.doc_analyze_by_custom_model:may_batch_image_analyze:278 - gpu_memory: 12 GB, batch_ratio: 8
2025-05-31 15:59:54.934 | INFO | magic_pdf.model.pdf_extract_kit:init:68 - DocAnalysis init, this may take some times, layout_model: doclayout_yolo, apply_formula: True, apply_ocr: True, apply_table: True, table_model: rapid_table, lang: None
2025-05-31 15:59:54.934 | INFO | magic_pdf.model.pdf_extract_kit:init:82 - using device: cuda
2025-05-31 15:59:54.934 | INFO | magic_pdf.model.pdf_extract_kit:init:86 - using models_dir: C:\Users\49631.cache\modelscope\hub\models\opendatalab\PDF-Extract-Kit-1___0\models
2025-05-31 16:00:05.001 | INFO | magic_pdf.model.pdf_extract_kit:init:164 - DocAnalysis init done!
2025-05-31 16:00:05.001 | INFO | magic_pdf.model.doc_analyze_by_custom_model:custom_model_init:116 - model init cost: 10.0718514919281
Layout Predict: 100%|████████████████████████████████████████████████████████████████████| 8/8 [00:01<00:00, 7.04it/s]
MFD Predict: 100%|███████████████████████████████████████████████████████████████████████| 8/8 [00:01<00:00, 5.72it/s]
MFR Predict: 0%| | 0/6 [00:00<?, ?it/s]C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\model\sub_modules\mfr\unimernet\unimernet_hf\unimer_mbart\modeling_unimer_mbart.py:653: UserWarning: 1Torch was not compiled with flash attention. (Triggered internally at ..\aten\src\ATen\native\transformers\cuda\sdp_utils.cpp:455.)
attn_output = torch.nn.functional.scaled_dot_product_attention(
MFR Predict: 100%|███████████████████████████████████████████████████████████████████████| 6/6 [00:00<00:00, 8.57it/s]
OCR-det Predict: 100%|███████████████████████████████████████████████████████████████████| 8/8 [00:04<00:00, 1.96it/s]
Table Predict: 0it [00:00, ?it/s]
OCR-rec Predict: 100%|███████████████████████████████████████████████████████████████| 233/233 [00:04<00:00, 57.71it/s]
Processing pages: 0%| | 0/8 [00:00<?, ?it/s]
2025-05-31 16:00:16.722 | ERROR | magic_pdf.tools.cli:parse_doc:134 - Due to a serious vulnerability issue in torch.load, even with weights_only=True, we now require users to upgrade torch to at least v2.6 in order to use the function. This version restriction does not apply when loading files with safetensors.
See the vulnerability report here https://nvd.nist.gov/vuln/detail/CVE-2025-32434
Traceback (most recent call last):

File "C:\Users\49631.conda\envs\MinerU\lib\runpy.py", line 196, in _run_module_as_main
return run_code(code, main_globals, None,
│ │ └ {'name': 'main', 'doc': None, 'package': '', 'loader': <zipimporter object "C:\Users\49631.conda\envs\Mi...
│ └ <code object at 0x0000016B68ABF9F0, file "C:\Users\49631.conda\envs\MinerU\Scripts\magic-pdf.exe_main.py", line 1>
└ <function _run_code at 0x0000016B687FF0A0>

File "C:\Users\49631.conda\envs\MinerU\lib\runpy.py", line 86, in run_code
exec(code, run_globals)
│ └ {'name': 'main', 'doc': None, 'package': '', 'loader': <zipimporter object "C:\Users\49631.conda\envs\Mi...
└ <code object at 0x0000016B68ABF9F0, file "C:\Users\49631.conda\envs\MinerU\Scripts\magic-pdf.exe_main.py", line 1>

File "C:\Users\49631.conda\envs\MinerU\Scripts\magic-pdf.exe_main_.py", line 7, in
sys.exit(cli())
│ │ └
│ └
└ <module 'sys' (built-in)>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\click\core.py", line 1442, in call
return self.main(*args, **kwargs)
│ │ │ └ {}
│ │ └ ()
│ └ <function Command.main at 0x0000016B6A7B7E20>
└

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\click\core.py", line 1363, in main
rv = self.invoke(ctx)
│ │ └ <click.core.Context object at 0x0000016B68B0C8B0>
│ └ <function Command.invoke at 0x0000016B6A7B7D00>
└

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\click\core.py", line 1226, in invoke
return ctx.invoke(self.callback, **ctx.params)
│ │ │ │ │ └ {'path': 'small_ocr.pdf', 'output_dir': './output', 'method': 'auto', 'lang': None, 'debug_able': False, 'start_page_id': 0, ...
│ │ │ │ └ <click.core.Context object at 0x0000016B68B0C8B0>
│ │ │ └ <function cli at 0x0000016B1D1CF370>
│ │ └
│ └ <function Context.invoke at 0x0000016B6A7B70A0>
└ <click.core.Context object at 0x0000016B68B0C8B0>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\click\core.py", line 794, in invoke
return callback(*args, **kwargs)
│ │ └ {'path': 'small_ocr.pdf', 'output_dir': './output', 'method': 'auto', 'lang': None, 'debug_able': False, 'start_page_id': 0, ...
│ └ ()
└ <function cli at 0x0000016B1D1CF370>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\tools\cli.py", line 155, in cli
parse_doc(Path(path))
│ │ └ 'small_ocr.pdf'
│ └ <class 'pathlib.Path'>
└ <function cli..parse_doc at 0x0000016B1D1CEEF0>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\tools\cli.py", line 121, in parse_doc
do_parse(
└ <function do_parse at 0x0000016B1D1CE8C0>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\tools\common.py", line 284, in do_parse
_do_parse(output_dir, pdf_file_name, pdf_bytes_or_dataset, model_list, parse_method, debug_able, start_page_id=start_page_id, end_page_id=end_page_id, lang=lang, layout_model=layout_model, formula_enable=formula_enable, table_enable=table_enable, f_draw_span_bbox=f_draw_span_bbox, f_draw_layout_bbox=f_draw_layout_bbox, f_dump_md=f_dump_md, f_dump_middle_json=f_dump_middle_json, f_dump_model_json=f_dump_model_json, f_dump_orig_pdf=f_dump_orig_pdf, f_dump_content_list=f_dump_content_list, f_make_md_mode=f_make_md_mode, f_draw_model_bbox=f_draw_model_bbox, f_draw_line_sort_bbox=f_draw_line_sort_bbox, f_draw_char_bbox=f_draw_char_bbox)
│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ └ False
│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ └ False
│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ └ False
│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ └ 'mm_markdown'
│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ └ True
│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ └ True
│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ └ True
│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ └ True
│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ └ True
│ │ │ │ │ │ │ │ │ │ │ │ │ │ └ True
│ │ │ │ │ │ │ │ │ │ │ │ │ └ True
│ │ │ │ │ │ │ │ │ │ │ │ └ None
│ │ │ │ │ │ │ │ │ │ │ └ None
│ │ │ │ │ │ │ │ │ │ └ None
│ │ │ │ │ │ │ │ │ └ None
│ │ │ │ │ │ │ │ └ None
│ │ │ │ │ │ │ └ 0
│ │ │ │ │ │ └ False
│ │ │ │ │ └ 'auto'
│ │ │ │ └ []
│ │ │ └ b'%PDF-1.7\r\n%\xa1\xb3\xc5\xd7\r\n1 0 obj\r\n<</Pages 2 0 R /Type/Catalog>>\r\nendobj\r\n2 0 obj\r\n<</Count 8/Kids[ 4 0 R ...
│ │ └ 'small_ocr'
│ └ './output'
└ <function _do_parse at 0x0000016B1D1CE830>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\tools\common.py", line 139, in _do_parse
pipe_result = infer_result.pipe_ocr_mode(
│ └ <function InferenceResult.pipe_ocr_mode at 0x0000016B2B5FE560>
└ <magic_pdf.operators.models.InferenceResult object at 0x0000016B2D696830>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\operators\models.py", line 144, in pipe_ocr_mode
res = self.apply(
│ └ <function InferenceResult.apply at 0x0000016B2B5FE440>
└ <magic_pdf.operators.models.InferenceResult object at 0x0000016B2D696830>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\operators\models.py", line 70, in apply
return proc(copy.deepcopy(self._infer_res), *args, **kwargs)
│ │ │ │ │ │ └ {'start_page_id': 0, 'end_page_id': None, 'debug_mode': True, 'lang': None}
│ │ │ │ │ └ (<magic_pdf.data.dataset.PymuDocDataset object at 0x0000016B1D1CB790>, <magic_pdf.data.data_reader_writer.filebase.FileBasedD...
│ │ │ │ └ [{'layout_dets': [{'category_id': 1, 'poly': [126, 395, 1045, 395, 1045, 911, 126, 911], 'score': 0.977}, {'category_id': 1, ...
│ │ │ └ <magic_pdf.operators.models.InferenceResult object at 0x0000016B2D696830>
│ │ └ <function deepcopy at 0x0000016B6A9FC550>
│ └ <module 'copy' from 'C:\Users\49631\.conda\envs\MinerU\lib\copy.py'>
└ <function InferenceResult.pipe_ocr_mode..proc at 0x0000016BA60848B0>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\operators\models.py", line 137, in proc
res = pdf_parse_union(*args, **kwargs)
│ │ └ {'start_page_id': 0, 'end_page_id': None, 'debug_mode': True, 'lang': None}
│ └ ([{'layout_dets': [{'category_id': 1, 'poly': [126, 395, 1045, 395, 1045, 911, 126, 911], 'score': 0.977, 'bbox': [126, 395, ...
└ <function pdf_parse_union at 0x0000016B2B5FE170>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\pdf_parse_union_core_v2.py", line 955, in pdf_parse_union
page_info = parse_page_core(
└ <function parse_page_core at 0x0000016B2B5FE0E0>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\pdf_parse_union_core_v2.py", line 873, in parse_page_core
sorted_bboxes = sort_lines_by_model(fix_blocks, page_w, page_h, line_height, footnote_blocks)
│ │ │ │ │ └ []
│ │ │ │ └ 32.5
│ │ │ └ 1800.0
│ │ └ 1225.5
│ └ [{'type': 'text', 'bbox': [141, 237, 1045, 385], 'lines': [{'bbox': [145, 246, 1042, 278], 'spans': [{'bbox': [145, 246, 1042...
└ <function sort_lines_by_model at 0x0000016B2B5FDD80>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\pdf_parse_union_core_v2.py", line 565, in sort_lines_by_model
model = model_manager.get_model('layoutreader')
│ └ <function ModelSingleton.get_model at 0x0000016B2B5FDBD0>
└ <magic_pdf.pdf_parse_union_core_v2.ModelSingleton object at 0x0000016B670AAD10>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\pdf_parse_union_core_v2.py", line 372, in get_model
self._models[model_name] = model_init(model_name=model_name)
│ │ │ │ └ 'layoutreader'
│ │ │ └ <function model_init at 0x0000016B2B5FDA20>
│ │ └ 'layoutreader'
│ └ {}
└ <magic_pdf.pdf_parse_union_core_v2.ModelSingleton object at 0x0000016B670AAD10>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\magic_pdf\pdf_parse_union_core_v2.py", line 341, in model_init
model = LayoutLMv3ForTokenClassification.from_pretrained(
│ └ <classmethod(<function PreTrainedModel.from_pretrained at 0x0000016B34AA05E0>)>
└ <class 'transformers.models.layoutlmv3.modeling_layoutlmv3.LayoutLMv3ForTokenClassification'>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\transformers\modeling_utils.py", line 309, in _wrapper
return func(*args, **kwargs)
│ │ └ {}
│ └ (<class 'transformers.models.layoutlmv3.modeling_layoutlmv3.LayoutLMv3ForTokenClassification'>, 'C:\Users\49631\.cache\mo...
└ <function PreTrainedModel.from_pretrained at 0x0000016B34AA0670>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\transformers\modeling_utils.py", line 4574, in from_pretrained
) = cls._load_pretrained_model(
│ └ <classmethod(<function PreTrainedModel._load_pretrained_model at 0x0000016B34AA3A30>)>
└ <class 'transformers.models.layoutlmv3.modeling_layoutlmv3.LayoutLMv3ForTokenClassification'>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\transformers\modeling_utils.py", line 4833, in _load_pretrained_model
load_state_dict(checkpoint_files[0], map_location="meta", weights_only=weights_only).keys()
│ │ └ True
│ └ ['C:\Users\49631\.cache\modelscope\hub\models\ppaanngggg\layoutreader\pytorch_model.bin']
└ <function load_state_dict at 0x0000016B34AA0AF0>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\transformers\modeling_utils.py", line 554, in load_state_dict
check_torch_load_is_safe()
└ <function check_torch_load_is_safe at 0x0000016B2EF06560>

File "C:\Users\49631.conda\envs\MinerU\lib\site-packages\transformers\utils\import_utils.py", line 1417, in check_torch_load_is_safe
raise ValueError(

ValueError: Due to a serious vulnerability issue in torch.load, even with weights_only=True, we now require users to upgrade torch to at least v2.6 in order to use the function. This version restriction does not apply when loading files with safetensors.
See the vulnerability report here https://nvd.nist.gov/vuln/detail/CVE-2025-32434
我的安装参数是： pip install --force-reinstall torch==2.3.1 torchvision==0.18.1 --index-url https://download.pytorch.org/whl/cu118
python 装的3.10版本 cuda是11.8版本 magic-pdf的version是1.3.12

[myhloli]

需要更新torch到2.6以上

[postman0001]

感谢回复，一开始用的是gitee里面看到的教程，cuda装的11.8的。后面用这边的装12.6就没问题了。就是有个小问题，用GPU的情况下 不能装cpu版本的Torch相关插件