ImportError: cannot import name CDM from pyHalo.preset_models (dependency/API mismatch)

[srividya-0001]

ImportError: cannot import name CDM from pyHalo.preset_models (dependency/API mismatch)

Hello maintainers,

While testing this repository on a clean Windows setup, I encountered a dependency/API mismatch that prevents running the Model_I simulation scripts.

Description

Running Model_I/sim_axion.py fails during import with:

ImportError: cannot import name 'CDM' from 'pyHalo.preset_models'

This happens before the script reaches execution, so it blocks cross-platform testing and reproducibility for new contributors.

Steps to Reproduce

1. Clone the repository
2. Create and activate a virtual environment (Windows) and install dependencies:

py -3.11 -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
pip install -e .

3. Run:

python Model_I/sim_axion.py

Observed Behavior

The script fails with:

ImportError: cannot import name 'CDM' from 'pyHalo.preset_models'

Environment:

• OS: Windows
• Python: 3.11

Installed pyHalo version:

python -c "import pyHalo; print(pyHalo.__version__)"

Output:

0.2.8

Additional check:

python -c "import pyHalo.preset_models as pm; print([x for x in dir(pm) if 'CDM' in x])"

Output:

[]

Expected Behavior

The simulation scripts should run successfully after installation.
Alternatively, the repository should pin a compatible pyHalo version and/or update the import statement according to the correct pyHalo API.

Thank you!

[Shrishagk]

Hi @srividya-0001 , I’ve addressed this in PR mwt5345/DeepLenseSim#1.

The fix resolves the import error by ensuring consistent initialization across models 1-3.
Please let me know if any additional changes are needed.

[srividya-0001]

Hi @Shrishagk, thanks for the update!

I re-tested on Windows (fresh venv, Python 3.11) after the changes in mwt5345/DeepLenseSim#1.
The hardcoded path issue from #99 looks resolved now.

However, end-to-end testing is still blocked by:
ImportError: cannot import name 'CDM' from pyHalo.preset_models

This occurs when running:
Model_I/sim_no_sub.py

Could you please confirm the expected pyHalo version/commit for this project, or should DeepLenseSim update the import to match the current pyHalo API?

[Shrishagk]

Hi! @srividya-0001 ,Thank you for the detailed feedback and for testing in a fresh environment.

I have updated the PR to resolve the ImportError you encountered. Here is a summary of the changes:

Version-Agnostic Imports: I modified deeplense/lens.py to use a try/except block. It now attempts to import CDM directly (for older pyHalo versions) and falls back to preset_model_from_name('CDM') for the newer API version you are using.

Path Portability: I've confirmed that all hardcoded absolute paths in the simulation scripts for Models I–IV have been replaced with os.path.join logic to ensure they work across Windows, Mac, and Linux.

Environment Sync: These changes should now match the current pyHalo factory pattern while remaining backwards compatible.

Please let me know if this allows the end-to-end testing to proceed! If you still encounter version-related blocks after these changes, it might be worth looping in the mentors to clarify the officially supported requirements.txt for the project, as there may be a larger API shift we need to align with.

[srividya-0001]

Hi @Shrishagk, great news — I re-tested on Windows (fresh venv, Python 3.11).

Your updated version-agnostic import logic (try/except + preset_model_from_name fallback) resolves the CDM import mismatch on Windows.
End-to-end execution now proceeds successfully (e.g., Model_I/sim_axion.py runs and completes sample generation).

Note: During testing on Windows, I found a few additional setup/dependency issues (e.g., matplotlib/mcfit not installed by default, and pyHalo 1.4.3 incompatibility with lenstronomy==1.9.2 due to missing pseudo_double_powerlaw). I opened an issue in ML4SCI/DeepLense to track this, and I also submitted a separate PR to mwt5345/DeepLenseSim to update setup.py dependencies for a more reproducible installation flow (your PR here is specifically the import/path fix PR).

Thanks again — this definitely unblocks the end-to-end testing!

I’m happy to help with any additional testing or fixes if needed — please feel free to tag me.