使用Olive转换OpenAI Whisper模型为ONNX时遇导入错误求助

操作流程 #

尝试用Olive转换OpenAI Whisper模型为ONNX并合并为单文件，执行命令如下：

python prepare_whisper_configs.py --model_name openai/whisper-tiny.en
python -m olive.workflows.run --config whisper_cpu_fp32.json --setup
python -m olive.workflows.run --config whisper_cpu_fp32.json

错误信息 #

执行时触发导入错误：

Traceback (most recent call last):
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\runpy.py", line 86, in _run_code
exec(code, run_globals)
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\workflows\run_main.py", line 17, in
run(**vars(args))
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\workflows\run\run.py", line 187, in run
return engine.run(
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\engine\engine.py", line 347, in run
run_result = self.run_accelerator(
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\engine\engine.py", line 412, in run_accelerator
return self.run_no_search(
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\engine\engine.py", line 483, in run_no_search
should_prune, signal, model_ids = self._run_passes(
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\engine\engine.py", line 887, in _run_passes
model_config, model_id = self._run_pass(
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\engine\engine.py", line 985, in _run_pass
output_model_config = host.run_pass(p, input_model_config, data_root, output_model_path, pass_search_point)
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\systems\local.py", line 32, in run_pass
output_model = the_pass.run(model, data_root, output_model_path, point)
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\passes\olive_pass.py", line 367, in run
output_model = self._run_for_config(model, data_root, config, output_model_path)
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\passes\onnx\conversion.py", line 58, in _run_for_config
return self._convert_model_on_device(model, data_root, config, output_model_path, "cpu")
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\passes\onnx\conversion.py", line 73, in _convert_model_on_device
component_model = model.get_component(component_name)
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\model_init.py", line 696, in get_component
user_module_loader = UserModuleLoader(self.model_script, self.script_dir)
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\common\user_module_loader.py", line 22, in init
self.user_module = import_user_module(user_script, script_dir)
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\common\import_lib.py", line 39, in import_user_module
return import_module_from_file(user_script)
File "C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\common\import_lib.py", line 27, in import_module_from_file
spec.loader.exec_module(new_module)
File "", line 883, in exec_module
File "", line 241, in _call_with_frames_removed
File "C:\OpenAI\Olive\examples\whisper\code\user_script.py", line 10, in
from olive.model import PyTorchModelHandler
ImportError: cannot import name 'PyTorchModelHandler' from 'olive.model' (C:\Users\User\anaconda3\envs\OliveEnv\lib\site-packages\olive\model_init.py)

环境信息 #

使用Conda环境OliveEnv，已安装包列表：

(OliveEnv) C:\OpenAI\Olive\examples\whisper>pip list
Package                Version
---------------------- ----------
alembic                1.13.1
certifi                2024.2.2
charset-normalizer     3.3.2
colorama               0.4.6
coloredlogs            15.0.1
colorlog               6.8.2
contextlib2            21.6.0
contourpy              1.2.0
cycler                 0.12.1
Deprecated             1.2.14
filelock               3.13.1
flatbuffers            23.5.26
fonttools              4.48.1
fsspec                 2024.2.0
greenlet               3.0.3
huggingface-hub        0.20.3
humanfriendly          10.0
idna                   3.6
Jinja2                 3.1.3
joblib                 1.3.2
kiwisolver             1.4.5
Mako                   1.3.2
MarkupSafe             2.1.5
matplotlib             3.8.2
mpmath                 1.3.0
networkx               3.2.1
neural-compressor      2.4.1
numpy                  1.26.4
olive-ai               0.3.2
onnx                   1.15.0
onnxruntime            1.17.0
opencv-python-headless 4.9.0.80
optuna                 3.5.0
packaging              23.2
pandas                 2.2.0
pillow                 10.2.0
pip                    23.3.1
prettytable            3.9.0
protobuf               3.20.3
psutil                 5.9.8
py-cpuinfo             9.0.0
pycocotools            2.0.7
pydantic               1.10.14
pyparsing              3.1.1
pyreadline3            3.4.1
python-dateutil        2.8.2
pytz                   2024.1
PyYAML                 6.0.1
regex                  2023.12.25
requests               2.31.0
safetensors            0.4.2
schema                 0.7.5
scikit-learn           1.4.0
scipy                  1.12.0
setuptools             68.2.2
six                    1.16.0
SQLAlchemy             2.0.25
sympy                  1.12
tabulate               0.9.0
threadpoolctl          3.2.0
tokenizers             0.15.1
torch                  2.2.0
torchmetrics           0.10.3
tqdm                   4.66.1
transformers           4.37.2
typing_extensions      4.9.0
tzdata                 2023.4
urllib3                2.2.0
wcwidth                0.2.13
wheel                  0.41.2
wrapt                  1.16.0

已尝试的操作 #

先后测试过Olive的0.3.1、0.3.2、0.4.0三个版本，问题均存在。

解决方案 #

1. 修正导入路径 #

PyTorchModelHandler在Olive 0.3+版本中已迁移到olive.model.handlers模块，修改C:\OpenAI\Olive\examples\whisper\code\user_script.py第10行的导入语句：

# 替换原导入语句
from olive.model.handlers import PyTorchModelHandler

若使用Olive 0.4.0，该类已更名为PyTorchModel，需改为：

from olive.model import PyTorchModel

同时需要同步修改脚本中所有使用PyTorchModelHandler的地方为PyTorchModel。

2. 匹配示例代码与Olive版本 #

拉取对应版本的Whisper示例代码，避免版本不兼容：

• 若用0.3.2版本，切换到仓库的v0.3.2标签；
• 若用0.4.0版本，切换到v0.4.0标签，确保示例代码与安装的Olive版本一致。

3. 清理缓存重新安装 #

卸载现有Olive并重新安装指定版本，避免缓存干扰：

pip uninstall olive-ai -y
# 安装0.4.0版本为例
pip install olive-ai==0.4.0 --no-cache-dir

4. 验证依赖兼容性 #

确保Torch、Transformers等依赖版本与Olive官方推荐版本匹配，Olive 0.4.0推荐Torch 2.0+、Transformers 4.30+，当前环境的Torch 2.2.0和Transformers 4.37.2符合要求，无需调整。

内容的提问来源于stack exchange，提问作者Rusty Nail