example docs is missing dependencies: ModuleNotFoundError: No module named 'onnxruntime'

[bmerkle]

Describe the bug
I am following the example doc in
https://learn.microsoft.com/en-us/azure/ai-foundry/foundry-local/how-to/how-to-compile-hugging-face-models?view=foundry-classic&tabs=PowerShell&source=docs

Unfortunately there seems to be some installation dependency missing in the tutorial docs.
when I try to invoke olive i get the following error:
ModuleNotFoundError: No module named 'onnxruntime'

To Reproduce

(venv) PS C:\work\microsoft\Foundry-Local> olive auto-opt --model_name_or_path meta-llama/Llama-3.2-1B-Instruct --trust_remote_code --output_path models/llama --device cpu --provider CPUExecutionProvider --use_ort_genai --precision int4 --log_level 1
Loading HuggingFace model from meta-llama/Llama-3.2-1B-Instruct
[2025-12-15 22:31:06,135] [INFO] [run.py:99:run_engine] Running workflow default_workflow
[2025-12-15 22:31:06,139] [WARNING] [run.py:111:run_engine] ORT log severity level configuration ignored since the module isn't installed.
[2025-12-15 22:31:06,144] [INFO] [cache.py:138:__init__] Using cache directory: C:\work\microsoft\Foundry-Local\.olive-cache\default_workflow
Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "C:\work\microsoft\Foundry-Local\venv\Scripts\olive.exe\__main__.py", line 7, in <module>
  File "C:\work\microsoft\Foundry-Local\venv\Lib\site-packages\olive\cli\launcher.py", line 66, in main
    service.run()
  File "C:\work\microsoft\Foundry-Local\venv\Lib\site-packages\olive\cli\auto_opt.py", line 173, in run
    return self._run_workflow()
           ^^^^^^^^^^^^^^^^^^^^
  File "C:\work\microsoft\Foundry-Local\venv\Lib\site-packages\olive\cli\base.py", line 44, in _run_workflow
    workflow_output = olive_run(run_config)
                      ^^^^^^^^^^^^^^^^^^^^^
  File "C:\work\microsoft\Foundry-Local\venv\Lib\site-packages\olive\workflows\run\run.py", line 178, in run
    return run_engine(package_config, run_config)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\work\microsoft\Foundry-Local\venv\Lib\site-packages\olive\workflows\run\run.py", line 131, in run_engine
    accelerator_spec = create_accelerator(
                       ^^^^^^^^^^^^^^^^^^^
  File "C:\work\microsoft\Foundry-Local\venv\Lib\site-packages\olive\systems\accelerator_creator.py", line 174, in create_accelerator
    system_config = normalizer.normalize()
                    ^^^^^^^^^^^^^^^^^^^^^^
  File "C:\work\microsoft\Foundry-Local\venv\Lib\site-packages\olive\systems\accelerator_creator.py", line 39, in normalize
    self.system_supported_eps = target.get_supported_execution_providers()
                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\work\microsoft\Foundry-Local\venv\Lib\site-packages\olive\systems\local.py", line 66, in get_supported_execution_providers
    return get_ort_available_providers()
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\work\microsoft\Foundry-Local\venv\Lib\site-packages\olive\common\ort_inference.py", line 72, in get_ort_available_providers
    import onnxruntime as ort
ModuleNotFoundError: No module named 'onnxruntime'

Expected behavior
following the instructions the example should work

Olive config
Add Olive configurations here.

Olive logs
Add logs here.

[bmerkle]

workaround/fix is to install onnxruntime. Please add this hint to the docs
pip install onnxruntime

[bmerkle]

if you post the URL on github for the docs, i can create a PR

[MaanavD]

@bmerkle https://github.com/MicrosoftDocs/azure-ai-docs/tree/main/articles/ai-foundry/foundry-local

Thanks for your contributions :)

[bmerkle]

@MaanavD a question before i make changes to the docs

the change would be in
https://github.com/MicrosoftDocs/azure-ai-docs/blob/main/articles/ai-foundry/foundry-local/how-to/how-to-compile-hugging-face-models.md

essentially I would add 2 things:

1. install missing dependencies

pip install olive-ai[auto-opt]
pip install transformers onnxruntime-genai

2. point the user to the repo where the recipes for olive models are collected
   https://github.com/microsoft/olive-recipes

I would explicitly mention this github repo as it contains a lot of usefull olive-recipes and also fixes my problem, with downloading and compiling the model. Is it ok to link to this github repo ?