OSError: [WinError 126] 找不到指定的模块。 Error loading "E:\programs\python\Python312\Lib\site-packages\intel_extension_for_pytorch\bin\intel-ext-pt-gpu.dll" or one of its dependencies.

[durongze]

Describe the bug

OSError: [WinError 126] 找不到指定的模块。 Error loading "E:\programs\python\Python312\Lib\site-packages\intel_extension_for_pytorch\bin\intel-ext-pt-gpu.dll" or one of its dependencies.

Versions

Traceback (most recent call last):
File "E:\programs\python\Python312\Lib\site-packages\torch_init_.py", line 2756, in import_device_backends
entrypoint = backend_extension.load()
^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\programs\python\Python312\Lib\importlib\metadata_init.py", line 205, in load
module = import_module(match.group('module'))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\programs\python\Python312\Lib\importlib_init_.py", line 90, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "", line 1387, in _gcd_import
File "", line 1360, in _find_and_load
File "", line 1331, in _find_and_load_unlocked
File "", line 935, in _load_unlocked
File "", line 999, in exec_module
File "", line 488, in call_with_frames_removed
File "E:\programs\python\Python312\Lib\site-packages\intel_extension_for_pytorch_init.py", line 124, in
raise err
OSError: [WinError 126] 找不到指定的模块。 Error loading "E:\programs\python\Python312\Lib\site-packages\intel_extension_for_pytorch\bin\intel-ext-pt-gpu.dll" or one of its dependencies.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
File "", line 1, in
File "E:\programs\python\Python312\Lib\site-packages\torch_init_.py", line 2784, in
import_device_backends()
File "E:\programs\python\Python312\Lib\site-packages\torch_init.py", line 2760, in _import_device_backends
raise RuntimeError(
RuntimeError: Failed to load the backend extension: intel_extension_for_pytorch. You can disable extension auto-loading with TORCH_DEVICE_BACKEND_AUTOLOAD=0.

[feng-intel]

Could you provide your code and steps how did you have this issue?

[rkilchmn]

I have same with ipex-llm[xpu_2.6], it happens when transformer is imported.
Have Intel iGPU Gen11 on Win11

#!/usr/bin/env python3
import os
import sys
import argparse
HERE: from transformers import pipeline

my requirements.txt:
transformers
--extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
ipex-llm[xpu_2.6]
torch
intel-extension-for-pytorch

There errorL

.conda\python.exe whisper.py -m distil-whisper/distil-large-v3 -q 4-bit -d xpu -P 8081
Traceback (most recent call last):
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\torch_init_.py", line 2756, in import_device_backends
entrypoint = backend_extension.load()
^^^^^^^^^^^^^^^^^^^^^^^^
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\importlib\metadata_init.py", line 205, in load
module = import_module(match.group('module'))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\importlib_init_.py", line 90, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "", line 1387, in _gcd_import
File "", line 1360, in _find_and_load
File "", line 1331, in _find_and_load_unlocked
File "", line 935, in _load_unlocked
File "", line 999, in exec_module
File "", line 488, in call_with_frames_removed
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\intel_extension_for_pytorch_init.py", line 124, in
raise err
OSError: [WinError 126] The specified module could not be found. Error loading "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\intel_extension_for_pytorch\bin\intel-ext-pt-gpu.dll" or one of its dependencies.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
File "c:\Users\Documents\project\openedai-whisper-ipex-llm\whisper.py", line 6, in
from transformers import pipeline
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers_init_.py", line 26, in
from . import dependency_versions_check
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers\dependency_versions_check.py", line 16, in
from .utils.versions import require_version, require_version_core
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers\utils_init_.py", line 25, in
from .chat_template_utils import DocstringParsingException, TypeHintParsingException, get_json_schema
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers\utils\chat_template_utils.py", line 40, in
from torch import Tensor
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\torch_init_.py", line 2784, in
import_device_backends()
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\torch_init.py", line 2760, in _import_device_backends
raise RuntimeError(
RuntimeError: Failed to load the backend extension: intel_extension_for_pytorch. You can disable extension auto-loading with TORCH_DEVICE_BACKEND_AUTOLOAD=0

[ZailiWang]

Would you try with the suggestion here?

[rkilchmn]

I tried: TORCH_DEVICE_BACKEND_AUTOLOAD=0 and got a bit further (some dependencies were missing).
But after that now again I ended up here. So it is same error, but a bit later in the execution

(c:\Users\Documents\project\openedai-whisper-ipex-llm.conda) c:\Users\Documents\project\openedai-whisper-ipex-llm>.conda\python.exe whisper.py -m distil-whisper/distil-large-v3 -q 4-bit -d xpu -P 8081
Traceback (most recent call last):
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers\utils\import_utils.py", line 1967, in get_module
return importlib.import_module("." + module_name, self.name)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\importlib_init.py", line 90, in import_module
return bootstrap.gcd_import(name[level:], package, level)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "", line 1387, in gcd_import
File "", line 1360, in find_and_load
File "", line 1331, in find_and_load_unlocked
File "", line 935, in load_unlocked
File "", line 999, in exec_module
File "", line 488, in call_with_frames_removed
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers\pipelines_init.py", line 49, in
from .audio_classification import AudioClassificationPipeline
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers\pipelines\audio_classification.py", line 21, in
from .base import Pipeline, build_pipeline_init_args
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers\pipelines\base.py", line 69, in
from ..modeling_utils import PreTrainedModel
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers\modeling_utils.py", line 158, in
import deepspeed
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed_init.py", line 25, in
from . import ops
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\ops_init.py", line 6, in
from . import adam
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\ops\adam_init.py", line 6, in
from .cpu_adam import DeepSpeedCPUAdam
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\ops\adam\cpu_adam.py", line 8, in
from deepspeed.utils import logger
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\utils_init.py", line 10, in
from .groups import *
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\utils\groups.py", line 28, in
from deepspeed import comm as dist
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\comm_init.py", line 7, in
from .comm import *
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\comm\comm.py", line 31, in
from deepspeed.comm.ccl import CCLBackend
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\comm\ccl.py", line 11, in
from deepspeed.ops.op_builder import NotImplementedBuilder
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\ops\op_builder_init.py", line 53, in
this_module.dict[member_name] = builder_closure(member_name)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\ops\op_builder_init_.py", line 41, in builder_closure
builder = get_accelerator().get_op_builder(member_name)
^^^^^^^^^^^^^^^^^
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\deepspeed\accelerator\real_accelerator.py", line 133, in get_accelerator
import intel_extension_for_pytorch as ipex
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\intel_extension_for_pytorch_init_.py", line 124, in
raise err
OSError: [WinError 126] The specified module could not be found. Error loading "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\intel_extension_for_pytorch\bin\intel-ext-pt-gpu.dll" or one of its dependencies.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
File "c:\Users\Documents\project\openedai-whisper-ipex-llm\whisper.py", line 12, in
from transformers import pipeline
File "", line 1412, in _handle_fromlist
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers\utils\import_utils.py", line 1955, in getattr
module = self._get_module(self._class_to_module[name])
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\transformers\utils\import_utils.py", line 1969, in _get_module
raise RuntimeError(
RuntimeError: Failed to import transformers.pipelines because of the following error (look up to see its traceback):
[WinError 126] The specified module could not be found. Error loading "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\intel_extension_for_pytorch\bin\intel-ext-pt-gpu.dll" or one of its dependencies.

[ZailiWang]

would you try with conda install libuv ?

[durongze]

我明天试一下，感谢回复。

[rkilchmn]

reinstalled everything with "conda install libuv" - still same error
File "c:\Users\Documents\project\openedai-whisper-ipex-llm.conda\Lib\site-packages\intel_extension_for_pytorch\bin\intel-ext-pt-gpu.dll" is in dir:

Mode LastWriteTime Length Name

---

-a---- 9/04/2025 8:34 AM 3006976 esimd_kernels.dll
-a---- 9/04/2025 8:34 AM 264237568 intel-ext-pt-gpu.dll
-a---- 9/04/2025 8:34 AM 460288 intel-ext-pt-python.dll
-a---- 9/04/2025 8:34 AM 1920765440 xetla_gemm.dll
-a---- 9/04/2025 8:34 AM 17577472 xetla_XeHpc_qmode0_bf16_1_1_1_128_1_1_0_0.dll
-a---- 9/04/2025 8:34 AM 18438144 xetla_XeHpc_qmode0_bf16_1_1_1_256_1_1_0_0.dll
-a---- 9/04/2025 8:34 AM 20814848 xetla_XeHpc_qmode0_bf16_1_1_1_512_1_1_0_0.dll
-a---- 9/04/2025 8:34 AM 16517120 xetla_XeHpc_qmode0_fp16_1_1_1_128_1_1_0_0.dll
(continues)

[Yu-ppy]

I also encountered your problem. The reason is that you installed torch+cpu in your environment. Uninstall the original torch and reinstall torch+xpu according to the document command.

[rkilchmn]

I am a bit unclear how to install latest ipex-llm for xpu - would you have a link to the documentation?

Is libuv still required? - I was reading somewhere that it is not required anymore. If not required anymore, can venv be used?

What Python version(s) are compatible with 2.6?

[rkilchmn]

Hi,

i fond all information here: https://pytorch-extension.intel.com/installation?platform=gpu&version=v2.7.10%2Bxpu&os=windows&package=pip

Will close ticket! Thanks for everybody contributing.

[rkilchmn]

Can't close - someone please close