🐛[BUG]: Import error: No module named 'physicsnemo.sym' after pip install the module

[ShawnLiaoyx]

Version

2.3.0

On which installation method(s) does this occur?

Pip

Describe the issue

Tried to run some examples under physicsnemo/examples/cfd/swe_nonlinear_pino that imports physicsnemo.sym. Program reported "Import error: No module named 'physicsnemo.sym'" even after installing the sym module.

The whole process:

1. git clone physicsnemo
2. under the root folder: uv sync
3. also under the root folder:

uv pip install torch torchvision
uv pip install Cython
uv pip install nvidia-physicsnemo-sym --no-build-isolation

4. installtion succeeded. uv pip show nvidia-physicsnemo-sym reports version number normally
5. run an example: uv run python examples/cfd/swe_nonlinear_pino/train_swe_nl_pino.py. Program aborted with following message:

Warp DeprecationWarning: The namespace `warp.context` will soon be removed from the public API. It can still be accessed from `warp._src.context` but might be changed or removed without notice.
Warp DeprecationWarning: The symbol `warp.context.Device` will soon be removed from the public API. Use `warp.Device` instead.
Traceback (most recent call last):
  File "/home/MY_USER_NAME/physicsnemo/examples/cfd/swe_nonlinear_pino/train_swe_nl_pino.py", line 33, in <module>
    from swe_nl_pde import SWE_NL
  File "/home/MY_USER_NAME/physicsnemo/examples/cfd/swe_nonlinear_pino/swe_nl_pde.py", line 17, in <module>
    from physicsnemo.sym.eq.pde import PDE
ModuleNotFoundError: No module named 'physicsnemo.sym'

### Minimum reproducible example

```shell
As described in the description.

Relevant log output

Environment details

running on WSL-Ubuntu (24.04 LTS). CUDA Toolkit version: 12.8.0

Other/Misc.

No response

[SinaAkhyani]

I tried the following installation commands and worked for me:

pip install --upgrade pip
pip install ninja
pip install torch==2.8.0 torchvision==0.23.0
pip install nvidia-physicsnemo
pip install nvidia-physicsnemo-sym --no-build-isolation

[ShawnLiaoyx]

I tried the following installation commands and worked for me:

pip install --upgrade pip pip install ninja pip install torch==2.8.0 torchvision==0.23.0 pip install nvidia-physicsnemo pip install nvidia-physicsnemo-sym --no-build-isolation

Thanks for replying. This way worked for me as well. The problem is that pip install packages are too old and are not compatible with examples right now. When installing the version as recommend by README.md, the problem I described in the issue happens.

I finally figured out that the error was because physicsnemo.sym installed from pip was not under physicsnemo, which was installed from uv. I added the following code to physicsnemo/__init__.py to solve the problem:

# in the beginning of physicsnemo/__init__.py
# to make physicsnemo.sym work
from pkgutil import extend_path
__path__ = extend_path(__path__, __name__)

Hopefully the dev team can help to make the installation of physicsnemo and physicsnemo.sym more unified. Incorporating my modification could be a way.

[ktangsali]

Closing as part of the archival of this repository. The key components have been upstreamed to NVIDIA/physicsnemo. Please refer to the v2.0 Migration Guide and feel free to re-open your issue there if it is still relevant.